Inversion of retinotopic coding between externally- and internally-oriented brain networks.

A. Population receptive field (pRF) modeling with fMRI. We established visual field preferences for each voxel using pRF modelling(38). Voxels with positive BOLD responses to the visual stimulus are classified as positive pRFs (+pRFs), and those with negative BOLD responses are classified as negative pRFs (-pRFs). B. Individualized resting-state network parcellation. Resting-state fMRI was collected in all participants (N=7; 34-102 minutes per participant) and used to derive individualized cortical network parcellations. Parcellations were generated using the multi-session hierarchical Bayesian modelling approach with the Yeo 15 HCP atlas as a prior(8, 42). C. PRF model variance explained is greater when fit to the pRF mapping task versus resting-state fMRI data for negative and positive pRFs in the DN and dATN. Histograms show voxel-wise variance explained for all pRFs fit to rest (dark) and the pRF modeling task fMRI data (light). All distributions were significantly shifted rightward for data fit to the pRF modeling task (dATN pRFs: K-S test: D-=0.655, p<0.001; Negative DN pRFs: D-=0.232, p<0.001; Positive DN pRFs: D-=0.124, p<0.001). Dashed lines indicate the 95th percentile of R2 within the null distribution used; we chose the threshold for visually-defined voxels for subsequent analyses (R2 > 0.14) to exceed these values. D. DN and dATN pRF amplitude and center location estimates are reliable. To assess reliability of pRF amplitude estimates, we compared voxel-wise signed amplitude by computing the dice coefficients of binarized pRF amplitude maps (i.e., positive or negative pRFs) between pairs of pRF modelling runs. To estimate center position, we calculated the distance between pairs of pRF centers iteratively between pRF modeling runs. For both metrics, chance was estimated by performing the same computation on a random sampling of voxels drawn from both pRFs and non-pRFs (5000 iterations for each metric). In all subjects, for dATN positive and DN positive and negative pRFs, pRF amplitude and center position exceeded what would be expected by chance (z-scored Dice/distance values > 2.9). Although only dATN positive pRFs were included in subsequent analyses, the dATN pRF amplitude was not reliable above chance in two participants; this suggests that negative amplitude pRFs are not stable in all cortical networks. E. The internally-oriented DN contains a larger proportion of negative pRFs than the externally-oriented dATN. Proportion of negative pRFs (of total pRFs that survive thresholding in that network) in all cortical networks are shown. Inset. DN-A and DN-B contain more negative pRFs than the dATN (t(6)=7.94, p<0.001), and there is no difference in the proportion of negative pRFs among sub-networks. F. DN and dATN pRFs networks show contralateral bias. Coverage plots across participants for positive dATN pRFs and positive and negative DN pRFs. Bar plots show the difference in coverage between contra-versus ipsi-lateral visual fields for each pRF type. Bar plots show mean across participants, connected points represent individual participant data.

Retinotopic coding organizes spontaneous interaction between internally and externally oriented brain networks.

A. Overall activity of the dorsal attention networks (dATN-A/B) and default networks (DN-A/B) were independent at rest. One individual’s dATN and DN are shown. Time series from all voxels in the DN and dATN subnetworks were averaged together, and we calculated the partial correlation(44) of these networks while accounting for variance of the other 11 networks within the group-prior parcellation(1, 8). The dATN and DN network correlation was not significantly different from zero, suggesting the activity of these networks is largely independent. Bar plot shows the average partial correlation across participants, each data point is the average of all runs for a single participant. B. Interaction between the DN and dATN differs by visual field preference of DN voxels. Splitting the DN voxels by visual responsiveness revealed distinct patterns of resting-state correlation. DN +pRFs had a positive correlation with the dATN (mean correlation = 0.22±0.144, t(6) = 3.99, p = 0.0072), while non-retinotopic DN voxels (i.e., pRF model R2 < 0.08) were not significantly correlated with the dATN (mean correlation = 0.02±0.115, t(6) = 0.37, p = 0.72). On the other hand, DN-pRFs were anti-correlated with the dATN (mean correlation = - 0.20±0.12, t(6) = 4.22, p = 0.0055). C. Determining spatially-matched pRFs in dATNs and DNs. We assessed the influence of retinotopic coding on the interaction between internally- and externally-oriented brain areas’ spontaneous activity during resting-state fMRI, by comparing the correlation in activation between pRFs in these networks that represent similar (vs. different) regions of visual space. For each DN-pRF, we established the top 10 closest positive dATN pRFs’ centers based on their x and y position (“matched”), and correlated the DN pRF time series with the average timeseries of these matched dATN pRFs. We compared these correlation values with the correlation of the DN pRF timeseries paired with 10 randomly selected pRFs from the 33% furthest (1000 iterations per pRF). We computed this for all DN pRFs for each all runs in each subject. Inset shows the pairwise distances between all pRFs in all participants, dotted line shows the 66%ile. D. Negative (upper) and positive (lower) DN pRFs show retinotopic specific interaction with the dATN. Each set of bars depicts the average correlation of the matched and randomly paired pRFs between the DN and dATN for each participant, with each run shown as connected points. All participants showed greater correlation (either positive or negative) for matched compared with randomly paired pRFs. E. Data shown averaging runs within each participant. For both positive and negative pRFs, the correlation between matched compared with randomly matched pRFs was significantly greater (consistent with their amplitude, ps<0.01). F, G. show the correlation between positive and negative pRFs in the DN with the dATN, split across the sub-networks of the DN-A and B. Both DN-A and B show a significant impact of matching, indicating that both subnetworks interact with the dATN via a retinotopic code (ps<0.05). *: p<0.05, **: p<0.01, ***:p<0.001.

