To better understand the response heterogeneity observed in IC neurons, we used transgenic lines in which the GCaMP reporter was selectively expressed in subpopulations. Two transgenic mouse lines were used: GP4.3, a Thy1-driven GCaMP6s expression line (Dana et al., 2014), and a cross between Gad2-IRES-Cre line (Taniguchi et al., 2011) and Ai96, a Cre-dependent GCaMP6s reporter line (Madisen et al., 2015). We will refer to the latter as Gad2;Ai96.

To confirm the neurochemical identity of GCaMP6s-positive neurons in the transgenic lines, we performed immunolabeling of GAD67. Figure 1 shows example staining and the proportion of GCaMP-positive neurons expressing GAD67 in each line. We did not distinguish between the different subregions of the IC in this analysis. The large majority (4651/4898; 95%) of GCaMP-positive neurons in the GP4.3 line were not positive for GAD67 (Figure 1A–D), showing that the Thy1-promoter preferentially expressed GCaMP6s in non-GABAergic cells in the IC. Not all non-GABAergic cells were GCaMP-positive, as shown by the NeuN-positive, GCaMP-negative cell bodies in Figure 1C (arrowheads). In particular, we observed that GCaMP6s-positive neurons are enriched at the border of neurochemical modules characterized by dense GAD67 terminals (Figure 1E–G).

Figure 1 with 1 supplement see all

Download asset Open asset

Figure 1—source data 1

CellCounter source data XML files generated by Cell counter plug-in in FIJI containing the actual counted datapoints of each cell type, with text files explaning the different cell types being counted.

https://doi.org/10.7554/eLife.49091.004

Download elife-49091-fig1-data1-v2.zip

In contrast, we detected GAD67 immunoreactivity in >93% (1702/1828) of GCaMP-positive neurons in the Gad2;Ai96 mice (Figure 1D,H–O). Not all GAD67-positive cells in this mouse line were labeled by GCaMP6s (e.g. arrowheads in Figure 1J).

In addition to GAD67, the calcium buffer parvalbumin (PV) is another marker for neurochemical modules in the rat (Chernock et al., 2004), while the calcium buffer calretinin (CR) shows a complementary expression (i.e. in the extramodular region) in early postnatal and juvenile mice (Dillingham et al., 2017). To get a better idea of the relative contribution of modular and extramodular regions to GCaMP6 responses, we investigated the co-expression of GCaMP6 with PV or CR in the two transgenic lines (Figure 1—figure supplement 1). We found that a large majority (94%) of GCaMP+ cells in GP4.3 did not show strong immunoreactivity for either marker (CR++: 2.2%; PV++: 2.9%; PV++CR++: 0.9%), whereas in the Gad2;Ai96 line a larger fraction of GCaMP+ cells were positive for CR or PV (CR++: 17%; PV++: 22%; PV++CR++: 7%; Figure 1—figure supplement 1G). On the other hand, GCaMP was expressed in 38% of CR++ cells and 70% of PV++ cells in the IC of Gad2;Ai96 mice, while a lower fraction of CR++ (9.1%) and PV++ (13.4%) cells were positive for GCaMP in GP4.3 mice (Figure 1—figure supplement 1H,I). At the single neuron level, the presence or absence of the two buffers was insufficient to indicate whether the neurons lay within a module. These results suggest that for both lines there is not a clear enrichment of GCaMP6+ neurons within the neurochemical modules.

Unanaesthetized, head-fixed mice with implanted cranial windows (Figure 2A–B) were imaged with two photon microscopy (Figure 2C–D) and subjected to 1 s pure tones of various frequencies and intensities (Figure 2E). Average fluorescence of regions-of-interest (ROIs) encompassing the cell body were analyzed, after subtracting the fluorescent change of a background region surrounding the cell body (Figure 2D). The use of longer stimuli allowed a detailed characterization of the kinetics of the fluorescence change (ΔF_GCaMP). Importantly, we observed both increases, which were either transient or sustained, and decreases in F_GCaMP during the presentation of stimuli, and some cells showed a sharp increase in F_GCaMP directly after the cessation of the sound (Figure 2E). The different fluorescence responses were classified as excitation, inhibitory and offset, respectively (see Materials and methods, Figure 3A–D). The excitatory responses were further divided into onset and sustained classes based on the kinetics of the fluorescence change (Figure 3A,B, Figure 3—figure supplement 1). The decrease in fluorescence upon sound stimulation was likely due to sound-evoked inhibition of spontaneous firing in these cells. Some cells showed a mixture of response kinetics, some even to the same stimulus. Particularly common was the combination of inhibitory and offset responses (e.g. Figure 3E,M). Another common combination were onset-offset cells in which a lower frequency elicited an onset/sustained response, while a slightly higher frequency elicited an offset response (e.g. Figure 3I). Inhibitory and offset responses were not reported in earlier imaging experiments performed in anesthetized mice (Ito et al., 2014; Barnstedt et al., 2015); their use of shorter stimuli did not allow a distinction between onset and sustained responses. A great variety of frequency response areas (FRAs) was observed (Figure 3F–M), with some cells showing responses to tones as high as 64 kHz, the highest frequency presented in this study (e.g. Figure 3G).

Figure 2 with 1 supplement see all

Download asset Open asset

Figure 3 with 2 supplements see all

Download asset Open asset

Figure 3—source data 1

Fluorescence kinetics source data CSV file containing fluorescence kinetics of ROIs, genotype of animal and type of FRA.

https://doi.org/10.7554/eLife.49091.010

Download elife-49091-fig3-data1-v2.zip

In GP4.3 mice, 64% of the cells (696 out of 1090) cells showed detectable fluorescence changes in response to 1 s pure tone stimuli. We classified the responses and the overall FRA of a cell based on the criteria described in the Materials and methods. Among the responsive cells, 30% (210 cells) showed excitatory FRAs (e.g. Figure 3—figure supplement 2), with fluorescence increasing during sound presentation; while 13% (89 cells) showed offset FRAs, with fluorescence increasing after sound presentation. Another 30% (207 cells) showed inhibitory FRAs, with fluorescence decreasing during sound. The FRAs of the remaining 27% (190 cells) showed mixtures of these response categories. The proportion of different FRA classes are summarized in Figure 4A. The proportion of cells showing sound-evoked inhibition was clearly higher than the 2–23% reported for anesthetized C57BL/6 mice in Willott et al. (1988a).

