Sulcal Depth in the Medial Ventral Temporal Cortex Predicts the Location of a Place-Selective Region in Macaques, Children, and Adults

Data Acquisition in Humans (Anatomical Scans)

For human subjects, participants were scanned using spin-echo inversion recovery with an echo-planar imaging (EPI), readout (SEIR-EPI), with a 3 T GE Signa scanner and a custom-built, phase array 32-channel, receive-only, head coil using methods as in prior studies (Mezer et al. 2013; Natu et al. 2016, 2019; Gomez et al. 2017, 2018; Nordt et al. 2018). This scan was done with a slab inversion pulse and spatial–spectral fat suppression. We used a 2 mm² in-plane resolution with a slice thickness of 4 mm, and the EPI readout was performed using 2× acceleration. QMRI parameters were measured from spoiled gradient echo images acquired with different flip angles (α = 4°, 10°, 20°, and 30°, TR = 14 ms, TE = 2.4 ms) and a voxel resolution of 0.8 × 0.8 × 1 mm, which was resampled to 1 mm³ isotropic. For SEIR-EPI, the TR was 3 s. The echo time was set to minimum full; inversion times were 50, 400, 1200, and 2400 ms. Anatomical data were aligned to the AC–PC plane. From the qMRI data, we generated the whole-brain T₁-weighted anatomy. The spoiled GE and the SEIR scans were processed using the mrQ software package in MATLAB to produce the T₁ maps. The mrQ analysis pipeline corrects for RF coil bias using SEIR-EPI scans, producing accurate T₁ fits across the brain. The full analysis pipeline can be found at https://github.com/mezera/mrQ (Mezer et al. 2013).

Data Acquisition in Macaques (Anatomical Scans)

For macaque anatomical scans, whole-brain structural volumes were acquired in a 3 T Skyra scanner while the animals were anesthetized with a combination of ketamine (4 mg/kg) and dexdomitor (0.02 mg/kg) and using a 15-channel transmit/receive knee coil. Monkeys were scanned using a magnetization-prepared rapid gradient echo (MPRAGE) sequence; 0.5 × 0.5 × 0.5 resolution; FOV = 128 mm; 256 × 256 matrix; TR = 2700 ms; TE = 3.35 ms; TI = 859 ms; flip angle = 9°). Three whole-brain T1-weighted anatomical images were collected in each animal.

Generation of Sulcal Depth Maps

For each participant (human and macaque), FreeSurfer’s automated algorithm was used to obtain sulcal depth maps. At each vertex along the surfaces, sulcal depth (in mm) is measured as the distance between the inflated and white surfaces obtained from FreeSurfer (Fischl, Sereno, Dale, et al. 1999a, Fischl, Sereno, Tootell, et al. 1999b) using the “mris_inflate” function. It is evaluated using a signed dot product of the displacement/movement vector with the outward pointing surface normal of the white matter surface during inflation. Thus, for a sulcus, the movement vector points outward (toward the pial surface), and therefore, the product is positive. Comparatively, for a gyral crown, the movement vector points inward and, therefore, is negative. Each participant’s sulcal depth maps (.sulc files) from the FreeSurfer’s auto-segmentation algorithm were used to obtain depth measures from anatomical parcellations of the collateral sulcus (CoS) in humans and the occipitotemporal sulcus (OTS) in macaques. We measured the depth of the sulcal pit in the medial VTC that overlapped with a place-selective region in macaques, as well as human children and adults. We then evaluated the developmental and species differences in the depth of the place-selective sulcal pit using 1) a two-way ANOVA with species (humans/macaques) and hemisphere (left/right) as factors and 2) a two-way ANOVA with age of subject (in humans: child/adults) and hemisphere (left/right) as factors.

Generation of Surface Area and Curvature Values

FreeSurfer’s segmentation algorithm was also used to generate the overall curvature of the CoS and its surface area at both pial and white matter surfaces. The overall curvature quantifies the average curvature magnitude of the entire sulcal or gyral fold.

