Integrative cross-species analysis of GABAergic neuron cell types and their functions in Alzheimer’s disease

The heterogeneity of GABAergic neurons in human, macaque, mouse, and pig

To perform a cross-species comparative study of the GABAergic neurons, we collected the snRNA-seq datasets of the cerebral cortex for human10,11, macaque12,13, mouse14,15, and pig16. After cell-type annotation and filtering out the excitatory neurons and non-neurons, the GABAergic neurons were used for heterogeneity, functional properties, and AD-related risk genes analysis across species (Fig. 1A, Table S1). Specifically, datasets from the same species were integrated using reciprocal principal component analysis (RPCA). After determining anchors between any two datasets using RPCA, we projected each dataset into the other PCA space and constrained the anchors by the same mutual neighborhood requirement. The datasets of the cerebral cortex for each species were integrated and visualized by t-SNE. The major cell types of oligodendrocyte progenitor cells, astrocytes, oligodendrocytes, endothelial cells, excitatory neurons, microglia, and inhibitory neurons were identified by the expression of canonical gene markers, suggesting the conservation of major cellular compositions across these four species (Figs. 1B, S1A).

Figure 1
figure 1

Cell type heterogeneity across species. (A) Pipeline of article design and data analysis. (B) The t-SNE plots show the major cell types in human, macaque, mouse, and pig. Exc, Excitatory neuron; Inc, Inhibitory neuron; MG, Microglia; Endo, Endothelial cells; OPC, Oligodendrocyte progenitor cells; Oligo, Oligodendrocytes; Ast, Astrocytes. (C) The t-SNE plots show the subtypes of GABAergic neurons in human, macaque, mouse, and pig. (D) The heatmaps show the correlation of different GABAergic neurons subtypes in human, macaque, mouse, and pig.

Subsequently, the GABAergic neurons were split from each species for cell-type heterogeneity analysis. The GABAergic neuron subclasses were annotated by differentially expressed genes (DEGs). Four major subclasses of GABAergic neurons were identified with significant DEGs of PVALB, SST, LAMP5, and VIP for each subclass. These results are consistent with previous reports22. However, some species-specific GABAergic neuron subclasses were defined, such as human Inc ECM1, mouse Inc Meis2, Inc Pde1a, Inc Reln, and pig Inc TRPC4. We determined the distinct cell classes for these four GABAergic neuron subclasses according to the DEGs in each subclass. For example, human Inc VIP OPRM1, Inc LAMP5 CA1, Inc PVALB SULF1, Inc SST NEFL highly expressed specific OPRM1, CA1, SULF1, and NEFL independently (Figs. 1C, S1B). We then split the GABAergic neuron subclasses to calculate the average gene expression of each cell class for four different species. The results showed that the cell types were linearly proportional to the average gene expression, which confirmed that the subclasses of GABAergic neurons are highly consistent with average gene expression patterns in each cell type (Fig. 1D).

Comparative analysis of GABAergic neuron subclasses across species

To analyze the GABAergic neurons between different mammalian species, we calculated the homologous genes of macaque, mouse, and pig according to humans (Table S2). The subclasses of GABAergic neurons were annotated by the shared homologous genes and integrated between non-human species (macaque, mouse, and pig) and humans. The results showed that more GABAergic subclasses were identified in human than in non-human species (Fig. S2A). We further analyzed the conservation and divergences of four subclasses of GABAergic neurons across these four species. In the Inc SST subclass, the identified subtypes in non-human species were highly conserved with human, while two subclasses of Inc SST NMBR and NEFL from human were discrete. However, Inc SST subtypes within each non-human species were various overlapped with human. For example, Inc SST RUNX1 and KLF5 from pig were more conserved than Inc SST HMGN4 compared to human Inc SST. Most of the Inc VIP subclass from macaque and mouse were highly conserved to the human, but the pig Inc VIP subclass was lowly conserved to human. The Inc PVALB subclasses from macaque and mouse were highly conserved with human, while pig Inc PVALB SULF1 was lowly conserved to human. For the Inc LAMP5 subclass, all the macaque, mouse, and pig subtypes were highly conserved to the human Inc LAMP5 CA1 subtype, and only macaque Inc LAMP5 FBN2 was conserved to the human Inc LAMP5 TPPC3 subtype. Interestingly, mouse-specific Inc Meis2 were highly conserved to human Inc LAMP5 CRH and Inc SST NEFL (Fig. 2A). To confirm the conservation of GABAergic neuron subclasses across four mammalian species, we analyzed the split inhibitory neuron cells by their DEGs. The results demonstrated that SST, LAMP5, PVALB, and VIP were highly expressed in the top 15 gene list, confirming the four major GABAergic neuron subclasses of Inc SST, Inc LAMP5, Inc PVALB, and Inc VIP across species (Figs. 2B, S2B). While the four major subclasses of GABAergic neurons were notably conserved across species, the differences in GABAergic neuron cell subclasses between different species were also depicted, which could be demonstrated by differential gene expression in the subclasses.