Figure 4

Download asset Open asset

Figure 4—source data 1

IC ROI Area source data CSV file containing fluorescence area and estimated diameter of ROI, genotype of animal and type of FRA.

https://doi.org/10.7554/eLife.49091.012

Download elife-49091-fig4-data1-v2.zip

In Gad2;Ai96 mice 73% (247 out of 342 cells) of GCaMP6s-positive IC neurons showed consistent fluorescence responses to 1 s pure tone stimuli. Among the responsive cells, 46% (115 cells) showed excitatory FRAs. Four percent (11 cells) showed offset FRAs. Another 26% (65 cells) showed inhibitory FRAs. The remaining 23% (56 cells) showed mixtures of the three response categories. The proportion of different FRA classes in Gad2;Ai96 mice are summarized in Figure 4B; it was similar to that of GP4.3 mice, but with a somewhat smaller fraction of offset and a larger fraction of excitatory FRAs.

GABAergic cells had a larger soma size, as reported previously (cat: Oliver et al., 1994; rat: Merchán et al., 2005). The estimated diameters for Gad2;Ai96 and GP4.3 cells, defined as 4/π times the square root of the ROI area in imaging experiments, were 21.9 ± 4.9 µm (mean ±s.d.) and 15.9 ± 2.6 µm, respectively (Figure 4C, p<0.001 Wilcoxon Rank Test). Within each genotype, soma size did not differ significantly among cells with different FRA classes. For GP4.3: onset cells 12.3 ± 2.6 µm, sustained 12.5 ± 2.0 µm, inhibited 13.0 ± 1.8 µm, offset 12.3 ± 2.0 µm (p=0.112, Kruskal-Wallis Rank test). For Gad2;Ai96: onset cells 16.6 ± 3.4 µm, sustained 17.4 ± 3.6 µm, inhibited 18.1 ± 3.3 µm, offset 15.6 ± 3.1 µm (Figure 4C; p=0.033, Kruskal-Wallis Rank test, followed by Dunn’s [Dunn-Holland-Wolfe] test where none of the comparisons reached critical value; only cells with non-mixed FRAs were compared because of the heterogeneity and small sample size of the different mixed response classes).

The temporal kinetics of the F_GCaMP response of each cell were assessed by averaging the fluorescence change across all stimuli that showed a significant response. Single exponential functions were fit to the onset (0–1 s re stimulus onset) and the offset/decay (0.5–4 s re stimulus offset) periods (Figure 3—figure supplement 1A,B). Time constants were restricted to be positive. Due to the limited stimulus duration, any fit resulting in an onset time constant (τ_onset) greater than 1000 ms to an increase in fluorescence was considered sustained activity and these fit constants were not used for averaging. For excitatory responses (onset, sustained and offset), a time constant of around 1 s (Figure 3—figure supplement 1D) was found for the decay period, similar to the decay time constant reported previously for GCaMP6s (Chen et al., 2013).

To relate the different fluorescence responses to spiking patterns, which are highly heterogeneous within the IC (e.g. Willott et al., 1988a; Willott et al., 1988b; Tan and Borst, 2007), we combined Ca²⁺ imaging with in vivo juxtacellular recordings in awake mice. We recorded from neurons showing onset, sustained and inhibitory tone-evoked fluorescence responses. For fluorescence responses that we classified as onset, electrophysiological recordings showed that this class can represent a cell whose firing rate quickly adapted during our one-second long stimuli (example in Figure 5A–C,D left). We observed sustained fluorescence responses that corresponded to sustained firing (Figure 5D middle) with different amount of adaptation. Inhibitory responses corresponded to cells that reduced their spontaneous firing upon sound stimulation (Figure 5D right).

Figure 5 with 3 supplements see all

Download asset Open asset

Figure 5—source data 1

Ground-truth model fitting source data CSV file containing model parameters from ground-truth data fitting and the variance explained, genotype of animal and type of FRA.

https://doi.org/10.7554/eLife.49091.017

Download elife-49091-fig5-data1-v2.zip

We further characterized the relationship between fluorescence responses and firing patterns by fitting a model to our data set (Figure 5E,F, Figure 5—figure supplements 1–3; see Materials and methods for equations). We found that on average, an action potential led to a median increase of 0.35 (GP4.3; n = 5 cells) or 0.14 (Gad2;Ai96; n = 12 cells) times the minimal fluorescence (F₀), with a median zero-to-peak rise time of 480 ms (GP4.3) or 560 ms (Gad2;Ai96) and a median decay time constant of 1.04 s (GP4.3) or 1.36 s (Gad2;Ai96). The model could explain between 50% and 95% of the variance in the fluorescence (median: 79%; average ±s.d.: 77 ± 13%).

The widespread and homogeneous expression of GCaMP6s in the IC of the transgenic mice allowed a good overview of its functional organization. We aligned the position of the cells from multiple animals (GP4.3: n = 7 mice; Gad2;Ai96: n = 6 mice) by anatomical landmarks (midline, anterior and posterior extent of the exposed IC, lateral extent of the exposed IC), and plotted the cells on this common anatomical coordinate system (Figure 6A; Figure 6—video 1).

Figure 6 with 9 supplements see all

Download asset Open asset

Figure 6—source data 1

IC CF Distribution source data CSV file containing 3D coordinate of ROIs, projected distance of ROIs, genotype of animal and type of FRA.

Source code for fit function used for fitting the spatial distribution.

https://doi.org/10.7554/eLife.49091.027

Download elife-49091-fig6-data1-v2.zip

Interestingly, we observed a central strip running in the caudomedial-rostrolateral orientation of cells responding to lower frequencies, while CF progressively increased in both the caudolateral and the rostromedial direction. We tried to estimate the orientation of the spatial organization by projecting the x,y-coordinates of all neurons onto an axis with a parametrized angle θ, while simultaneously fitting the log-transformed CF using a polynomial model (see Materials and methods). We used a polynomial model fit as a way to extract any general direction along which CFs may diverge, without imposing any presumption about the spatial CF dependence. We found that a fourth order polynomial captured the variance maximally; further increasing its order did not lead to better fits. The results of the fit are presented as contour lines in Figure 6A, with the best orientation (θ) 50° from the medial-lateral (x) axis. A plot along this orientation shows a high fraction of low CF (~4 kHz) neurons near the 625 μm position, with the mode going towards ~20 kHz at the 1300 μm end (Figure 6B). The increase in CFs on the rostromedial side was not as pronounced, likely due to the low number of neurons sampled in this region. The solid line in Figure 6B shows the best polynomial fit, which explained around 13% of the variance in CFs. This tonotopic organization can also be appreciated in individual animals in which a wide extent of the dorsal IC was imaged (Figure 6—figure supplements 1–3). Additional examples of the tonotopic organization in individual animals are shown in Figure 6—figure supplements 4–6. A similar spatial distribution of CFs was observed for GABAergic and glutamatergic neurons (Figure 6C) and for cells showing different response classes (Figure 6D).

