Transversal functional connectivity and scene-specific processing in the human entorhinal-hippocampal circuitry

  1. Xenia Grande  Is a corresponding author
  2. Magdalena M Sauvage
  3. Andreas Becke
  4. Emrah Düzel
  5. David Berron  Is a corresponding author
  1. Institute of Cognitive Neurology and Dementia Research, Otto von Guericke University Magdeburg, Germany
  2. German Center for Neurodegenerative Diseases, Germany
  3. Functional Architecture of Memory Department, Leibniz-Institute for Neurobiology, Germany
  4. Functional Neuroplasticity Department, Otto-von-Guericke University Magdeburg, Germany
  5. Center for Behavioral Brain Sciences, Otto-von-Guericke University Magdeburg, Germany
  6. Institute of Cognitive Neuroscience, University College London, United Kingdom
  7. Clinical Memory Research Unit, Department of Clinical Sciences Malmö, Lund University, Sweden

Abstract

Scene and object information reach the entorhinal-hippocampal circuitry in partly segregated cortical processing streams. Converging evidence suggests that such information-specific streams organize the cortical – entorhinal interaction and the circuitry’s inner communication along the transversal axis of hippocampal subiculum and CA1. Here, we leveraged ultra-high field functional imaging and advance Maass et al., 2015 who report two functional routes segregating the entorhinal cortex (EC) and the subiculum. We identify entorhinal subregions based on preferential functional connectivity with perirhinal Area 35 and 36, parahippocampal and retrosplenial cortical sources (referred to as ECArea35-based, ECArea36-based, ECPHC-based, ECRSC-based, respectively). Our data show specific scene processing in the functionally connected ECPHC-based and distal subiculum. Another route, that functionally connects the ECArea35-based and a newly identified ECRSC-based with the subiculum/CA1 border, however, shows no selectivity between object and scene conditions. Our results are consistent with transversal information-specific pathways in the human entorhinal-hippocampal circuitry, with anatomically organized convergence of cortical processing streams and a unique route for scene information. Our study thus further characterizes the functional organization of this circuitry and its information-specific role in memory function.

Editor's evaluation

Grande and colleagues provide important new insights into how different regions of the entorhinal cortex functionally interact with specific cortical brain areas and how, in turn, subregions of the entorhinal cortex interact with the hippocampus during 'scene' and 'object' processing. The study is well-motivated, well-designed, and provides convincing evidence using appropriate methodology. This paper is relevant to cognitive neuroscientists with an interest in the entorhinal cortex – hippocampal pathways and 'scene' and 'object' representation in the medial temporal lobe.

https://doi.org/10.7554/eLife.76479.sa0

Introduction

Entorhinal and hippocampal subregions form a critical functional circuitry that binds cortical information into cohesive representations (Eichenbaum et al., 2007; Ritchey et al., 2015). The interaction of the entorhinal-hippocampal circuitry with large-scale cortical information streams and the circuitry’s inner communication are key to the formation of these cohesive representations. Here, we advance insight into how the human entorhinal cortex (EC) receives information from cortical streams and how information proceeds between the EC and the transversal axis of hippocampal subiculum and CA1 (here referred to as transversal sub/CA1 axis). These insights are relevant to our understanding of the circuitry’s fundamental role in cognitive functions such as episodic memory.

Large-scale cortical information streams, that originate in the visual ‘Where’ and ‘What’ pathways and process scene and object information (Berron et al., 2018; Haxby et al., 1991; Ranganath and Ritchey, 2012; Ritchey et al., 2015; Ungerleider and Haxby, 1994), map onto the EC in a complex manner and define functional EC subregions. [Note, in light of confusing nomenclature, here we adhere to scene and object information - elsewhere referred to as contextual, spatial or "Where" and content, non-spatial, item or "What" information, respectively.] Recent rodent research updates the former conception of a parallel mapping of scene and object information via parahippocampal and perirhinal cortices onto medial versus lateral EC subregions (cf. posterior-medial versus anterior-lateral EC subregions as the human homologues; Maass et al., 2015; Navarro Schröder et al., 2015). Instead of a strict parallel mapping, profound cross-projections exist from the parahippocampal cortex towards the perirhinal cortex and the lateral EC (Nilssen et al., 2019). In accordance, information seems to converge in the rodent lateral EC (Doan et al., 2019). The update, thus, implies a more complex functional organization than parallel scene and object information mapping. Moreover, this advance highlights the retrosplenial cortex as an additional source to convey information directly from the cortical scene processing stream onto the EC. The retrosplenial cortex projects to the medial EC and, like the parahippocampal cortex, is part of the scene processing stream (e.g. involved in scene translation; Vann et al., 2009; Nilssen et al., 2019; Witter et al., 2017). The update, furthermore, evokes the question how cortical sources of information uniquely map onto the EC and which kind of information is processed in the resulting functional EC subregions.

Within the entorhinal-hippocampal circuitry, an important direct way of communication exists between the EC and hippocampal subiculum and CA1. How functional EC subregions communicate towards the transversal sub/CA1 axis in humans is, however, unclear. Similarly, the extent to which specific scene and object information processing routes might emerge, despite information convergence in the EC, is unknown. On one hand, rodent research indicates a transversal organization where scene and object information is processed along two anatomically wired routes, the medial EC – distal subiculum – proximal CA1 route and the lateral EC – proximal subiculum – distal CA1 route, respectively (Witter et al., 2017; note sparse functional evidence in the subiculum: Ku et al., 2017; Cembrowski et al., 2018; but frequent reports in the rodent CA1 region: Henriksen et al., 2010; Nakamura et al., 2013; Igarashi et al., 2014; Nakazawa et al., 2016; Beer et al., 2018). Initial functional and structural connectivity data also indicate such a transversal connectivity profile in humans (Maass et al., 2015; Syversen et al., 2021). In accordance, scene information seems to be preferentially processed in the distal subiculum (Dalton et al., 2018; Dalton and Maguire, 2017; Zeidman et al., 2015) and hints exist for preferential object processing at the subiculum/CA1 border (Dalton et al., 2018). On the other hand, anatomical projections in the monkey show a longitudinal profile on top of the transversal profile with mainly the anterior-lateral and posterior-lateral entorhinal portions projecting to the distal subiculum – proximal CA1 and proximal subiculum – distal CA1, respectively (Witter and Amaral, 2020). According to information convergence in the EC, a recent report finds convergence along the rodent transversal CA1 axis (Vandrey et al., 2021). In humans, visual stream projections towards the entorhinal-hippocampal circuitry similarly suggest convergence of scene and object information in the subiculum/CA1 border region but preserved scene processing in the distal subiculum (Dalton and Maguire, 2017). A detailed examination of the latter hypothesis is, however, lacking. The diversity of findings emerging from the literature calls for a thorough investigation to elucidate whether multiple transversal processing routes exist within the human entorhinal-hippocampal circuitry.

To summarize, our conception of how information travels towards the entorhinal-hippocampal circuitry underwent key changes which warrant an extensive exploration of the circuitry’s functional organization. First, rodent research shows that there is no strict parallel mapping of cortical information from the perirhinal and parahippocampal cortex towards the EC. Second, information seems to converge already before the hippocampus. These changes add to several knowledge gaps. First, it is unclear in which subregions of the entorhinal-hippocampal circuitry scene and object information are processed. The general connectivity patterns in the human entorhinal-hippocampal circuitry have not yet been directly related to information processing. Moreover, it is unclear how scene information from the retrosplenial cortex maps onto the human EC as a critical source of the cortical scene processing stream. Hence, it is also unclear how retrosplenial information is communicated between the EC and the hippocampus. Finally, it remains elusive whether a transversal functional segregation can be extended towards the human CA1 region in analogy to the rodent literature.

Here, we leverage ultra-high field 7 Tesla functional imaging (fMRI) data and advance the earlier findings on human entorhinal subregions and a transversal intrinsic functional connectivity pattern in the subiculum (Maass et al., 2015). With a combination of functional connectivity and information processing analyses, we seek to answer two sets of questions. Regarding functional connectivity, we ask where the parahippocampal, perirhinal, and retrosplenial cortical sources uniquely map onto the human EC and how these functionally connected routes continue between EC subregions and the transversal sub/CA1 axis. Regarding information processing, we ask whether and where scene and object information are specifically processed in the EC and along the transversal sub/CA1 axis. We test the hypotheses of (1) a transversal functional connectivity pattern and (2) multiple information processing routes within the entorhinal-hippocampal circuitry. Thus, following the updated conception of a non-parallel cortical scene and object information mapping onto the EC in rodents, we will show how cortical information streams map onto the EC in humans. This mapping will then be our detailed starting point to investigate the functional connectivity and information processing within the entorhinal-hippocampal circuitry.

Results

We seek to comprehensively investigate functional connectivity within the entorhinal-hippocampal circuitry and the contribution of cortical scene and object information processing streams. In an initial step, we identified where cortical sources map uniquely onto the entorhinal cortex (building upon Maass et al., 2015). The identified entorhinal subregions are based on their voxel’s preferred intrinsic functional connectivity with the retrosplenial cortex, parahippocampal cortex, perirhinal Area 35 or Area 36 regions (‘sources’). Note, that we evaluate both perirhinal subregions, Area 35 and Area 36 as separate sources as accumulating evidence suggests their structural and functional distinction (e.g. Berron et al., 2021; Burwell, 2000; Suzuki and Naya, 2014; van Strien et al., 2009). Next, we evaluated the continuation of the functional connectivity streams within the entorhinal-hippocampal circuitry and examined the intrinsic functional connectivity pattern between the identified entorhinal subregions (‘seeds’) and hippocampal subiculum and CA1 in the hippocampal body. Therefore, temporal fluctuations of BOLD signal were correlated in a seed-to-voxel manner within each participant. The resulting statistical correlational maps were aligned between participants. Repeated measures ANOVAs were calculated on connectivity preferences with seeds and transversal segments as factors to determine statistical differences in connectivity topography. All functional connectivity analyses were performed on the dataset after task-related effects have been regressed out, creating a dataset that resembles resting-state data (Gavrilescu et al., 2008; Maass et al., 2015).

Finally, we identified the corresponding bias in scene (here operationalized with room stimuli) and object information processing within entorhinal subregions and along the transversal sub/CA1 axis using the same dataset. Therefore, we extracted parameter estimates from a mnemonic discrimination task with scene and object conditions from aligned statistical maps across participants. Repeated measures ANOVAs were calculated on parameter estimates in entorhinal subregions and transversal sub/CA1 segments to determine biases in information processing within the entorhinal-hippocampal circuitry.

In the following, we first describe the four obtained entorhinal seeds and display the intrinsic functional connectivity pattern with the entorhinal seed regions along the transversal sub/CA1 axis. Thereafter, we report the information processing characteristics of the entorhinal and hippocampal subregions. Note that all results have been obtained with independent analyses in the left and right hemispheres. The largely similar left hemisphere results can be found in appendix 1. Source data and statistical maps are provided under Grande, X., Berron, D. (2022). Open Science Framework. ID 9v3qp. Source Data from Functional Connectivity and Information Processing in the Entorhinal-Hippocampal Circuitry. https://osf.io/9v3qp.

Four cortical sources divide the EC in retrosplenial-, parahippocampal, Area 35- and Area 36-based seeds

The four entorhinal subregions that we later used as seeds to determine the topography of entorhinal-hippocampal connectivity are based on intrinsic functional connectivity preferences with either the parahippocampal cortex, the retrosplenial cortex, perirhinal Area 36 or Area 35. These cortical regions are in general concordance with Maass et al., 2015 but consider recent advances that put forward the retrosplenial cortex as a critical source from the cortical scene processing stream (Nilssen et al., 2019) and evaluate perirhinal Area 35 and 36 separately.

Based on functional connectivity preferences with the four sources - parahippocampal cortex (Source code 8), retrosplenial cortex (Source code 7, Area 36 (Source code 6), and Area 35 (Source code 5) - we obtained four entorhinal seeds. The seeds refer to different parts of the EC whose voxels expressed preferential functional connectivity to either cortical source. For the ECPHC-based seed, the majority of voxels can roughly be described as clustering in the posterior-medial entorhinal portion, for the ECRSC-based seed in the anterior-medial portion, for the ECArea35-based seed in the anterior-lateral portion and for the ECArea36-based seed in the posterior-lateral entorhinal portion (see appendix 2 for exact voxel counts). Note that both perirhinal-based entorhinal seeds extended along the anterior to posterior axis such that the ECArea35-based progressed more along the outer EC (i.e. laterally, with a main focus anteriorly) and the ECArea36-based along the inner EC (i.e. medially, with a main focus posteriorly, see Figure 1 and the medial reflection of the EC seeds). It is important to note that these are rough qualitative descriptions of the main clusters, without quantification or an established relationship to coherent cytoarchitectonic regions. We will therefore continue to refer to them as ECRSC-based, ECPHC-based, ECArea35-based and ECArea36-based seeds.

Entorhinal seed regions based on connectivity preferences to cortical regions.

Displayed is the right EC as a 3D image with colored seed regions. The seed regions have been identified based on a source-to-voxel functional connectivity analysis and resulting connectivity preference to either the right retrosplenial cortex (RSC, green), parahippocampal cortex (PHC, blue), Area 36 (A36, purple), or Area 35 (A35, pink) sources. Note that preferences to Area 36 are best visible from a medial perspective on the EC as depicted in the medial reflection. Seed regions have been determined based on the thresholded (T>3.1) maximum voxels across four one-sample T-tests at group level, one per source, sample size n = 32. M – medial; L – lateral; A – anterior; P – posterior.

Distal subiculum is functionally connected with the ECPHC-based seed while the subiculum/CA1 border is connected with ECRSC-based and ECArea35-based seeds

Following the characterization of entorhinal seeds, we focused on the functional connectivity between these entorhinal subregions and hippocampal subiculum and CA1 to test the hypothesis of a transversal functional connectivity pattern. We predicted that while some EC subregions have a preference to functionally connect with the subiculum/CA1 border, others preferentially connect with the distal subiculum and proximal CA1. In the previous step, we identified EC subregions based on unique cortical source contributions. Therefore, our predictions remained in accordance with Maass et al., 2015: We expected that the EC subregion preferentially connected with the parahippocampal cortex (ECPHC-based seed) maps towards the distal subiculum and EC subregions connected with the perirhinal cortex (ECArea35-based seed, ECArea36-based seed) map towards the proximal subiculum, a mapping that we predicted to be extended towards the distal CA1.

When extracting estimates of connectivity preferences across individuals from proximal and distal hippocampal subfield segments for either entorhinal seed, repeated measures ANOVAs revealed significant seed X segments interaction effects along the transversal sub/CA1 axis (see Figure 2; subiculum: F(12,372) = 19.561; p<0.001; CA1: F(6,186) = 3.212; p=0.024).

Functional connectivity preferences to entorhinal seeds along the transversal axis of subiculum and CA1.

Displayed are the results of a seed-to-voxel functional connectivity analysis between the displayed right entorhinal seeds and the right subiculum and CA1 subregion. The 3D figure displays voxel-wise connectivity preferences to the entorhinal seeds (color coded to refer to the respective entorhinal seed [E]) on group level ([A] - subiculum; [B] - CA1; maps for connectivity preferences: Source code 13 - ECArea35-based, pink; Source code 14 - ECArea36-based, purple; Source code 16 - ECPHC-based, blue; Source code 15 - ECRSC-based seed, green). Note that preferences to the ECArea35-based seed (pink) are located mainly in the inferior subiculum and CA1 and are therefore best visible in the inferior reflection. To display mean connectivity preferences across participants along the transversal sub/CA1 axis, beta estimates were extracted and averaged from equally sized segments from proximal to distal ends (five segments in subiculum [A], three segments in CA1 [B]; schematized in white on the 3D figures) on each coronal slice and averaged along the longitudinal axis. Repeated measures ANOVAs revealed significant differences in connectivity estimates along the transversal axis of CA1 [D] and subiculum [C] with interaction effects in the subiculum. Displayed significances were obtained by FDR-corrected post-hoc tests and refer to p<0.05. Shaded areas in the graphs refer to standard errors of the mean, sample size n = 32. EC – entorhinal; M – medial; L – lateral; A – anterior; P – posterior; prox – proximal; dist – distal. Figure 2—source data 1 contains individual connectivity estimates per subregion (Sub – subiculum and CA1, respectively) and seed (ECRSC-based – RSCECseed, ECArea35-based – A35ECseed, ECPHC-based – PHCECseed, ECArea36-based – A36ECseed) for each transversal segment (1–5 or 1–3, respectively from proximal to distal).

Figure 2—source data 1

Individual functional connectivity estimates to right entorhinal seeds, extracted from right subiculum and CA1 transversal segments.

https://cdn.elifesciences.org/articles/76479/elife-76479-fig2-data1-v3.zip

In the subiculum, additional repeated measures ANOVAs showed that the ECArea35-based (F(4,124) = 8.913; pFDR <0.001), ECRSC-based (F(4,124) = 10.538; pFDR <0.001) and ECPHC-based (F(4,124) = 42.201; pFDR <0.001) seeds displayed a significant main effect across the transversal subiculum segments. These differential functional connectivity preferences across the transversal axis of the subiculum interacted significantly in a subsequent repeated measures ANOVA (ECPHC-based versus ECRSC-based seed preference interaction: F(4,124) = 46.452; pFDR <0.001; ECPHC-based versus ECArea35-based seed preference interaction: F(4,124) = 35.208; pFDR <0.001). This pattern provides statistical evidence for an increase in preferential functional connectivity with the ECPHC-based seed towards the distal portion of the subiculum while the preferential functional connectivity with the ECArea35-based as well as the ECRSC-based seeds rather increased towards the proximal portion of the subiculum.

In hippocampal CA1, additional repeated measures ANOVAs showed that the connectivity preference towards the ECRSC-based seed displays a significant main effect across the transversal axis of CA1 (F(2,62) = 10.489; pFDR <0.001). In distal CA1, the preferential functional connectivity with the ECRSC-based seed was higher than in the proximal portion of CA1. In right CA1, a similar but weaker transversal pattern was observed for connectivity preferences with the ECArea35-based (F(2,62) = 4.146; pFDR = 0.041; note in the left hemisphere a comparable transversal pattern was observed for the ECPHC-based and ECRSC-based portions, see appendix 1).

Thus, in the entorhinal-hippocampal circuitry, voxels in the distal subiculum were preferentially functionally connected with the ECPHC-based portion whereas voxels in the subiculum/CA1 border were preferentially connected with more anterior EC portions (ECRSC-based and ECArea35-based).

Distal subiculum and ECPHC-based exhibit higher functional activity in the scene condition while other subregions show no significant difference between conditions

Besides the intrinsic functional connectivity patterns within the entorhinal-hippocampal circuitry, we also examined the characteristics of scene and object information processing to test the hypothesis of multiple information processing routes within the entorhinal-hippocampal circuitry. We predicted a route of specific scene processing and another route of convergent information processing. Following the proposal by Dalton and Maguire, 2017 and the updated cross-projections from the scene to the object information processing stream (Nilssen et al., 2019), we expected scene processing in the distal subiculum. The updated parahippocampal cross-projections imply convergence wherever specific object processing had been expected previously. Thus, we explored whether any entorhinal-hippocampal subregions still process object information specifically. However, we largely expected to find evidence consistent with convergent processing of scene and object information within the entorhinal-hippocampal circuitry.

We first focused on the entorhinal seed regions. When extracting task-related parameter estimates from object and scene conditions, a repeated measures ANOVA showed a significant interaction between region and information type (object versus scene, F(3,93) = 20.9267; p<0.001). Post-hoc t-tests revealed that only in the ECPHC-based seed region functional activity in the scene condition was significantly higher than in the object condition (pFDR<0.001), while in the remaining three entorhinal seed regions no significant difference between scene and object conditions existed (ECArea35-based: pFDR = 0.9129; ECArea36-based: pFDR = 0.9129; ECRSC-based: pFDR = 0.5646; see Figure 3).

Functional activity during scene and object conditions in entorhinal seed regions.

Displayed are the extracted parameter estimates for the object condition versus baseline contrast (‘object information processing’, red) and the scene condition versus baseline contrast (‘scene information processing’, cyan) from each entorhinal seed region per individual (dots) and summarized across individuals (lines). A schematic depiction of the respective entorhinal seed regions is displayed by a 3D drawing of the right EC. A repeated measures ANOVA revealed a significant interaction between condition and seed region. The displayed significant difference is obtained with FDR-corrected post-hoc tests and refers to p<0.05. During the object condition, participants were presented with 3D rendered objects on screen, during the scene condition with 3D rendered indoor rooms and during the baseline condition they saw scrambled pictures. The shaded area around the lines refers to standard errors of the mean, sample size n = 32. EC – entorhinal cortex; M – medial; L – lateral; A – anterior; P – posterior. Figure 3—source data 1 contains extracted parameter values per individual and EC seed (isthmuscingulate – ECRSC-based, Area 35 – ECArea35-based, Area 36 – ECArea36-based, PHC – ECPHC-based seed) for the object versus baseline and scene versus baseline contrasts.

Figure 3—source data 1

Individual paramenter estimates for scene and object processing in right entorhinal seed regions.

https://cdn.elifesciences.org/articles/76479/elife-76479-fig3-data1-v3.zip

When extracting task-related parameter estimates for scene and object conditions from proximal and distal segments of hippocampal subregions within each participant, we found a significant interaction between transversal segments and information type only in the subiculum (F(4,124) = 15.994; p<0.001) and not in CA1 (F(2,62) = 2.553; p = 0.105) as revealed by repeated measures ANOVAs. Post-hoc T-tests showed significantly higher functional activity in the scene than object condition (both pFDR <0.001) only in the distal subiculum segments. In all other segments along the subiculum transversal axis, there was no significant difference in functional activity related to scene and object conditions (all pFDR = 0.1222; see Figure 4).

Functional activity during scene and object conditions along the transversal axis of subiculum and CA1.

Displayed are the extracted parameter estimates for the object condition versus baseline contrast (‘object information processing’, red) and the scene condition versus baseline contrast (‘scene information processing’, cyan) from the respective transversal segments in the subiculum ([A] grey) and CA1 ([B] blue) per individual (dots) and summarized across individuals (lines). A schematic depiction of the respective transversal segment is displayed by a 3D drawing of the right subiculum and CA1 subregion. Repeated measures ANOVAs revealed a significant interaction between condition and seed region in the subiculum only. The displayed significant difference is obtained with FDR-corrected post-hoc tests and refers to p<0.05. During the object condition, participants were presented with 3D rendered objects on screen, during the scene condition with 3D rendered indoor rooms and during the baseline condition they saw scrambled pictures. The shaded area around the lines refers to standard errors of the mean, sample size n = 32. Figure 4—source data 1 contains extracted parameter values for each subregion (Sub – subiculum and CA1, respectively) per individual and transversal segment (1–5 and 1–3, respectively from proximal to distal) for the object versus baseline and scene versus baseline contrasts.