Figure 2
figure 2

The conserved and diverse GABAergic subclasses across species and their functional features. (A) Heatmap showing the cluster overlaps of GABAergic subclasses between human and the other three species. (B) Heatmap showing top 15 DEGs of conserved GABAergic neuron types containing Inc SST, Inc LAMP5, Inc PVALB, and Inc VIP. (C) The GO terms are enriched from conserved and specific genes in the four species. Blue: Human; Red: Mouse; Brown: Macaque; Green: Pig.

Changes in gene expression patterns drive individual phenotypic differences and the evolution of new phenotypes between species25,26. To investigate cross-species functional conservation and divergences for GABAergic neurons, we performed GO (Gene Ontology) term enrichment analysis of four GABAergic neuron subclasses in human, macaque, mouse, and pig. While most terms across species showed highly conserved in four species, in Inc SST subclass, the cell types from pig are more divergent than the other three species, especially for the functions of axonogenesis, regulation of GTPase activity. However, some of functional terms are conserved in different species, such as regulation of intracellular transport terms detecting in macaque and pig, regulation of trans-synaptic signaling, modulation of chemical synaptic transmission, synapse organization detecting in all four species. In Inc PVALB subclass, the functional terms from three non-human species showed more discrete than human. For example, the cell types from pig showed some species-specific functions of regulation of GTPase activity, dendrite development, dendrite morphogenesis, positive regulation of GTPase activity; macaque showed some specific functions of transport along the microtubule, axo-dendritic transport, positive regulation of neurogenesis, cytoskeleton-dependent intracellular transport; and mouse demonstrated species-specific functions of Golgi vesicle transport, proteasome-mediated ubiquitin-dependent protein catabolic process, and organelle transport along microtubule. In the Inc LAMP5 subclass, the most divergences of cellular functional features from mouse to human are RNA splicing, regulation mRNA metabolic process, et al. In the Inc VIP subclass, human and pig showed more species-specific cell type functions than macaque. In addition, some functional terms are conserved in four species, such as modulation of chemical synaptic transmission, synapse organization. (Fig. 2C). These results showed the conservation of GABAergic neuron subclasses with functional diversity across human, macaque, mouse, and pig.

Conservation and divergences of regulatory modules for GABAergic neurons across species

Comparative studies of gene expression levels and the evolution of gene regulation provide the compelling evidence that most gene regulatory patterns evolve under selective constraint27. Regulons, consisting of transcriptional factors (TFs) and cofactors regulating each other and their downstream target genes, are associated with characteristic functional transcriptional activities28,29. Based on the homologous genes among the four species, we identified 9277 homologous genes for constructing gene regulatory networks. Finally, 182 regulons activated in at least one cell which were used for downstream analysis. To find out distinct regulons in different GABAergic subtypes, we calculated the percentage of activated cells in each cluster. All regulons were clustered into eight groups with specific activated patterns, conserved in part of species or all four species. For example, the regulons in group 1 were not enriched in any species; the regulons in group 2 were conserved in macaque; the regulons in group 3 demonstrated in mouse and part of macaque; the regulons in group 4 were enriched in human; the regulons in group 5 partly conserved in human and pig. Interestingly, the regulons in groups 6 and 7 were highly conserved in all four species. However, the regulons in group 8 were conserved in most species, excluding mouse (Fig. 3A). These results demonstrated that the regulons exhibited the conserved and divergent features across species.

Figure 3
figure 3

Conserved and specific regulons among four species. (A) Heatmap showing the percentage of regulon-activated cells in each subtype in different species. The top annotation indicated species and major cell types. Major regulons were clustered as 8 groups based on activated patterns. (B, C) TF-targets regulon networks in groups 4 and 7. Red: TFs; Blue: target genes. (D) GO terms enriched from genes of group2-8.