Based on a clustering analysis, Barnstedt et al. (2015) suggested that the central strip of low frequency neurons were from the most dorsal end of the central nucleus, and we did observe a small but significant correlation between depth and log(CF) (Figure 6—figure supplement 7; correlation coefficient = −0.22; n = 799; p=3 × 10⁻¹⁰). However, a plot of CF along the best orientation from the polynomial fit as a function of depth illustrates that most cells were located within 100 µm from the surface, and that the low frequency neurons were also among the most superficial cells (Figure 6E). When we limited the fit to neurons < 50 μm from the pia surface, the same tonotopic pattern appeared (Figure 6—figure supplement 8). There was an over-representation of low CF neurons in the deepest regions (121–160 μm deep, Figure 6E,F). We attribute this to a sampling bias in which the low frequency region coincided with the center of the cranial window, where it was possible to image somewhat deeper than towards the edges (Figure 6E). Objects in the center of the window profit from utilization of the full NA for focusing excitation light, which is especially critical for two-photon excitation, and in the center both excitation and fluorescence emission light is least obstructed (Figure 2A). Restricting the analysis to cells in the central strip (525–725 μm of projected area) reduced the correlation coefficient for the relation between depth and log(CF), but it remained significant (Figure 6—figure supplement 7; correlation coefficient = −0.17, n = 295; p=0.003), hinting at differences between deeper and the most superficial cells. While we do not exclude the possibility of imaging into the central nucleus of the IC, we believe the observed tonotopy to be a good representation of the IC shell region because all CFs were represented among the most superficial neurons (Figure 6E,F).

We next asked whether there was any obvious spatial organization of the different response classes. Figure 7A shows the proportion of different response classes along our presumed tonotopic axis, which looked rather homogenous, without any obvious difference between cells rostromedial and caudolateral to the CF minimum. The same pattern holds for the orthogonal direction (Figure 7B).

Figure 7

Download asset Open asset

Figure 7—source data 1

IC Response Distribution source data CSV file containing 3D coordinate of ROIs, projected distance of ROIs, genotype of animal and type of FRA.

https://doi.org/10.7554/eLife.49091.030

Download elife-49091-fig7-data1-v2.zip

Interestingly, we observed cells that showed fluorescence transients in the absence of sound presentation (Figure 8A,B). Numerous studies have shown both ascending and descending somatosensory projections to the IC (Aitkin et al., 1981; Künzle, 1998; Jain and Shore, 2006; Zhou and Shore, 2006; Lesicko et al., 2016; Patel et al., 2017), as well as from nuclei that are upstream from IC such as the dorsal cochlear nucleus, which also receive somatosensory inputs (Wu et al., 2014). In addition, the IC receives inputs from the motor cortex (Cooper and Young, 1976; Olthof et al., 2019). This prompted us to investigate whether these ‘spontaneous’ activities could be attributed to non-auditory inputs. Our recordings were performed while the animal was passively awake, which allowed us to observe and correlate the voluntary movement of the animal to simultaneously recorded calcium transients.

Figure 8 with 1 supplement see all

Download asset Open asset

Figure 8—source data 1

IC Movement cells source data CSV file containing 3D coordinate of ROIs, projected distance of ROIs, genotype of animal, type of FRA and parameters related to movement-related response.

https://doi.org/10.7554/eLife.49091.033

Download elife-49091-fig8-data1-v2.zip

We found that the onset of many of these spontaneous transients coincided with movement events of the animal (paw and facial movements in Figure 8A). By correlating movement and fluorescence, we detected 165 (out of 1359) cells that showed a positive correlation (r > 0.25) with facial movements during the spontaneous recording period. To exclude potential false-positive detection due to motion artefacts (i.e. cells moving into and out of focus), we excluded cells with fluorescence transients that did not show the typical ~1 s exponential decay kinetics, which last much longer than the brief image shifts that could accompany movement events (image shifts in Figure 8A, Figure 8—figure supplement 1). Since animal movement inevitably produces sound that may activate the IC neurons, leading to an apparent movement sensitivity, we also excluded the cells that showed excitatory or offset responses in their FRA. In the end, 41 cells showed spontaneous calcium transients that correlated with animal movement and that could not be explained by their FRA, which was either inhibited-type or showed no clear pure tone evoked responses (example FRAs in Figure 8F). These cells seemed to be enriched at the caudolateral side of the dorsal IC (Figure 8C–D). Four of the 41 cells were found in the Gad2;Ai96 line, in agreement with a recent report that inputs from motor or somatosensory cortex can directly target GABAergic cells (Olthof et al., 2019), and the remainder in the GP4.3 line, making them 1% of GABAergic and 4% of investigated glutamatergic cells, respectively.

Figure 9A and E show two relatively superficial brain sections (within 80 μm and 120 μm from dorsal surface, respectively) from one Gad2;Ai96 and one GP4.3 animal stained for GAD67 after two photon imaging. We overlaid the line representing the neurons with the lowest CF and compared its location with the border of LCIC and DCIC traced from the latest Allen Reference Atlas (CCFv3) and from the classical reference atlas by Paxinos and Franklin (2001). The minimum frequency line aligns with the demarcation from the Allen Reference Atlas, while that by Paxinos and Franklin lays in the orthogonal direction.

Figure 9 with 7 supplements see all

Download asset Open asset

Neurochemical modules with a high density of GAD67-positive terminals have been reported as a hallmark for the LCIC (Chernock et al., 2004; Lesicko et al., 2016; Dillingham et al., 2017). We indeed found modules with higher density of GAD67 terminals in our animals as well (arrowheads in Figure 9A,E,a clear example in Figure 9C). In the most superficial sections, these modules were not as well defined, but they were contiguous to the more densely stained modules in deeper sections (Figure 9—figure supplements 1 and 4; 3D reconstruction in Figure 9—video 1), and the superficial modules also had reduced calretinin staining (Figure 9—figure supplements 2 and 5), in line with previous findings (Dillingham et al., 2017). These modules are the predominant target of the somatosensory projections from the cerebral cortex as well as other brainstem areas (Lesicko et al., 2016). We also observed a central strip of dense GABAergic staining (enlarged in Figure 9B), which was not contiguous with the modules (Figure 9—figure supplements 1 and 4). GCaMP expression in these two example animals are presented for comparison in Figure 9—figure supplements 3 and 6.