Figure 4—source data 1

Individual parameter estimates for scene and object processing in right transversal subiculum and CA1 segments.

https://cdn.elifesciences.org/articles/76479/elife-76479-fig4-data1-v3.zip

Discussion

This study aims to advance insight into the organizational principles of information processing within the entorhinal-hippocampal circuitry and the circuitry’s embedding in designated cortical processing streams. Leveraging ultra-high field 7 Tesla fMRI, we find a resemblance between the intrinsic functional connectivity pattern and subregional biases in scene information processing in the entorhinal-hippocampal circuitry. In the EC, we observe a topographical mapping of regions from the cortical scene and object information processing streams, including the retrosplenial, parahippocampal and perirhinal Area 35 and Area 36 cortices. This mapping continues to determine a transversal organization of information processing routes between the EC and the human hippocampal circuitry. Our results unify previous evidence and uncover novel features in the human brain that can be a window into the circuitry’s critical role in memory function and decline.

Scene information is processed within an ECPHC-based – distal subiculum route

We identified regions in the entorhinal-hippocampal circuitry that are dedicated to process scene information. These regions consisted of two functionally connected portions: the ECPHC-based and the distal subiculum. The subiculum showed a transversal difference in intrinsic functional connectivity with a preference to the ECPHC-based in its distal portions (of note: the ECPHC-based was defined by entorhinal voxels with preferential functional connectivity to the parahippocampal cortex). Importantly, the distal subiculum and the ECPHC-based were the only studied entorhinal-hippocampal subregions that exhibited functional activity specifically in the scene condition (see appendix 5 for information processing in cortical source regions).

These findings provide clear evidence for a hypothesized transversal difference in scene information processing within the human subiculum. Our data also replicate the earlier functional and structural connectivity reports in humans as well as anatomical findings of a route between posterior-medial EC (based on parahippocampal connectivity) and distal subiculum (Maass et al., 2015; Syversen et al., 2021; Witter et al., 2000). The scene information processing bias has mainly been previously reported for the EC (in rodents, operationalized by spatial processing conditions: Neunuebel et al., 2013; Keene et al., 2016; in humans, operationalized by scene stimulus conditions: Berron et al., 2018; Navarro Schröder et al., 2015; Reagh and Yassa, 2014; Schultz et al., 2015). In animal studies, the importance of the subiculum as a translator of hippocampal information towards the entorhinal and other cortical structures is increasingly acknowledged (O’Mara, 2006; Roy et al., 2017). We here contribute to the sparse investigations regarding the nature of information processed along the transversal axis of the subiculum (see Cembrowski et al., 2018; Ku et al., 2017). Our observation is in line with the hypothesis that the distal subiculum is more involved in processing scenes than objects based on previous findings in the human brain. While the subiculum in general was associated with scene discrimination (Hodgetts et al., 2017), a growing body of evidence relates particularly the medial hippocampus to scene processing. This entails two medial areas, the pre- and parasubiculum, that we attribute to the distal subiculum in our current segmentation. Especially the area that resembles the pre- (or here: distal) subiculum has been shown to be involved in scene construction (Dalton et al., 2018; Zeidman et al., 2015). Recently, a gradient with coarser voxel-wise autocorrelation signals in the medial hippocampus has been reported, a finding that implies larger representations in the distal subiculum (Bouffard et al., 2022). In the latter two studies, however, the authors did not specifically extract data from the transversal axis of hippocampal subfields. Our joint investigation of functional entorhinal-subiculum connectivity and type of information processing along the full transversal axis of the subiculum, is the first to show a clear preference of scene information towards the distal portion, in comparison to more proximal portions.

Information processing is consistent with convergence within the anterior entorhinal portions – subiculum/CA1 border route

Our data revealed a route that did not show differences in scene and object information processing. Both, the ECArea35-based and the ECRSC-based portion exhibited preferential functional connectivity with the subiculum/CA1 border. Comparable levels of functional activity in scene and object conditions along these entorhinal-hippocampal routes are consistent with information convergence.

While we again confirm earlier findings and previously stated hypotheses, several features in our data are fundamentally novel. First, we provide initial human evidence for a functional connection between the ECArea35-based and the subiculum/CA1 border. Non-primate and primate anatomical data as well as ex vivo and in vivo structural connectivity data in humans show the possibility of information flow along that route (Syversen et al., 2021; Witter et al., 2017; Witter and Amaral, 1991; Witter and Amaral, 2020). Our results now underpin a functional relevance of that connection beyond the subiculum (for the subiculum see Maass et al., 2015). Our findings are derived based on a voxel-wise analysis, unconstrained by a priori selection of regions-of-interest. We thereby confirm the long-held proposal of a transversal functional organization in human subiculum and CA1.

Convergence of scene and object information is compatible with recent rodent work that shows joint coding of scene and object information (notably operationalized as spatial and non-spatial information, respectively) along CA1 and within the lateral EC (Deshmukh, 2014; Doan et al., 2019; Vandrey et al., 2021; Wilson et al., 2013; Yeung et al., 2019). Note that a supplemental analysis of information processing in the cortical source regions showed indeed, specific object processing in perirhinal source regions (see appendix 5). The lack of increased object processing in the anterior EC subregions and subiculum/CA1 border is thus likely not a result of increased noise in the object condition. Instead, increased object processing in perirhinal cortical source regions indicates subsequent convergence in entorhinal-hippocampal subregions, as hypothesized based on the updated cortical mapping scheme onto the EC.

Our results cannot confirm previous reports about higher functional activity for object than scene processing within these areas in the human brain (Reagh and Yassa, 2014; Navarro Schröder et al., 2015; Berron et al., 2018; also indicated in Dalton et al., 2018 and Schultz et al., 2015). Neither did we observe proximodistal differences in CA1 for object versus scene information processing as suggested by several rodent studies (Beer et al., 2018; Henriksen et al., 2010; Nakamura et al., 2013; Nakazawa et al., 2016). Differences in experimental design and contrasts could have contributed to these discrepancies (i.e. specific object information processing versus convergence). Previous studies used a variety of different conditions to tackle scene and object information processing (e.g. objects in time versus objects in space in Beer et al., 2018 or imagined objects on a 2D versus 3D grid in Dalton et al., 2018). In contrast to the current data, previous human studies did not derive functional data from specific, functionally defined entorhinal portions in the same dataset. As most previous studies were conducted in the light of the ‘parallel mapping hypothesis’, the related assumptions influenced the examined subregions, which may have altered the extracted measures.

Regarding the human proximal CA1, a firm conclusion is limited with our data. First, the functional connectivity results varied between hemispheres. In both hemispheres, proximal CA1 showed a different connectivity profile compared to distal CA1. However, even though statistically not significant, the preferences at the group level indicated increased functional connectivity with the ECPHC-based portion in the right proximal CA1 but with the ECArea35-based portion in the left hemisphere. Second, we do not prove similar information processing along the transversal CA1 axis. Instead, we find no significant difference in information processing along the transversal CA1 axis. As indicated in the previous paragraph, we cannot rule out that our object versus scene processing conditions may not have been sensitive enough to tackle functional differences in CA1. Thus, future research will have to identify defining characteristics of information processing along the transversal CA1 axis in a less constraint manner to allow conclusions on distinct information processing in proximal CA1.

In addition, we observed an unreported resemblance in functional connectivity profiles of ECRSC-based and ECArea35-based portions in the anterior EC. The sources of these entorhinal portions are part of cortical scene and object processing streams, respectively (see also appendix 5 that shows increased scene processing in the retrosplenial and increased object processing in perirhinal cortical source regions). To our knowledge, the ECRSC-based portion has not yet been identified in earlier investigations. While anatomical projections from the retrosplenial to deep medial EC layers have been confirmed in rodents, they appear in the posterior EC (Czajkowski et al., 2013; Sugar et al., 2011). Very recently, Syversen et al., 2021 found structural connectivity between the human retrosplenial cortex and the medial EC, but again not in the anterior part of the EC. The EC segmentation of Syversen and colleagues, however, followed different rules which may have contributed to differences in the topographical evaluation of the region. Also, structural and functional connectivity methods may yield different results, in particular as we identified EC subregions with a different set of cortical source regions. Under the assumption that retrosplenial connectivity defines the medial EC (Witter et al., 2017), the mapping of the ECRSC-based to the subiculum/CA1 border opposes conventional views that the medial EC communicates with the distal subiculum and proximal CA1 (based on rodent anatomy – see e.g. Nilssen et al., 2019). Whether species differences exist in the retrosplenial cortex – EC – hippocampus connectivity pattern or whether functional and structural connectivity diverge needs further investigation in the future.

Relevance of the current findings for the functional organization of the entorhinal-hippocampal circuitry

The current findings (summarized in Figure 5) advance our insight into the organization of the entorhinal-hippocampal circuitry on multiple levels and contribute to cognitive and clinical research. Recent efforts to understand how the human entorhinal-hippocampal circuitry accomplishes conjunction and segregation of information largely focused on the longitudinal hippocampal axis (e.g. Brunec et al., 2018; Brunec et al., 2020; Robin and Moscovitch, 2017). The transversal axis of the hippocampus has been approached by studies in humans that did not directly relate connectivity findings to information processing and did not assess subfield-specific organization (Vos de Wael et al., 2018; Plachti et al., 2019; Kharabian Masouleh et al., 2020; Paquola et al., 2020; Bouffard et al., 2022; for an overview see Genon et al., 2021). Dalton and Maguire, 2017, however, made a relevant proposal based on visual processing pathways and information processing. In correspondence to our results, they proposed the subiculum/CA1 border as a point of convergence between scene and object information processing streams. While their conclusion was based on direct parahippocampal, retrosplenial and perirhinal connections to the hippocampus, we found that both, the ECArea35-based (that is connected with the cortical object processing stream) and the ECRSC-based (that is connected with the cortical scene processing stream) show connectivity with the subiculum/CA1 border (see also appendix V for information processing in cortical source regions). Convergence is potentially also achieved via recurrency within the entorhinal-hippocampal system and cortical regions (cf. Koster et al., 2018 for evidence on recurrency). These considerations are an exciting future research avenue and remain speculative based on the current data due to insufficient temporal resolution. We nevertheless hypothesize the existence of two processing routes: one that processes converged object and scene information and one that processes scene information specifically. Thus, scene and object information processing might converge before the hippocampus. This presumably occurs within the anterior EC, given object-specific and scene-specific processing take place in the cortical source regions of the ECArea35-based and ECRSC-based portions, respectively (see appendix 5). Here, objects may be bound together with their defining scene-like or contextual features (akin to the ‘object-in-location’ idea in Connor and Knierim, 2017; Knierim et al., 2014). In addition, the dedicated scene processing that we observe along the ECPHC-based – distal subiculum route, may functionally underpin ideas about an anatomically graded contextual scaffold that the hippocampus utilizes to incorporate detailed information from the object-in-scene route into meaningful chunks of cohesive memory representations (‘events’; Behrens et al., 2018; Clewett et al., 2019; Robin, 2018; Robin and Olsen, 2019).

Summary of current results on the functional connectivity and information processing within the entorhinal-hippocampal circuitry.

Displayed is a schematic overview of our results on the functional connectivity and information processing within the entorhinal-hippocampal circuitry with four entorhinal seed regions and a focus on the transversal axis of hippocampal subiculum and CA1. The four entorhinal seed regions are derived from preferential functional connectivity to retrosplenial (RSC, green), parahippocampal (PHC, blue) and perirhinal Area 36 (A36, purple) and Area 35 (A35, pink) sources. Routes of preferred functional connectivity are depicted with dashed lines and the preferred information processed in the connected areas is depicted with symbolizing icons (scene – blue and object – red; stimuli from the task performed by the participants). M – medial; L – lateral; A – anterior; P – posterior; prox – proximal; dist – distal.

For completeness, we noted differences in functional connectivity along the longitudinal axis of the subiculum. We observed, for instance, more widespread functional connectivity of the ECArea35-based in the posterior subiculum whereas functional connectivity with the ECPHC-based portion seems more prominent in the anterior subiculum. The latter is consistent with previous reports (Dalton et al., 2019). The former, however, needs to be explored further by taking different segmentation protocols and seed regions into account. Note, that Maass et al., 2015 did not report longitudinal differences in connectivity strength between the EC and the subiculum. Future work needs to investigate how these observations relate to the reported gradient in functional connectivity and information resolution along the hippocampal longitudinal axis (e.g. Brunec et al., 2018 but many more). Altogether, the functional organization indicates that when a memory is to be formed, some degree of information convergence happens already before the hippocampus, nevertheless keeping specific aspects of scene information separated. This conclusion is in accordance with the updated cortical mapping scheme onto the EC (Nilssen et al., 2019). The topographical specificity of our results supports the necessity of functionally assessing the entorhinal-hippocampal circuitry with high spatial resolution and investigate memory function at a subregional level (Lee et al., 2020). The features we identified can inform future hypotheses on how the hippocampus achieves the formation of cohesive representations that serve memory function.

From a clinical research perspective, it is remarkable that the current functional connectivity pattern resembles the topology of early cortical tau pathology in Alzheimer’s disease (Lace et al., 2009). An influential hypothesis suggests tau progression in Alzheimer’s disease along functionally connected pathways in the human brain (Franzmeier et al., 2020; Vogel et al., 2020). Earliest cortical tau pathology in Alzheimer’s disease accumulates in perirhinal Area 35 (also referred to as transentorhinal region) and the anterior-lateral EC before it can be found along the subiculum/CA1 border (Braak and Braak, 1995; Berron et al., 2021; Kaufman et al., 2018; Lace et al., 2009). The topology of early tau pathology in Alzheimer’s disease thus mirrors the regions that we find biased towards ECArea35-based connectivity (Braak and Braak, 1991; Lace et al., 2009; Roussarie et al., 2020). Tau pathology in Alzheimer’s disease is associated with memory impairment (Bejanin et al., 2017; Berron et al., 2021; Nelson et al., 2012) and information processing might be affected accordingly as reports have shown an association between Alzheimer’s related tau pathology and object memory in early disease stages (Berron et al., 2019; Maass et al., 2019). However, given our finding of activity patterns consistent with object – scene convergence in those subregions of the hippocampal-entorhinal circuitry that are affected by early tau pathology, object-in-scene memory tasks might have increased sensitivity to memory impairment. Moreover, both, the entorhinal portion based on retrosplenial connectivity (ECRSC-based) and the entorhinal portion based on Area 35 connectivity (ECArea35-based), are functionally connected to the subiculum/CA1 border. This overlapping functional connectivity pattern in the hippocampus might be a way along which tau and amyloid pathologies in Alzheimer’s disease could interact. This is consistent with early hypometabolism and cortical tau progression in the retrosplenial cortex and early amyloid in posterior parietal regions (Grothe et al., 2017; Palmqvist et al., 2017; Ziontz et al., 2021). The revealed functional connectivity and information processing profile may guide future hypotheses on the propagation of Alzheimer’s pathology and related functional and cognitive impairment.

Limitations

This study has a number of limitations. First, the biases in seed connectivity in the left hemisphere were generally weaker and proximal CA1 results were less consistent across hemispheres. We conducted all analyses independently for both hemispheres to allow internal replication of our findings, however, whether partially different effects indeed signal a lateralization of the entorhinal-hippocampal organization in humans or whether the task or another parameter influenced these observations, will require further research.

Second, while it is unlikely that our functional connectivity pattern is the result of spatial proximity, increased correlation between spatially adjacent regions is an inherent problem of functional connectivity analyses. Distances between seed and target regions differ and may determine patterns in the functional connectivity data. To diminish the influence of proximity, our smoothing kernel was smaller than two times the voxel size. It is important to stress moreover, that the pattern of our results is not easily explainable by spatial distance between seed and target regions. The ECArea35-based or ECRSC-based, for instance, are not adjacent to the subiculum/CA1 border. Furthermore, we observed roughly comparable results for neighboring seeds and targets (e.g. ECPHC-based and distal subiculum) when we performed the functional connectivity analyses with seed and source regions in the contralateral hemisphere.

Third, our perspective was entirely functional and we cannot conclude on the directionality of our results. Also, it was beyond the scope of this study to examine direct connectivity between the cortical sources and hippocampal subregions. To what extent there is correspondence to structural connectivity (Syversen et al., 2021) remains to be determined, considering differences in the experimental task constraints and contrasts. Note that as a first step towards an understanding of the system’s functional organization and to increase comparability with earlier studies, we assessed functional connectivity and information processing within the entorhinal-hippocampal circuitry with univariate methods. These allow relative comparisons between functional activity levels in different conditions. Consequently, we are neither able to assess what the EC is processing during the baseline condition, meaning the absolute level of functional activity, nor are we able to verify that information processing is similar across conditions in for example the ECArea35-based seed. Univariate methods, moreover, average the signal over regions of interest. To capture hidden voxel-wise patterns of activity that scale with the processing of certain representations, future studies could examine information pathways with multivariate methods that evaluate informational content in the activity pattern of voxels instead of in an averaged manner (Kragel et al., 2018; Kriegeskorte et al., 2008). Moreover, recent methodological advances can be employed in the future that study functional connectivity based on the underlying content representations between regions (Basti et al., 2020).

Fourth, our study was originally conducted within the assumption that (functional) connectivity profiles reveal functional subregions. Based on that approach, the medial EC is identified based on i.a. retrosplenial connectivity. We, therefore conclude a surprisingly anterior yet medial EC mapping of the retrosplenial cortex. This approach has been followed by Maass et al., 2015 and also in numerous anatomical connectivity studies in animals (see Witter et al., 2017). It is possible that species differences lead our ECRSC-based to be more anterior than one would expect based on animal studies. However, given that the medial subregion in the primate EC remains posterior (cf. posterior-medial EC homologue in Maass et al., 2015), another possibility is that our retrosplenial functional connectivity cluster maps onto the human anterior-lateral EC. Our data does not allow us to verify this latter option. It is unclear, however, why functional subregions in line with predictions from animal research can be identified for some cortical source-to-EC mappings (like the parahippocampal cortex) but not for others. In combination with closely matched histological or structural magnetic resonance imaging data, future work can further reveal the nature of retrosplenial mapping on the human EC.

In general, the quantification of the transversal connectivity pattern should be considered with some caution from the anatomist’s perspective. The segmentation of subregions on functional data is an approximation because the anatomical ground truth cannot be captured by any segmentation protocol (even histological data can lead to divergent opinions). This shortcoming is amplified by group comparisons that do not account for participant-specific anatomy. Future research is needed to evaluate how the functionally derived entorhinal seeds in this study relate to histologically derived entorhinal subregions (Oltmer et al., 2022) or entorhinal subregions based on structural connectivity (Syversen et al., 2021). For a dedicated comparison of subregions, it is essential to pay close attention to the segmentation of the EC itself. Note moreover, that we excluded the head and the tail of the hippocampus from our investigation. The head is highly complex in its subfield topography (Ding and Van Hoesen, 2015; Berron et al., 2017) and prevents clear hypotheses regarding a transversal pattern. For the tail we lack an established segmentation protocol (de Flores et al., 2020; DeKraker et al., 2018). In the future, advanced segmentation methods and evaluations in the participant-space will improve this issue and reveal the organization in more detail.

Conclusion

In sum, leveraging ultra-high field functional imaging, we provide a comprehensive in vivo exploration of the functional organization within the human entorhinal and hippocampal subregions and the circuitry`s embedding within cortical information processing streams. Within the entorhinal and hippocampal subiculum, our data partially support a continuation of cortical object and scene information processing with convergence in anterior and lateral entorhinal portions (ECArea35-based, ECRSC-based, ECArea36-based), proximal subiculum and CA1, while the posterior-medial entorhinal portion (ECPHC-based) and distal subiculum process scene information specifically. Topographically, this organization of information processing overlaps with our identified pattern of functional connectivity. The data yield spatially organized information processing along functionally connected subregions in the human EC and transversal sub/CA1 axis. Our high-resolution approach revealed unknown characteristics of functional connectivity and scene processing within the human entorhinal-hippocampal circuitry. These aid our understanding of how cortical information comes together and is further communicated within the entorhinal-hippocampal circuitry, underpinning the formation of cohesive memory representations. We provide essential insights for basic and clinical research that we believe to be crucial for the development of future hypotheses on memory function and decline.

Methods

The current data is part of a larger study that examines exercise effects on cognition. The data that is subject to the current study have been acquired during the baseline measurement before any intervention took place. In the following, we focus on the study setup and methodological aspects of direct relevance for the current questions and data analyses.

Participants

In total, 32 healthy participants (15 female) with a mean age of 25.5 years (range 19–35 years, standard deviation 4.3 years) were included in the current data analyses. All participants were right-handed, finished education on A-level (German Abitur or comparable) and reported absence of any neurological or psychiatric diseases. General exclusion criteria determined by the 7 Tesla MR scanning procedure were applied (e.g. metallic implants, tinnitus, known metabolic disorders). All participants gave informed consent prior to participation and received a monetary compensation. The study received approval by the ethics committee of Otto-von-Guericke University, Magdeburg (Germany) under reference number 128/14.

Task

While functional images were acquired, participants engaged in a mnemonic discrimination task (see Berron et al., 2018). The object-scene task consisted of 64 objects and 64 rooms. In two runs, participants encoded always two stimuli, two 3D rendered objects in the object condition or two 3D rendered rooms in the scene condition and subsequently identified the following two same or similar stimuli as novel or old. Ten scrambled images were presented in blocks at the beginning and end of each run and served as baseline condition. All stimuli were presented for three seconds. In the recognition phase, participants had to respond during that time. Each stimulus was followed by a noise stimulus to prevent after-image and pop-out effects. The short alternating encoding/recognition sequences were embedded in an event-related design.

Data acquisition

All MRI data was acquired with a 7 Tesla Siemens MR machine (Erlangen, Germany) using a 32-channel head coil. First structural images were obtained. A whole-brain MPRAGE volume was acquired with isotropic voxel size of 0.6 mm, TR 2500ms; TE 2.8ms, 288 slices in an interleaved manner (FOV 384x384x288). Thereafter, a partial structural T2*- weighted volume (TR 8000ms; TE 76ms, interleaved, 55 slices, FOV 512x512x55), orientated orthogonal to the main longitudinal hippocampal axis was obtained with a resolution of 0.4x0.4 mm in-plane and a slice thickness of 1 mm.

The subsequent acquisition of functional data took place in two runs à 14 min (332 volumes each) employing echo-planar imaging (EPI). The volumes were partial (40 slices, TR 2400ms, TE 22ms, FOV 216x216x40, interleaved slice acquisition), oriented along the longitudinal axis of the hippocampus and acquired with an isotropic voxel size of 1 mm.

All EPIs were distortion corrected with a point-spread function method and motion corrected during online reconstruction (Zaitsev et al., 2004).

Data analyses

Preprocessing

Preprocessing and statistical modeling of fMRI data was performed with SPM12 (Wellcome Department of Cognitive Neuroscience, University College, London UK; Penny et al., 2011). The individual functional images were slice time corrected and smoothed with a full-width half-maximum Gaussian kernel of 1.5 mm. To preserve a high level of anatomical specificity, smoothing was performed with a kernel smaller than two times the voxel size. The artifact detection toolbox ARTrepair (Mozes & Whitfield-Gabrieli, 2011) was subsequently used to identify outliers regarding mean image intensity and motion between scans (threshold in global intensity: 1.3%; movement threshold: 0.3 mm). Identified outliers are included as spike regressors in subsequent statistical modeling.

Task effects in the functional data were removed by fitting general linear models (with regressors for all task conditions, outliers and movement parameters) to the data. The obtained residual images were saved for the intrinsic functional connectivity analyses. Note that task-related parameter estimates were extracted for the final information processing analysis, as described later.

