Fig 1.
Manual annotation of calbindin-positive (CB+) neurons across the cerebral cortex.
(A) Representative types of CB+ neurons (from top left to bottom right): bipolar, multipolar, small pyramidal, and neurogliaform neurons, imaged at ×20 magnification (0.4974 μm per pixel). The neurogliaform neurons were identified according to criteria described in studies of the macaque cortex [22,24], and their neuronal identity has been confirmed by immunofluorescence co-staining with NeuN and GABA in other sections (S1 Fig). (B) Locations of 163 image strips sampled in case CJ1741, visualized against a semi-transparent rostrolateral view of the 3D reconstructed left hemisphere (opaque model presented as a thumbnail for clarity). Different colors of the image strips correspond to the areas they were derived from (see e.g. S1 File), and the black contour indicates the coronal level of the section presented in panel C. (C) Coronal cross-section taken approximately at the interaural +13.0 mm [44] containing five image strips (black rectangles) from areas A24c, A6M, A6DR, A8C, ProM, clockwise. (D, E) Strips from areas 6M and 8C shown at high magnification. Counting boxes are represented with red (gray matter) or green (white matter) squares of 150 μm in size. The thin, dashed band in the middle of the strips indicates the boundaries of layer 4, while the bottom dashed line shows the border between layer 6 and the white matter. The laminar boundaries were determined by comparison with adjacent Nissl-stained sections.
Fig 2.
Training and evaluation of the U-Net Convolutional Neural Network (U-Net CNN).
(A) Examples of counting boxes manually annotated by neuroanatomists (red circles of various sizes). The entire counting box is 150 × 150 μm while the area annotated with the red square is a 256 × 256 px (127 μm × 127 μm) image patch used for training the CNN. A total of 4,072 counting boxes were defined, including boxes representing various densities of neurons (top left and top middle), parts of the tissue containing blood vessels or artifacts (top right and bottom left), as well as samples of white matter (bottom center) or samples that do not depict the brain tissue (bottom right). (B) Density map generated based on the counting box indicated by the gray arrow. A single Gaussian blob corresponds to a single CB+ neuron, and 43.89 neurons are located within the indicated image patch (red square). Note that non-integer cell number estimates are possible, as explained in Automated detection of CB+ neurons Methods section. Upon training the U-Net CNN counting boxes not previously presented to the network can be turned into a density maps, and the total number of neurons within an image patch can be computed. (C) Locations of nine areas selected for a comparison between the CB+ densities estimated by the U-Net CNN and multiple human annotators (V1, V2, MT, AuA1, A3b, A3a, A4ab, A8C, and A13M, see S1 File for a list of areas, color coding and abbreviations). (D) Per-area (i.e. average values for all boxes sampled from a given area) densities for the three analyzed hemispheres. Different symbols show the results for individual human experts (dM1 to dM3), an average of the three expert observers (), and the densities obtained with the U-Net CNN (dCNN). The mean of the differences between dCNN and the dM densities is statistically indifferent from zero (see S3F and S3G Fig for statistical details), indicating that the automatic and the average manual counts are indistinguishable. (E) The results obtained by the U-Net CNN reflect the consensus between the expert annotators. (rows, top to bottom) Examples of individual CB+ neurons marked by all, two, and only one expert, respectively, and a proportional density estimate by the U-Net CNN. The proposed method helps alleviate the interindividual variability of manual cell counting. Box size: 50 μm. (F) We applied the procedures for estimating the density of CB+ neurons to all CB-stained sections in all three hemispheres studied. Here, results for an example section (CJ1741-r16c) are presented. From left to right: section image, segmentation of the cortex into individual areas based on manual parcellation and coregistration to the reference template [35] (see Materials and Methods for details), and density map of CB+ neurons (see G for scale). Note that the quantification of the results is performed only in the cortical areas, while the subcortical regions are not considered. (G) Example three-dimensional reconstruction of a CB+ density map constituting the basis of the flat maps illustrated in other figures. The black line indicates the location of the section presented in panel F, and the black patches correspond to the parts of cortex excluded from the analyses due to staining artifacts or corrupted tissue. The datasets are available for download from https://www.marmosetbrain.org/whole_brain_cb_maps.
Fig 3.
Variation in the density of calbindin-positive (CB+) neurons across the marmoset cortical areas.