Probabilistic Group Maps of Place Selectivity and Sulcal Depth

To visualize the probabilistic location of place-selective functional ROIs (fROIs) on the cortical sheet, we generated probabilistic group maps of CoS-places in children (N = 26) and adults (N = 28) separately using cortex-based alignment (CBA) to the FreeSurfer average cortical surface (Fischl, Sereno, Dale, et al. 1999a, Fischl, Sereno, Tootell, et al. 1999b) (Fig. 1). We then calculated the percentage of subjects in which CoS-places overlapped at each vertex on the surface (Fig. 1b,c). To visualize how the sulcal depth of the CoS varied on the cortical sheet within participants of an age group, we generated probabilistic group maps of the sulcal depth of the CoS also using CBA separately for children and adults. Here, we 1) transformed each participant’s sulcal depth maps into FreeSurfer’s average cortical space, 2) normalized the transformed maps based on the deepest points of the CoS, 3) added the normalized values at each point of the surface, and 4) divided them by the total number of participants in each group. Then, we obtained the percentage of subjects included in each vertex on the surface. In macaques (N = 9), the probability maps of place-selective fROIs (Fig. 1a) and sulcal depth maps of the OTS were generated using a macaque average cortical surface template (NMT template from 31 independent macaque brains; Seidlitz et al. 2018) and a procedure identical to that used for humans.

Open in a separate window

Figure 1

Probabilistic maps of a functionally defined place-selective region in macaques and humans are located within a sulcus in the medial ventral temporal cortex (VTC) across species. (a) Inflated cortical surface reconstruction of the right hemisphere of a macaque (NMT; Seidlitz et al. 2018). The map indicates the most probable location of the lateral place patch (LPP) in nine macaques. The most probable location (yellow) is within the occipitotemporal sulcus (OTS; blue). (b) Inflated cortical surface reconstruction of the right hemisphere of a human anatomical template (the FreeSurfer average brain from 39 independent adult brains (Fischl, Sereno, Dale, et al. 1999a, Fischl, Sereno, Tootell, et al. 1999b) showing the probabilistic location of CoS-places (also known as the PPA; Epstein and Kanwisher 1998) in 26 children. The most probable location (yellow) of CoS-places is within the collateral sulcus (CoS; green). (c) Same as (b), but for 28 human adults. Like children, the most probable location (yellow) of CoS-places is within the CoS (green). Warmer colors in the probabilistic maps represent greater overlap of place selectivity across participants within a group. Abbreviations: RH, right hemisphere; CoS, collateral sulcus; OTS, occipital temporal sulcus. Brain sizes not to scale.

For visualization, we also overlapped each place-selective region with each individual subject’s normalized depth maps (normalized to the deepest point in the sulcal fold) thresholded at 80%. We implemented this approach as it is common practice to localize a basin, or cluster, of the deepest points together as the sulcal pit (as opposed to a single point) with MR/cortical surface reconstructions as in previous studies (Im and Grant 2019, Meng et al. 2014, Auzias et al. 2015; Leroy et al. 2015; Bodin et al. 2018).

The Deepest Sulcal Points (Sulcal Pit) in the Medial Ventral Temporal Cortex Are in Similar Positions across Individuals

Given the consistency in the location of the place-selective region within a sulcal fold in the medial VTC, we next sought to test if the deepest points of the sulcal fold (known as the sulcal pit) are also anatomically consistent. We aligned each individual participant’s sulcal depth map (Fig. 2a, for sample macaque and human sulcal depth maps) to the average surface template of their respective species and generated a probabilistic map of sulcal depth across participants. This approach revealed that within each group, the locations of the deep sulcal points were well aligned across participants (Fig. 2b, warmer colors indicate greater overlap of deep sulcal points across each group). Projecting the location of the sulcal pit (e.g., coordinates of the vertex with the deepest depth) from individual subjects to the average cortical surface revealed that the locations of the deep sulcal pits are largely clustered in close proximity across individual subjects (Fig. 2b, see inset and Supplementary Fig. 1 for both hemispheres). Individual sulcal pits were largely constrained (macaque: N = 8/9 (RH), N = 9/9 (LH), children: N = 22/26 (RH), N = 22/26 (LH); adults: N = 25/28 (RH), N = 27/28 (LH)) within the probabilistic place-selective region in both species (dotted white lines in Fig. 2b, inset), which we further quantify in subsequent sections.

Open in a separate window

Figure 2

A place-selective region in the medial VTC overlaps the deepest sulcal points in macaques, children, and adults. (a) Coronal slices showing depth maps obtained from FreeSurfer’s auto-segmentation algorithm in sample macaque and human adult brains (warmer colors represent deeper). For simplicity, the map has been thresholded to only show positive depth values representing sulcal folds, walls, and pits. (b) Inflated cortical surfaces of a right hemisphere showing probabilistic maps (percent overlap) of sulcal depth across participants within each group (macaque, children, and adults) at each point of the occipitotemporal sulcus (OTS) in macaques and the collateral sulcus (CoS) in humans (magenta dotted outlines). The locations of the sulcal pit from each individual subject (colored circles) is shown in the inset on the right of each group surface. In each group, deep sulcal points (warmer colors) within the middle of the sulcus overlap the most across individuals. Additionally, probabilistic locations of the place-selective region (dotted white lines obtained from maps in Fig. 1) overlapped with the probabilistic sulcal depth maps (warm colors) in macaques, children, and adults. (c) Four sample macaque (top row), child (middle row), and adult (bottom row) ventral temporal surfaces, showing the overlap between the functionally defined place-selective region (white dotted lines) and sulcal depth maps (warm colors). To visualize the deep points of the sulcus, the depth maps in each individual subject are thresholded at 80% depth (100% is the deepest point within the sulcus in each individual). Warmer colors indicate deeper sulcal points. All individual surfaces are shown in Supplementary Figures 2–7. Brain sizes are not to scale.