How do motion-sensitive cells associate with the GAD67-dense modules? Figure 9D2 and F show two motion-sensitive neurons retrieved in histology (cells 3 and 4, marked by arrows), both of which resided in close proximity to a GAD67-dense module. We observed that although the somata of both cells seemed to reside in the extramodular regions, both possessed dendrites that extended into a GAD67-dense module (Figure 10). Figure 10A and B show a dendritic extension of cell 3. For cell 4, a pixel-wise correlation of the in vivo somatic fluorescence change (Junek et al., 2009) nicely revealed its dendritic arbor at the same focal plane (Figure 10C,E). Aligning this to post hoc histological staining (Figure 10D), we can demonstrate that cell four has dendritic branches extending both inside and outside of a GAD67 module (Figure 10F). We suggest that this may be a functional connection scheme for extramodular neurons to enable them to integrate somatosensory or motor with auditory inputs to the IC (Figure 10G).

Figure 10

Download asset Open asset

Because of the prominence of GABAergic neurons and their substantial contribution to ascending projections (Schofield and Beebe, 2019), we compared the responses and functional organization of GABAergic and glutamatergic neurons in the dorsal IC using in vivo two-photon calcium imaging. We confirmed that virtually all neurons expressing GCaMP6s in the GP4.3 line were glutamatergic, but not all glutamatergic neurons were labeled. There may have been a preponderance of labeled cells around the modules in the LCIC (Figure 1D), but a systematic screen of markers would be needed to test for a selective enrichment of glutamatergic subtypes in the GP4.3 line. The large majority of GCaMP6s-positive neurons in the IC of mice from the Gad2;Ai96 line expressed GAD67, in agreement with findings in a related mouse line (Gay et al., 2018). In our hands, these two transgenic lines had the advantage that the kinetics of the responses were relatively homogeneous and well approximated by a linear model (Figure 5B), suggesting that the expression levels, which are a major determinant of the Ca²⁺ kinetics (e.g. Éltes et al., 2019), were similar across cells. Moreover, in contrast to pilot experiments in which we expressed GCaMP6 using adeno-associated viruses, we found little evidence for toxicity. The more sparse expression facilitated the isolation of the responses of individual neurons.

A substantial fraction of cells did not respond to tones. Low probability responses may have been below detection, since the juxtacellular recordings indicated that single APs could not always be detected. Some cells may respond preferentially to more complex sound stimuli, as previously found for neurons in the shell region (Ehret and Moffat, 1985; Aitkin et al., 1994). As discussed below, some cells may prefer somatosensory or motor stimuli instead of sound stimuli.

In both transgenic lines, we observed four different tone-evoked fluorescence response patterns in awake mice: onset, sustained, offset and inhibitory. Our simultaneous juxtacellular recordings showed corresponding patterns. These types were typically also observed in several earlier electrophysiological studies (e.g. Willott et al., 1988a; Willott et al., 1988b; Jain and Shore, 2006; Xie et al., 2007). Apart from a possibly lower fraction of GABAergic neurons with pure offset responses (Figure 4B), our results are thus in line with a recent study that showed that GABAergic and glutamatergic IC neurons have similar response properties throughout the IC (Ono et al., 2017).

The prominence of inhibitory responses in the IC (~44% of sound-responsive neurons) in both transgenic lines was a striking finding in our recordings. An important factor contributing to their prominence is probably that our experiments were done in awake animals, as a recent single unit recording study showed the presence of inhibitory responses as well as higher spontaneous spike rates in IC neurons of awake mice compared to mice under urethane anesthesia (Duque and Malmierca, 2015), and synaptic inhibition has also been shown to be prominently present in whole-cell recordings in awake bats (Xie et al., 2007).

Offset responses can be found throughout the auditory system including the IC, and they may have an important role in perceptual grouping or in duration discrimination (Kopp-Scheinpflug et al., 2018). However, studies in the IC of anesthetized mice, rats and chinchillas showed that offset responses are generally restricted to cells showing band-pass duration tuning, which only responded to much shorter tones than the 1 s employed here (Chen, 1998; Brand et al., 2000; Pérez-González et al., 2006). Another striking finding was therefore that we saw clear evidence for offset responses in 3–11% of sound-responsive cells. A whole-cell study in the mouse IC showed that in most cases offset responses are inherited from upstream, but that they may also be generated de novo as a rebound from inhibition (Kasai et al., 2012). Similar to the inhibition class, we therefore suggest that the prominence of offset responses may be related to the increased impact of synaptic inhibition in the IC of awake animals. A study using communication calls suggests that offset responses may be even more prominent following more complex sound stimuli, such as communication calls (Akimov et al., 2017), possibly by a summation of inhibition at different frequency bands (Sanchez et al., 2008).

Following pooling of the data from several experiments, we observed a tonotopic gradient that ran along a gradient from caudolateral to rostromedial. The gradient reversed at the site where the CFs were the lowest. This reversal is in line with the banding patterns observed with epifluorescent calcium imaging in unanesthetized mice before and at hearing onset; the pre-hearing spontaneous neural activities originate from the cochlea, thus regions activated synchronously likely receive input from the same tonotopic area (Babola et al., 2018). In addition, they observed a double band pattern with higher frequency stimulation in the juvenile (p15) mice. Interestingly, our data are also compatible with two earlier two-photon calcium imaging studies (Ito et al., 2014; Barnstedt et al., 2015). Ito et al. (2014) reported gradients in the medial regions of the dorsal IC, which ran from lateral (low-frequency) to medial (high-frequency), whereas Barnstedt et al. (2015) reported gradients in a mediorostral to laterocaudal direction.

We found considerable variability in CFs within a band. Imprecisions in the alignment of imaging areas and variability between animals may have contributed to the variability, but substantial variability was also observed within a single animal, in line with previous results (Ito et al., 2014; Barnstedt et al., 2015). Two-photon calcium imaging studies in the mouse auditory cortex have met with variable amount of microheterogeneity, apparently depending on layer, calcium dye, or the use of anesthetics (Bandyopadhyay et al., 2010; Rothschild et al., 2010; Winkowski and Kanold, 2013; Issa et al., 2014; Kato et al., 2017; Tischbirek et al., 2019).