Top-down vs. bottom-up neural events detected in spontaneous resting-state dynamics show evidence for retinotopically-specific suppression.

A. Event detection and analysis procedure and example events from a single resting-state run. To detect events, we extracted the time series from each pRF in the source regions (top-down: positive or negative pRFs in the DN; bottom-up: positive pRFs in the dATN) and isolated time points where the z-scored time series exceeded 2.4 s.d. (99th percentile). We then examined the activity of 10 best matched and 10 worst matched (anti-matched) pRFs from the target region in this peri-event time frame (6 TRs (8 s) before and after the event). B. Example top-down event in the representative time series shows elevated source pRF activity (DN-pRF) and corresponding target area activity (dATN +pRF). C. Same as B, but for bottom-up activity with the event detected in a dATN +pRF. D, E. Evidence for retinotopically-specific suppression of ongoing activity during events top-down events detected in DN-pRFs (D) and bottom-up events detected in dATN +pRFs (E). D. Time series show the peri-event activity of matched and anti-matched dATN +pRFs averaged across all participants for events detected in DN-pRFs (N=78,978 events). E. Time series shows the peri-event activity in DN-pRFs during events detected in dATN +pRFs (N=541,546). F. For DN-pRFs paired with dATN +pRFs, target area shows retinotopically-specific suppression of activity for both top-down and bottom-up events. Bars show the average activation at event time of the target areas’ matched and anti-matched pRFs for each participant. Activity was lower in matched compared with anti-matched pRFs for top-down and bottom-up events (ts>3.8, ps<0.009). G, H. Evidence for retinotopically-specific excitation of ongoing activity during events top-down events detected in DN +pRFs (G) and bottom-up events detected in dATN +pRFs (H). G. Grand average time series for matched and anti-matched dATN +pRFs during events detected in DN +pRFs for all participants. H. Grand average time series for matched and anti-matched DN +pRFs during events detected in dATN +pRFs. I. For DN +pRFs paired with dATN +pRFs, target area shows retinotopically-specific excitation for both top-down and bottom-up events. Bars show the average activation at event time of the target areas’ matched and anti-matched pRFs for each participant. Activity was higher in matched compared with anti-matched pRFs for top-down and bottom-up events (ts>8.75, ps<0.001). Interestingly, the activity of anti-matched pRFs was greater during top-down compared with bottom-up events (t(6)=4.92, p=0.0026), suggesting that top-down input from DN +pRFs to the dATN +pRFs may be more global and less retinotopically-targeted compared with bottom-up signaling between these networks. For time series plots, solid bars along the x-axis indicate periods with a significant difference between matched and anti-matched pRF activity within the target area. *: p<0.05, **: p<0.01, ***:p<0.001. -DN: Negative DN pRFs; +DN: positive DN pRFs; +dATN: Positive dATN pRFs.

Cortical network parcellations for individual participants.

Networks were parcellated based on a 15-network group prior(1, 8) using the multi-session hierarchical Bayesian model (MS-HBM) method(42).

PRF amplitude maps from all participants.

Only voxels with greater than 8% variance explained by the pRF model are shown for visualization only.

Negative pRFs are clustered in DN-A and DN-B.

Brain shows the consensus signed pRF amplitude map, which is the sum of the signed pRF amplitude across participants. Consensus was calculated by binarizing each vertex based on the sign of the visual response: participants with a positive response at a vertex were assigned a +1, participants with a negative response at a vertex were assigned a −1, and participants that did not respond (i.e., had variance explained below our 0.14 threshold) were assigned a 0. We then summed these maps across participants. Vertices with positive values were positive in the majority of participants, vertices with negative values were negative in the majority of participants, and vertices with 0 value showed no consensus. Outlines show DN-A (purple), DN-B (pink), dATN A (light green), and dATN-B (dark green).

Eccentricity and pRF size differs between DN and dATNs.

A. Both positive and negative DN pRFs tend to be more eccentric than dATN pRFs (positive dATN v negative DN pRFs: t(6)=2.27, p=0.06; positive dATN v positive DN pRFs: t(6)=3.04, p=0.023; negative DN v positive DN pRFs: t(6)=0.41, p=0.69). B. Both positive and negative DN pRFs tend to be smaller than dATN pRFs (positive dATN v negative DN pRFs: t(6)=2.36, p=0.056; positive dATN v positive DN pRFs: t(6)=2.46, p=0.049; negative DN v positive DN pRFs: t(6)=0.16, p=0.87) (Fig. S4A). Importantly, these results are averaged across the whole network. Results may differ when individual regions of each network are considered. *=p<0.05, +=p<0.1, n.s. = non-significant.