Structural data processing and segmentation

Structural template calculation (T1-weighted) and segmentation

To examine and illustrate group-level results later on, a group specific T1-weighted template was calculated using ANTS buildtemplateparallel.sh (Avants et al., 2010). For illustration purposes and to aid group analyses, in addition, the T1 template was manually segmented into subregions subiculum and CA1 within the hippocampal body with ITK-SNAP (Yushkevich et al., 2006) based on the segmentation rules described in Berron et al., 2017. The first slice in each hemisphere that did not contain the uncus anymore, served as start of the hippocampal body in all hippocampal subregions. The last segmented slice was the one at which both, the inferior and superior colliculi had completely disappeared, applied for each hemisphere separately. Moreover, to evaluate results across the transversal sub/CA1 axis, the subiculum masks in each hemisphere were cut in five equally wide segments from medial to lateral within each coronal image. As the CA1 region gets more and more tilted towards the hippocampal tail, the three transversal CA1 segments were determined based on manual segmentation following a geometrical rule. Therefore, the two outer borders along the transversal axis of CA1 were connected with a line. From the middle point of that line, two straight lines were drawn in a 60° angle to determine roughly equally sized transversal CA1 segments within each coronal slice and hemisphere (a figure displaying the cuts, the procedure for the CA1 segments and the numbers of voxels within each segment can be found in appendix 9). Related to the overall size of the subregions, we opted to build five subiculum and three CA1 segments along the transversal axis from proximal to distal ends.

Segmentation of individual regions of interest

We manually segmented regions of interest (ROI) in the medial temporal lobe according to the segmentation protocol by Berron et al., 2017. Based on individual T2-weighted images, the parahippocampal cortex, Area 35, Area 36 and the EC are delineated (see appendix 8 for quality assurance measures). Moreover, we ran a Freesurfer 6.0 segmentation on the group T1 template to segment the isthmus cingulate cortex as retrosplenial mask (Desikan et al., 2006; Fischl, 2012). Note here that Syversen et al., 2021 used a similar region, however excluded the most superior part. For individual retrosplenial masks, the obtained mask was co-registered from the group T1 template space to the individual T1 space by making use of the alignment matrices obtained during above described T1 group template calculation (see appendix 7 for co-registration procedure and alignment assessment). For this alignment process we used ANTS WarpImageMultiTransform.sh (Avants et al., 2011). The retrosplenial, parahippocampal and perirhinal Area 36 and Area 35 regions served as cortical source regions for an initial functional connectivity analysis that we conducted to obtain functional subregions within the entorhinal cortex (see upcoming paragraphs and appendix 2).

Co-registration of individual structural data to functional data space

For later functional data extraction, the individual T1-weighted and T2-weighted structural images were co-registered and resliced to the echo-planar images. Therefore, ANTS was used to transfer the T2-weighted structural image to the participant’s T1 space (Avants et al., 2011). For the co-registration between individual T1-weighted and EPIs, FSL epi_reg was applied (Jenkinson and Smith, 2001). All subsequently segmented individual masks were co-registered to the participant’s functional (echo-planar) images using the obtained warping matrices (see appendix 7 for co-registration procedure and alignment assessment). ANTS WarpImageMultiTransform.sh was applied for T2 to T1 co-registration and FSL flirt was used for T1 to echo-planar image co-registration (Avants et al., 2011; Jenkinson and Smith, 2001).

ROI preparation for seed regions in functional connectivity analyses

All masks that served as source and seed regions throughout the functional connectivity analyses (retrosplenial, parahippocampal, perirhinal Area 36 and Area 35 and the later defined entorhinal subregions) were thresholded according to mean intensity to prevent signal dropout and thus a distortion of the average functional signal extracted from seed regions for the connectivity analysis. Therefore, we followed Libby et al., 2012 and Maass et al., 2015, to remove all voxels from each ROI that showed a mean intensity over time of less than two standard deviations from the mean intensity across all voxels. The thresholding was performed before each seed-to-voxel functional connectivity analysis.

Functional connectivity analyses at the participant level

Two different functional connectivity analyses were performed that build upon the approach by Maass et al., 2015. The first analysis served to identify functional subregions (‘seeds’) within the entorhinal cortex that uniquely connect with functionally and clinically relevant cortical sources. The second, core analysis, then evaluated the intrinsic functional connectivity pattern between these entorhinal seeds and hippocampal subiculum and CA1. Both functional connectivity analyses were performed on residuals of task-related functional data, creating a dataset that resembles resting-state data (Gavrilescu et al., 2008; Maass et al., 2015). In the following we describe the analysis procedure in detail. Note, that all analyses were conducted independently in both hemispheres.

To determine functional entorhinal seed regions we first performed a seed-to-voxel semipartial correlation analysis (Whitfield-Gabrieli and Nieto-Castanon, 2012) between the individually extracted residuals from retrosplenial, parahippocampal and perirhinal Area 36 and Area 35 sources as well as entorhinal voxels. The regions we call cortical sources served as seeds in that analysis. Note that the semipartial correlations calculate the variance in a voxel that is uniquely explained by the source, excluding contributions from other sources. Please refer to appendix 6 for more details on this functional semipartial correlation analysis. To obtain entorhinal seeds for the core functional connectivity analysis, first, we calculated one-sample T-tests across participants on the individually obtained and aligned, standardized beta maps for each source, respectively. Second, the four resulting statistical maps (one for each source) have been thresholded at T>3.1. Each entorhinal voxel now was attributed to be preferentially connected with one of the four source regions, based on the voxel’s maximum T value across the thresholded one-sample T-test maps. Those voxels that did not reach the threshold of T>3.1 in any of the four statistical maps have not been attributed to be preferentially connected with any of the four cortical sources. Finally, across hemispheres we selected for each source preference an equal number of these highest preference voxels across all T-tests (the number is determined by the hemisphere with the lowest relevant number of voxels). This procedure yielded four entorhinal subregions, one containing the entorhinal voxels that preferentially functionally connect with the retrosplenial (1530 voxels), one containing the entorhinal voxels that preferentially functionally connect with the parahippocampal cortex (145 voxels) and one each that contained the preferentially functionally connected voxels with perirhinal Area 35 (298 voxels) and Area 35 (751 voxels), respectively. All four entorhinal seed masks were determined on group level and co-registered to each participant. They then served as seed regions for the core functional connectivity analysis between entorhinal cortex seeds and hippocampal subregions.

For the core functional connectivity analysis (entorhinal seeds-to-hippocampal subregion voxels), an analogous seed-to-voxel semipartial correlation analysis was performed on the individual residual functional imaging data using the CONN toolbox (Whitfield-Gabrieli and Nieto-Castanon, 2012). Note again that the semipartial correlations calculate the variance in a voxel that is uniquely explained by the seed, excluding contributions from other seeds. Now the four entorhinal subregions served as seeds and functional connectivity was examined with the whole brain (later masked by the hippocampal subregion masks). For each functional connectivity analysis, mean time series were extracted from the respective seed region and entered as regressor of interest. White matter and CSF time series, realignment parameters and outliers served as regressor of no interest. The functional data from the residuals was band-pass filtered (0.01–0.1 Hz) and semipartial correlations were obtained between the seed timeseries and all other brain voxel’s timeseries. The obtained beta maps contained Fisher-transformed correlation coefficients and were used for subsequent group analyses.

Alignment between participants

To be able to perform group statistics on the resulting topography in the beta maps, the individual data was aligned to group space. Here, the T1 template image served as reference space. Using the inverse of the previously obtained individual warping matrices from individual T1 to EPI, first the standardized beta maps were co-registered from EPI to individual T1 space. In a further step, the statistical maps were then aligned between the individual T1 space and the group T1 template space, by making use of the alignment matrices obtained during above described T1 group template calculation. For this alignment process we used ANTS WarpImageMultiTransform.sh (Avants et al., 2011).

Functional connectivity analysis at group level

To investigate the functional connectivity profile between the four entorhinal seeds and the subiculum and CA1 subregion across individuals, we evaluated connectivity preferences to either seed within all transversal segments of the subiculum and CA1 target regions. Therefore, mean values for connectivity estimates to either entorhinal cortex seed were extracted from the group aligned but participant-specific beta maps out of each transversal segment, averaged along all coronal slices. Note, that segment-based extraction is necessary due to the varying number of sagittal slices that cover the respective regions along the longitudinal axis of the hippocampal body. Based on these participant-level connectivity results, connectivity preference plots for all four entorhinal seeds have been created to depict tendencies along the transversal sub/CA1 axis.

A hierarchical repeated-measures ANOVA testing procedure was employed to reveal significant differences in the transversal hippocampal connectivity patterns between entorhinal seed regions. Therefore, in a first step, an overall repeated measures ANOVA (4 seed X transversal segments) was performed per target region (subiculum and CA1 in both hemispheres) to reveal whether significant differences in seed connectivity estimates exist across the transversal axis of the respective target region (subiculum or CA1). If the overall seed X transversal segment interaction effect was significant (false-discovery-rate corrected according to Benjamini and Hochberg, 1995), in a second step, one-way repeated measures ANOVAs have been performed for each seed to identify those entorhinal seeds that indeed show a differential connectivity pattern across the transversal axis of the target region (all false-discovery-rate corrected according to Benjamini and Hochberg, 1995). If more than one seed main effect was significant, finally we determined whether these seeds exhibit an opposing connectivity pattern across the transversal axis of subiculum or CA1, respectively, by evaluating the pair-wise seed X transversal segment interaction effects on the extracted connectivity estimates.

For a more detailed topographical display of the entorhinal-hippocampal connectivity results, we calculated one-sample t-tests on the aligned, standardized beta maps that we obtained in the first-level analyses for each seed respectively. Crucially, the resulting group-level one-sample T-test statistical maps were only used to display results but not for any further statistical inference. To depict the topography of the respective voxel-wise seed preferences, the resulting group-level T-maps were thresholded with T > |0.001| and masked with the respective subregion of interest. To depict general tendencies in the connectivity profile, for each voxel in the region of interest the preferred seed connectivity was determined by attributing it to the seed with the highest T value across the one-sample T-test maps. The resulting maps were depicted in 3D plots, generated with ITK-SNAP (Yushkevich et al., 2006) that provide an overview of each voxel’s preference for the respective seed functional connectivity at a glance.

Functional analysis of content-related activity at participant level

To investigate whether scene and object information is differentially processed within entorhinal seed regions and along the transversal sub/CA1 axis, the results from the initially fitted general linear models (used to remove task effects) were examined. Contrast estimates were calculated between the beta estimates obtained from task conditions in which individuals saw indoor rooms (scene) versus objects on the screen and conditions in which individuals saw the scrambled stimuli (baseline). The resulting contrast value maps for object > baseline and scene > baseline were then co-registered to the T1 group template space. Subsequently, individual mean contrast estimates have been extracted from the four entorhinal seed regions and from those transversal segments that had previously been used for the evaluation of the intrinsic functional connectivity results (three or five segments in CA1 and subiculum, respectively).

With repeated measures ANOVAs (content condition X entorhinal region or content condition X transversal hippocampal segment), we investigated whether contrast estimates differed under scene and object conditions in the respective regions. Effect of interest thus, was the interaction between the content condition and the subregion or segment, respectively. Post-hoc paired-samples T-test were performed if the respective interaction effect was significant, to reveal in which subregion or segment functional activity between scene and object processing conditions differed significantly from each other (all false-discovery-rate corrected according to Benjamini and Hochberg, 1995).

Appendix 1

Left hemisphere results

Four cortical sources divide the left entorhinal cortex (EC) in retrosplenial-, parahippocampal-, Area 35- and Area 36-based seeds

Based on functional connectivity preferences to the sources parahippocampal cortex (Source code 4), retrosplenial cortex (Source code 3), Area 36 (Source code 2) and Area 35 (Source code 1), we obtained four entorhinal seeds. The majority of voxels can roughly be described as clustering in the posterior-medial entorhinal portion for the ECPHC-based, the anterior-medial (and posterior-medial) portion for the ECRSC-based seed, the anterior-lateral portion for the ECArea35-based and the posterior-lateral portion for the ECArea36-based seed (see appendix 2 for exact voxel counts). Note that both perirhinal-based entorhinal seeds extend along the anterior to posterior axis such that the ECArea35-based progresses more along deep entorhinal portions (with a main focus anteriorly) and the ECArea36-based along superficial entorhinal portions (with a main focus posteriorly, see Appendix 1—figure 1 and the medial reflection of the EC seeds).

Appendix 1—figure 1
Left entorhinal seed regions based on connectivity preferences to cortical regions.

Displayed is the left entorhinal cortex as a 3D image with colored seed regions. The seed regions have been identified based on a source-to-voxel functional connectivity analysis and resulting connectivity preference to either the left retrosplenial cortex (RSC, green), parahippocampal cortex (PHC, blue), Area 36 (A36, purple) or Area 35 (A35, pink) sources. Note that preferences to Area 36 are best visible from a medial perspective on the entorhinal cortex as depicted in the medial reflection. Seed regions have been determined based on the maximum voxels across four one-sample T-tests at group level, one per source, sample size n = 32. M – medial; L – lateral; A – anterior; P – posterior.

Left distal subiculum is functionally connected with the ECPHC-based seed while the subiculum/CA1 border is connected with ECRSC-based and ECArea35-based seeds

When extracting estimates of connectivity preferences across individuals from proximal and distal hippocampal subfield segments for either entorhinal seed, repeated measures ANOVAs revealed significant seed X segments interaction effects along the transversal axis of the left subiculum and CA1 (subiculum: F(12,372) = 4.609; p<0.001; CA1: F(6,186) = 2.458; p=0.047; see Appendix 1—figure 2).

Appendix 1—figure 2
Functional connectivity preferences to entorhinal seeds along the subiculum and CA1 transversal axis, left hemisphere.

Displayed are the results of a seed-to-voxel functional connectivity analysis between the displayed left entorhinal seeds and the left subiculum and CA1 subregion. The 3D figure shows voxel-wise connectivity preferences to the entorhinal seeds (color coded to refer to the respective entorhinal seed [E]) on group level ([A] - subiculum; [B] - CA1; maps for connectivity preferences: Source code 9 - ECArea35-based, pink; Source code 10 - ECArea36-based, purple; Source code 12 - ECPHC-based, blue; Source code 11 - ECRSC-based seed, green). Note that preferences to the ECArea35-based seed (pink) are located mainly in the inferior subiculum and CA1 and are therefore visible in the inferior reflection. To display mean connectivity preferences across participants along the transversal axis, beta estimates were extracted and averaged from equally sized segments from proximal to distal ends (five segments in subiculum [A], three segments in CA1 [B]; schematized in white on the 3D figures) on each coronal slice and averaged along the longitudinal axis. Repeated measures ANOVAs revealed significant differences in connectivity estimates along the transversal axis in CA1 [D] and subiculum [C] with interaction effects in the subiculum. Displayed significances obtained by FDR-corrected post-hoc tests and refer to p<0.05. Empty asterisks refer to effects that did not reach significance under FDR-correction. Shaded areas in the graphs refer to standard errors of the mean, sample size n = 32. EC – entorhinal; M – medial; L – lateral; A – anterior; P – posterior; prox – proximal; dist – distal. Appendix 1—figure 2—source data 1 contains individual connectivity estimates per subregion (Sub – subiculum and CA1, respectively) and seed (ECRSC-based – RSCECseed, ECArea35-based – A35ECseed, ECPHC-based – PHCECseed, ECArea36-based – A36ECseed) for each transversal segment (1–5 or 1–3, respectively from proximal to distal).

Appendix 1—figure 2—source data 1

Individual functional connectivity estimates to left entorhinal seeds, extracted from left subiculum and CA1 transversal segments.

https://cdn.elifesciences.org/articles/76479/elife-76479-app1-fig2-data1-v3.zip

In the left subiculum, additional repeated measures ANOVAs showed that the ECArea35-based (F(4,124) = 4.489; pFDR = 0.025), and ECPHC-based (F(4,124) = 8.701; pFDR <0.001) seeds displayed a significant main effect across the transversal subiculum axis. Here, the transversal preference to the ECRSC-based entorhinal seed does not survive FDR correction (F(4,124) = 4.489; Huynh-Field uncorrected p=.05), shows however the same tendency as in the right hemisphere. The differential functional connectivity preferences for the ECArea35-based and ECPHC-based seed interacted significantly across the transversal axis, as shown in a subsequent repeated measures ANOVA (F(4,124) = 10.795; pFDR <0.001).

In the left CA1, additional repeated measures ANOVAs showed that the connectivity preference towards the ECRSC-based seed displayed a significant main effect across the transversal CA1 axis (F(2,62) = 6.753; p=0.024). In the distal CA1, the preferential functional connectivity with the ECPHC-based seed was higher than in the proximal portion of CA1. In the left CA1 a similar but weaker transversal pattern was observed for connectivity preferences with the ECArea36-based (F(2,62) = 3.841; pFDR = 0.051) and ECPHC-based seed regions (F(2,62) = 3.468; pFDR = 0.051).

Left distal subiculum and ECPHC-based exhibit higher functional activity in the scene condition while other subregions show no significant difference between conditions

For the characteristics of information processing, we first focus on the left entorhinal seed regions. When extracting task-related parameter estimates for object and scene information conditions, a repeated measures ANOVA showed a significant interaction between region and information type (object versus scene; F(3,93) = 9.772; p<0.001). Post-hoc t-tests revealed that only in the ECPHC-based seed region functional activity for scene information was significantly higher than for object information (pFDR = 0.003), while in the remaining three left entorhinal seed regions no significant difference between object and scene conditions existed (all pFDR = 0.5776; see Appendix 1—figure 3).

Appendix 1—figure 3
Functional activity during scene and object conditions in entorhinal seed regions, left hemisphere.

Displayed are the extracted parameter estimates for the object versus baseline contrast (‘object information processing’, red) and the scene versus baseline contrast (‘scene information processing’, cyan) from each left entorhinal seed region per individual (dots) and summarized across individuals (lines). A schematic depiction of the respective entorhinal seed regions is displayed by a 3D drawing of the left EC. A repeated measures ANOVA revealed a significant interaction between condition and seed region. The displayed significant difference is obtained with FDR-corrected post-hoc tests and refers to p<0.05. During the object condition, participants were presented with 3D rendered objects on screen, during the scene condition with 3D rendered rooms and during the baseline condition they saw scrambled pictures. The shaded area around the lines refers to standard errors of the mean, sample size n = 32. EC – entorhinal; M – medial; L – lateral; A – anterior; P – posterior. Appendix 1—figure 3—source data 1 contains extracted parameter values per individual and EC seed (isthmuscingulate – ECRSC-based, Area 35 – ECArea35-based, Area 36 – ECArea36-based, PHC – ECPHC-based seed) for the object versus baseline and scene versus baseline contrasts.

Appendix 1—figure 3—source data 1

Individual parameter estimates for scene and object processing in left entorhinal seed regions.

https://cdn.elifesciences.org/articles/76479/elife-76479-app1-fig3-data1-v3.zip

In the left hippocampal subregions, extracting the task-related parameter estimates for object and scene conditions from proximal and distal segments within each participant showed a significant interaction between transversal segments and information type in the subiculum (F(4,124) = 7.697; p<0.001), not however in CA1 (F(2,62) = 1.1925; p = 0.3042) as revealed by repeated measures ANOVAs. Post-hoc T-tests showed that only in the distal subiculum segments and in the middle segment significantly more scene than object information was processed (from most distal to middle segment pFDR <0.001; pFDR = 0.0015; pFDR = 0.0274). In all other segments along the transversal axis, no significant difference in functional activity related to object and scene conditions existed (from medial to proximal: pFDR = 0.1009; pFDR = 0.2435; see Appendix 1—figure 4).

Appendix 1—figure 4
Functional activity during scene and object conditions along the transversal axis of subiculum and CA1, left hemisphere.

Displayed are the extracted parameter estimates for the object versus baseline contrast (red) and the scene versus baseline contrast (cyan) from the respective transversal segments in the subiculum ([A] grey) and CA1 ([B] blue) per individual (dots) and summarized across individuals (lines). A schematic depiction of the respective transversal segment is displayed by a 3D drawing of the left subiculum and CA1 subregions. Repeated measures ANOVAs revealed a significant interaction between condition and transversal segment in the subiculum only. The displayed significant difference was obtained with FDR-corrected post-hoc tests and refers to p<0.05. During the object condition, participants were presented with 3D rendered objects on screen, during the scene condition with 3D rendered rooms and during the baseline condition they saw scrambled pictures. The shaded area around the lines refers to standard errors of the mean, sample size n = 32. Appendix 1—figure 4—source data 1 contains extracted parameter values for each subregion (Sub – subiculum and CA1, respectively) per individual and transversal segment (1–5 and 1–3, respectively from proximal to distal).

Appendix 1—figure 4—source data 1

Individual parameter estimates for scene and object processing in left transversal subiculum and CA1 segments.

https://cdn.elifesciences.org/articles/76479/elife-76479-app1-fig4-data1-v3.zip

Appendix 2

Quantitative assessment of entorhinal seeds

To assess the main location of each cortical source preferences within the EC, we cut the left and right EC in four quadrants. This was performed in T1 template space. First, the middle slice of all coronal slices which capture the EC was determined separately for each hemisphere. This slice was used to cut the EC in quadrants I, III and II, IV. Second, the middle slice of all axial slices which capture the EC was determined. This slice served to cut the EC in quadrants I, II and III, IV (see Appendix 2—figure 1). Note, to determine the most superior axial slice, the most posterior coronal level of the EC was used. Subsequently, we counted the number of voxels that have been assigned to each of the four cortical source regions after the initial functional connectivity analyses (that served to determined EC seeds). Averaged across hemispheres, most voxels assigned to the retrosplenial source are in EC quadrant I, most voxels assigned to the Area 35 source in EC quadrant II, most voxels assigned to the parahippocampal cortex in EC quadrant III and most voxels assigned to Area 36 in EC quadrant IV (see Appendix 2—table 1 for detailed voxel counts). Note that these quadrants do not refer to anatomically defined EC subregions.

Appendix 2—figure 1
Entorhinal cortex cut in four quadrants.

Illustrated is the schematic entorhinal cutting in four quadrants (I, II, III and IV) in the right hemisphere. Stippled lines illustrate approximate cuts. M – medial, L – lateral, A – anterior, P – posterior.

Appendix 2—table 1
Number of voxels attributed to have a preferred functional connectivity to either cortical source (RSC, PHC, A35, A36) within each EC quadrant (I.-IV.).

Bold voxel numbers refer to the highest number across EC quadrants. EC – entorhinal cortex, RSC – retrosplenial cortex, PHC – parahippocampal cortex, A35 – perirhinal Area 35, A36 – perirhinal Area 36.

EC quadrantI.II.III.IV.
RSC-source599421337173
PHC-source1313200
A35-source7180433167
A36-source1035139201

Appendix 3

Functional connectivity gradients by source and seed region

Appendix 3—figure 1
Entorhinal functional connectivity with isolated cortical sources.