The four response types showed a similar tonotopic organization, but considerable microheterogeneity. This microheterogeneity was larger than in an earlier study in which bulk loading of an organic Ca indicator was used (Ito et al., 2014), which may have made it more difficult to isolate responses from individual cells. Our results are in line with patch-clamp results showing considerable heterogeneity between adjacent cells within the dorsal IC, which extended to heterogeneity in their inputs (Geis et al., 2011). The microheterogeneity was similar for GABAergic and glutamatergic neurons. Our results thus differ from auditory cortex, where response patterns are different and local microheterogeneity is much smaller for (parvalbumin-positive) GABAergic neurons than for glutamatergic neurons (Maor et al., 2016; Liang et al., 2018; Liu et al., 2019). We conclude that despite the shared input and firing behavior during development (Babola et al., 2018), functional heterogeneity dominates within bands in the dorsal IC of adult mice. Whereas some of the mechanisms that underlie the microheterogeneity within the auditory cortex are being elucidated (Kato et al., 2017; Tao et al., 2017; Vasquez-Lopez et al., 2017), they remain to be explored for the dorsal IC.

We monitored spontaneous whisker and general body movements, and used strict criteria to exclude movement artefacts. As we restricted the analysis to neurons that were not excited by tones, we consider it unlikely that the observed movement-related responses were caused by self-generated sounds. Despite these severe restrictions, we did find that a few percent of the cells were excited during whisker or body movements, suggesting a substantial role for somatosensory or motor inputs. Little is known about the functional characteristics of motor inputs that are not related to eye movements (Gruters and Groh, 2012). For somatosensory inputs, the dominant effect appears to be inhibitory and only a minority of cells have been shown to respond to unimodal tactile stimuli (Aitkin et al., 1978; Aitkin et al., 1981; Zhou and Shore, 2006), which would not help detection based on our selection criteria. We did not further discriminate between different movements as they tended to be heavily correlated. Moreover, it is known that neurons in the LCIC have broad tuning for somatosensory input (Aitkin et al., 1981). Movement-sensitive neurons were not found in the high-frequency DCIC regions, but only in the putative CNIC region and the LCIC.

A defining feature of the LCIC is the presence of neurochemical modules, which have been shown in many species, including rats (Chernock et al., 2004; Choy Buentello et al., 2015), but also in both adult (Choy Buentello et al., 2015; Lesicko et al., 2016; Patel et al., 2017) and developing mice (Dillingham et al., 2017). While it is still unclear how neurons in these modules and extramodular zones connect with each other, there is evidence that somatosensory inputs project predominantly to the modules, while auditory input projects to the extramodular region (Lesicko et al., 2016). The main sources for somatosensory input to the LCIC are the spinal trigeminal nucleus, the dorsal column nuclei and the somatosensory cortex (reviewed in Gruters and Groh, 2012). These inputs target the layer two modules, which constitute only a small fraction of the LCIC volume. Moreover, neurons in the modules do not seem to target extramodular IC neurons (Lesicko and Llano, 2019). This raises the question how so many cells in the LCIC can be sensitive to non-auditory inputs. We obtained anecdotal evidence that neurons at the border of the modules can extend dendrites into the modules (Figure 10A,F), which would allow them to sample non-auditory inputs. Indeed, neurons in LCIC can have extensive dendritic trees (Smith, 1992), and Golgi stainings show that cells in layer 3 of LCIC can send dendrites up to layer 1 (Meininger et al., 1986; Malmierca et al., 2011). The CNIC has few somatosensory inputs, but at least half of the cells in CNIC are affected by stimulation of the dorsal column nuclei (Gruters and Groh, 2012). The CNIC may get somatosensory inputs indirectly via the dorsal cochlear nucleus, which is innervated by the spinal trigeminal nucleus and the dorsal column nuclei, or in the form of intracollicular input from the LCIC. Direct stimulation of the different non-auditory inputs would be needed to obtain a better idea of their importance for the observed facial or body movement-related responses.

There is little agreement on the precise borders of the different nuclei of the IC, and its parcellation has been different based on whether cell morphology, inputs, or physiological properties were chosen as the main classifier (Oliver, 2005). Here we used physiological properties as the main criterion to look at the borders between the DCIC, LCIC and CNIC in the dorsal IC. Our findings suggest that the laterocaudal area with high frequencies is the dorsal edge of the LCIC, whereas the mediorostral part of the dorsal IC belongs to the DCIC. The tonotopic reversal would thus provide a functional demarcation between the dorsal and lateral cortices of the IC. Earlier tracing studies of intracollicular projections in the guinea pig (Malmierca et al., 1995) and the rat (Saldaña and Merchán, 1992), as well as ascending projections from the cochlear nucleus in the rat (Malmierca et al., 2002), have shown interesting V-shaped arrangements in coronal sections, where the medial side corresponds to isofrequency laminae in CNIC and DCIC, while the lateral side corresponds to those in the LCIC. These V-shaped axon terminal plexuses, if extended to the dorsal surface of the IC, provide an anatomical explanation for the observed tonotopic reversal. This is also in agreement with the latest anatomical framework published by the Allen Brain Institute (Figure 9A; Allen Mouse Common Coordinate Framework, 2015), previous imaging studies (Ito et al., 2014; Barnstedt et al., 2015; Babola et al., 2018), the known somatosensory inputs to the LCIC, and both the descending and ascending inputs to the IC (Aitkin et al., 1981; Künzle, 1998; Jain and Shore, 2006; Zhou and Shore, 2006; Lesicko et al., 2016; Patel et al., 2017). Surprisingly, we observed GAD67-dense modules in both the medial and the lateral aspect of the IC (arrowheads Figure 9A,G), that is within both LCIC and DCIC in either atlas. The medial modules could not be imaged within this study, so their association with somatosensory inputs is currently unclear.

The central strip with neurons tuned to low frequencies in between the LCIC and the DCIC may be the most dorsal extension of the dorsolateral low frequency area of the CNIC (Stiebler and Ehret, 1985; Barnstedt et al., 2015). This strip contains the large density of GABAergic inputs (Figure 9A,B) that is characteristic of the CNIC (Choy Buentello et al., 2015). Lemniscal inputs are known to extend quite dorsally, almost up to the surface, and neurons with short latency sound responses have been found in the dorsal IC, although their exact location was not studied (Geis and Borst, 2013). The low tuning in the central strip extended close to the surface, but a more systematic study of their properties would be needed to be able to assign the most superficial layers to either the DCIC or another region. From our data it thus appears that the three main areas of the IC, LCIC, CNIC, DCIC, are readily accessible at the dorsal IC. Their accessibility for imaging studies will thus help to further delineate their functions and the role of their inputs in the future.