To identify the target genes of the regulons in each group, we performed the GRNs analysis for seven groups except non-conserved regulons in any species of the first group. The results showed that the conserved transcription factors (TFs) specifically regulated functional gene modules for each group, such as the regulons conserved in human regulating functional modules gene in group 4 and the conserved TF-target pairs across all four species in group 7. The conserved TF-targets were also investigated in other groups (Figs. 3B,C, S3A, Table S3. We further performed the functional enrichment analysis using the TF-target pairs enriched in different groups. The enriched terms in the identified groups showed the functional similarities and differences of TF-target pairs across species (Fig. 3D).

GABAergic neuron cell-type-specific AD-associated risk gene module

Genetic abnormalities in the central nervous system can lead to neurological disorders. AD is the most common neurodegenerative disease and the main cause of adult-onset symptoms of dementia. 2911 genes (Table S4) were associated with AD, which has been demonstrated by genome-wide association studies (GWAS) (www.ebi.ac.uk/gwas/)30. Analysis of gene co-expression modules by weighted gene co-expression network analysis (WGCNA) found that genes in the AD risk gene lists converged on seven modules. All modules exhibited noticeable changes in most cell types between normal and disease samples (Fig. 4A). Five of these seven modules were enriched for significant association with AD risk (P < 0.05, hypergeometric test). The module of particular interest intensively enriched for AD-associated risk genes was turquoise (Table S5), whose expression was evenly enriched in four inhibitory neuron subtypes (Inc PVALB, Inc VIP, Inc LAMP5, and Inc SST) irrespective of their origin (normal or disease samples). Next, we obtained the intersection between turquoise module genes and AD risk genes (119 genes, Fig. S4, Table S6), which were selected to understand the conservation of each cell type genes associated with AD. We compared the cell-type expression of these AD susceptibility genes between human and each of the other three species (Fig. 4B). Cell-type expression of the AD-associated genes in macaque displayed a high correlation with that in human. Except for the Inc LAMP5, the other three cell-type expressions of AD genes in pig showed moderate correlation with human AD susceptibility genes. Nevertheless, expression of AD-associated genes in four mouse cell types showed a very low correlation with those in human.

Figure 4
figure 4

Critical AD-risk modules and GABAergic subtypes among four species. (A) Heatmap (left) showed relative expression of 7 WGCNA modules among different cell types. Each module expression was calculated as an average expression of all genes in the corresponding module and scaled by columns. Heatmap (right) showed the number of overlaps between module genes and AD high-risk genes, and the significance (P value) of the hypergeometric distribution test identified the module in which AD risk genes were over-represented (P value: 0.01–0.05, *; 0.001–0.01, **). (B) Spearman correlation of snRNA-seq expression of AD susceptibility genes showing the conservation of each cell type genes associated with AD between human and each of three species on four cell types. (C) Venn plots showing DEG overlaps in each cell type between AD and control, AD risk genes, and turquoise module genes. (D) The above heatmap displayed the DEGs in four GABAergic subtype neurons, in which blue represented significantly DEGs (adjusted P value < 0.05). Expression patterns of genes (columns) were associated with AD by cell type (rows) in four species. The degree of disease-associated gene enrichment in cell type was calculated by the Wilcoxon Rank Sum Test. *P < 0.05.

Finally, we obtained the intersection among turquoise module, AD risk genes, and DEGs (between control and AD samples). We found 19, 25, 45, and 22 AD-related disease genes in four inhibitory neuron cell types, respectively (Fig. 4C). We further assessed the cell-type expression of these genes in four species. We also found that the global expressional patterns of the four AD-related genes in four cell types were very similar between human and macaque (Fig. 4D). Many genes were conservatively expressed in four cell types of these two species but not in the other two species, such as SGCZ and CTNNA2 (Fig. 4D), which was in good accordance with the abundant localization in all layers of the adult rhesus monkey dorsolateral prefrontal cortex by situ hybridization31. Intriguingly, a few genes such as CLU and APP in most cell types of macaque showed a higher expression level than those in human (Fig. 4D). Compared with mouse, pig showed more similar expression patterns to those in human (Fig. 4D).

Read more here: Source link