Displayed are the voxel-wise functional connectivity values (T values) of the EC with the respective cortical sources [A] retrosplenial cortex (RSC, green, left: Source code 3, right: Source code 7), [B] perirhinal Area 36 (A36, purple, left: Source code 2, right: Source code 6), [C] parahippocampal cortex (PHC, blue, left: Source code 4, right: Source code 8) and [D] perirhinal Area 35 (A35, pink, left: Source code 1, right: Source code 5). Results from left and right hemisphere one-sample T-tests for the functional connectivity with the respective source are displayed alongside each other for each cortical source, sample size n = 32. The smaller entorhinal cortex maps in the middle of each rectangle are medial reflections of the respective results. Colorbars reflect the range of T values. Grey areas refer to T values of T<0.1. L – lateral; M – medial; A – anterior; P – posterior.

Appendix 3—figure 2
Subiculum/CA1 functional connectivity with isolated entorhinal (EC) seeds.

Displayed are the voxel-wise functional connectivity values (T values) of the subiculum and CA1 to the respective [A] green (ECRSC-based, left: Source code 11, right: Source code 15) [B] purple (ECArea36-based, left: Source code 10, right: Source code 14), [C] blue (ECPHC-based, left: Source code 12, right: SourceSource code 16) and [D] pink (ECArea35-based, left: Source code 9, right: Source code 13) EC seeds. The respective seeds are illustrated in the lower panel. Results from left and right hemisphere one-sample T-test for the functional connectivity with the respective seed are displayed alongside each other, sample size n = 32. The lower subiculum/CA1 maps within each rectangle are inferior reflections of the respective results. Colorbars reflect the range of T values. Grey areas refer to T values of T<0.1. L – lateral; M – medial; A – anterior; P – posterior.

Appendix 4

Superior and inferior view on voxel-wise functional connectivity preferences to entorhinal seeds

Appendix 4—figure 1
Functional connectivity preferences to entorhinal seeds along the subiculum and CA1 transversal axis.

Displayed are the results of a seed-to-voxel functional connectivity analysis between entorhinal seeds and the left and right subiculum [A] and CA1 [B] subregion. Voxel-wise connectivity preferences to the entorhinal seeds on group level are shown from a superior and an inferior perspective on the respective subregion. The figure displays the same data as in Appendix 1—figure 2 and Figure 2 and is based on Source code 9–16. The color coding refers to the respective entorhinal seed: green - ECRSC-based; purple - ECArea36-based; blue - ECPHC-based and pink - ECArea35-based seed. M – medial; L – lateral; A – anterior; P – posterior; prox – proximal; dist – distal.

Appendix 5

Object and scene processing in cortical source regions

To examine whether lower parameter estimates for object processing could be due to increased noise in this condition, we evaluated object and scene processing in the four cortical source regions. Therefore, we extracted parameter estimates for the object versus baseline and the scene versus baseline contrast from the retrosplenial and parahippocampal cortex and from perirhinal Area 36 and Area 35, respectively. All parameter estimates were extracted from the previously segmented regions of interests, coregistered to the individual EPI space.

Repeated-measures ANOVAs in both hemispheres showed a significant interaction effect between condition and region (right: F(3,93) = 60.4229; p<0.001; left: F(3,93) = 47.3421; p<0.001). Subsequent paired-samples T-tests show significantly more functional activity in the object than scene condition in Area 36 (bilateral: pFDR < 0.001) and the left Area 35 (pFDR= 0.0011). No significant difference between object and scene conditions is observed in the right Area 35 (pFDR = 0.9821). There is a significant effect of more functional activity in the scene than object condition in the parahippocampal (bilateral: pFDR<0.001) and retrosplenial cortex (bilateral: pFDR<0.001, see Appendix 5—figure 1).

Appendix 5—figure 1
Functional activity during scene and object conditions in cortical source regions.

Displayed are the extracted parameter estimates for the object versus baseline contrast (red) and the scene versus baseline contrast (cyan) from four cortical source regions in the [A] left and [B] right hemisphere, per individual (dots) and summarized across individuals (lines). Repeated measures ANOVAs revealed a significant interaction between condition and cortical source region in both hemispheres. The displayed significant differences (asterisks) were obtained with FDR-corrected post-hoc tests and refer to p<0.05, sample size n = 32. During the object condition, participants were presented with 3D rendered objects on screen, during the scene condition with 3D rendered rooms and during the baseline condition they saw scrambled pictures. The shaded area around the lines refer to standard errors of the mean. PHC – parahippocampal cortex (blue), RSC – retrosplenial cortex (green), A35 – perirhinal Area 35 (pink), A36 – perirhinal Area 36 (purple). Appendix 5—figure 1—source data 1 contains extracted parameter values from cortical source regions (left – lSources, right – rSources, isthmuscingulate – retrosplenial) for the object versus baseline and scene versus baseline conditions per individual.

Appendix 5—figure 1—source data 1

Individual parameter estimates for scene and object processing in cortical source regions.

https://cdn.elifesciences.org/articles/76479/elife-76479-app5-fig1-data1-v3.zip

The increased object processing in adjacent cortical source regions indicates that noise differences across conditions are not likely to cause the lack of increased object processing within entorhinal seed regions and hippocampal subregions.

Appendix 6

Functional connectivity analysis to determine entorhinal seeds

Before performing the core functional connectivity analysis between entorhinal seeds and hippocampal voxels, we had to determine the entorhinal seeds, that is, the functional subregions of the EC. We largely followed Maass et al., 2015 approach to assure comparability of results. The seeds were determined based on their functional connectivity with functionally relevant sources from the cortical object and scene information processing streams, that are the perirhinal Area 35 and Area 36, the parahippocampal cortex and the retrosplenial cortex (see Nilssen et al., 2019).

The CONN toolbox (Whitfield-Gabrieli and Nieto-Castanon, 2012) was applied to perform a seed-to-voxel semipartial correlation analysis on the residual fMRI data between the retrosplenial, parahippocampal, Area 35 and Area 36 sources and the voxels within the segmented EC mask of each individual (see the description of the core functional connectivity analysis for the precise parameters). The resulting z-transformed correlation maps were then aligned for each participant to the group template T1 space and subjected to four one-sample T-tests (one for each source preference map) to reveal significant clusters of entorhinal connectivity preferences per source across all other entorhinal seeds, respectively. The functional subregions in the EC that we identify on group level generally overlap for the preferences towards the perirhinal cortex (Area 35 and Area 36) and towards the parahippocampal cortex with the findings by Maass et al., 2015. The exact procedure to determine the entorhinal seeds for further analysis is described in the main article.

Appendix 7

Co-registration procedure and alignment assessment

Appendix 7—figure 1
Co-registration procedure.

[A] Medial temporal lobe regions of interest (ROIs) were segmented on individual T2 images. Displayed is an example region (perirhinal Area 35, A35, pink) on a representative example individual T2 image. [B] Individual T2 images (blue overlay) were co-registered to whole-brain individual T1 images (upper image). The resulting warping matrices were applied to transfer the segmented ROIs from individual T2 space to individual T1 space. The co-registration procedure was manually evaluated based on landmarks (lower two images and [C]). [C] Displayed is the same example region Area 35 of the same example individual on corresponding coronal (left) and sagittal (right) slices on the individual’s T1 image (upper two images). [D] Individual echo-planar images (EPI, yellow overlay) have been co-registered to the whole-brain individual T1 images as well. [E] The inverse warping matrices were applied to warp segmented ROIs from individual T1 space to the individual EPI space. Displayed is the same example region Area 35 of the same example individual on corresponding coronal (left) and sagittal (right) slices on the individual’s EPI image (upper two images). The warping result was manually evaluated based on landmarks. [F] To evaluate results on group level, all individual T1 images were averaged to create a sample-specific T1 template. Displayed is the overlay (red boundaries) of an individual T1 image on the sample-specific T1 template. The resulting warping matrices were applied to move segmented ROIs from the individual T1 space to the sample-specific T1 template. [G] The retrosplenial cortex (RSC, green) ROI was originally segmented on the sample-specific T1 template. Respective (inverse) warping matrices were applied to move the retrosplenial ROI from the sample-specific T1 template to the individual T1 ([C], lower image) and EPI ([E], lower image) spaces. Landmark-based manual evaluation was applied to all co-registration steps. Displayed is the retrosplenial ROI (green) of an example individual on corresponding coronal slices on the individual T1 image ([C], lower image) and EPI ([E], lower image).

Appendix 8

Quality assurance measures of manually segmented regions-of-interest

The individual regions of interest were segmented by the same two experienced raters that also segmented a subsample of our data (24 hemispheres of 22 participants) for a previous publication (Berron et al., 2017). Quality assurance measures were calculated for that subsample. Regarding intra-rater reliability, the dice similarity coefficients are above 0.88 for all segmented regions (region-specific means (SD) are as follows: PHC 0.93 (0.03); Area 36 0.91 (0.02); Area 35 0.88 (0.02); EC 0.91 (0.01)). The intraclass-correlation coefficients for intra-rater reliability are all above 0.95 (PHC 0.99; Area 36 0.96; Area 35 0.97; EC 0.98). For the inter-rater reliability, dice similarity coefficients are above 0.84 for all segmented regions (region-specific means (SD) are as follows: PHC 0.86 (0.12); Area 36 0.91 (0.02); Area 35 0.84 (0.05); EC 0.87 (0.02)). The intraclass-correlation coefficients for inter-rater reliability are all above 0.78 (PHC 0.94; Area 36 0.88; Area 35 0.87; EC 0.94; see Berron et al., 2017).

Appendix 9

Metrics for transversal subiculum and CA1 segments

Transversal subiculum and CA1 segments were cut on the group template T1 images. The average number of voxels contained in each subiculum segment was 460.8 voxels for the left subiculum (standard deviation 104.36) and 458 voxels for the right subiculum (standard deviation 75.09). For the left CA1 the average equals 360 voxels (standard deviation 27.58) and 335 voxels for the right CA1 segments (standard deviation 3.56, see Appendix 9—table 1 for segment-specific values and Appendix 9—figure 1 for an illustration).

Appendix 9—figure 1
Transversal subiculum and CA1 segments.

[A] Displayed are segments cut along the transversal subiculum (red and yellow) and CA1 (cyan and dark blue) axis in the right hemisphere. Segments were cut on coronal images (as displayed in the example image) on the study-specific T1 template. [B] To cut CA1 segments, the endpoints of the transversal CA1 axis (a and b) were connected. From the middle point of that line CA1 was cut into three segments by two lines oriented in 60° angles from the line that connected a and b.

Appendix 9—table 1
Number of voxels in transversal subiculum and CA1 segments for each hemisphere.
left hemisphere (distal to proximal segments)right hemisphere (distal to proximal segments)
subiculum340419511396638338465451460575
CA1399341340337330338

Data availability

Source data that contain numerical data used to generate Figure 2, Figure 3, Figure 4, Appendix 1 Figure 2, Appendix 1 Figure 3, Appendix 1 Figure 4, Appendix 5 Figure 1 as well as group-level statistical maps (referred to as Source Code 1-16) that underlie Figure 1, Figure 2, Appendix 1 Figure 1, Appendix 1 Figure 2, Appendix 3 Figure 1, Appendix 3 Figure 2 and Appendix 4 Figure 1 have been provided under: Open Science Framework. ID 9v3qp. https://osf.io/9v3qp.

The following data sets were generated
    1. Grande X
    2. Berron D
    (2022) Open Science Framework
    ID 9v3qp. Source Data from Functional Connectivity and Information Processing in the Entorhinal-Hippocampal Circuitry.

References

  1. Book
    1. Deshmukh SS
    (2014) Spatial and nonspatial representations in the lateral entorhinal cortex
    In: Derdikman D, Knierim JJ, editors. Space,Time and Memory in the Hippocampal Formation. Springer Vienna. pp. 127–152.
    https://doi.org/10.1007/978-3-7091-1292-2_6
  2. Book
    1. Penny WD
    2. Friston KJ
    3. Ashburner JT
    4. Kiebel SJ
    5. Nichols TE
    (2011)
    Statistical Parametric Mapping: The Analysis of Functional Brain Images
    Elsevier.

Decision letter

  1. Muireann Irish
    Reviewing Editor; University of Sydney, Australia
  2. Laura L Colgin
    Senior Editor; University of Texas at Austin, United States
  3. Marshall A Dalton
    Reviewer; The University of Sydney, Australia
  4. Menno P Witter
    Reviewer; Norwegian University of Science and Technology, Norway

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Functional connectivity and information pathways in the human entorhinal-hippocampal circuitry" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Laura Colgin as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Marshall A Dalton (Reviewer #1); Menno P Witter (Reviewer #2).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

Introduction/terminology:

1. There are many different terms used interchangeably throughout the manuscript and this is confusing since it is not immediately apparent whether they cover similar notions. Throughout the paper the phrase transversal axis is used differently in many different places; usage varies from in the abstract: 'cortical – entorhinal interaction and the circuitry's inner communication along the transversal axis'; in the introduction: 'from the EC towards the transversal hippocampal axis', or 'information continues to flow in a segregated manner along the transversal human entorhinal-hippocampal axis'; in the beginning of the discussion, page 14 line 9: the 'transversal entorhinal-hippocampal axis' and there are many more. It seems that you really refer to the transverse/transversal axis of CA1/subiculum so define that, give it a name or an abbreviation if you want, and use that consistently.

2. On page 6 line 15 you state/define that you are working with two parallel input streams traditionally associated with information on object ("item") versus scene ("contextual"). In the discussion on page 14 you start with a heading that carries 'contextual information. In the following text you talk about scene representation, which is ok if one remembers that you have defined the two as synonyms, but it becomes confusing when you suddenly jump to spatial or contextual information (line 31). So, I suggest that again you start with defining the terms you want to use and use them throughout; scene=context=space and object = item.

3. In general, the Introduction could be streamlined and the research question(s) described more clearly. The authors should incorporate hypotheses/predictions in the Introduction given that numerous theoretical models make very specific predictions regarding the neuroanatomical underpinnings of scene/object processing (i.e., see Dalton and Maguire, 2017. Curr Opin Behav Sci. 17: 34-40 for an example relating to human hippocampal subfields).

4. On page 18, lines 15 to 20, there follows a paragraph that is so dense of information, that the two concepts that I think the authors are trying to convey both might get lost. You mention 'contextual features of an item' that might be the result of convergence taking place in the anterior portion of EC, i.e. outside the hippocampus, and you contrast that with a pure contextual loop that allows for subsequent converge (with what) in the hippocampus. That second loop is associated with the distal subiculum, whereas the first is actually associated with the proximal subiculum/distal CA1 border. You might want to emphasize that more deliberately. Moreover, what is the evidence that this assumed convergence in EC actually occurs, i.e. what is the evidence that anterior-lateral EC and anterior-medial EC communicate with one another; actually available data in rodents and monkeys argue against that and favor a predominant interaction along the AP-axis. Note that the summary in the abstract adds to the confusion referred to at the second bullet: you mention that the anterior-medial EC, defined through the RSC source processing scene information, whereas the second loop processes only contextual information, implying that scene and context are two different things.

Methods:

5. Considering a major strength of the paper is the detail with which anatomical ROIs were created, I'm curious to know how the authors accounted for variability in MTL anatomy across participants on steps that required warping of ROIs into a group space or, in the case of the RSC mask, warping from group template space to individual participant space. Importantly, how were the ROIs assessed for their accuracy of alignment on the EPI images?

6. Related to this, no quality assurance measures have been presented for the manually created ROIs. More information relating to these should be included in the methods section. For example, could the authors describe how many people were involved in manually segmenting the ROI's? If only one person, could the authors provide the results of intra-rater reliability analyses using, for example, the DICE overlap metric for each ROI? If more than one person was involved, could the authors also provide the results of inter-rater reliability analyses? In addition, considering it can sometimes be difficult to warp small ROIs such as hippocampal subfields from structural to EPI images with sufficient accuracy, could the authors provide some visual representations (perhaps as a supplemental figure) showing, for example, that the subiculum and CA1 ROIs created on the structural images were well aligned with the subiculum and CA1 on EPI images.

7. For ROIs whose borders lie immediately adjacent each other (e.g., medial EC and distal subiculum; anterior PHC and posterior EC), did the authors take steps to reduce the possibility of fMRI signal 'bleeding' between ROIs by, for example, ensuring the distance between borders of each ROI was greater than the spatial smoothing kernel used on the functional data? If so, could you please describe these steps in the methods section?

8. On page 20, lines 29-30, the authors describe the anatomical landmark used to create the anterior most slice of the hippocampal subfield masks but do not describe the landmark used to demarcate the posterior most extent of the masks. This should also be described.

9. The authors state that the subiculum and CA1 subfields were sagittally cut into "equally wide" segments (page 20; line 32) but no metrics relating to this are provided. Could the authors provide some metrics to support this? For example, was the mean number of functional voxels contained within each segment equivalent or did these substantially differ?

Results:

10. Although this is described in the methods section, perhaps the authors could briefly mention early in the Results section that the fMRI dataset used in Analysis 1 and 2 are derived from a task-based dataset where task-related effects have been regressed out to create a "rest-like" dataset. It may also be helpful to readers interested in (or cautious of) this particular method, to cite previous work showing that these "rest-like" datasets yield similar results to those obtained from resting-state data (e.g., Gavrilescu et al., Hum Brain Mapp 29:1040-1052, 2008 or similar).

11. Functional subregions of the EC associated with RSC, PHC, A35 and A36 are displayed very clearly in Figure 1. It appears that the EC seed relating to A36 is less 'localised' than the other seeds and displays roughly equal size clusters in both the posterior lateral and anterior medial EC. However, in the text, the authors suggest that A36 shows preferential connectivity with the posterior lateral EC (page 7; lines 7-8). How did the authors come to preferentially associate A36 with the posterior lateral EC rather than with the anterior medial EC? Did the authors conduct any additional quantitative assessments to determine this?

12. While I understand the primary focus of the subiculum and CA1 analyses relate to their transverse axis, it seems clear when looking at Figure 2, that there may also be interesting anterior-posterior differences along the body of these subfields. For example, functional clusters are largely absent in anterior portions of segment 3 of the subiculum but posterior portions of segment 3 contain functional clusters. Observations relating to the anterior-posterior axis are not discussed in text but may be informative to those with an interest in anterior-posterior variations in hippocampal function.

13. All the data are described in terms of preferred connectivity, which is what the data show. However, by mapping the preferred connectivity as absolute color maps with sharp borders, a lot of the complexities, i.e. gradients are lost. It would be great if the authors could provide supplementary maps of each source in isolation and how it maps across the surface of EC and the transverse axis of CA1/subiculum.

Interpretation:

14. The authors created ROI masks for the subiculum and refer to medial portions of this mask as 'distal subiculum'. However, the mask appears to encompass the entire 'subicular complex' inclusive of the subiculum, presubiculum and parasubiculum. Importantly, the area referred to as 'distal subiculum' throughout the manuscript (specifically, segments 1 and 2 of the subiculum mask) most likely aligns with the location of the pre- and parasubiculum in addition to containing portions of the distal subiculum. This is important considering the growing body of evidence that the pre- and parasubiculum are preferentially engaged during scene-based cognition. While this is briefly touched on in the discussion (page 13; line 5-6), I feel the interpretation of results as they relate to the 'distal subiculum' could benefit from being placed more firmly in the context of this growing body of evidence relating to the involvement of medial aspects of the hippocampus (inclusive of the pre- and parasubiculum) in scene-based cognition.

15. I am afraid that the central premise of the 'two segregated memory streams' is somewhat outdated, certainly at the level of the entorhinal cortex and hippocampus (see Doan et al. 2019). Regarding the rodent homologue of the parahippocampal cortex (referred to as POR, or postrhinal cortex) Doan et al. write: 'Postrhinal cortex preferably targets lateral instead of medial entorhinal cortex', and that '..dorsolateral parts of LEC receive inputs from both POR and PER and that both these projections show very similar topological features..'

16. Nevertheless, in the present study the authors claim that 'The lateral EC preferentially communicates with the perirhinal cortex' (page 3, line 20). In the same paragraph they write that context processing should be specific to the MEC, in contrast to the LEC that would rather process non-spatial and item information. Strikingly, they then go on to cite Doan et al. 2019 and effectively contradict themselves within the same paragraph '…parahippocampal cortex communicates with the EC along its full extend' (as Doan et al. write, the LEC is similarly connected to POR (PHC) and PER). This means that contextual information from PHC should not uniquely be passed on to the MEC. On the contrary, it should even be more strongly passed on to the LEC than the MEC.

17. I am rather puzzled why the authors cite Nilssen et al. and Doan et al. and still uphold the segregated stream hypothesis of entorhinal and hippocampal subregions. In my reading, those studies are entirely incompatible with the view of two parallel, segregated memory streams involving the LEC and MEC. Ironically, the null finding of category specific processing in the lateral cluster of EC voxels, and associated regions, might be due to the convergent, multimodal input patterns. Unfortunately, the two segregated stream hypothesis seems to be a central tenet in the current study.

18. There seems to be no main effect of activations to object stimuli (difference to zero) in any region. Nevertheless the authors claim that regions show processing of object-related information. This is not warranted based on the results. For example in the abstract: 'The regions of another route, that connects the anterior-lateral EC and a newly identified retrosplenial-based anterior-medial EC subregion with the CA1/subiculum border, process object and scene information similarly'. If there is no evidence of any level of processing, it is not warranted to claim 'similar processing'.

19. Based on the introduction, a double dissociation between object and scene processing in two segregated sets of regions is expected, but the authors don't provide evidence for this. In my view the interaction effect (and post-hoc tests) for higher scene processing in segment 1and2 in Figure 4 is a very interesting and meaningful finding. However, neither main effects nor interactions are presented to support claims of any level of 'object processing'. How can we be sure that we are not looking at pure noise in the object condition? Absence of evidence is not equal to evidence of absence (of different object/ scene processing).

20. No 'information flow' is actually assessed in this study. In my view, this would require a directed connectivity analysis such as dynamic causal modeling or transfer entropy. The research question or hypotheses should pertain to what is actually being done. For example, 'we predict functional connectivity between X-Y to reflect the structural connectivity described in rodents/ humans previously', and 'we expect specific sensitivity to scene stimuli in distal subiculum, because of connectivity to the MEC region in rodents'. This would be 'consistent with information flow' between those regions, but it is not directly or conclusively showing it.

21. The approach to segment the EC into sub-clusters based on known connectivity to other regions seems fine in general. However, this should be guided by gold-standard, a priori knowledge of entorhinal subregions such as the rodent MEC and LEC (or finer cyto/ myeloarchitectonic subdivisions, for example described in Kimer et al. 1997). In the present study, the authors select four cortical seed regions to segment entorhinal subregions based on functional connectivity, without providing appropriate justification how some of those regions would uniquely connect to specific, cytoarchitectonically-defined entorhinal subregions. For example, BA35, BA36, and PHC all project to the rodent LEC, and PHC also project (albeit weaker) to the MEC. Which a-priori defined entorhinal subregions should be uniquely identified with these four seed regions? Following the semipartial correlation analyses, clusters of EC voxels are then segmented and labeled posterior-medial, posterior-lateral, anterior-medial and anterior-lateral EC. Those names are then used quasi synonymously with coherent and established, cytoarchitectonically defined subregions such as the rodent LEC and MEC. This seems egregious, not only because of the questionable choice of initial seed regions, but more so because of the discontinuous topography of the segmentations shown in Figure 1. What is the ground-truth (cytoarchitecture, myeloarchitecture, tracing studies, gene expression) evidence for a salt-and-pepper organization of entorhinal subregions? The clusters can't correspond to coherent cytoarchitectonic regions, so why would they be referred to as such.

