Interspecies activation correlations reveal functional correspondences between marmoset and human brain areas

Abstract

The common marmoset has enormous promise as a nonhuman primate model of human brain functions. While resting-state functional MRI (fMRI) has provided evidence for a similar organization of marmoset and human cortices, the technique cannot be used to map the functional correspondences of brain regions between species. This limitation can be overcome by movie-driven fMRI (md-fMRI), which has become a popular tool for noninvasively mapping the neural patterns generated by rich and naturalistic stimulation. Here, we used md-fMRI in marmosets and humans to identify whole-brain functional correspondences between the two primate species. In particular, we describe functional correlates for the well-known human face, body, and scene patches in marmosets. We find that these networks have a similar organization in both species, suggesting a largely conserved organization of higher-order visual areas between New World marmoset monkeys and humans. However, while face patches in humans and marmosets were activated by marmoset faces, only human face patches responded to the faces of other animals. Together, the results demonstrate that higher-order visual processing might be a conserved feature between humans and New World marmoset monkeys but that small, potentially important functional differences exist.

Keywords: cortex; functional MRI; marmoset; naturalistic movie.

Fig. 1.

Intersubject correlation map of brain activity during movie viewing. Spatial maps of correlated brain activity across 13 human subjects (A–C) and eight marmoset subjects (D–F), mapped on each flattened cortex (A and D) and left cortical surface [lateral-medial view (B and E) and dorsal-ventral view (C and F)]. Approximate locations of parietal, auditory, and frontal regions are indicated by green, purple, and white dashed lines on the flat maps, respectively. The VOIs in visual-related areas were manually created based on the multimodal cortical parcellation atlas (21) for humans and Paxinos atlas for marmosets (22) so as not to include the low correlation areas among scans (G for humans and H for marmosets). Then, the time courses were extracted from two VOIs (e.g., PIT in humans and FST in marmosets), and the cross-correlation coefficient was calculated between them (I). See SI Appendix, Fig. S1 for the right hemisphere. LO1-3: area lateral occipital 1 to 3; VVC: ventral visual complex; A19DI: area 19 of cortex dorsointermediate part; TLO: temporal area TL occipital part; TPO: temporo-parieto-occipital association area.

Fig. 2.

Correlation maps (z-score maps) in human (A) and marmoset (B) brains with human face-specific (PeEc indicated by pink area in C), body-specific (TE2p indicated by light blue area in C), and scene-specific (POS1 indicated by green area in C) areas in the right hemisphere were presented on each flattened map. C and D show the same data focusing around the occipital and temporal regions (areas surrounded by blue squares in A and B) in right hemispheres. White lines indicate the borders of the multimodal cortical parcellation atlas (21) for humans and the Paxinos atlas for marmosets (22). The correlation coefficient maps with human VOIs in the left hemisphere are shown in SI Appendix, Figs. 3–6.

Fig. 3.

The peak locations of human (A) and marmoset (B) correlation maps with the time course in each human VOI (presented by color labels). Note that 11 out of 26 human VOIs (V3, V3A, V3B, V3CD, V4, V8, PH, PHA, ProSt, VMV, and VVC) were not presented due to the differences of the peak locations between the left and right hemispheres. The maps were focused on the areas around the occipital and temporal regions in the left hemispheres. Black lines indicate the borders of the multimodal cortical parcellation atlas (21) for humans and the Paxinos atlas for marmosets (22).

Fig. 4.

Correlations of the time course in each face or body patch with the pseudoevent-related design. We first identified which animals (marmosets, other animals, or no animals) we can find in each TR (1.5 s) in the movie clip, and they were classified as 0 (OFF) and 1 (ON). These pseudoevent-related designs were convolved by hemodynamic response function and were applied as regressors of interest (A). Correlation coefficients between the predicted designs and the time courses in each face and body patch in human (B) and marmoset (C). The locations of the clusters are shown on the cortical surfaces (F1 to F4 for face-specific clusters and B1 for body-specific clusters), which is the same data as shown in Fig. 2 C and D. The black, slash, and white bars indicate the correlation coefficient values with the features of marmosets, other animals, and no animals, respectively. An asterisk and a cross indicate significant differences of P < 0.05 and P < 0.01 using ANOVA with Bonferroni post hoc correction, respectively. Error bars indicate the SDs.

Fig. 5.

The peak locations of human (A) and marmoset (B) correlation maps with the time course in each marmoset parietal VOI (presented by red or blue labels). White lines indicate the borders of the multimodal cortical parcellation atlas (21) for humans and the Paxinos atlas for marmosets (22).