Detailed two-photon imaging experiments were conducted on seven GP4.3 transgenic animals (Dana et al., 2014) (three in C57BL/6J background and four in B6CBAF1/J background) and six F1 progeny between Gad2-IRES-cre (Taniguchi et al., 2011), and Ai96 reporter line (B6;129S6-Gt(ROSA)26Sor^{tm96(CAG-GCaMP6s)Hze}/J) (Madisen et al., 2015). We will refer to the Gad2-IRES-Cre x Ai96 cross as Gad2;Ai96. Postnatal age at recordings ranged between 11–35 weeks (median: 19; Q1: 15; Q3: 23). Ground-truth juxtacellular recordings were performed on four GP4.3 and three Gad2;Ai96 animals. Immunohistochemistry for cell counting was performed on three GP4.3 and three Gad2;Ai96 animals.

GP4.3 animals were originally obtained from the Jackson Laboratory (C57BL/6J-Tg(Thy1-GCaMP6s)GP4.3Dkim/J; JAX stock #024275); they were maintained in a heterozygous state by backcrossing to C57BL/6J from Charles Rivers (JAX C57BL/6J). To create GP4.3 animals with B6CBAF1/J background, heterozygous GP4.3 in C57BL/6J background were crossed with CBA/JRj mice from Janvier. Gad2-IRES-Cre (originally STOCK Gad2^tm2(cre)Zjh/J, the Jackson Laboratory) was maintained in homozygous state after >10 generations of backcrossing to C57BL/6J, and re-backcrossed to C57BL/6J every 4–5 generations. Ai96 mice were obtained from Jackson Laboratory with already 3 generations of backcross to C57BL/6J (N3), and were subsequently maintained by backcrossing to C57BL/6J for 5–7 generations. All experiments complied with the ethical guidelines for laboratory animals within our institute and with European guidelines, and were approved by the animal ethical committee of the Erasmus MC.

Animals were anaesthetised through respiratory intake of isoflurane and maintained at surgical level of anaesthesia, assessed through the hind limb withdrawal reflex. A heating pad with rectal feedback probe (40-90-8C; FHC, Bowdoinham, ME, USA) was used to maintain body core temperature at 36–37°C. Eye ointment (Duratears; Alcon Nederland, Gorinchem, The Netherlands) was used to keep the eyes moist during surgery. A bolus of buprenorphine (0.05 mg/kg; Temgesic, Merck Sharp and Dohme, Inc, Kenilworth, NJ, USA) was injected subcutaneously at the beginning of surgery. The skin overlying the IC was incised. Lidocaine (Xylocaine 10%; AstraZeneca, Zoetermeer, The Netherlands) was applied before removing the periosteum and cleaning the skull. After etching the bone surface with phosphoric acid gel (Etch Rite; Pulpdent Corporation, Watertown, MA, USA), a titanium head plate was glued to the cleaned bone above the left IC using dental adhesive (OptiBond FL; Kerr Italia S.r.l., Scafati, SA, Italy) and further secured with dental composite (Charisma; Heraeus Kulzer GmbH, Hanau, Germany).

Through an opening in the head plate, a craniotomy of 3 mm diameter centered at one of the ICs was made by thinning and removing the skull bone. A cranial window, made by gluing a 3 mm cover slip (CS-3R-0; Warner Instrument Inc, Hamden, CT, USA) on a custom built, 500 µm thick steel ring with UV-cured optical adhesive (NOA68; Norland Products), was installed over the exposed brain surface and secured with superglue. Each animal was allowed to recover for at least two days before the first measurements. For in vivo electrophysiology, the cranial window construct was gently removed, and the dura mater covering the IC and part of the cerebellum was carefully punctured and removed with a pair of fine forceps.

After all recordings had been done, animals received an intraperitoneal injection of pentobarbital (300 mg/kg) and were perfused transcardially, first with physiological saline (Baxter Healthcare, Zurich, Switzerland), followed by 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer (PB; 4% PFA in PB, pH 7.4; Merck).

For two-photon imaging of the IC, a 20X water-immersion objective (LUMPlanFI/IR, 20X, NA: 0.95; Olympus Corporation, Tokyo, Japan) on a custom-built two-photon microscope was used, except for simultaneous juxtacellular recordings, for which a long working distance 40x objective (LUMPlanFl/IR, 40X, NA: 0.80, Olympus Corporation, Tokyo, Japan) was used. Excitation light was provided by a MaiTai Ti:Sapphire laser (Spectra Physics Lasers, Mountain View, CA, USA) tuned to a wavelength of 920 nm. A layer of Ringer solution was put between the objective and the cranial window. GCaMP6s fluorescence was captured by a photomultiplier tube (H6780-20, Hamamatsu, Japan) after a barrier filter at 720 nm (FF01-720/SP-25; Semrock), a secondary dichroic at 558 nm and a green bandpass filter centered at 510 nm (bandwidth: 84 nm; FF01-510/84-25; Semrock); AlexaFluor 594 fluorescence was captured at the second channel with a red bandpass filter centered at 630 nm (bandwidth: 60 nm; D630/60; Chroma). Images (256 × 128 pixels) were collected at 9 Hz (2 μs/pixel; 1–2 μm/pixel). To minimize acoustic noise from scanning, a sinusoidal waveform was used for the X Galvo-scanner (for acoustic spectra see Figure 2—figure supplement 1). Data were acquired in the middle 80% of the sinusoidal waveform to minimize nonlinearity. Multiple regions were imaged sequentially in multiple sessions in awake, head-fixed animals. For depth estimation, a Z-stack was acquired over a 512 × 512 µm area at 1 × 1 × 1 µm voxel size. Depth of each imaged area was estimated by measuring the Z-distance from the pia surface. Imaged neurons lay between 15–155 μm from the pia surface. The relative position of each region was tracked through the micromanipulator (MP-285, Sutter Instrument) that controlled the microscope objective. The positions of imaged areas were further aligned across animals to a common coordinate using the midline, lateral extreme and anterior-posterior extremes of the IC as anatomical landmarks in bright field images (SZ 61,Olympus) of the cranial window.

Frame timing of the scanner, timing of the sound stimuli, and animal movements were digitized using a Digidata 1440A (Molecular Devices, Sunnyvale, CA, USA) with Clampex v. 10.3 (Molecular Devices, Sunnyvale, CA, USA).