22. Why did the authors not simply use the ROIs from their previous identification of the human homologues of the MEC and LEC (Maass et al. 2015) to address their current research questions? In summary, this could be reconciled by a consistently used naming scheme for the entorhinal seeds that avoids confusion with cytoarchitectonic subregions.

23. In the results you describe data on proximal CA1, but you do not mention them anywhere in the paper explicitly. If your data do not allow you to functionally 'interpret' proximal CA1, you might come back to that in the discussion, state this and mention that in the rodent literature there is a gradient along transverse CA1 with more precise spatial information in proximal than in distal. Any ideas of why that does not seem to hold in humans? In particular on page 15, line 31 you make a very general statement about the transverse organization of CA1, so that might be the place to elaborate a bit more.

Clinical relevance:

24. The reported findings may have clinical relevance, but this is to be determined by future studies. An entire paragraph in the Introduction is dedicated to this topic (starting on page 4, line34). The authors state that the findings show that tau pathology spreads through 'functionally connected' regions. Functionally connected regions must also be structurally connected (through mono or polysynaptic connections). It seems the authors are insinuating that functional and structural connections are independent. The prediction of functional connectivity between parahippocampal, perirhinal and entorhinal cortices, and the subiculum is already abundantly well founded on previous findings of tracing studies in rodents and primates, and even the authors' previously published functional MRI findings (Maass et al. 2015). The fact that the spreading of tau pathology follows structural and functional connections provides no additional predictive value to inform the research question (in my understanding these are: (1) is there a specific functional connectivity pattern between regions, (2) do specific regions show differential fMRI activations for object vs scene stimuli). It appears unwarranted and misleading to portray the clinical findings as a formal motivation (meaning, previous findings that form the basis for the research hypotheses) for the present study – page 5, line 8. If this was not the intention, then I feel this needs to be formulated more clearly.

Reviewer #1 (Recommendations for the authors):

I'd first like to congratulate the authors for their fine and detailed work investigating entorhinal-hippocampal information pathways. It is exciting to see functional imaging experiments with such a focus on anatomical detail and I look forward to reading future work from these authors. The paper does a very good job at underscoring the importance of characterising the functional organisation of entorhinal-hippocampal pathways and their relationship to surrounding extra-hippocampal cortices. The authors present exciting new evidence that extra-hippocampal cortical areas display preferential patterns of functional connectivity with subregions of the entorhinal cortex (EC). These EC subregions, in turn, display preferential patterns of functional connectivity along the transverse axis of the subiculum and area CA1 of the hippocampus. In addition, the authors show that the posterior medial EC and distal subiculum are preferentially engaged during 'scene' processing. In contrast, anterior portions of the EC and the CA1/subiculum border engage equally during both 'scene' and 'object' processing. Overall, this work has important implications for our understanding of information transfer between the entorhinal cortex and hippocampus.

My suggestions below, rather than criticisms, largely reflect potential issues/questions that I believe, if addressed, may improve the clarity and interpretation of results.

1) Considering a major strength of the paper is the detail with which anatomical ROIs were created, I'm curious to know how the authors accounted for variability in MTL anatomy across participants on steps that required warping of ROIs into a group space or, in the case of the RSC mask, warping from group template space to individual participant space. Importantly, how were the ROIs assessed for their accuracy of alignment on the EPI images?

2) Related to this, no quality assurance measures have been presented for the manually created ROIs. More information relating to these should be included in the methods section. For example, could the authors describe how many people were involved in manually segmenting the ROI's? If only one person, could the authors provide the results of intra-rater reliability analyses using, for example, the DICE overlap metric for each ROI? If more than one person was involved, could the authors also provide the results of inter-rater reliability analyses? In addition, considering it can sometimes be difficult to warp small ROIs such as hippocampal subfields from structural to EPI images with sufficient accuracy, could the authors provide some visual representations (perhaps as a supplemental figure) showing, for example, that the subiculum and CA1 ROIs created on the structural images were well aligned with the subiculum and CA1 on EPI images. I believe such a representation would increase reader confidence.

3) For ROIs whose borders lie immediately adjacent each other (e.g., medial EC and distal subiculum; anterior PHC and posterior EC), did the authors take steps to reduce the possibility of fMRI signal 'bleeding' between ROIs by, for example, ensuring the distance between borders of each ROI was greater than the spatial smoothing kernel used on the functional data? If so, could you please describe these steps in the methods section?

4) The authors created ROI masks for the subiculum and refer to medial portions of this mask as 'distal subiculum'. However, the mask appears to encompass the entire 'subicular complex' inclusive of the subiculum, presubiculum and parasubiculum. Importantly, the area referred to as 'distal subiculum' throughout the manuscript (specifically, segments 1 and 2 of the subiculum mask) most likely aligns with the location of the pre- and parasubiculum in addition to containing portions of the distal subiculum. This is important considering the growing body of evidence that the pre- and parasubiculum are preferentially engaged during scene-based cognition. While this is briefly touched on in the discussion (page 13; line 5-6), I feel the interpretation of results as they relate to the 'distal subiculum' could benefit from being placed more firmly in the context of this growing body of evidence relating to the involvement of medial aspects of the hippocampus (inclusive of the pre- and parasubiculum) in scene-based cognition.

5) Functional subregions of the EC associated with RSC, PHC, A35 and A36 are displayed very clearly in Figure 1. It appears that the EC seed relating to A36 is less 'localised' than the other seeds and displays roughly equal size clusters in both the posterior lateral and anterior medial EC. However, in the text, the authors suggest that A36 shows preferential connectivity with the posterior lateral EC (page 7; lines 7-8). How did the authors come to preferentially associate A36 with the posterior lateral EC rather than with the anterior medial EC? Did the authors conduct any additional quantitative assessments to determine this?

6) I am a little confused by the colour scheme in the 3D hippocampal subfield models presented in Figures 2, 4 and S2. No EC seeds are colour coded brown but functional clusters colour coded brown are present in both the subiculum and CA1. Does brown represent an overlap between clusters? If not, could the authors please explain what these brown functional clusters correspond to?

7) While I understand the primary focus of the subiculum and CA1 analyses relate to their transverse axis, it seems clear when looking at Figure 2, that there may also be interesting anterior-posterior differences along the body of these subfields. For example, functional clusters are largely absent in anterior portions of segment 3 of the subiculum but posterior portions of segment 3 contain functional clusters. Observations relating to the anterior-posterior axis are not discussed in text but may be informative to those with an interest in anterior-posterior variations in hippocampal function.

8) The authors state that the subiculum and CA1 subfields were sagittally cut into "equally wide" segments (page 20; line 32) but no metrics relating to this are provided. Could the authors provide some metrics to support this? For example, was the mean number of functional voxels contained within each segment equivalent or did these substantially differ?

9) For Analysis 3, the authors report the results of contrasts between the 'object' condition vs 'baseline and the 'scene' condition vs 'baseline'. Is there a reason the authors did not report results relating to the direct contrast of 'scene' vs. 'object' conditions? While acknowledging the visual stimuli between these two conditions would not be equally matched, it would nonetheless be interesting to see if, for 'scenes' > 'objects', the posterior medial EC – distal subiculum effect remains. It would also be interesting if the 'object' > 'scene' contrast revealed effects not observed for the contrast of 'object' > 'baseline'. Could the authors offer some insights into why these direct contrasts were not analysed?

10) Although this is described in the methods section, perhaps the authors could briefly mention early in the Results section that the fMRI dataset used in Analysis 1 and 2 are derived from a task-based dataset where task-related effects have been regressed out to create a "rest-like" dataset. It may also be helpful to readers interested in (or cautious of) this particular method, to cite previous work showing that these "rest-like" datasets yield similar results to those obtained from resting-state data (e.g., Gavrilescu et al., Hum Brain Mapp 29:1040-1052, 2008 or similar).

11) It may also be helpful to specify early in the Results section that hippocampal subfield analyses were conducted only on the body of the hippocampus.

Reviewer #2 (Recommendations for the authors):

The authors are to be complimented with an impressive combinatorial attempt to understand the organization of the entorhinal cortex and its position as part of the cortico-hippocampal processing streams in the human brain, which are critical components of the human episodic memory system. This is an important and as yet unresolved issue, largely due to the lack of proper imaging technology. Even in experimental animal studies with all the current versatile experimental tools to selectively manipulate parts of the circuits, the organization of these parallel cortical streams is as yet not clear. The conclusions are supported by strong data, obtained through experiments that, as far as I can evaluate, are sound. In particular the use of several carefully chosen cortical sources to differentiate between four 'functionally' different entorhinal seeds that subsequently are mapped onto CA1 and subiculum is a major contribution to the field.

Having said this, I strongly feel that the manuscript could improve substantially; a critical reappraisal on how the experimental data are presented and how the overall topic is introduced and handled is recommended. There are many different terms used interchangeably throughout the manuscript and this is confusing since it is not immediately apparent whether they actually cover similar notions:

• Throughout the paper you use transversal axis in many different phrases and context resulting in massive confusion; usage varies from in the abstract: 'cortical – entorhinal interaction and the circuitry's inner communication along the transversal axis'; in the introduction: 'from the EC towards the transversal hippocampal axis', or 'information continues to flow in a segregated manner along the transversal human entorhinal-hippocampal axis'; in the beginning of the discussion, page 14 line 9: the 'transversal entorhinal-hippocampal axis' and there are many more. It seems that you really refer to the transverse/transversal axis of CA1/subiculum so define that, give it a name or an abbreviation if you want, and use that consistently. This is important since EC has a transverse axis of its own, which you apparently refer to as the lateral-to-medial axis, but that does not simply relate to the transversal hippocampal axis.

• On page 6 line 15 you state/define that you are working with two parallel input streams traditionally associated with information on object ("item") versus scene ("contextual"). In the discussion on page 14 you start with a heading that carries 'contextual information. IN the following text you talk about scene representation, which is ok if one remembers that you have defined the two as synonyms, but it becomes confusing when you suddenly jump to spatial or contextual information (line 31). So, I suggest that again you start with defining the terms you want to use and use them throughout; scene=context=space and object = item. If you want to use all these different concepts interchangeable, you might want to consider to expand this even further; although you have not tested sequence coding, would you be inclined to extend object=item=sequence and how do the two sets of terms relate to the proposal of allocentric versus egocentric representations as proposed among others by Knierim in some of his recent papers. To further complicate this potential confusion: objects may be part of a context or a scene. (Note that I do not have the solution for this nomenclatural nightmare either, but try to keep it as simple as possible in this manuscript).

• On page 18, lines 15 to 20, there follows a paragraph that is so dense of information, that the two concepts that I think the authors are trying to convey both might get lost. You mention 'contextual features of an item' that might be the result of convergence taking place in the anterior portion of EC, i.e. outside the hippocampus, and you contrast that with a pure contextual loop that allows for subsequent converge (with what) in the hippocampus. That second loop is associated with the distal subiculum, whereas the first is actually associated with the proximal subiculum/distal CA1 border. You might want to emphasize that more deliberately. Moreover, what is the evidence that this assumed convergence in EC actually occurs, i.e. what is the evidence that anterior-lateral EC and anterior-medial EC communicate with one another; actually available data in rodents and monkeys argue against that and favor a predominant interaction along the AP-axis. Note that the summary in the abstract adds to the confusion referred to at the second bullet: you mention that the anterior-medial EC, defined through the RSC source processing scene information, whereas the second loop processes only contextual information, implying that scene and context are two different things.

Four more general comments:

• In the results you do describe data on proximal CA1, but you do not mention them anywhere in the paper explicitly. If your data do not allow you to functionally 'interpret' proximal CA1, you might come back to that in the discussion, state this and mention that in the rodent literature there is a gradient along transverse CA1 with more precise spatial information in proximal than in distal. Any ideas of why that does not seem to hold in humans? In particular on page 15, line 31 you make a very general statement about the transverse organization of CA1, so that might be the place to elaborate a bit more.

• I found it of interest that in the most medial parts, likely the ambiens gyrus area in both the left and right hemisphere there seems to be an indication of a second representation of four seeds associated with the four sources. Is this correct and would the authors care to comment on this?

• You divide the hippocampus along the transverse axis and EC along the lateral-medial (transverse) axis and the anterior-posterior axis. The hippocampus also has an AP axis and even though I learned from the method section that you only sampled the body, leaving out the uncal portion and the tail, for good reasons, there is still a pretty substantial AP axis left to sample from; any differences along the AP axis? If not, it might be relevant to mention that explicitly.

• All the data are described in terms of preferred connectivity, which is what the data show. However, by mapping the preferred connectivity as absolute color maps with sharp borders, a lot of the complexities, i.e. gradients are lost. It would be great if the authors could provide supplementary maps of each source in isolation and how it maps across the surface of EC and the transverse axis of CA1/subiculum.

I have several more detailed, smaller comments listed below in sequence as they appear in the paper (not relevance):

Page 5, line 16/17: The way the EC communicates with different regions of these two cortical streams implies topographical differences in information processing within the EC. Please rephrase and make it more concrete.

Page 6, line 29: This is to our knowledge…; does this refer to Syvertsen et all or to the present study?

Page 6, last line you mention tau pathology as being specifically occurring in this part of the brain. Of course, that is correct in AD but not in many other tau-opathies so you might want to add AD, which your first do only 7 lines further down.

Page 12, line 3: Please carefully read this heading: it states the opposite of the sentence in lines 10/11. The latter is likely correct as this is repeated throughout the paper.

Page 15, line 2: insert full stop between 'Our'

Page 16 lines 15-19, please rephrase. The projections from RSC to deep layers on EC in the rodent are really limited to the area defined as MEC based on connectivity and that is true in the monkey as well. The area defined here as the RSC connected area seems a lot more anterior, and has been 'considered' to be the medial part of LEC. So this needs better wording to avoid further confusion.

Likewise, page 16, lines 23 to 25 are a bit misleading, again as the result of the complexity of terminology and the risk of circular definitions/statements. If RSC is a defining input of MEC, then the statement 'The mapping of the anterior-medial EC (identified by retrosplenial connectivity) to the subiculum/CA1 border opposes conventional views that the medial EC communicates with the distal subiculum and proximal CA1 (based on rodent anatomy – see e.g. Nilssen et al., 2019)' is correct and it shows that in the human the connectivity or anterior medial entorhinal cortex might be different. However, the next sentence: 'It is feasible that complex interactions within the EC underlie this observation' opens up for another interpretation, namely that what the authors define as the anterior-medial portion of EC is not MEC but might be LEC and that the resulting signal correlations to not indicate real anatomical connectivity.

Though I credit the authors that they clearly defined the potential methodological difficulties, in parts like the above, they might critically reappraise their writing, as to avoid possible confusion in the readership.

Figure 1: the L-M axis is the wrong way around, flip L and M.

Figure 2: add in the figure text indicating that left is subiculum, right is CA1 and explain the numbers so indicate the proximal-distal axis in the two 3D figures 1 is distal in sub but proximal in CA1. Without such indications the figure is incomprehensible for non-anatomical experts and one needs to filter through the methods to get that. Do the same in figure 4.

In both figures 2 and 4 the overlap of various colors results in some vague brown coloring that is hard to interpret. Please indicate which inputs result in this color pattern.

Reviewer #3 (Recommendations for the authors):

– There seems to be no main effect of activations to object stimuli (difference to zero) in any region. Nevertheless the authors claim that regions show processing of object-related information. This is not warranted based on the results. For example in the abstract: 'The regions of another route, that connects the anterior-lateral EC and a newly identified retrosplenial-based anterior-medial EC subregion with the CA1/subiculum border, process object and scene information similarly'. If there is no evidence of any level of processing, it is not warranted to claim 'similar processing'. Based on the introduction, a double dissociation between object and scene processing in two segregated sets of regions is expected, but the authors don't provide evidence for this. In my view the interaction effect (and post-hoc tests) for higher scene processing in segment 1and2 in Figure 4 is a very interesting and meaningful finding. However, neither main effects nor interactions are presented to support claims of any level of 'object processing'. How can we be sure that we are not looking at pure noise in the object condition? Absence of evidence is not equal to evidence of absence (of different object/ scene processing).

– The introduction left me wondering which research questions are actually being pursued. In general the introduction could be streamlined, as it seemed to contain apparently irrelevant information. Most importantly, the research question(s) need to be described much more clearly. The closest thing to a research question was written on page 4 line 17-18: '..the hypothesis from rodent research that information continues to flow in a segregated manner along the transversal human entorhinal-hippocampal axis.' I take issue with a couple of things here.

1) No 'information flow' is actually being assessed in this study. In my view, this would require a directed connectivity analysis such as dynamic causal modeling or transfer entropy. The research question or hypotheses should pertain to what is actually being done. For example, 'we predict functional connectivity between X-Y to reflect the structural connectivity described in rodents/ humans previously', and 'we expect specific sensitivity to scene stimuli in distal subiculum, because of connectivity to the MEC region in rodents'. This would be 'consistent with information flow' between those regions, but it is not directly or conclusively showing it.

2) More importantly, I am afraid that the central premise of the 'two segregated memory streams' is by now somewhat outdated, certainly at the level of the entorhinal cortex and hippocampus (see Doan et al. 2019). Regarding the rodent homologue of the parahippocampal cortex (referred to as POR, or postrhinal cortex) Doan et al. write: 'Postrhinal cortex preferably targets lateral instead of medial entorhinal cortex', and that '..dorsolateral parts of LEC receive inputs from both POR and PER and that both these projections show very similar topological features..'

Nevertheless, in the present study the authors claim that 'The lateral EC preferentially communicates with the perirhinal cortex' (page 3, line 20). In the same paragraph they write that context processing should be specific to the MEC, in contrast to the LEC that would rather process non-spatial and item information. Strikingly, they then go on to cite Doan et al. 2019 and effectively contradict themselves within the same paragraph '…parahippocampal cortex communicates with the EC along its full extend' (as Doan et al. write, the LEC is similarly connected to POR (PHC) and PER). This means that contextual information from PHC should not uniquely be passed on to the MEC. On the contrary, it should even be more strongly passed on to the LEC than the MEC.

Nilssen et al. 2019 (whom the authors also cite) discuss the multimodal inputs to the LEC, including from the parahippocampal cortex (POR in rodents). They write 'In line with this shared input, we propose that it is the PER/LEC interface that provides the optimal substrate to detect changes in the context.'

I am rather puzzled why the authors cite Nilssen et al. and Doan et al. and still uphold the segregated stream hypothesis of entorhinal and hippocampal subregions. In my reading, those studies are entirely incompatible with the view of two parallel, segregated memory streams involving the LEC and MEC. Ironically, the null finding of category specific processing in the lateral cluster of EC voxels, and associated regions, might be due to the convergent, multimodal input patterns. Unfortunately, the two segregated stream hypothesis seems to be a central tenet in the current study. But it is entirely possible that I am overemphasizing or misunderstanding the premise of the study, in which case a critical rewriting could clear out the issues pointed out above.

– The approach to segment the EC into sub-clusters based on known connectivity to other regions seems fine in general. However, this should be guided by gold-standard, a priori knowledge of entorhinal subregions such as the rodent MEC and LEC (or finer cyto/ myeloarchitectonic subdivisions, for example described in Kimer et al. 1997). In the present study, the authors select four cortical seed regions to segment entorhinal subregions based on functional connectivity, without providing appropriate justification how some of those regions would uniquely connect to specific, cytoarchitectonically-defined entorhinal subregions. For example, BA35, BA36, and PHC all project to the rodent LEC, and PHC also project (albeit weaker) to the MEC. Which a-priori defined entorhinal subregions should be uniquely identified with these four seed regions? Following the semipartial correlation analyses, clusters of EC voxels are then segmented and labeled posterior-medial, posterior-lateral, anterior-medial and anterior-lateral EC. Those names are then used quasi synonymously with coherent and established, cytoarchitectonically defined subregions such as the rodent LEC and MEC. This seems egregious, not only because of the questionable choice of initial seed regions, but more so because of the discontinuous topography of the segmentations shown in Figure 1. What is the ground-truth (cytoarchitecture, myeloarchitecture, tracing studies, gene expression) evidence for a salt-and-pepper organization of entorhinal subregions? The clusters can't correspond to coherent cytoarchitectonic regions, so why would they be referred to as such. The BA36-related cluster seems to be all over the place. Why is it referred to as 'posterior-lateral EC'?. The discontinuities are arguably the result of sources of noise and error in fMRI. I am not saying that this general approach is necessarily useless, but the naming of the clusters should not be misleading. E.g. they could be referred to as 'RSC-connected voxels in EC' etc., or their predominant anatomical location could be used, but after actual quantification of their predominant location and including the specification that it is a discontinuous cluster of voxels – not a coherent subregion that has cytoarchitectonic counterpart. Why did the authors not simply use the ROIs from their previous identification of the human homologues of the MEC and LEC (Maass et al. 2015) to address their current research questions? In summary, this could be reconciled by a different, consistently used naming scheme for the entorhinal seeds that avoids confusion with cytoarchitectonic subregions.

https://doi.org/10.7554/eLife.76479.sa1

Author response

Essential revisions:

introduction/terminology:

1. There are many different terms used interchangeably throughout the manuscript and this is confusing since it is not immediately apparent whether they cover similar notions. Throughout the paper the phrase transversal axis is used differently in many different places; usage varies from in the abstract: 'cortical – entorhinal interaction and the circuitry's inner communication along the transversal axis'; in the introduction: 'from the EC towards the transversal hippocampal axis', or 'information continues to flow in a segregated manner along the transversal human entorhinal-hippocampal axis'; in the beginning of the discussion, page 14 line 9: the 'transversal entorhinal-hippocampal axis' and there are many more. It seems that you really refer to the transverse/transversal axis of CA1/subiculum so define that, give it a name or an abbreviation if you want, and use that consistently.

We thank you for that helpful observation and we have changed our wording towards “transversal sub/CA1 axis” throughout the entire manuscript.

We introduce the name on page 3, line 7:

“…the transversal axis of hippocampal subiculum and CA1 (here referred to as transversal sub/CA1 axis).”

2. On page 6 line 15 you state/define that you are working with two parallel input streams traditionally associated with information on object ("item") versus scene ("contextual"). In the discussion on page 14 you start with a heading that carries 'contextual information. In the following text you talk about scene representation, which is ok if one remembers that you have defined the two as synonyms, but it becomes confusing when you suddenly jump to spatial or contextual information (line 31). So, I suggest that again you start with defining the terms you want to use and use them throughout; scene=context=space and object = item.

Indeed, we see and agree that it eases readability if we stick to one set of terms. We appreciate your comment and decided to use “object versus scene information” throughout. This wording remains closest to our operationalization. Nevertheless, we mention that other terms have been used to define these representations. See therefore the footnote in the introduction on page 3, line 11:

1 In light of confusing nomenclature, here we adhere to scene and object information (elsewhere referred to as contextual, spatial or ‘Where’ and content, non-spatial, item or ‘What’ information, respectively).”