Spatial distribution of positive dATN pRFs matched with negative DN pRFs in cortex across all participants.

Vertices are colored based on number of matched pRFs. Only vertices with greater than 7 matched pRFs (i.e., at least one for each subject) are shown. Group prior maps(1, 8) dATN A-B and DN A-B are shown for reference. Note that pRFs were matched in individualized dATN of each participant and are not expected to correspond precisely with the group network maps.

A. Time series show the peri-event cross-covariance of matched and anti-matched dATN pRFs averaged across all participants for events detected in negative DN pRFs (N=78,978 events). B. Time series shows the peri-event cross-covariance in negative DN pRFs during events detected in dATN pRFs (N=541,546). C. Bars show the average cross-covariance at event time of the target areas’ matched and anti-matched pRFs for each participant. Although this difference did not reach significance (F(1,24)=3.94, p=0.058), the results qualitatively support our conclusion – that matched pRFs in the target region have a stronger interaction with the source pRFs. D, E. Evidence for retinotopically-specific excitation of ongoing activity during events top-down events detected in positive DN pRFs (D) and bottom-up events detected in dATN pRFs (E). D. Grand average cross-covariance for matched and anti-matched dATN pRFs during events detected in positive pRFs for all participants. E. Grand average cross-covariance for matched and anitmatched positive DN pRFs during events detected in dATN pRFs. I. For positive DN pRFs paired with dATN pRFs, target area shows retinotopically-specific positive covariance for both top-down and bottom-up events. Bars show the average activation at event time of the target areas’ matched and anti-matched pRFs for each participant. We observed a strong influence of match status on covariance (F(1,25)=25.324, p<0.001; t(6)=7.73, p<0.001). Similar to our findings when examining activation at event time, bottom-up events had a greater influence from retinotopic coding compared with top-down events, evidenced by a greater difference between matched versus anti-matched pRF covariation at event time (F(1,24)=6.22, p=0.02, t(6)=3.99, p=0.007). For time series plots, solid bars along the x-axis indicate periods with a significant difference between matched and anti-matched pRF activity within the target area.

Events are widely distributed in time and across voxels.

A,C,E. Number of events per time point combined for top down and bottom up events for dATN, positive DN, and negative DN pRFs. Bars indicate the number of voxel events detected at a given time point within a resting-state run, aggregated across runs. Events were sparsely distributed: most time points contained no events (Proportion = 0.66, bar not shown), with the number of events decreasing approximately exponentially. This is further evidence for a wide temporal distribution of events. B,D,F. Proportion of voxels with events in each run for dATN, positive DN, and negative DN pRFs. No voxel had more than 4 events per run. Data are combined between top down and bottom-up events.

Retinotopic coding structures the spontaneous interaction between functionally-coupled mnemonic and perceptual areas during resting-state fMRI.

A. Isolating functionally coupled internally- and externally-oriented brain areas within the DNs and dATNs. We identified brain areas that were specialized in two domains: scenes and faces processing. Specifically, we focused on the lateral place memory area (LPMA; from (36)), white), a memory area in the domain of scene perception at the posterior edge of the DN-A (purple; from (8)). We examined LPMA’s relationship to three different of category-selective visual areas in the dATN (green; from (8)), 1) the occipital place area (OPA; from (36)), an area within the domain of scene perception, along with 2) the occipital face area (OFA) and 3) the fusiform face area (FFA), two areas involved in the domain of face perception (white; from (101)). B-C. We localized LPMA in all participants by contrasting the correlation in resting-state activity between anterior and posterior parahippocampal place area (PPA) (B). This yielded a region in lateral occipital-parietal cortex that overlapped with the LPMA defined in an independent group of participants (C). D-E. Consistent with prior work, the connectivity-defined LPMA had greater concentration of negative pRFs compared to OPA (D), and exhibited a lower visual field bias to OPA (E), consistent with an opponent interaction between these areas during perception. F. We assessed the influence of retinotopic coding on the interaction between negative pRFs in mnemonic and positive pRFs in perceptual areas using the same pRF matching and correlation procedure described above. We compared pRFs within functional domain (scene memory x perception – LPMA to OPA) as well as across domains (scene memory x face perception – LPMA to the occipital face area (OFA) and fusiform face area (FFA)). G. We found that LPMA and OPA were interlocked in a retinotopically-grounded opponent interaction. Resting-state activity of -LPMA and +OPA pRFs was reliably negatively correlated after accounting for variance associated with +LPMA pRFs, and, critically, this negative correlation was stronger for matched compared with randomly matched pRFs (t(6)=3.012, p=0.024). We next tested whether this opponent dynamic was modified by functional domain (i.e., the scene memory area LPMA paired with the face perception areas FFA and OFA), and we observed no significant difference between the distribution of correlation values for matched and randomly-matched pRFs between -LPMA and OFA (t(6)=0.59, p=0.57) and -LPMA and FFA (t(6)=2.01, p=0.09). However, when we compared the difference of matched and randomly-matched pRFs across domains, we found no significant interaction (F(1,19)=0.99,p=0.33).