In vivo juxtacellular recordings were made under two-photon guidance (Kitamura et al., 2008). Glass pipettes were pulled from 1.5 mm, thick-walled borosilicate capillaries (Hilgenberg, Malsfeld, Germany) to 1–2 µm tip diameter (P-97; Sutter Instrument, Novato, CA) and filled with internal solution containing (in mM): potassium gluconate 138, KCl 8, Na₂-phosphocreatine 10, Mg-ATP 4, Na₂-GTP 0.3, EGTA 0.5, HEPES 10, (pH 7.2 with KOH; Merck). The internal solution also contained 40 µM Alexa Fluor 594 hydrazide. Ringer solution containing 1–2% agarose and (in mM): NaCl 135, KCl 5.4, MgCl₂ 1, CaCl₂ 1.8, HEPES 5, was applied on the brain surface to reduce movement artefacts. A positive pressure of about 300 mbar was maintained before penetration of the pia surface, and reduced to about 30 mbar upon pia entry; pressure was removed upon cell approach. Electrode resistances were constantly monitored and recording started when resistance increased to >25 MΩ. Juxtacellular potentials were acquired with a MultiClamp 700A amplifier (Molecular Devices, Sunnyvale, CA, USA) in current-clamp mode. Signals were low-pass filtered at 10 kHz (four-pole Bessel filter) and digitized at 25 kHz (Digidata 1322A). Data were recorded with pCLAMP 9.2 (Molecular Devices). In some cases a large current injection (1–6 nA) was used to elicit (positive current) or suppress (negative current) spikes of the cell being recorded (nanostimulation; for example Houweling et al., 2010), usually for cells that did not show obvious change in spike rate upon pure-tone stimulations. These data were also used for fitting the fluorescence ground-truth model.

Tone stimuli were generated in MATLAB v7.6.0 (The MathWorks, Natick, MA, USA) and played back via a TDT System3 setup (RX6 processor, PA5 attenuator, ED1 electrostatic speaker driver and two EC1 electrostatic speakers; Tucker Davis Technologies, Alachua, FL, USA). Sound stimuli were presented bilaterally in open field. Sound intensities were calibrated using a condenser microphone (ACO pacific Type 7017; ACO Pacific, Inc, Belmont, CA, USA) connected to a calibrated pre-amplifier and placed at the position of the pinnae.

For measurement of FRAs, 1 s tones (including 2.5 ms cosine-squared rise/decay times) with frequencies between 1 and 64 kHz with three steps per octave were presented at intensities between 30 and 80 dB sound pressure level (dB SPL) in steps of 10 dB. The set of stimuli was presented 6–10 times per experiment each in a pseudorandom order at an inter-stimulus interval of 1.5 s. For juxtacellular recordings, due to the limited recording time tones were only presented at one or two intensities (70 dB SPL; or 50 and 70 dB SPL).

Visual recordings were made using an RS Miniature CCD Camera (RS Components, Corby, UK) at 3 Hz. The camera was aimed at the animal’s head from a roughly right lateral perspective, in order to clearly record the animal’s right eye and whiskers. Facial movement was detected by calculating the root-mean-square intensity change between successive frames in a rectangular region behind the whisker pad (facial ROI in Figure 8B). General movement of the animal was registered using a piezo-electric motion sensor under the front paws of the mouse (pressure sensor in Figure 8B), digitized without further amplification using an AD channel of the Digidata 1440A.

Gelatin-embedded, 40 μm sections of PFA-fixed mouse brains were stained with the following antibodies: mouse anti-Gad67 (cat. no.: MAB5406; Millipore; 1:1000; RRID: AB_2278725); chicken anti-GFP (cat. no.: GFP-1020; Aves; 1:1000; RRID: AB_10000240); rabbit anti-NeuN (cat. no.: ABN78; Millipore; 1:1000; RRID: AB_10807945); mouse anti-parvalbumin (cat. no.: 235; Swant; 1:7000; RRID: AB_10000343); rabbit anti-calretinin (cat. no.: 7699/4; Swant; 1:5000; RRID: AB_2313763); and AlexaFluor-, Cy3- or Cy5-conjugated secondary antibodies (Invitrogen or Jackson ImmunoResearch). To ensure a more homogeneous GAD67 fluorescence for the post-hoc immunostaining of imaged brains, the brain slices were incubated twice in the primary antibody solution, each for 1 week at 4°C. The secondary antibody was also applied twice, but overnight at room temperature. Cell counting was performed manually in FIJI using the Cell Counter plug-in on confocal z-stacks. Confocal images were acquired using a Zeiss LSM700 microscope. Overview epifluorescence images were acquired using a Zeiss AxioImagerM2 equipped with a Zeiss Axiocam 503 mono camera.

Data analysis was performed with Igor Pro (WaveMetrics, Inc, Lake Oswego, OR, USA) using custom written procedures. Two-photon images and behavior video were aligned to ClampEx data using stimulus timing. Whisking behavior was assessed by calculating the root-mean-square of the frame to frame intensity difference in an area at the whisker pad.

Movement artefacts in two-photon images were corrected based on the built-in ImageRegistration operation in Igor Pro, which is based on a published algorithm (Thévenaz et al., 1998). Neuronal cell bodies were identified visually based on the average image of the motion-corrected image series, and a higher sampling Z-stack of the area (0.5 × 0.5 µm pixels; 1 µm z steps) taken directly after each experiment. Regions-of-interests (ROIs) were drawn around cell bodies. Average fluorescence values for individual ROIs were corrected for background fluorescence, which was defined as the change in average fluorescence in a 2 µm wide contour surrounding the ROI, excluding any pixel directly belonging to another ROI (Figure 2D).

Individual baseline fluorescence values of an ROI were measured for each trial by averaging the fluorescence in the 1 s preceding each stimulus onset. To mitigate carry over effects from earlier stimuli caused by the slow kinetics of GCaMP6s, the distribution of these baseline values was fitted with a Gaussian distribution. Any trial with a more extreme average baseline fluorescence than $\mu \pm 3\sigma$ was excluded from further analysis.

To analyze the FRA of each ROI, we isolated the stimulus-related response by extracting the signal autocorrelation (Geis et al., 2011) by calculating the average Pearson correlation coefficient among fluorescence waveforms to the same stimulus; the fluorescence trace within 1 s of the beginning or end of each stimulus was taken as the stimulus related waveform ($F_i\left(t\right)$ for stimulus i), that is between -1 s and +2 s of the stimulus onset. For a stimulus that was repeated n times, we calculated the signal autocorrelation (${\stackrel{-}{\rho }}_{auto}$) by averaging the Pearson’s correlation coefficient (${\rho }_{i,j}$) for the $\frac{n(n-1)}{2}$ possible pairs of responses:

where ${\rho }_{i,j}$ was calculated from fluorescence waveforms at trial i and j with the StatsCorrelation function in Igor Pro using the standard formula:

Statistical significance of signal correlation was done by a bootstrap method: for each ROI, a distribution of average Pearson correlation coefficient was constructed by drawing 30,000 random samples of n fluorescence trace segments within an experimental session, where n is again the number of stimulus repetitions. The p-value of each stimulus was calculated as the fraction of n-member samples having a greater (for frequency autocorrelation area; Geis et al., 2011) signal autocorrelation than that of the stimulus. The ranked p-values were then tested for significance with α = 0.05 and Holm-Bonferroni correction for the number of different stimuli presented (19 frequencies × 6 intensities=114 for FRA measurements). The characteristic frequency (CF) of an ROI was defined as the sound frequency at which the lowest intensity evoked a significant response. If multiple frequencies evoked significant responses at the lowest level, their geometric mean was taken as CF.