3. In general, the Introduction could be streamlined and the research question(s) described more clearly. The authors should incorporate hypotheses/predictions in the Introduction given that numerous theoretical models make very specific predictions regarding the neuroanatomical underpinnings of scene/object processing (i.e., see Dalton and Maguire, 2017. Curr Opin Behav Sci. 17: 34-40 for an example relating to human hippocampal subfields).

We appreciate your helpful advice. To streamline the introduction better (and likewise incorporate comment #15 #17) we deleted the rather historical overview on the former parallel mapping hypothesis and point earlier to the problem at hand.

See therefore page 3, line 10:

“Large-scale cortical information streams, that originate in the visual ‘Where’ and ‘What’ pathways and process scene and object information (Berron et al., 2018; Haxby et al., 1991; Ranganath and Ritchey, 2012; Ritchey et al., 2015; Ungerleider and Haxby, 1994), map onto the EC in a complex manner and define functional EC subregions. Recent rodent research updates the former conception of a parallel mapping of scene and object information via parahippocampal and perirhinal cortices onto medial versus lateral EC subregions (cf. posterior-medial versus anterior-lateral EC subregions as the human homologues; Maass, Berron et al., 2015; Navarro Schröder et al., 2015). Instead of a strict parallel mapping, profound cross-projections exist from the parahippocampal cortex towards the perirhinal cortex and the lateral EC (Nilssen et al., 2019). In accordance, information seems to converge in the rodent lateral EC (Doan et al., 2019). The update, thus, implies a more complex functional organization than parallel scene and object information mapping. Moreover, this advance highlights the retrosplenial cortex as an additional source to convey information directly from the cortical scene processing stream onto the EC. The retrosplenial cortex projects to the medial EC and, like the parahippocampal cortex, is part of the scene processing stream (e.g. involved in scene translation; Vann et al., 2009; Nilssen et al., 2019; Witter et al., 2017). The update, furthermore, evokes the question how cortical sources of information uniquely map onto the EC and which kind of information is processed in the resulting functional EC subregions.

Within the entorhinal-hippocampal circuitry, an important direct way of communication exists between the EC and hippocampal subiculum and CA1. How functional EC subregions communicate towards the transversal sub/CA1 axis in humans is, however, unclear. Similarly, the extent to which specific scene and object information processing routes might emerge, despite information convergence in the EC, is unknown.”

In addition, we directly point out that evidence on transversal information processing in the hippocampal subiculum and CA1 is mixed. We here also incorporated the work by e.g. Dalton and Maguire (2017). See page 3, line 32:

“On one hand, rodent research indicates a transversal organization where scene and object information is processed along two anatomically wired routes, the medial EC – distal subiculum – proximal CA1 route and the lateral EC – proximal subiculum – distal CA1 route, respectively (Witter et al., 2017; note sparse functional evidence in the subiculum: Ku et al., 2017; Cembrowski et al., 2018; but frequent reports in the rodent CA1 region: Henriksen et al., 2010; Nakamura et al., 2013; Igarashi et al., 2014; Nakazawa et al., 2016; Beer et al., 2018). Initial functional and structural connectivity data also indicate such a transversal connectivity profile in humans (Maass, Berron et al., 2015; Syversen et al., 2021). In accordance, scene information seems to be preferentially processed in the distal subiculum (Dalton et al., 2018; Dalton and Maguire, 2017; Zeidman et al., 2015) and hints exist for preferential object processing at the subiculum/CA1 border (Dalton et al., 2018). On the other hand, anatomical projections in the monkey show a longitudinal profile on top of the transversal profile with mainly the anterior-lateral and posterior-lateral entorhinal portions projecting to the distal subiculum – proximal CA1 and proximal subiculum – distal CA1, respectively (Witter and Amaral, 2020). According to information convergence in the EC, a recent report finds convergence along the rodent transversal CA1 axis (Vandrey et al., 2021). In humans, visual stream projections towards the entorhinal-hippocampal circuitry similarly suggest convergence of scene and object information in the subiculum/CA1 border region but preserved scene processing in the distal subiculum (Dalton and Maguire, 2017). A detailed examination of the latter hypothesis is, however, lacking. The diversity of findings emerging from the literature calls for a thorough investigation to elucidate whether multiple transversal processing routes exist within the human entorhinal-hippocampal circuitry.”

We also make the specific problem more apparent and clearly state which questions we aim to answer.

See therefore page 4, line 20:

“To summarize, our conception of how information travels towards the entorhinal-hippocampal circuitry underwent key changes which warrant an extensive exploration of the circuitry’s functional organization. First, rodent research shows that there is no strict parallel mapping of cortical information from the perirhinal and parahippocampal cortex towards the EC. Second, information seems to converge already before the hippocampus.”

Page 4, line 35:

“With a combination of functional connectivity and information processing analyses, we seek to answer two sets of questions. Regarding functional connectivity, we ask where the parahippocampal, perirhinal and retrosplenial cortical sources uniquely map onto the human EC and how these functionally connected routes continue between EC subregions and the transversal sub/CA1 axis. Regarding information processing, we ask whether and where scene and object information are specifically processed in the EC and along the transversal sub/CA1 axis.”

We, moreover, clearly state our hypotheses in the introduction on page 5, line 5:

“We test the hypotheses of (1) a transversal functional connectivity pattern and (2) multiple information processing routes within the entorhinal-hippocampal circuitry. Thus, following the updated conception of a non-parallel cortical scene and object information mapping onto the EC in rodents, we will show how cortical information streams map onto the EC in humans. This mapping will then be our detailed starting point to investigate the functional connectivity and information processing within the entorhinal-hippocampal circuitry.”

To keep the introduction concise, we specifically outline our predictions based on previous literature at the respective places in the results. See page 7, line 13:

“Following the characterization of entorhinal seeds, we focused on the functional connectivity between these entorhinal subregions and hippocampal subiculum and CA1 to test the hypothesis of a transversal functional connectivity pattern. We predicted that while some EC subregions have a preference to functionally connect with the subiculum/CA1 border, others preferentially connect with the distal subiculum and proximal CA1. In the previous step we identified EC subregions based on unique cortical source contributions. Therefore, our predictions remained in accordance with Maass, Berron et al. (2015): We expected that the EC subregion preferentially connected with the parahippocampal cortex (ECPHC-based seed) maps towards the distal subiculum and EC subregions connected with the perirhinal cortex (ECArea35-based seed, ECArea36-based seed) map towards the proximal subiculum, a mapping that we predicted to be extended towards the distal CA1.”

Page 10, line 4:

“Besides the intrinsic functional connectivity patterns within the entorhinal-hippocampal circuitry, we also examined the characteristics of scene and object information processing to test the hypothesis of multiple information processing routes within the entorhinal-hippocampal circuitry. We predicted a route of specific scene processing and another route of convergent information processing. Following the proposal by Dalton and Maguire (2017) and the updated cross-projections from the scene to the object information processing stream (Nilssen et al., 2019), we expected scene processing in the distal subiculum. The updated parahippocampal cross-projections imply convergence wherever specific object processing had been expected previously. Thus, we explored whether any entorhinal-hippocampal subregions still process object information specifically. However, we largely expected to find evidence consistent with convergent processing of scene and object information within the entorhinal-hippocampal circuitry.”

Finally, we excluded the reference to potential clinical implications from the introduction and moved this aspect entirely to the discussion.

4. On page 18, lines 15 to 20, there follows a paragraph that is so dense of information, that the two concepts that I think the authors are trying to convey both might get lost. You mention 'contextual features of an item' that might be the result of convergence taking place in the anterior portion of EC, i.e. outside the hippocampus, and you contrast that with a pure contextual loop that allows for subsequent converge (with what) in the hippocampus. That second loop is associated with the distal subiculum, whereas the first is actually associated with the proximal subiculum/distal CA1 border. You might want to emphasize that more deliberately. Moreover, what is the evidence that this assumed convergence in EC actually occurs, i.e. what is the evidence that anterior-lateral EC and anterior-medial EC communicate with one another; actually available data in rodents and monkeys argue against that and favor a predominant interaction along the AP-axis. Note that the summary in the abstract adds to the confusion referred to at the second bullet: you mention that the anterior-medial EC, defined through the RSC source processing scene information, whereas the second loop processes only contextual information, implying that scene and context are two different things.

Thank you for this thoughtful comment. We realize that the paragraph is not only dense but that our wording was also misleading. Indeed, we are not aware of explicit evidence that anterior-lateral EC and anterior-medial EC communicate with each other. What we observe, however, is that functionally adjacent cortical regions do not show convergence but specific object versus scene processing (perirhinal and retrosplenial source regions, respectively). As we interpret our functional activity pattern in the two anterior EC seeds ECArea35-based and ECRSC-based as consistent with convergence, we thus conclude that convergence happens before the hippocampus – and it may occur in the anterior EC. Future research can obtain the exciting investigation of how and where convergence happens.

We hope that the following changes to that paragraph make our point clearer on page 17, line 11:

“Dalton and Maguire (2017), however, made a relevant proposal based on visual processing pathways and information processing. In correspondence to our results, they proposed the subiculum/CA1 border as a point of convergence between scene and object information processing streams. While their conclusion was based on direct parahippocampal, retrosplenial and perirhinal connections to the hippocampus, we found that both, the ECArea35-based (that is connected with the cortical object processing stream) and the ECRSC-based (that is connected with the cortical scene processing stream) show connectivity with the subiculum/CA1 border (see also appendix V for information processing in cortical source regions). Convergence is potentially also achieved via recurrency within the entorhinal-hippocampal system and cortical regions (cf. Koster et al., 2018 for evidence on recurrency). These considerations are an exciting future research avenue and remain speculative based on the current data due to insufficient temporal resolution. We nevertheless hypothesize the existence of two processing routes: one that processes converged object and scene information and one that processes scene information specifically. Thus, scene and object information processing might converge before the hippocampus. This presumably occurs within the anterior EC, given object-specific and scene-specific processing take place in the cortical source regions of the ECArea35-based and ECRSC-based subregions, respectively (see appendix V). Here, objects may be bound together with their defining scene-like or contextual features (akin to the “object-in-location” idea in Connor and Knierim, 2017; Knierim et al., 2014). In addition, the dedicated scene processing that we observe along the ECPHC-based – distal subiculum route, may functionally underpin ideas about an anatomically graded contextual scaffold that the hippocampus utilizes to incorporate detailed information from the object-in-scene route into meaningful chunks of cohesive memory representations ("events"; Behrens et al., 2018; Clewett et al., 2019; Robin, 2018; Robin and Olsen, 2019).”

To avoid confusion, we moreover changed the wording in the abstract on page 2, line 9:

“Our data show specific scene processing in the functionally connected ECPHC-based and distal subiculum. Another route, that functionally connects the ECArea35-based and a newly identified ECRSC-based with the subiculum/CA1 border, however, shows no selectivity between object and scene conditions. Our results are consistent with transversal information-specific pathways in the human entorhinal-hippocampal circuitry, with anatomically organized convergence of cortical processing streams and a unique route for scene information.”

Methods:

5. Considering a major strength of the paper is the detail with which anatomical ROIs were created, I'm curious to know how the authors accounted for variability in MTL anatomy across participants on steps that required warping of ROIs into a group space or, in the case of the RSC mask, warping from group template space to individual participant space. Importantly, how were the ROIs assessed for their accuracy of alignment on the EPI images?

Thank you for raising that important point. To assess accurate warping, we performed careful manual checks on the co-registered and warped single-subject data based on landmark overlays. To illustrate the process, we now incorporated Figure 1 in Appendix 7. The figure shows the stepwise co-registration process between group and individual participant space. It provides examples of individual region of interest alignment.

However, anatomical variability is particularly challenging for the segmentation of the perirhinal cortex due to variable sulcal patterns and the variability in the depth of the collateral sulcus (see e.g. Berron, Vieweg et al., 2017, Ding and van Hoesen, 2010). Thus, all analyses including perirhinal Area 35 and Area 36 were performed in individual participant space. Within the hippocampus, the hippocampal body is mostly consistent across individuals with respect to major anatomical landmarks (we excluded the hippocampal head and tail from analyses).

In addition, please note that we utilized a study-specific group template. This decreases the amount of variability incorporated into the group space. For an appropriate study-specific template, good results can be achieved in averaging structural images when the sample size exceeds ten participants.

We refer to the appendix figure in the main text on page 24, line 20 and line 32.

Please see Appendix 7 – Figure 1 and figure caption:

6. Related to this, no quality assurance measures have been presented for the manually created ROIs. More information relating to these should be included in the methods section. For example, could the authors describe how many people were involved in manually segmenting the ROI's? If only one person, could the authors provide the results of intra-rater reliability analyses using, for example, the DICE overlap metric for each ROI? If more than one person was involved, could the authors also provide the results of inter-rater reliability analyses? In addition, considering it can sometimes be difficult to warp small ROIs such as hippocampal subfields from structural to EPI images with sufficient accuracy, could the authors provide some visual representations (perhaps as a supplemental figure) showing, for example, that the subiculum and CA1 ROIs created on the structural images were well aligned with the subiculum and CA1 on EPI images.

We appreciate this important request. Indeed, the regions have been segmented by two experienced raters. As the regions-of-interest (ROIs) have been part of a previously published study (Berron,Vieweg et al., 2017) for which the authors evaluated quality assurance measures, we report these as follows in appendix VIII now:

“The individual regions of interest were segmented by the same two experienced raters that also segmented a subsample of our data (24 hemispheres of 22 participants) for a previous publication (Berron, Vieweg et al., 2017). Quality assurance measures were calculated for that subsample. Regarding intra-rater reliability, the dice similarity coefficients are above 0.88 for all segmented regions (region-specific means (SD) are as follows: PHC 0.93 (0.03); Area 36 0.91 (0.02); Area 35 0.88 (0.02); EC 0.91 (0.01)). The intraclass-correlation coefficients for intra-rater reliability are all above 0.95 (PHC 0.99; Area 36 0.96; Area 35 0.97; EC 0.98). For the inter-rater reliability, dice similarity coefficients are above 0.84 for all segmented regions (region-specific means (SD) are as follows: PHC 0.86 (0.12); Area 36 0.91 (0.02); Area 35 0.84 (0.05); EC 0.87 (0.02)). The intraclass-correlation coefficients for inter-rater reliability are all above 0.78 (PHC 0.94; Area 36 0.88; Area 35 0.87; EC 0.94; see Berron, Vieweg et al., 2017).”

Please note that we falsely reported CA1 and subiculum segmentation in individual space and removed this now from page 24, line 13 in the methods section.

To assure accurate co-registration and warping, careful manual assessment based on anatomical landmarks was performed. To illustrate the procedure, we now include Figure 1 in appendix VII. This figure shows not only the coregistration process but also illustrates successful coregistration with images from an example participant. Note that this figure illustrates the coregistration process for Area35 and RSC. We applied the same resulting warping matrices to all other subregions of interest respectively.

7. For ROIs whose borders lie immediately adjacent each other (e.g., medial EC and distal subiculum; anterior PHC and posterior EC), did the authors take steps to reduce the possibility of fMRI signal 'bleeding' between ROIs by, for example, ensuring the distance between borders of each ROI was greater than the spatial smoothing kernel used on the functional data? If so, could you please describe these steps in the methods section?

Thank you for raising attention to this point. We, indeed, cannot rule out that signal from adjacent regions leaks into the regions of interest. This is yet another reason for why we stress the importance of future studies employing different methods to solidify our findings even more. See page 20, line 21:

“Future research is needed to evaluate how the functionally derived entorhinal seeds in this study relate to histologically derived entorhinal subregions (Oltmer et al., 2022) or entorhinal subregions based on structural connectivity (Syversen et al., 2021). For a dedicated comparison of subregions, it is essential to pay close attention to the segmentation of the EC itself.”

Page 20, line 14:

“In combination with closely matched histological or structural magnetic resonance imaging data, future work can further reveal the nature of retrosplenial mapping on the human EC.”

To diminish the influence of neighbouring regions on signal in target regions, we used a smoothing kernel smaller than two times the voxel size. Most seed and target regions are further than a voxel apart from each other. We established that functional connectivity analyses with source and seed regions in the contralateral hemisphere, provided roughly comparable results to analyses with ipsilateral sources and seeds performed across spatially proximal voxels (see especially the extracted estimates from individual participants and statistical effects in Author response images 1- 4).

We address this aspect as follows in the limitations section of the discussion in the manuscript on page 19, line 10:

“Second, while it is unlikely that our functional connectivity pattern is the result of spatial proximity, increased correlation between spatially adjacent regions is an inherent problem of functional connectivity analyses. Distances between seed and target regions differ and may determine patterns in the functional connectivity data. To diminish the influence of proximity, our smoothing kernel was smaller than two times the voxel size. It is important to stress moreover, that the pattern of our results is not easily explainable by spatial distance between seed and target regions. The ECArea35-based or ECRSC-based, for instance, are not adjacent to the subiculum/CA1 border. Furthermore, we observed roughly comparable results for neighboring seeds and targets (e.g. ECPHC-based and distal subiculum) when we performed the functional connectivity analyses with seed and source regions in the contralateral hemisphere.”

Author response image 1
Functional connectivity preferences to contralateral entorhinal seeds along the transversal axis of subiculum and CA1.

Displayed are the results of a seed-to-voxel functional connectivity analysis between the displayed right entorhinal seeds and the left subiculum and CA1 subregion. The 3D figure displays voxel-wise connectivity preferences to the entorhinal seeds (color coded to refer to the respective entorhinal seed [E]) on group level ([A] – subiculum; [B] – CA1). To display mean connectivity preferences across participants along the transversal sub/CA1 axis, β estimates were extracted and averaged from equally sized segments from proximal to distal ends (five segments in subiculum [A], three segments in CA1 [B]; schematized in white on the 3D figures) on each coronal slice and averaged along the longitudinal axis. Repeated measures ANOVAs revealed significant differences in connectivity estimates along the transversal axis of the subiculum ([C]; overall effect of seed by transversal segment: F(12,372) = 4.554; p <.001; main transversal effect of seed ECArea35-based: F(4,124) = 5.856, pFDR = .0012; ECPHC-based: F(4,124) = 3.147; pFDR = .037; ECRSC-based: F(4,124) = 7.828, pFDR <.001; ECArea36-based: F(4,124) = 0.856, pFDR = 0.493) with interaction effects (between ECPHC-based and ECRSC-based: F(4,124) = 9.249, pFDR <.001; between ECArea35-based and ECPHC-based: F(4,124) = 5.923, pFDR <.001; between ECArea35-based and ECRSC-based: p = 0.051 (uncorrected)) and in CA1 ([D] main seed by transversal effect: F(6,186) = 2.722, p = .034; main transversal effect of seed ECRSC-based: F(2,62) = 13.782; pFDR <.001; ECArea35-based: F(2,62) = .221; pFDR = .821; ECArea36-based: F(2,62) = 3.598, pFDR = .069; ECPHC-based: F(2,62) = .198, pFDR = .821). Displayed significances obtained by FDR-corrected post-hoc tests and refer to p <.05. Shaded areas in the graphs refer to standard errors of the mean. EC – entorhinal; M – medial; L – lateral; A – anterior; P – posterior; prox – proximal; dist – distal.

Author response image 2
Functional connectivity preferences to contralateral entorhinal seeds along the transversal axis of subiculum and CA1.

Displayed are the results of a seed-to-voxel functional connectivity analysis between the displayed left entorhinal seeds and the right subiculum and CA1 subregion. The 3D figure displays voxel-wise connectivity preferences to the entorhinal seeds (color coded to refer to the respective entorhinal seed [E]) on group level ([A] – subiculum; [B] – CA1). To display mean connectivity preferences across participants along the transversal sub/CA1 axis, β estimates were extracted and averaged from equally sized segments from proximal to distal ends (five segments in subiculum [A], three segments in CA1 [B]; schematized in white on the 3D figures) on each coronal slice and averaged along the longitudinal axis. Repeated measures ANOVAs revealed significant differences in connectivity estimates along the transversal axis of the subiculum ([C]; overall effect of seed by transversal segment: F(12,372) = 6.273; p <.001; main transversal effect of seed ECPHC-based: F(4,124) = 16.660; pFDR <.001; ECRSC-based: F(4,124) = 3.543, pFDR = .018; ECArea35-based: F(4,124) = 2.374, pFDR = .086; ECArea36-based: F(4,124) = 1.436, pFDR = .226) but not in CA1 ([D] main seed by transversal effect: F(6,186) = 1.812, p = .145). Displayed significances obtained by FDR-corrected post-hoc tests and refer to p <.05. Shaded areas in the graphs refer to standard errors of the mean. EC – entorhinal; M – medial; L – lateral; A – anterior; P – posterior; prox – proximal; dist – distal.

Author response image 3
Entorhinal seed regions based on connectivity preferences to contralateral cortical regions.

Displayed is the left EC as a 3D image with colored subregions. The subregions have been identified based on a source-to-voxel functional connectivity analysis and resulting connectivity preference to either the right retrosplenial (RSC, green) cortex, parahippocampal cortex (PHC, blue), Area 36 (A36, purple) or Area 35 (A35, pink) sources. Subregions have been determined based on the thresholded (T > 3.1) maximum voxels across four one-sample t-tests at group level, one per source. M – medial; L – lateral; A – anterior; P – posterior.

Author response image 4
Entorhinal seed regions based on connectivity preferences to contralateral cortical regions.

Displayed is the right EC as a 3D image with colored subregions. The subregions have been identified based on a source-to-voxel functional connectivity analysis and resulting connectivity preference to either the left retrosplenial (RSC, green) cortex, parahippocampal cortex (PHC, blue), Area 36 (A36, purple) or Area 35 (A35, pink) sources. Subregions have been determined based on the thresholded (T > 3.1) maximum voxels across four one-sample t-tests at group level, one per source. M – medial; L – lateral; A – anterior; P – posterior.

8. On page 20, lines 29-30, the authors describe the anatomical landmark used to create the anterior most slice of the hippocampal subfield masks but do not describe the landmark used to demarcate the posterior most extent of the masks. This should also be described.

Thank you for pointing out that missing information. We added the following specification on page 23, line 30:

“The last segmented slice was the one at which both, the inferior and superior colliculi had completely disappeared, applied for each hemisphere separately.”

9. The authors state that the subiculum and CA1 subfields were sagittally cut into "equally wide" segments (page 20; line 32) but no metrics relating to this are provided. Could the authors provide some metrics to support this? For example, was the mean number of functional voxels contained within each segment equivalent or did these substantially differ?

We appreciate your comment to provide more details here. The way we created the segments has been described in the methods section on page 23, line 32:

“Moreover, to evaluate results across the transversal sub/CA1 axis, the subiculum masks in each hemisphere were cut in five equally wide segments from medial to lateral within each coronal image. As the CA1 region gets more and more tilted towards the hippocampal tail, the three transversal CA1 segments were determined based on manual segmentation following a geometrical rule. Therefore, the two outer borders along the transversal axis of CA1 were connected with a line. From the middle point of that line, two straight lines were drawn in a 60° angle to determine roughly equally sized transversal CA1 segments within each coronal slice and hemisphere (a figure displaying the cuts, the procedure for the CA1 segments and the numbers of voxels within each segment can be found in appendix IX).”