There is a Consistent Anatomical–Functional Relationship between Sulcal Depth and the Location of the Place-Selective Region in Humans and Macaques

In both species, the sulcus within which the place-selective region is located extends nearly from the occipital to the temporal pole. Given the functional consistency of the place-selective regions across species (Fig. 1), as well as the anatomical consistency of the deepest sulcal points across individuals within each species (Fig. 2b), we examined the overlap between each place-selective region and sulcal depth map across groups, as well as in each individual. In all groups (macaques, children, and adults), the place-selective region overlaps the deepest points of the sulcus in the medial VTC (Fig. 2b, white dotted line). Specifically, 94.4% of the OTS pits in macaques, 84.6% of the CoS pits in human children, and 92.9% of the CoS pits in human adults are located within the probabilistic place-selective region across both hemispheres (Fig. 2b, see inset and Supplementary Fig. 1). Importantly, this structural–functional relationship was also present at the individual subject level (Fig. 2c for sample individual subjects; Supplementary Figs 2–4 (for the right hemisphere) and Supplementary Figs 5–7 (for the left hemisphere) for all subjects). Across both left and right hemispheres, in 88.8% of monkeys and in 76.9% of humans (children and adults), the sulcal pit is located within their individual place-selective region. Interestingly, however, the structural–functional overlap was more consistently observed in human adults (85.7%) than in children (68.0%). Thus, the sulcal pit is an anatomical marker that is a good predictor for the location of place selectivity along the anterior–posterior axis of the OTS in macaques and CoS in humans even as the degree of coupling increases from childhood to adulthood.

Sulcal–Functional Relationship Strengthens from Childhood to Adulthood with Additional Evidence that it is Preserved between Macaques and Humans

To further explore the relationship between sulcal depth and place selectivity across development and species, we generated depth and place selectivity profiles along a posterior–anterior axis of each respective sulcus (CoS for humans and OTS for macaques; Materials and Methods). Qualitative examination of the average profiles revealed three main findings. First, we observed that in both macaques and humans, sulcal depth (Fig. 3a,c) and place selectivity (Fig. 3b,d) have relatively smooth group average profiles, with a single peak (represented by ^*). Second, average selectivity and average depth profiles are well aligned in macaques, children, and adults. Specifically, mean place selectivity increases as the sulcus deepens (Fig. 3 and Supplementary Fig. 8 for the right and left hemispheres, respectively). Further, in both adults and macaques, the peak of selectivity (represented by ^* in Fig. 3b,d and Supplementary Fig. 8b,d) is within the deepest region in the sulcus (represented by the gray bar in Fig. 3a–d and Supplementary Fig. 8a,c). Third, there are clear developmental differences across human children and adults: In children, but not adults, the mean peak place selectivity was outside but immediately adjacent to the sulcal pit (Fig. 3d, peak of the blue line vs. gray bar).

Open in a separate window

Figure 3

Sulcal depth and place selectivity in the medial VTC are correlated in both humans and macaques. (a, c) Mean (dark line) and standard deviation (lighter shading) of sulcal depth along the anterior–posterior length of the right OTS in nine macaques (a) and the right CoS in humans (c). In (c), data are averaged across 26 children (blue) and 28 adults (orange). Star ^(*)represents deepest sulcal point. (b, d) Mean place selectivity (t-value) in macaques (b, green) and humans (d, blue/orange). Across macaques and humans, the profiles are well-matched along the anterior–posterior axis in which there is a functional peak (represented by ^* in b, d) around the deepest sulcal points (represented by gray bar). Left hemisphere data in Supplementary Fig. 8. (e) Normalized depth (black lines) and selectivity (colored dotted lines) values as a function of distance along the posterior (P) to anterior (A) axis in four sample macaques (top row), human children (middle row), and human adults (bottom row). Correlation (R) and significance values below each subplot relate the depth and selectivity curves to one another. All individual profiles and their respective correlations are in Supplementary Figures 9–11.