The FRA of a cell is generally continuous in frequency-intensity space. Making use of this, we refined the FRA by quantifying the similarity of responses to “adjacent” stimuli by analogously calculating the signal crosscorrelation (${\stackrel{-}{\rho }}_{cross}$).

where n and m are the number of repetitions of the two adjacent stimuli, leading to n×m possible pairs of responses and ${\rho }_{i,j}$ is again the Pearson correlation coefficient of the two responses. Note that, different from the cell to cell comparison described in Geis et al. (2011), none of the response pairs here were simultaneous. If the maximum ${\stackrel{-}{\rho }}_{cross}$ calculated against up to 8 (i.e. including diagonals) neighbors of a particular stimulus was >0.12, the stimulus was included for classification of its response type. For this purpose, the fluorescence response was calculated for four periods: onset (0-500 ms re onset), steady (500-1000 ms re onset), offset (0-500 ms re offset) and off-late (500-1000 ms re offset). We used the following scheme (in pseudocode) to classify individual responses within an FRA as excitatory, inhibitory or offset:

If at least 25% of all significant responses to frequency-intensity combinations belonged to the same class (e.g. inhibition), we named the FRA after that response (e.g. inhibition FRA). Typically an ROI showed a dominant response class in the majority of its significant responses: 20% of cells showed a dominant class in all significant responses; 54% of cells in ≥75% of significant responses; 92% of cells in ≥50% of significant responses. If more than one class reached the threshold of 25% of the significant responses, the FRA of an ROI was classified as mixed (e.g. Figure 3I,M).

Finally, to distinguish between onset and sustained responses, we averaged all excitatory responses of an ROI that were not immediately preceded or followed by any significant responses. The kinetics of sound-evoked responses were fitted by a single exponential function with the form $A(1-\mathrm{e}\mathrm{x}\mathrm{p}(-t/{\tau }_{\text{onset}}\left)\right)$. If this onset time constant, ${\tau }_{\text{onset}}$, was >1 s, the ROI response type was designated as sustained, else as onset.

To find the most prominent direction of CF gradient, we parametrized the location of each neuron as a projected distance $r$ from the anatomical origin along a line with an angle $\theta$ with the medial-lateral orientation:

where x and y are coordinates (in micrometers) of cells along the medial-lateral and anterior-posterior axes, respectively. We then model CF as a polynomial function of $r$:

This whole function was then fitted to the logarithm of the CF values of the recorded neurons using the Levenberg-Marquardt least-squares method, implemented in Igor Pro, that is:

A 4^th order polynomial was used in Figure 6 because increasing the number of exponents did not increase the explained variance further. The best fit parameters were ${\displaystyle \theta }$ = 0.875 rad (≈50°), ${\displaystyle a_0}$= 6.95, ${\displaystyle a_1}$= −0.0135 dec μm⁻¹, ${\displaystyle a_2}$= 1.95 × 10⁻⁵ dec μm⁻², ${\displaystyle a_3}$= −1.11 × 10⁻⁸ dec μm⁻³, ${\displaystyle a_4}$= 2.17 × 10⁻¹² dec μm⁻⁴. Cross-validation (70% fitting/30% test) indicated a slight overfitting. We also considered 2D polynomial fits in the form ${\displaystyle \sum a_{ij}{x}^i{y}^j}$, where ${\displaystyle 0\le i+j\le n}$, and found that a 3^rd order (${\displaystyle n=3}$) fit would be an optimal compromise between bias (underfitting) and variance (overfitting). However, we opted for the single dimension polynomial to avoid potential overinterpretation of the complex contour created by the 2D polynomial.

Traces from juxtacellular recording were first subjected to a digital DC remove filter which subtracts at each point in time the average potential within ±1 ms to remove DC drift or offset introduced by nanostimulation (Houweling et al., 2010). Traces were blanked around the start and end of each current injection (2 ms before and 3 ms after) to remove stimulus artefacts. Spikes were then detected by a simple thresholding procedure, with spike times defined as the peak time of the spikes, which presumably corresponds to the maximum rate of rise of the action potential.

For spike vs fluorescence model, spike times were converted to number of spikes over time (${\displaystyle n\left(t\right)}$) at the same sampling rate of the imaging (114.4 ms bins). It was then convolved with a simple ramp-decay kernel (${\displaystyle g\left(t\right)}$) according to the equations:

where F_1AP is the amplitude of the fluorescence change per action potential, t_rise is the rise time of the fluorescence and τ_decay is the decay time constant for the fluorescence. An offset (F₀) is added, which represents the minimal fluorescence at zero spike rate.

This 4-parameter model was fitted to the fluorescence trace using a genetic fitting routine in Igor Pro (Gencurvefit XOP, kindly provided by Andrew Nelson, Australian Nuclear Science and Technology Organization).

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

We thank Kees Donkersloot and Elize Haasdijk for their excellent technical support and Alba Membrilla Esteban for performing part of the immunohistochemistry analysis, Zhenyu Gao for sharing the design of the chronic cranial window, and Travis Babola and Dwight Bergles for helpful discussions on tonotopy. We thank Douglas Kim and the GENIE Project at the Janelia Research Campus for making the GP4.3 mouse line available.

This work was supported by a NeuroBasic PharmaPhenomics grant (Agentschap NL of the Ministry of Health, Welfare and Sports of the Netherlands) and a TOP subsidy (#91218033) from ZonMW to JGGB. ABW was supported by an Individual Fellowship (660157-OPTIMAPIC) through the Marie Skłodowska-Curie actions (European Commission).

Animal experimentation: All experiments in this study were performed in accordance with the ethical guidelines for laboratory animals within our institute and with European guidelines. The study was carried out under the project license (AVD2016789) approved by the Centrale Commissie Dierproeven (CCD) and the animal ethical committee (Instantie voor Dierenwelzijn; IvD) of the Erasmus MC. All recovery surgeries were performed under general isoflurane anesthesia supplemented with lidocaine, carprofen and buprenorphine as peri-operative analgesics. Terminal transcardiac perfusion was performed under pentobarbital anesthesia. Every effort was made to minimize suffering.

© 2019, Wong and Borst

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.