Inferotemporal face patches are histo-architectonically distinct

Abstract

An interconnected group of cortical regions distributed across the primate inferotemporal cortex forms a network critical for face perception. Understanding the microarchitecture of this face network can refine mechanistic accounts of how individual areas function and interact to support visual perception. To address this, we acquire a unique dataset in macaque monkeys combining fMRI to localize face patches in vivo and then ex vivo histology to resolve their histo-architecture across cortical depths in the same individuals. Our findings reveal that face patches differ based on cytochrome oxidase (CO) and, to a lesser extent, myelin staining, with the middle lateral (ML) face patch exhibiting pronounced CO staining. Histo-architectonic differences are less pronounced when using probabilistic definitions of face patches, underscoring the importance of precision mapping integrating in vivo and ex vivo measurements in the same individuals. This study indicates that the macaque face patch network is composed of architectonically distinct components.

Keywords: CP: Neuroscience; anatomical histo-architecture; fMRI; face patches; histology; inferior temporal cortex.

Figure 1.. Projection of fMRI-defined face patches onto histological sections

(A) Sequential sagittal histological sections from a right hemisphere in monkey M1. Sections were alternately stained for cytochrome oxidase (CO) and myelin. The histological sections were combined to construct a 3D histological volume (gray). The top representative sections correspond to the red sections in the 3D histological volume underneath (section 1, section 10, etc.). (B) To align and project fMRI-defined face patches onto histological sections, each monkey’s 3D anatomical MRI was coregistered to their 3D histological volume. Left: face patches on M1’s right hemisphere cortical surface. Face patches ML, AL, and AF/AD are shown in red, green, and blue, respectively. Center and right: face patches projected onto a single slice from M1’s anatomical MRI volume (center) and the corresponding histological section stained for CO (right). Videos scrolling in sagittal and coronal directions through the 3D histological volume and their aligned MR volume are provided (Videos S1 and S2).

Figure 2.. Architectonic profiles of fMRI-defined V1 and face patch ML from histological data in the same individual

(A) Primary visual cortex (V1) and ML localized in the same histological section as shown in Figure 1B, with magnifications of V1 (pink boxes) and ML (orange boxes). Dashed colored lines illustrate the definitions of V1 and ML from fMRI data. (B) Cortical depth maps of V1 (pink) and surrounding cortex. (C) CO depth profiles across V1 sections. The pale pink curves illustrate CO profiles of each section, and the bolded pink curve illustrates the mean staining across sections. The black curve shows the depth profile for the section shown in (A). Higher numbers indicate darker staining intensity. The peak of the CO staining at mid-cortical depths (black arrow) corresponds to the visible thick band in V1 (A, black arrow). (D) Cortical depth map of ML (orange) and surrounding cortex. Conventions are consistent with (B). (E) CO staining across ML sections (n = 11). The pale orange curves illustrate the CO intensity of each section, and the bolded orange curve illustrates the mean intensity across sections. The black curve shows the depth profile for the section shown in (A). The dark staining toward mid-depths (red arrow) corresponds to the visible band of dark staining in ML (A, red arrow).

Figure 3.. ML is architectonically distinct from other IT face patches

CO stain intensities are plotted from the cortical surface, gray matter boundary (top) to the gray/white matter boundary (bottom). The curved lines represent the CO depth profiles of the three face patches measured in the present study (ML, red; AL, green; AF/AD, blue) in each hemisphere of M1 and M2. The shaded area represents ±1 SEM across histological sections. Higher numbers indicate darker staining intensity. The horizontal bar graphs to the right of each depth profile show the mean stain intensities across outer layers (top seven depth bins) and inner layers (bottom eight depth bins). The mean stain intensities across all layers are shown below each depth profile (error bars, ±1 SEM). Asterisks indicate significant differences between face patches (p < 0.05, false discovery rate [FDR] corrected). The statistics of all comparisons are shown in Table S1.

Figure 4.. Architectonic differences between ML and other face patches are strongest for CO in outer layers in individual participants

(A) Image intensity differences between face patches and neighboring STS regions outside of face patches (non-face patch). Data points (N = 4 hemispheres) shown as filled and empty circles and squares, with bars representing the mean across hemispheres. Intensity differences are significant for ML vs. AL and ML vs. AF/AD in all hemispheres (ts > 2.692, ps < 0.001 for CO and myelin, using unpaired, two-sample t tests [FDR corrected for multiple comparisons]). See Figure S2 for individual hemispheric data. (B) Mean effect sizes (Cohen’s d) for CO and myelin across all hemispheres/monkeys (error bars, ±1 SEM) for ML vs. all other face patches (magenta), ML vs. AL (n = 4, dark gray), and ML vs. AF/AD (n = 4, light gray). Cohen’s d for CO was significantly higher compared to myelin in 7 of 8 pairs for each comparison (Figure S5A). (C) Mean Cohen’s d for outer and inner depths across all hemispheres/monkeys (error bars, ±1 SEM) for CO (left) and myelin (right) for ML vs. all other face patches (cyan). Cohen’s d was significantly higher in the outer layers compared to the inner layers for CO in 7 of 8 pairs and 5 of 8 pairs for myelin (Figure S5B). The depth bins corresponding to the outer and inner depths are shown in Figure 3. The other conventions are consistent with (B). (D) Cohen’s d was significantly higher for individually defined compared to probabilistically defined face patches in CO and myelin (both significant for 7 of 8 pairs; Figure S5C). Each line represents Cohen’s d collapsed across hemispheres and monkeys for individually and probabilistically defined face patches, shown in gold. The other conventions are consistent with (C). All statistics for each pairwise comparison for (B)–(D) are shown in Tables S2, S3, and S4.