To directly quantify this structural–functional correspondence, we calculated the correlation between each individual’s depth and selectivity profiles. This approach revealed that across hemispheres, functional–structural profiles were significantly (P < 0.05) positively correlated in 88.9% of the macaques (N_left = 8/9, N_right = 8/9; 83.3% when corrected for multiple comparisons). In humans, the functional–structural profiles were significantly and positively correlated in 85.7% of the adults (N_left = 24/28; N_right = 24/28; 82.1% when corrected for multiple comparisons) and 67.3% of the children (N_left = 16/26; N_right = 19/26; 55.7% children when corrected for multiple comparisons; see Supplementary Figs 9–11 for individual macaque and human functional selectivity and depth profiles, as well as their correlations).

To test if these developmental and species differences are statistically significant, we Fisher z-transformed all correlation coefficients and then compared these correlations using an ANOVA. When collapsing across age groups, the correspondence between sulcal depth and place selectivity was not significantly different between species (macaques (M_{zscore ± sem}): 0.47 ± 0.09; humans: 0.55 ± 0.04, F_1,122 = 0.56, P > 0.05. Additionally, within humans, the correspondence between sulcal depth and place selectivity significantly differed between children and adults (children: 0.46 ± 0.06 adults: 0.63 ± 0.05; significant main effect of age: F_1,104 = 5.12, P < 0.05). Adults showed stronger coupling between sulcal depth and place selectivity, as well as less variability in the structural–functional coupling compared to children. We find no hemispheric effect across development or species (Fs < 1.17, Ps > 0.05). In macaques, while they did vary in age, we were unable to test for developmental effects due to the relatively small number of subjects (N = 9).

To determine if the structural–functional coupling in and around the sulcal pit is driving the correlation, we used a sliding window approach to calculate the correlation of selectivity and depth into three separate bins: 1) posterior to the sulcal pit, 2) within and surrounding the sulcal pit, and 3) anterior to the sulcal pit (no sulcal point overlapped across the three bins). The correlation between sulcal depth and place selectivity was strongest within and surrounding the sulcal pit and then decreased as the distance from the sulcal pit increased on either side of the pit for both hemispheres and species (Supplementary Fig. 12 for left and right hemispheres; macaque: R_{mean ± std}: M_close = 0.62 ± 0.3, M_away = 0.24 ± 0.5; children: C_close = 0.54 ± 0.49, C_away = 0.01 ± 0.49; adults: A_close = 0.58 ± 0.45, A_away = 0.15 ± 0.48). Importantly, we evaluated if these differential effects at deep sulcal depths versus their surrounding areas were driven by a potentially heightened functional response related to greater signal to noise ratio (SNR)—due to presumably less partial voluming effects—at the sulcal fundus (Koo et al. 2009). Measuring the time series signal-to-noise ratio (tSNR = mean(time-series)/std(time-series)) across species showed that this was not the case. Regions around the deep sulcal depths do not show the highest tSNR in our data (Supplementary Fig. 13). Furthermore, in humans, although we find stronger sulcal–functional coupling in adults compared to children (which would predict higher tSNR in adults compared to children), we find the opposite: There is larger tSNR in children than adults along the entire length of the CoS (Supplementary Fig. 13d).

Finally, to examine a stringent sulcal–functional mapping not commonly implemented in previous studies (Im et al. 2010; Meng et al. 2014; Leroy et al. 2015; Auzias et al. 2015; Bodin et al. 2018), we also tested if the single deepest point predicted the single point with the highest place selectivity. To do so, we measured the distance between the single deepest sulcal point and the single voxel with the highest place selectivity using the FreeSurfer function “mris_pmake,” which calculates the distance between two vertices on a cortical surface map. The approximate peak-to-peak distance is ~0.5 cm in macaques (M = 4.97 mm, Std: 2.61 mm), ~1.5 cm in human children (M: 15.39 mm, Std: 10.13 mm), and ~1 cm in human adults (M: 10.5 mm, Std: 7.04 mm). An ANOVA with species (humans/monkeys) and hemisphere (left/right) as factors revealed a main effect of species (F_1,122 = 16.2, P < 0.05). Additionally, an ANOVA with age group (child/adult) and hemisphere as factors also showed a main effect of age (F_1,104 = 5.5, P < 0.05), in which there was a tighter coupling between the deepest point and most selective voxel in adults compared to children. Altogether, these results reveal that the anatomical landmark of the sulcal pit is functionally relevant within the medial VTC across species. Furthermore, within humans, the structural–functional coupling between sulcal depth and place selectivity becomes stronger and more stable from childhood to adulthood.