Please note that the segments have been created in the T1 group template space. To illustrate, we now provide an illustration of the transversal subiculum and CA1 segments and the procedure for CA1 cuts, exemplified on the right hemisphere in the appendix. In addition, we provide the number of voxels within each segment and the following summary in appendix IX:

“Transversal subiculum and CA1 segments were cut on the group template T1 images. The average number of voxels contained in each subiculum segment was 460.8 voxels for the left subiculum (standard deviation 104.36) and 458 voxels for the right subiculum (standard deviation 75.09). For the left CA1 the average equals 360 voxels (standard deviation 27.58) and 335 voxels for the right CA1 segments (standard deviation 3.56, see Appendix 9 -Table 1 for segment-specific values and Figure 1 for an illustration).”

Results:

10. Although this is described in the methods section, perhaps the authors could briefly mention early in the Results section that the fMRI dataset used in Analysis 1 and 2 are derived from a task-based dataset where task-related effects have been regressed out to create a "rest-like" dataset. It may also be helpful to readers interested in (or cautious of) this particular method, to cite previous work showing that these "rest-like" datasets yield similar results to those obtained from resting-state data (e.g., Gavrilescu et al., Hum Brain Mapp 29:1040-1052, 2008 or similar).

Thank you for pointing out that we need to be more specific here. We now added the following lines to the manuscript on page 5 line 27 and on page 25, line 15:

“All functional connectivity analyses were performed on the dataset where task-related effects have been regressed out before, creating a dataset that resembles resting-state data (Gavrilescu et al., 2008; Maass, Berron et al., 2015).”

“Both functional connectivity analyses were performed on residuals of task-related functional data, creating a dataset that resembles resting-state data (Gavrilescu et al., 2008; Maass, Berron et al., 2015).”

11. Functional subregions of the EC associated with RSC, PHC, A35 and A36 are displayed very clearly in Figure 1. It appears that the EC seed relating to A36 is less 'localised' than the other seeds and displays roughly equal size clusters in both the posterior lateral and anterior medial EC. However, in the text, the authors suggest that A36 shows preferential connectivity with the posterior lateral EC (page 7; lines 7-8). How did the authors come to preferentially associate A36 with the posterior lateral EC rather than with the anterior medial EC? Did the authors conduct any additional quantitative assessments to determine this?

We appreciate this thoughtful comment. We based our conclusion on the following, additional metric: We evaluated for each source in which entorhinal part it yields the highest number of preferentially connected voxels.

Therefore, we cut the EC into quadrants (see Appendix 2 – Figure 1).

We then counted the number of voxels that have been assigned to be preferentially connected to each cortical source within each quadrant.

When we assign each source to the quadrant in which the respective source yields most functionally connected voxels we find the following organization, averaged across hemispheres:

‘anterior-medial’ (ECRSC-based) – quadrant number 1

‘posterior-medial’ (ECPHC-based) – quadrant number 2

‘anterior-lateral’ (ECA35-based) – quadrant number 3

‘posterior-lateral’ (ECA36-based) – quadrant number 4

However, we understand the concern that these functional connectivity clusters in the EC are not all locally constrained. Given reviewer comments number #10 and #20 and our methodology’s lack of anatomical precision, we now decided to name the entorhinal seed regions in a non-biased manner that does not reflect their anatomical position.

In accordance, we changed the naming of the EC seed regions throughout the entire manuscript into: ECRSC-based, ECPHC-based, ECArea35-based and ECArea36-based. Future studies with a stronger anatomical focus can then determine the exact location of entorhinal subregions.

To clarify these aspects, we added the following sentences to the manuscript on page 6, line 23:

For the ECPHC-based seed, the majority of voxels can roughly be described as clustering in the posterior-medial entorhinal portion, for the ECRSC-based seed in the anterior-medial portion, for the ECArea35-based seed in the anterior-lateral portion and for the ECArea36-based seed in the posterior-lateral entorhinal portion (see appendix II for exact voxel counts). Note that both perirhinal-based entorhinal seeds extended along the anterior to posterior axis such that the ECArea35-based progressed more along the outer EC (i.e. laterally, with a main focus anteriorly) and the ECArea36-based along the inner EC (i.e. medially, with a main focus posteriorly, see Figure 1 and the medial reflection of the EC seeds). It is important to note that these are rough qualitative descriptions of the main clusters, without quantification or an established relationship to coherent cytoarchitectonic regions. We will therefore continue to refer to them as ECRSC-based, ECPHC-based, ECArea35-based and ECArea36-based seeds.”

Please also note that the ECArea36-based focus in the posterior entorhinal portion, particularly in the left hemisphere, becomes visible when looking at the EC from a different angle. To point this out better, we therefore included another viewpoint in our 3D figure and mirror the EC.

We include the quadrant-based counting in appendix II. The supplement reads as follows:

“To assess the main location of each cortical source preferences within the EC, we cut the left and right EC in four quadrants. This was performed in T1 template space. First, the middle slice of all coronal slices that capture the EC was determined separately for each hemisphere. This slice was used to cut the EC in quadrants I, III and II, IV. Second, the middle slice of all axial slices that capture the EC was determined. This slice served to cut the EC in quadrants I, II and III, IV (see Appendix 2 – Figure 1). Note, to determine the most superior axial slice, the most posterior coronal level of the EC was used. Subsequently, we counted the number of voxels that have been assigned to each of the four cortical source regions after the initial functional connectivity analyses (that served to determined EC seeds). Averaged across hemispheres, most voxels assigned to the retrosplenial source are in EC quadrant I, most voxels assigned to the Area 35 source in EC quadrant II, most voxels assigned to the parahippocampal cortex in EC quadrant III and most voxels assigned to Area 36 in EC quadrant IV (see Appendix 2 – Table 1 for detailed voxel counts). Note that these quadrants do not refer to anatomically defined EC subregions.”

12. While I understand the primary focus of the subiculum and CA1 analyses relate to their transverse axis, it seems clear when looking at Figure 2, that there may also be interesting anterior-posterior differences along the body of these subfields. For example, functional clusters are largely absent in anterior portions of segment 3 of the subiculum but posterior portions of segment 3 contain functional clusters. Observations relating to the anterior-posterior axis are not discussed in text but may be informative to those with an interest in anterior-posterior variations in hippocampal function.

We agree that the longitudinal effects are interesting and important to the scientific community and thank you for raising that point. While we feel it would stretch the scope of a single paper too much to also thoroughly analyse the longitudinal axis statistically, we now added a qualitative description of longitudinal effects to the discussion on page 17, line 34:

“For completeness, we noted differences in functional connectivity along the longitudinal axis of the subiculum. We observed, for instance, more widespread functional connectivity of the ECArea35-based in the posterior subiculum whereas functional connectivity with the ECPHC-based portion seems more prominent in the anterior subiculum. The latter is consistent with previous reports (Dalton et al., 2019). The former, however, needs to be explored further by taking different segmentation protocols and seed regions into account. Note, that Maass, Berron et al. (2015) did not report longitudinal differences in connectivity strength between the EC and the subiculum. Future work needs to investigate how these observations relate to the reported gradient in functional connectivity and information resolution along the hippocampal longitudinal axis (e.g. Brunec et al., 2018 but many more).”

13. All the data are described in terms of preferred connectivity, which is what the data show. However, by mapping the preferred connectivity as absolute color maps with sharp borders, a lot of the complexities, i.e. gradients are lost. It would be great if the authors could provide supplementary maps of each source in isolation and how it maps across the surface of EC and the transverse axis of CA1/subiculum.

We fully agree that gradient maps are a very interesting supplement and provide the respective maps now in appendix III:

Interpretation:

14. The authors created ROI masks for the subiculum and refer to medial portions of this mask as 'distal subiculum'. However, the mask appears to encompass the entire 'subicular complex' inclusive of the subiculum, presubiculum and parasubiculum. Importantly, the area referred to as 'distal subiculum' throughout the manuscript (specifically, segments 1 and 2 of the subiculum mask) most likely aligns with the location of the pre- and parasubiculum in addition to containing portions of the distal subiculum. This is important considering the growing body of evidence that the pre- and parasubiculum are preferentially engaged during scene-based cognition. While this is briefly touched on in the discussion (page 13; line 5-6), I feel the interpretation of results as they relate to the 'distal subiculum' could benefit from being placed more firmly in the context of this growing body of evidence relating to the involvement of medial aspects of the hippocampus (inclusive of the pre- and parasubiculum) in scene-based cognition.

Thank you for that insightful comment. We agree and have now added the following paragraph in the discussion of our results on page 13, line 28:

“Our observation is in line with the hypothesis that the distal subiculum is more involved in processing scenes than objects based on previous findings in the human brain. While the subiculum in general was associated with scene discrimination (Hodgetts et al., 2017), a growing body of evidence relates particularly the medial hippocampus to scene processing. This entails two medial areas, the pre- and parasubiculum, that we attribute to the distal subiculum in our current segmentation. Especially the area that resembles the pre- (or here: distal) subiculum has been shown to be involved in scene construction (Dalton et al., 2018, Zeidman et al., 2015).”

15. I am afraid that the central premise of the 'two segregated memory streams' is somewhat outdated, certainly at the level of the entorhinal cortex and hippocampus (see Doan et al. 2019). Regarding the rodent homologue of the parahippocampal cortex (referred to as POR, or postrhinal cortex) Doan et al. write: 'Postrhinal cortex preferably targets lateral instead of medial entorhinal cortex', and that '..dorsolateral parts of LEC receive inputs from both POR and PER and that both these projections show very similar topological features..'

We appreciate that concern. Indeed, our rather historical approach was due to our manuscript considered as advancing the Maass et al., 2015 publication. To circumvent the impression that the two parallel entorhinal-hippocampal stream hypothesis is the current state of research, we now rephrased the introduction considerably.

From the very beginning, we make clear that this hypothesis was recently updated on page 3, line 10:

“Large-scale cortical information streams, that originate in the visual ‘Where’ and ‘What’ pathways and process scene and object information (Berron et al., 2018; Haxby et al., 1991; Ranganath and Ritchey, 2012; Ritchey et al., 2015; Ungerleider and Haxby, 1994), map onto the EC in a complex manner and define functional EC subregions. Recent rodent research updates the former conception of a parallel mapping of scene and object information via parahippocampal and perirhinal cortices onto medial versus lateral EC subregions (cf. posterior-medial versus anterior-lateral EC subregions as the human homologues; Maass, Berron et al., 2015; Navarro Schröder et al., 2015). Instead of a strict parallel mapping, profound cross-projections exist from the parahippocampal cortex towards the perirhinal cortex and the lateral EC (Nilssen et al., 2019). In accordance, information seems to converge in the rodent lateral EC (Doan et al., 2019). The update, thus, implies a more complex functional organization than parallel scene and object information mapping.”

We state this again, before sketching our study on page 4, line 20:

“To summarize, our conception of how information travels towards the entorhinal-hippocampal circuitry underwent key changes which warrant an extensive exploration of the circuitry’s functional organization. First, rodent research shows that there is no strict parallel mapping of cortical information from the perirhinal and parahippocampal cortex towards the EC. Second, information seems to converge already before the hippocampus.”

16. Nevertheless, in the present study the authors claim that 'The lateral EC preferentially communicates with the perirhinal cortex' (page 3, line 20). In the same paragraph they write that context processing should be specific to the MEC, in contrast to the LEC that would rather process non-spatial and item information. Strikingly, they then go on to cite Doan et al. 2019 and effectively contradict themselves within the same paragraph '…parahippocampal cortex communicates with the EC along its full extend' (as Doan et al. write, the LEC is similarly connected to POR (PHC) and PER). This means that contextual information from PHC should not uniquely be passed on to the MEC. On the contrary, it should even be more strongly passed on to the LEC than the MEC.

Thank you for this important comment, we apologize if we raised the impression that the parallel stream connection and related information segregation are the current state of research. This overview was rather meant as a brief historical introduction. We hope that our rephrased introduction makes this clear. For example on page 3, line 17:

“Instead of a strict parallel mapping, profound cross-projections exist from the parahippocampal cortex towards the perirhinal cortex and the lateral EC (Nilssen et al., 2019). In accordance, information seems to converge in the rodent lateral EC (Doan et al., 2019). The update, thus, implies a more complex functional organization than parallel scene and object information mapping. Moreover, this advance highlights the retrosplenial cortex as an additional source to convey information directly from the cortical scene processing stream onto the EC. The retrosplenial cortex projects to the medial EC and, like the parahippocampal cortex, is part of the scene processing stream (e.g. involved in scene translation; Vann et al., 2009; Nilssen et al., 2019; Witter et al., 2017). The update, furthermore, evokes the question how cortical sources of information uniquely map onto the EC and which kind of information is processed in the resulting functional EC subregions.”

When we refer to current literature we also incorporate functional predictions based on the parahippocampal cross-projections on page 4, line 12:

“According to information convergence in the EC, a recent report finds convergence along the rodent transversal CA1 axis (Vandrey et al., 2021). In humans, visual stream projections towards the entorhinal-hippocampal circuitry similarly suggest convergence of scene and object information in the subiculum/CA1 border region but preserved scene processing in the distal subiculum (Dalton and Maguire, 2017). A detailed examination of the latter hypothesis is, however, lacking. The diversity of findings emerging from the literature calls for a thorough investigation to elucidate whether multiple transversal processing routes exist within the human entorhinal-hippocampal circuitry.”

We clearly summarize this in the introduction now on page 4, line 22:

“First, rodent research shows that there is no strict parallel mapping of cortical information from the perirhinal and parahippocampal cortex towards the EC. Second, information seems to converge already before the hippocampus.”

Likewise our predictions now state clearly that we expect convergence due to the parahippocampal cross-projections on page 7, line 13 and on page 10, line 4:

“Following the characterization of entorhinal seeds, we focused on the functional connectivity between these entorhinal subregions and hippocampal subiculum and CA1 to test the hypothesis of a transversal functional connectivity pattern. We predicted that while some EC subregions have a preference to functionally connect with the subiculum/CA1 border, others preferentially connect with the distal subiculum and proximal CA1. In the previous step we identified EC subregions based on unique cortical source contributions. Therefore, our predictions remained in accordance with Maass, Berron et al. (2015): We expected that the EC subregion preferentially connected with the parahippocampal cortex (ECPHC-based seed) maps towards the distal subiculum and EC subregions connected with the perirhinal cortex (ECArea35-based seed, ECArea36-based seed) map towards the proximal subiculum, a mapping that we predicted to be extended towards the distal CA1.”

“Besides the intrinsic functional connectivity patterns within the entorhinal-hippocampal circuitry, we also examined the characteristics of scene and object information processing to test the hypothesis of multiple information processing routes within the entorhinal-hippocampal circuitry. We predicted a route of specific scene processing and another route of convergent information processing. Following the proposal by Dalton and Maguire (2017) and the updated cross-projections from the scene to the object information processing stream (Nilssen et al., 2019), we expected scene processing in the distal subiculum. The updated parahippocampal cross-projections imply convergence wherever specific object processing had been expected previously. Thus, we explored whether any entorhinal-hippocampal subregions still process object information specifically. However, we largely expected to find evidence consistent with convergent processing of scene and object information within the entorhinal-hippocampal circuitry.”

17. I am rather puzzled why the authors cite Nilssen et al. and Doan et al. and still uphold the segregated stream hypothesis of entorhinal and hippocampal subregions. In my reading, those studies are entirely incompatible with the view of two parallel, segregated memory streams involving the LEC and MEC. Ironically, the null finding of category specific processing in the lateral cluster of EC voxels, and associated regions, might be due to the convergent, multimodal input patterns. Unfortunately, the two segregated stream hypothesis seems to be a central tenet in the current study.

We appreciate your comment. Our intention was not to uphold an outdated hypothesis. Given that most published research so far was conducted in light of this hypothesis, we however wanted to give an overview of the resulting ideas in our introduction. This applies in particular to information processing. While we agree that both hypotheses are not aligned with each other, we still think that at least for the MEC, the segregated stream hypothesis and research in light of that provides valuable information. Moreover, future research will need to explicitly investigate why and how previous studies could confirm the segregated stream hypothesis. As our research is an advancement to a previous study conducted in the light of the segregated stream hypothesis, our aim was to provide a historical introduction to the entangled research about functional connectivity and information processing and current idea of heavy cross-projections before the entorhinal cortex.

One goal of the current study was to show what is “left” of the original segregated stream hypothesis and to focus on where unique cortical information is mapping on the entorhinal cortex.

To prevent the impression that we uphold an outdated hypothesis, we now rephrased our introduction considerably.

In the very beginning, we mention that the parallel mapping hypothesis has been updated on page 3, line 13:

“Recent rodent research updates the former conception of a parallel mapping of scene and object information via parahippocampal and perirhinal cortices onto medial versus lateral EC subregions (cf. posterior-medial versus anterior-lateral EC subregions as the human homologues; Maass, Berron et al., 2015; Navarro Schröder et al., 2015). Instead of a strict parallel mapping, profound cross-projections exist from the parahippocampal cortex towards the perirhinal cortex and the lateral EC (Nilssen et al., 2019). In accordance, information seems to converge in the rodent lateral EC (Doan et al., 2019). The update, thus, implies a more complex functional organization than parallel scene and object information mapping.”

We summarize the recent changes and their implications later on page 4, line 20:

“To summarize, our conception of how information travels towards the entorhinal-hippocampal circuitry underwent key changes which warrant an extensive exploration of the circuitry’s functional organization. First, rodent research shows that there is no strict parallel mapping of cortical information from the perirhinal and parahippocampal cortex towards the EC. Second, information seems to converge already before the hippocampus.”

Then, we state how we aim to advance the existing research that has largely been based on the parallel mapping hypothesis on page 4, line 35:

“With a combination of functional connectivity and information processing analyses, we seek to answer two sets of questions. Regarding functional connectivity, we ask where the parahippocampal, perirhinal and retrosplenial cortical sources uniquely map onto the human EC and how these functionally connected routes continue between EC subregions and the transversal sub/CA1 axis. Regarding information processing, we ask whether and where scene and object information are specifically processed in the EC and along the transversal sub/CA1 axis. We test the hypotheses of (1) a transversal functional connectivity pattern and (2) multiple information processing routes within the entorhinal-hippocampal circuitry. Thus, following the updated conception of a non-parallel cortical scene and object information mapping onto the EC in rodents, we will show how cortical information streams map onto the EC in humans. This mapping will then be our detailed starting point to investigate the functional connectivity and information processing within the entorhinal-hippocampal circuitry.”

In the predictions that we outline now in each result’s respective section, we illustrate how the update to the parallel mapping hypothesis shapes our current expectations on page 7, line 13 and on page 10, line 4:

“Following the characterization of entorhinal seeds, we focused on the functional connectivity between these entorhinal subregions and hippocampal subiculum and CA1 to test the hypothesis of a transversal functional connectivity pattern. We predicted that while some EC subregions have a preference to functionally connect with the subiculum/CA1 border, others preferentially connect with the distal subiculum and proximal CA1. In the previous step we identified EC subregions based on unique cortical source contributions. Therefore, our predictions remained in accordance with Maass, Berron et al. (2015): We expected that the EC subregion preferentially connected with the parahippocampal cortex (ECPHC-based seed) maps towards the distal subiculum and EC subregions connected with the perirhinal cortex (ECArea35-based seed, ECArea36-based seed) map towards the proximal subiculum, a mapping that we predicted to be extended towards the distal CA1.”

“Besides the intrinsic functional connectivity patterns within the entorhinal-hippocampal circuitry, we also examined the characteristics of scene and object information processing to test the hypothesis of multiple information processing routes within the entorhinal-hippocampal circuitry. We predicted a route of specific scene processing and another route of convergent information processing. Following the proposal by Dalton and Maguire (2017) and the updated cross-projections from the scene to the object information processing stream (Nilssen et al., 2019), we expected scene processing in the distal subiculum. The updated parahippocampal cross-projections imply convergence wherever specific object processing had been expected previously. Thus, we explored whether any entorhinal-hippocampal subregions still process object information specifically. However, we largely expected to find evidence consistent with convergent processing of scene and object information within the entorhinal-hippocampal circuitry.”

18. There seems to be no main effect of activations to object stimuli (difference to zero) in any region. Nevertheless the authors claim that regions show processing of object-related information. This is not warranted based on the results. For example in the abstract: 'The regions of another route, that connects the anterior-lateral EC and a newly identified retrosplenial-based anterior-medial EC subregion with the CA1/subiculum border, process object and scene information similarly'. If there is no evidence of any level of processing, it is not warranted to claim 'similar processing'.

We appreciate this comment and understand the concern.

Out additional analysis of cortical source regions could show significantly increased activity to object stimuli in the perirhinal cortex (Area 36 and left Area 35; see the following comment #19 for a figure). In the EC, subiculum and CA1, we however do not find significantly increased activity to object stimuli anywhere. We believe that the nature of our functional data does not allow us to conclude that no (object/scene) information processing occurs there at all, instead we agree that we can only compare between different information processing conditions (object/scene). Based on this comparison we also agree that we do not see differences in information processing between object and scene conditions. We see that our wording is potentially misleading and therefore changed it to “no differences in information processing conditions”.

To elaborate on this important point further:

The difference to zero refers to a difference from the baseline condition (scrambled pictures). As functional MRI is a relative measure, at least with our methodological set up, the baseline condition may not suffice as an indicator of “zero activation” or “no level of processing”. Instead, it is a relative condition towards which we could relate both, the object and the scene condition. Technically, we did not assess whether, for example, the EC does not activate at all during the presentation of scrambled pictures (baseline). Given the setup of our study, we have no means to assess this question. It is possible that individuals engaged in some sort of functional processing, even of the scrambled pictures (e.g. they may have tried to “see” something in them). While we cannot assess this question, we can say that based on our data, the relationship between object processing and baseline compared to scene processing and baseline appears to be similar in several regions of interest.

To stress that point, we included a paragraph in the “limitations section” of the manuscript on page 19, line 24 (new sentences in bold):

“Note that as a first step towards an understanding of the system’s functional organization and to increase comparability with earlier studies, we assessed functional connectivity and information processing within the entorhinal-hippocampal circuitry with univariate methods. These allow relative comparisons between functional activity levels in different conditions. Consequently, we are neither able to assess what the EC is processing during the baseline condition, meaning the absolute level of functional activity, nor are we able to verify that information processing is similar across conditions in for example the ECArea35-based seed. Univariate methods, moreover, average the signal over regions of interest. To capture hidden voxel-wise patterns of activity that scale with the processing of certain representations, future studies could examine information pathways with multivariate methods that evaluate informational content in the activity pattern of voxels instead of in an averaged manner (Kragel et al., 2018; Kriegeskorte et al., 2008). Moreover, recent methodological advances can be employed in the future that study functional connectivity based on the underlying content representations between regions (Basti et al., 2020).”

Functional activation levels from the object condition and from the scene condition are both contrasted with the baseline condition, mainly for plotting purposes. It allowed us to show object and scene processing with separate lines. We realize that our figures may potentially have been misleading in that sense and now we excluded the stippled line indicating “zero”:

19. Based on the introduction, a double dissociation between object and scene processing in two segregated sets of regions is expected, but the authors don't provide evidence for this. In my view the interaction effect (and post-hoc tests) for higher scene processing in segment 1and2 in Figure 4 is a very interesting and meaningful finding. However, neither main effects nor interactions are presented to support claims of any level of 'object processing'. How can we be sure that we are not looking at pure noise in the object condition? Absence of evidence is not equal to evidence of absence (of different object/ scene processing).