(A) Density (neurons·mm-3) of CB+ neurons across areas, grouped according to the classification proposed by [45]. For each box plot, the center line indicates the median, the box limits the upper and lower quartiles, the whiskers 1.5× the interquartile range, and the annotated points outliers). Abbreviations: DLP: dorsolateral prefrontal cortex; VLP: ventrolateral prefrontal cortex; OFC: orbitofrontal cortex; MPC: medial prefrontal cortex; MPM: motor and premotor cortex; INS: insular cortex; SSC: somatosensory cortex; AUD: auditory cortex; LIT: lateral and inferior temporal cortex; VTC: ventral temporal cortex (encompassing parahippocampal, perirhinal and entorhinal areas); PCR: posterior cingulate and retrosplenial cortex; PPC: posterior parietal cortex; VIS: visual cortex. (B) Top left: the relation between the absolute density of CB+ neurons and total neuronal density (NeuN staining) in different areas. Top right and bottom: relations between relative and absolute densities of CB+ neurons in 3 animals. (C) Percentages of CB+ neurons across groups of areas, classified as in panel A. (D) Flat map representation of the CB+ neuronal density (mean of all 3 animals) in different areas. Some of the cortical areas are identified for orientation. (E, F) The relation between the absolute and relative densities of CB+ neurons and hierarchical level derived from laminar patterns of connections between cortical areas [36]. (G) Differences in relative density of CB+ neurons between areas according to the degree of lamination (adapted from [43]). Abbreviations: Agr: agranular areas; Dys: dysgranular areas; Eu1, Eu2, Eu3: eulaminate areas with increasing levels of laminar differentiation; Kon: koniocortical areas. Colors in B, E, and F correspond to those in A and C; see S1 File for a full list of areas, color coding and abbreviations.
Fig 4.
Densities of CB+ neurons in cortical areas of three marmosets.
For this figure, areas were grouped according to the classification proposed by [45] (modified from [44]). Abbreviations (groups of areas): DLP: dorsolateral prefrontal cortex; VLP: ventrolateral prefrontal cortex; OFC: orbitofrontal cortex; MPC: medial prefrontal cortex; MPM: motor and premotor cortex; INS: insular cortex; SSC: somatosensory cortex; AUD: auditory cortex; LIT: lateral and inferior temporal cortex; VTC: ventral temporal cortex (encompassing parahippocampal, perirhinal and entorhinal areas); PCR: posterior cingulate and retrosplenial cortex; PPC: posterior parietal cortex; VIS: visual cortex. See S1 File for a full list of areas, color coding and abbreviations.
Fig 5.
Maps of the distribution of CB+ neurons in three animals, shown in unfolded representations of a left hemisphere.
For each map, the upper scale indicates densities in neurons·mm-3, and the lower scale shows the same data normalized to the peak density of a given individual.
Fig 6.
Laminar trends in the distribution of calbindin-positive (CB+) neurons in areas of the marmoset cortex.
(A) Ratios of the density of CB+ neurons in supragranular versus infragranular layers (d2,3/ d5,6). (B) Ratios of the density of CB+ neurons in supragranular layers versus layer 4 (d2,3/ d4). (C, D) Flat map representations of the distributions of d2,3/ d5,6 and d2,3/ d4, summarizing the data shown in (A) and (B). Areas without a clear homolog of layer 4 were excluded from analysis, indicated in gray.
Fig 7.
Relations between the laminar distribution of CB+ neurons and large-scale trends in cortical organization.
(A) The d2,3/ d5,6 ratios were strongly correlated with the area’s hierarchical level, derived from the laminar pattern of connections. This demonstrates that the bias towards higher density of CB+ neurons in supragranular layers was most marked in low-level (e.g. purely sensory) areas, and more subtle in association and premotor areas. (B) In parallel, the d2,3/ d5,6 ratios varied systematically with respect to the classification of an area according to the degree of laminar differentiation. Analyses of the laminar distribution of CB+ neurons. (C) In contrast with (A), there was no significant correlation between d2,3/ dL4 and hierarchical level. (D) Analysis of d2,3/ d4 relative to the degree of cortical lamination indicated significantly lower d2,3/ d4 ratios (i.e. higher relative concentrations in layer 4) in koniocortex and eulaminate 3 areas, compared to others.