Sulcal Features as Functional Landmarks in the High-Level Visual Cortex in Gyrencephalic Brains: Methodological, Behavioral, and Theoretical Implications

Our results showing that there is a tight correspondence between the location of sulcal pits in the medial VTC and place selectivity in both macaques and humans, as well as in children and adults, are surprising because these place-selective regions are located in high-level visual cortex. Previous theories of cortical folding (Richman et al. 1975; Welker 1990; Armstrong et al. 1995; Van Essen 1997; Toro and Burnod 2005; Zilles et al. 2013) would predict a tight sulcal–functional relationship between a primary sulcus in the visual cortex (i.e., the calcarine sulcus) and a primary visual area (V1). It is unclear that such sulcal–functional relationships would exist in high-level visual regions where sulcal folds are more complex and vary across individuals and primate species. Thus, these theories would not predict that 1) deep sulcal points located in what is considered one of the last stages of the visual processing hierarchy would predict the strongest selectivity of a high-level region selective for visual scenes, 2) this sulcal–functional coupling would occur in both macaques and humans, or c) this sulcal–functional coupling would strengthen from childhood to adulthood—each of which we show in the present study. This latter point is particularly striking as one might expect structure–function relationships to be established early in development (potentially by genetic markers) and, thus, largely fixed.

The combination of these results further illustrates the utility of sulcal features as functional landmarks in high-level visual cortex in gyrencephalic brains. For example, previous research in humans showed that 1) the anterior tip of the mid-fusiform sulcus accurately identifies a region selective for faces (Weiner and Grill-Spector 2010: Weiner et al. 2014), as well as 2) a sulcal intersection between two sulci in the medial VTC accurately identifies the location of a region selective for visual places and scenes (Weiner et al. 2018). One might question the role of relatively small sample sizes when identifying these structural–functional relationships. In our previous work, we showed that a predicted location of CoS-places predicted the cortical location of voxels exhibiting the highest functional place selectivity in over 500 participants from the Human Connectome Project (HCP) (Weiner et al. 2018). Furthermore, a recent study by Abbasi et al. (2020) examined fMRI data of 424 adults (212 twin pairs) also from the HCP database to determine the relationship between genetics and the organization of category-selective visual cortex. They found that genetic voxels were preferentially concentrated in regions with higher cortical curvature: sulcal fundi for place-selective regions (consistent with the present results reported here) and gyral crowns for face-selective regions in the ventral temporal cortex. Given these consistent structural–functional links observed in the relatively diverse and large-scale HCP dataset, we anticipate that the strong sulcal–functional relationship we report here will generalize to a randomly selected, large-scale, and diverse group of participants. Furthermore, our work extends the structural–functional relationship identified previously (Weiner et al. 2018) to a specific morphological feature of sulcal depth, which is consistent with recent work showing that a sulcal pit within the human STS predicts the location of a functional region processing human voices (Bodin et al. 2018). Together, these findings suggest the possibility that a subset of functional regions in gyrencephalic brains are likely identifiable simply from anatomical features such as sulcal depth maps instead of functional localizer experiments, which can be examined in future work.

Additional future work will likely build on this correspondence between sulcal features and functional selectivity to also incorporate aspects of behavior and cognition. For instance, sulcal interruptions in human lateral VTC have been shown to predict the location of regions selective for words as well as correlate with reading ability (Cachia et al. 2018). Furthermore, the depth of sulcal pits is correlated with different aspects of cognition such as verbal IQ (Im et al. 2011), and differences in the cortical depths and morphology of sulci are also well-known markers for developmental neurocognitive disorders such as autism (Auzias et al. 2014; Brun et al. 2016). Finally, significant hemispheric asymmetries in the frequency and distribution of sulcal pits in the superior temporal sulcus (STS) are considered to be human-specific cortical landmarks of language and socio-communication abilities (Im et al. 2010; Auzias et al. 2015; Leroy et al. 2015).

Given this growing body of work showing a relationship between sulcal depth and functional cortical representations, as well as sulcal depth and human cognition, we believe that our results are theoretically meaningful rather than epiphenomenal. Nevertheless, the commonalities between the functional arealization of the visual cortex more generally across species despite vast differences in brain size, cortical folding, and gyrification (Van Essen 1997; Rosa and Tweedale 2005; Hill et al. 2010; Zilles et al. 2013; Arcaro and Kastner 2015) cannot be ignored as they implicate that sulci may not be necessary for the development of category-selective regions. Thus, while it is beyond the scope of the present study, future theoretical and modeling work of cortical folding relative to functional areas across species should also consider the functional role of sulcal pits in the emergence, development, and evolution of functionally specialized areas specifically in gyrencephalic brains.