Indeed, we did not see any processing biased towards the object condition in any of our tested regions of interest. Your comment is very thoughtful and important and we agree that absence of evidence is not equal to evidence of absence. We performed an additional set of analyses regarding information processing in the cortical source regions and include it in the appendix. Here, we find evidence for significantly more object than scene processing in the bilateral perirhinal Areas 36 and the left Area 35 while in the retrosplenial and parahippocampal cortex scene processing is significantly higher than object processing. This additional finding suggests that we are not looking at pure noise in the object condition. We think it rather provides an additional hint towards multimodal object and scene processing in the lateral entorhinal regions.

We refer to the analysis in the supplement in the main manuscript on page 14, line 27:

“Note that a supplemental analysis of information processing in the cortical source regions showed indeed, specific object processing in perirhinal source regions (see appendix V). The lack of increased object processing in the anterior EC subregions and subiculum/CA1 border is thus likely not a result of increased noise in the object condition. Instead, increased object processing in perirhinal cortical source regions indicates subsequent convergence in entorhinal-hippocampal subregions, as hypothesized based on the updated cortical mapping scheme onto the EC.”

We describe the analysis in the appendix V as follows:

“To examine whether lower parameter estimates for object processing could be due to increased noise in this condition, we evaluated object and scene processing in the four cortical source regions. Therefore, we extracted parameter estimates for the object versus baseline and the scene versus baseline contrast from the retrosplenial and parahippocampal cortex and from perirhinal Area 36 and Area 35, respectively. All parameter estimates were extracted from the previously segmented regions of interests, coregistered to the individual EPI space.

Repeated-measures ANOVAs in both hemispheres showed a significant interaction effect between condition and region (right: F(3,93) = 60.4229; p <.001; left: F(3,93) = 47.3421; p <.001). Subsequent paired-samples T-tests show significantly more functional activity in the object than scene condition in Area 36 (bilateral: pFDR <.001) and the left Area 35 (pFDR =.0011). No significant difference between object and scene conditions is observed in the right Area 35 (right: pFDR = 0.9821). There is a significant effect of more functional activity in the scene than object condition in the parahippocampal (bilateral: pFDR <.001) and retrosplenial cortex (bilateral: pFDR <.001, see Appendix 5 – Figure 1).

The increased object processing in adjacent cortical source regions indicates that noise differences across conditions are not likely to cause the lack of increased object processing within entorhinal seed regions and hippocampal subregions.”

20. No 'information flow' is actually assessed in this study. In my view, this would require a directed connectivity analysis such as dynamic causal modeling or transfer entropy. The research question or hypotheses should pertain to what is actually being done. For example, 'we predict functional connectivity between X-Y to reflect the structural connectivity described in rodents/ humans previously', and 'we expect specific sensitivity to scene stimuli in distal subiculum, because of connectivity to the MEC region in rodents'. This would be 'consistent with information flow' between those regions, but it is not directly or conclusively showing it.

We agree with this potential point of misunderstanding and changed our wording accordingly throughout the manuscript on page 2, line 12:

“Our results are consistent with transversal information-specific pathways”

Page 19, line 20:

“Third, our perspective was entirely functional and we cannot conclude on the directionality of our results. […] Note that as a first step towards an understanding of the system’s functional organization and to increase comparability with earlier studies, we assessed functional connectivity and information processing within the entorhinal-hippocampal circuitry with univariate methods.”

Page 21, line 7:

“Our high-resolution approach revealed unknown characteristics of functional connectivity and scene processing within the human entorhinal-hippocampal circuitry.”

21. The approach to segment the EC into sub-clusters based on known connectivity to other regions seems fine in general. However, this should be guided by gold-standard, a priori knowledge of entorhinal subregions such as the rodent MEC and LEC (or finer cyto/ myeloarchitectonic subdivisions, for example described in Kimer et al. 1997). In the present study, the authors select four cortical seed regions to segment entorhinal subregions based on functional connectivity, without providing appropriate justification how some of those regions would uniquely connect to specific, cytoarchitectonically-defined entorhinal subregions. For example, BA35, BA36, and PHC all project to the rodent LEC, and PHC also project (albeit weaker) to the MEC. Which a-priori defined entorhinal subregions should be uniquely identified with these four seed regions? Following the semipartial correlation analyses, clusters of EC voxels are then segmented and labeled posterior-medial, posterior-lateral, anterior-medial and anterior-lateral EC. Those names are then used quasi synonymously with coherent and established, cytoarchitectonically defined subregions such as the rodent LEC and MEC. This seems egregious, not only because of the questionable choice of initial seed regions, but more so because of the discontinuous topography of the segmentations shown in Figure 1. What is the ground-truth (cytoarchitecture, myeloarchitecture, tracing studies, gene expression) evidence for a salt-and-pepper organization of entorhinal subregions? The clusters can't correspond to coherent cytoarchitectonic regions, so why would they be referred to as such.

Thank you for raising that important concern. Indeed, we realized that our naming of entorhinal seed regions may potentially be misleading. Our findings capture functional connectivity but as we raised in the limitations, this does not necessarily correspond to anatomically defined regions. See page 20, line 21:

“Future research is needed to evaluate how the functionally derived entorhinal seeds in this study relate to histologically derived entorhinal subregions (Oltmer et al., 2022) or entorhinal subregions based on structural connectivity (Syversen et al., 2021). For a dedicated comparison of subregions, it is essential to pay close attention to the segmentation of the EC itself.”

This is also illustrated in a new paragraph of the limitations section on page 20, line 3:

“Fourth, our study was originally conducted within the assumption that (functional) connectivity profiles reveal functional subregions. Based on that approach, the medial EC is identified based on i.a. retrosplenial connectivity. We, therefore conclude a surprisingly anterior yet medial EC mapping of the retrosplenial cortex. This approach has been followed by Maass, Berron et al. (2015) and also in numerous anatomical connectivity studies in animals (see Witter et al., 2017). It is possible that species differences lead to our ECRSC-based to be more anterior than one would expect based on animal studies. However, given that the medial subregion in the primate EC remains posterior (cf. posterior-medial EC homologue in Maass, Berron et al., 2015), another possibility is that our retrosplenial functional connectivity cluster maps onto the human anterior-lateral EC. Our data does not allow us to verify this latter option. It is unclear, however, why functional subregions in line with predictions from animal research can be identified for some cortical source-to-EC mappings (like the parahippocampal cortex) but not for others. In combination with closely matched histological or structural magnetic resonance imaging data, future work can further reveal the nature of retrosplenial mapping on the human EC.”

To prevent raising the impression that our seed regions relate to cytoarchitectonically defined regions, we have renamed the entorhinal seed regions according to their functional source regions throughout the manuscript on page 6, line 23:

“For the ECPHC-based seed, the majority of voxels can roughly be described as clustering in the posterior-medial entorhinal portion, for the ECRSC-based seed in the anterior-medial portion, for the ECArea35-based seed in the anterior-lateral portion and for the ECArea36-based seed in the posterior-lateral entorhinal portion (see appendix II for exact voxel counts). Note that both perirhinal-based entorhinal seeds extended along the anterior to posterior axis such that the ECArea35-based progressed more along the outer EC (i.e. laterally, with a main focus anteriorly) and the ECArea36-based along the inner EC (i.e. medially, with a main focus posteriorly, see Figure 1 and the medial reflection of the EC seeds). It is important to note that these are rough qualitative descriptions of the main clusters, without quantification or an established relationship to coherent cytoarchitectonic regions. We will therefore continue to refer to them as ECRSC-based, ECPHC-based, ECArea35-based and ECArea36-based seeds”

To explain why we initially opted the anatomical terms for the seed regions, we point out that we find a considerable topographical organization of our EC seed regions in four quadrants of the entorhinal cortex. We included this information now in appendix II:

“To assess the main location of each cortical source preferences within the EC, we cut the left and right EC in four quadrants. This was performed in T1 template space. First, the middle slice of all coronal slices that capture the EC was determined separately for each hemisphere. This slice was used to cut the EC in quadrants I, III and II, IV. Second, the middle slice of all axial slices that capture the EC was determined. This slice served to cut the EC in quadrants I, II and III, IV (see Appendix 2 – Figure 1). Note, to determine the most superior axial slice, the most posterior coronal level of the EC was used. Subsequently, we counted the number of voxels that have been assigned to each of the four cortical source regions after the initial functional connectivity analyses (that served to determined EC seeds). Averaged across hemispheres, most voxels assigned to the retrosplenial source are in EC quadrant I, most voxels assigned to the Area 35 source in EC quadrant II, most voxels assigned to the parahippocampal cortex in EC quadrant III and most voxels assigned to Area 36 in EC quadrant IV (see Appendix 2 – Table 1 for detailed voxel counts). Note that these quadrants do not refer to anatomically defined EC subregions.”

As our manuscript is an advancement of the Maass et al. (2015) publication, we opted for a comparable terminology. Therefore, we initially stretched their anterior-lateral (for the EC voxels preferentially connected to the perirhinal seed) and anterior-medial (for the EC voxels preferentially connected to the parahippocampal seed) terminology further and applied it to the additional two EC seed regions we acquire. We agree, however, that this may falsely imply more profound anatomical grounding of our results. Hence, our new naming of EC seed regions. We hope, this prevents further misunderstandings.

22. Why did the authors not simply use the ROIs from their previous identification of the human homologues of the MEC and LEC (Maass et al. 2015) to address their current research questions? In summary, this could be reconciled by a consistently used naming scheme for the entorhinal seeds that avoids confusion with cytoarchitectonic subregions.

We appreciate that comment. The human homologues of the MEC and LEC were defined within the concept of the two parallel streams hypothesis (in Maass et al., 2015). Each voxel was identified by either preferential functional connection to the perirhinal or the parahippocampal cortex. This approach neither acknowledged further cortical sources like the retrosplenial cortex nor did it allow to test unique mappings of the perirhinal and parahippocampal cortices. In addition, we opted to separate perirhinal Area 35 and Area 36 given reports about early cortical tau pathology in Alzheimer’s disease that appears specifically in Area 35 (Lace et al., 2009).

As we understand the concern regarding the anatomical correspondence of our entorhinal seed regions, we renamed them as follows:

ECRSC-based

ECPHC-based

ECArea36-based

ECArea35-based

23. In the results you describe data on proximal CA1, but you do not mention them anywhere in the paper explicitly. If your data do not allow you to functionally 'interpret' proximal CA1, you might come back to that in the discussion, state this and mention that in the rodent literature there is a gradient along transverse CA1 with more precise spatial information in proximal than in distal. Any ideas of why that does not seem to hold in humans? In particular on page 15, line 31 you make a very general statement about the transverse organization of CA1, so that might be the place to elaborate a bit more.

Thank you for that interesting point. We added the following discussion of that issue now to the manuscript on page 15, line 11:

“Regarding the human proximal CA1, a firm conclusion is limited with our data. First, the functional connectivity results varied between hemispheres. In both hemispheres, proximal CA1 showed a different connectivity profile compared to distal CA1. However, even though statistically not significant, the preferences at the group level indicated increased functional connectivity with the ECPHC-based portion in the right but with the ECArea35-based portion in the left hemisphere. Second, we do not prove similar information processing along the transversal CA1 axis. Instead, we find no significant difference in information processing along the transversal CA1 axis. As indicated in the previous paragraph, we cannot rule out that our object versus scene processing conditions may not have been sensitive enough to tackle functional differences in CA1. Thus, future research will have to identify defining characteristics of information processing along the transversal CA1 axis in a less constraint manner to allow conclusions on distinct information processing in proximal CA1.”

Future projects may evaluate longitudinal effects that potentially average out our analysis of the transversal axis. In addition, individual variability in the subregion’s borders along the longitudinal axis may be captured by analyses methods that can be applied independently of segmentation protocols and that focus on single-voxel results only.

Clinical relevance:

24. The reported findings may have clinical relevance, but this is to be determined by future studies. An entire paragraph in the Introduction is dedicated to this topic (starting on page 4, line34). The authors state that the findings show that tau pathology spreads through 'functionally connected' regions. Functionally connected regions must also be structurally connected (through mono or polysynaptic connections). It seems the authors are insinuating that functional and structural connections are independent. The prediction of functional connectivity between parahippocampal, perirhinal and entorhinal cortices, and the subiculum is already abundantly well founded on previous findings of tracing studies in rodents and primates, and even the authors' previously published functional MRI findings (Maass et al. 2015). The fact that the spreading of tau pathology follows structural and functional connections provides no additional predictive value to inform the research question (in my understanding these are: (1) is there a specific functional connectivity pattern between regions, (2) do specific regions show differential fMRI activations for object vs scene stimuli). It appears unwarranted and misleading to portray the clinical findings as a formal motivation (meaning, previous findings that form the basis for the research hypotheses) for the present study – page 5, line 8. If this was not the intention, then I feel this needs to be formulated more clearly.

Thank you for raising that important concern. We agree, that the spreading of tau pathology does not provide additional predictive value to inform our research questions. However, we would like to emphasize that according to our view it makes sense from a clinical point of view to investigate our research questions. It has previously been shown that tau pathology may spread in an activity-dependent manner across connected regions (Adams et al., 2019; Berron et al., 2020; Berron et al., 2021; Franzmeier et al., 2020; Maass et al., 2019; Vogel et al., 2020). Therefore, knowledge on these potential routes and hence on functional connectivity and related information processing, as we observe it, is clinically relevant. Thus, we do not see the clinical literature as holding predictions (besides a potential different functional role for Area 35 and Area 36 due to differences in tau pathology here).

Still, in our view it appears logical to us to investigate functional architecture, even from a mere clinical perspective. Clinical research can rely on our basic findings for future, more precise clinical predictions. This was likewise the case for the previous Maass et al. (2015) publication (as well as Navarro Schröder et al., 2015) which informed subsequent publications in the clinical population by the functional architecture they revealed (e.g. Tran et al., 2022; Berron et al., 2022; Berron et al., 2021; Maass et al., 2020; Reagh et al., 2018; Olsen et al., 2017; Bastin et al., 2019; Yeung et al., 2019). Therefore, we do see the clinical literature as a strong motivator to investigate functional architecture. Hence, we opted to refer to this aspect and thereby also make our basic findings accessible to the clinical research community.

We hope, that this clarification helps to prevent any potential misunderstandings in this regard.

Moreover, we stress in the manuscript, that the clinical perspective is yet another important motivator for our study, but does not provide any predictive value. We therefore deleted the paragraph on our clinically-driven motivation in the introduction.

When interpreting our findings, we refer to the clinical relevance at the end of the Discussion section, when we give a more general outlook on the applications of our basic results on page 18, line 15:

“From a clinical research perspective, it is remarkable that the current functional connectivity pattern resembles the topology of early cortical tau pathology in Alzheimer’s disease (Lace et al., 2009). An influential hypothesis suggests tau progression in Alzheimer’s disease along functionally connected pathways in the human brain (Franzmeier et al., 2020; Vogel et al., 2020). Earliest cortical tau pathology in Alzheimer’s disease accumulates in perirhinal Area 35 (also referred to as transentorhinal region) and the anterior-lateral EC before it can be found along the subiculum/CA1 border (Braak and Braak , 1995; Berron et al., 2021; Kaufman et al., 2018; Lace et al., 2009). The topology of early tau pathology in Alzheimer’s disease thus mirrors the regions that we find biased towards ECArea35-based connectivity (Braak and Braak, 1991; Lace et al., 2009; Roussarie et al., 2020). Tau pathology in Alzheimer’s disease is associated with memory impairment (Bejanin et al., 2017; Berron et al., 2021; Nelson et al., 2012) and information processing might be affected accordingly as reports have shown an association between Alzheimer’s related tau pathology and object memory in early disease stages (Berron et al., 2019; Maass et al., 2019). However, given our finding of activity patterns consistent with object – scene convergence in those subregions of the hippocampal-entorhinal circuitry that are affected by early tau pathology, object-in-scene memory tasks might have increased sensitivity to memory impairment. Moreover, both, the entorhinal portion based on retrosplenial connectivity (ECRSC-based) and the entorhinal portion based on Area 35 connectivity (ECArea35-based), are functionally connected to the subiculum/CA1 border. This overlapping functional connectivity pattern in the hippocampus might be a way along which tau and amyloid pathologies in Alzheimer’s disease could interact. This is consistent with early hypometabolism and cortical tau progression in the retrosplenial cortex and early amyloid in posterior parietal regions (Grothe et al., 2017; Palmqvist et al., 2017; Ziontz et al., 2021). The revealed functional connectivity and information processing profile may guide future hypotheses on the propagation of Alzheimer’s pathology and related functional and cognitive impairment.”

In case you continue to see a potentially misleading aspect in our introduction, please let us know as it is not our intention.

https://doi.org/10.7554/eLife.76479.sa2

Article and author information

Author details

  1. Xenia Grande

    1. Institute of Cognitive Neurology and Dementia Research, Otto von Guericke University Magdeburg, Magdeburg, Germany
    2. German Center for Neurodegenerative Diseases, Magdeburg, Germany
    Contribution
    Conceptualization, Formal analysis, Visualization, Writing - original draft, Writing – review and editing, Data Interpretation
    For correspondence
    xenia.grande@dzne.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-2486-3201
  2. Magdalena M Sauvage

    1. Functional Architecture of Memory Department, Leibniz-Institute for Neurobiology, Magdeburg, Germany
    2. Functional Neuroplasticity Department, Otto-von-Guericke University Magdeburg, Magdeburg, Germany
    3. Center for Behavioral Brain Sciences, Otto-von-Guericke University Magdeburg, Magdeburg, Germany
    Contribution
    Funding acquisition, Writing – review and editing, Data interpretation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7586-6410
  3. Andreas Becke

    1. Institute of Cognitive Neurology and Dementia Research, Otto von Guericke University Magdeburg, Magdeburg, Germany
    2. German Center for Neurodegenerative Diseases, Magdeburg, Germany
    Contribution
    Investigation, Project administration
    Competing interests
    No competing interests declared
  4. Emrah Düzel

    1. Institute of Cognitive Neurology and Dementia Research, Otto von Guericke University Magdeburg, Magdeburg, Germany
    2. German Center for Neurodegenerative Diseases, Magdeburg, Germany
    3. Institute of Cognitive Neuroscience, University College London, London, United Kingdom
    Contribution
    Funding acquisition, Writing – review and editing
    Competing interests
    No competing interests declared
  5. David Berron

    1. German Center for Neurodegenerative Diseases, Magdeburg, Germany
    2. Center for Behavioral Brain Sciences, Otto-von-Guericke University Magdeburg, Magdeburg, Germany
    3. Clinical Memory Research Unit, Department of Clinical Sciences Malmö, Lund University, Lund, Sweden
    Contribution
    Conceptualization, Supervision, Funding acquisition, Investigation, Project administration, Writing – review and editing, Data Interpretation
    For correspondence
    david.berron@dzne.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1558-1883

Funding

Deutsche Forschungsgemeinschaft (Project-ID 42589994)

  • Magdalena M Sauvage
  • Emrah Düzel

HORIZON EUROPE Marie Sklodowska-Curie Actions (843074)

  • David Berron

European Union’s Horizon 2020 Framework Programme for Research and Innovation (785907 (HBP SGA2) and 945539 (HBP SGA3))

  • Emrah Düzel

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank the Leibniz Institute for Neurobiology in Magdeburg for providing access to the 7 Tesla MRI Scanner. We are grateful for the support of Anne Hochkeppler and Regina Schwarzer with manual segmentations of the medial temporal lobe subregions and for insightful discussions regarding functional connectivity with Yi Chen. This work was supported by the European Union’s Horizon 2020 Research and Innovation Programme under grant agreements 785907 (HBP SGA2) and 945539 (HBP SGA3) as well as the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – Project-ID 42589994. DB has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No. 843074.

Ethics

Human subjects: Informed consent and consent to publish was obtained from human participants. The study received approval by the ethics committee of Otto-von-Guericke University, Magdeburg (Germany) under reference number 128/14.

Senior Editor

  1. Laura L Colgin, University of Texas at Austin, United States

Reviewing Editor

  1. Muireann Irish, University of Sydney, Australia

Reviewers

  1. Marshall A Dalton, The University of Sydney, Australia
  2. Menno P Witter, Norwegian University of Science and Technology, Norway

Publication history

  1. Preprint posted: December 18, 2021 (view preprint)
  2. Received: January 11, 2022
  3. Accepted: October 11, 2022
  4. Accepted Manuscript published: October 12, 2022 (version 1)
  5. Accepted Manuscript updated: October 13, 2022 (version 2)
  6. Version of Record published: November 11, 2022 (version 3)

Copyright

© 2022, Grande et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 600
    Page views
  • 237
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Xenia Grande
  2. Magdalena M Sauvage
  3. Andreas Becke
  4. Emrah Düzel
  5. David Berron
(2022)
Transversal functional connectivity and scene-specific processing in the human entorhinal-hippocampal circuitry
eLife 11:e76479.
https://doi.org/10.7554/eLife.76479
  1. Further reading

Further reading

    1. Neuroscience
    Kevin Vinberg, Jörgen Rosén ... Fredrik Ahs
    Research Article Updated

    Understanding the neural basis for individual differences in the skin conductance response (SCR) during discriminative fear conditioning may inform on our understanding of autonomic regulation in fear-related psychopathology. Previous region-of-interest (ROI) analyses have implicated the amygdala in regulating conditioned SCR, but whole brain analyses are lacking. This study examined correlations between individual differences in SCR during discriminative fear conditioning to social stimuli and neural activity throughout the brain, by using data from a large functional magnetic resonance imaging study of twins (N = 285 individuals). Results show that conditioned SCR correlates with activity in the dorsal anterior cingulate cortex/anterior midcingulate cortex, anterior insula, bilateral temporoparietal junction, right frontal operculum, bilateral dorsal premotor cortex, right superior parietal lobe, and midbrain. A ROI analysis additionally showed a positive correlation between amygdala activity and conditioned SCR in line with previous reports. We suggest that the observed whole brain correlates of SCR belong to a large-scale midcingulo-insular network related to salience detection and autonomic-interoceptive processing. Altered activity within this network may underlie individual differences in conditioned SCR and autonomic aspects of psychopathology.

    1. Neuroscience
    Rong Zhao, Stacy D Grunke ... Joanna L Jankowsky
    Research Article

    Neurodegenerative diseases are characterized by selective vulnerability of distinct cell populations; however, the cause for this specificity remains elusive. Here we show that entorhinal cortex layer 2 (EC2) neurons are unusually vulnerable to prolonged neuronal inactivity compared with neighboring regions of the temporal lobe, and that reelin+ stellate cells connecting EC with the hippocampus are preferentially susceptible within the EC2 population. We demonstrate that neuronal death after silencing can be elicited through multiple independent means of activity inhibition, and that preventing synaptic release, either alone or in combination with electrical shunting, is sufficient to elicit silencing-induced degeneration. Finally, we discovered that degeneration following synaptic silencing is governed by competition between active and inactive cells, which is a circuit refinement process traditionally thought to end early in postnatal life. Our data suggests that the developmental window for wholesale circuit plasticity may extend into adulthood for specific brain regions. We speculate that this sustained potential for remodeling by entorhinal neurons may support lifelong memory but renders them vulnerable to prolonged activity changes in disease.