Functional variation among mesenchymal stem cells derived from different tissue sources

Study approval

Amniotic membrane, chorionic villi, and umbilical cord MSCs were sourced from full-term placentas of healthy women post-vaginal delivery or cesarean section, without any evident pregnancy complications or infections such as HIV, syphilis, or hepatitis B virus. ADMSCs were sourced from the lower abdominal fat of healthy individuals without obesity, chronic illnesses, or infections such as HIV, syphilis, or hepatitis B virus. Table S1 provides sample details. This study was approved by the ethics committee of China Human Genetic Resources Management Office ([2022] BC0102). All procedures involving human participants were performed in accordance with the ethical standards of the research committee. Informed consent was obtained from all participants.

Isolation and cultivation of human MSCs under normoxic and hypoxic conditions

Following tissue collection, the samples were promptly transported to the laboratory of Yuanpin Cell Biotechnology Co., Ltd. (Hunan, China) within 12 h for further processing, maintaining a continuous cold chain at 4 °C throughout the entire process.

For amniotic membrane MSCs, amniotic membrane tissue was separated from the placenta, cut into 1 × 1 cm² pieces, and then digested with an equal volume of 0.4 U/mL Collagenase NB6 (N0002779; Nord Mark, Plymouth, MN, USA) for 1 h at 37 °C with agitation. After digestion, the sample was filtered through a 70 µm cell sieve, followed by centrifugation at 1,000 × g for 5 min. The supernatant was removed, and the cells were seeded in a T75 culture flask with MSC-T4 (Baso 04304P05) complete culture medium.

For umbilical cord MSCs, umbilical cord tissue was transferred to a 15 cm culture dish, both ends were trimmed with ophthalmic scissors to remove any remaining blood, and the sample was transferred to a new 15 cm culture dish. The umbilical cord was cut into 2–3 cm segments, longitudinally split along the vein, and the umbilical vein, artery, and Wharton’s jelly were removed. The tissue was placed in a 10 cm culture dish. Wharton’s jelly was cut into small tissue blocks (0.2–0.4 cm²) and transferred to T75 cell culture bottles, with approximately 40 tissue blocks per bottle. The culture bottles were transferred to a CO₂ incubator and allowed to stand for 1 h to allow the tissue to adhere to the bottom. Then, 8 mL of MSC-T4 (Baso 04304P05) complete culture medium was slowly added to ensure that the tissue was completely immersed, and the culture was continued in a CO₂ incubator.

For chorionic villus MSCs, placental tissue was transferred to a 10 cm culture dish, a portion of the tissue was trimmed into the dish using ophthalmic scissors, and the outer surface of the placenta was rinsed to remove blood. The rinsed placenta was then placed in a new 10 cm culture dish and immersed in 0.9% saline solution for 3 min. After soaking, the placenta was washed with 0.9% saline solution 2–3 times to separate the chorionic membrane layer from the amniotic membrane and maternal tissue blocks. The separated chorionic membrane was placed in a 10 cm culture dish, cut into 2 mm tissue blocks, and transferred to T75 cell culture bottles with approximately 50–60 tissue blocks per bottle, filling the bottle completely. The bottles were transferred to a CO₂ incubator to allow the tissue to adhere to the bottom of the bottle for 1 h. Then, 8 mL of MSC-T4 (Baso 04304P05) complete culture medium was slowly added to ensure complete immersion of the tissue, and the culture was continued in a CO₂ incubator.

For adipose-derived MSCs, adipose tissue was washed three times with phosphate-buffered saline (PBS) (A12858-01; Gibco, Waltham, MA, USA) containing 1% penicillin-streptomycin and then transferred to a 15 mL centrifuge tube, where it was minced using surgical scissors. Next, an equal volume of 0.2% type I collagenase (C1639; Sigma, St. Louis, MO, USA) was added, and the mixture was mixed thoroughly by pipetting. The mixture was placed on a 37 °C shaker at 80 rpm to ensure digestion. After digestion was complete, an equal volume of MSC-T4 (Baso 04304P05) complete culture medium was added to terminate digestion, and the mixture was pipetted to ensure uniform mixing. The mixture was then centrifuged at 1,200 rpm for 10 min, the supernatant was removed, and the cell pellet was resuspended in 1 mL of MSC-T4 (Baso 04304P05) complete culture medium. Cells were seeded at a density of 2.50 × 10⁴/cm² in T25 culture flasks and incubated in a CO₂ incubator.

MSCs from these four distinct tissue sources were cultured at 37 °C with 5% CO₂ under normal atmospheric oxygen conditions (approximately 21% oxygen concentration) and hypoxic conditions (1% oxygen). The hypoxic condition was generated using a hypoxia chamber (catalog no. 4131F0; Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer’s instructions. In brief, the cultures were placed in the chamber and flushed with a gas mixture (95% nitrogen and 5% carbon dioxide) for 3 min. After the flushing period, the chamber was sealed to prevent the ingress of external air. The final level of hypoxia was maintained at 1%, as specified by the manufacturer (Rosová et al., 2008).

Flow cytometry sorting and analysis

The MSCs derived from individual donors were subjected to trypsinization using TrypLE Express (Thermo Fisher Scientific, Waltham, MA, USA) and subsequently washed twice with PBS at either the first or second passage. Flow cytometry was employed to isolate and identify the MSCs based on specific surface markers (i.e., positive for CD90, CD44, CD73, and CD105 and negative for CD11b, CD19, CD34, CD45, and HLA-DR), as per the instructions outlined in the BD Stemflow^™ Human MSC Analysis Kit. A minimum of 10,000 events were recorded and analyzed using a FACSVerse instrument (BD Bioscience, Franklin Lakes, NJ, USA).

Lineage differentiation in vitro

Stem cell chondrogenic differentiation was conducted in a 24-well plate, with cells labeled for identification and controls. Passage 5 (P5) stem cells (2.0 × 10⁵) were placed in the wells and cultured in a 37 °C, 5% CO₂ incubator. Chondrogenic differentiation medium (A1007101; Gibco, Waltham, MA, USA), prepared by mixing 10% Chondrogenesis Supplement with 90% Chondrogenesis Differentiation Basal medium, was applied every 3 days after cells reached 60–70% confluence. To confirm chondrogenic differentiation, Alcian Blue (G2541; Solarbio, Beijing, China) staining was performed on days 14–21. After fixation with 4% PFA and washing, cells were stained with Alcian Blue and observed under a fluorescent microscope to evaluate aggregation.

Passage 5 (P5) MSCs were cultured in 24-well plates and adipogenic differentiation was induced (A1007001; Gibco, Waltham, MA, USA) when they reached 70–80% confluence. The culture medium was changed every 3 days. To confirm adipogenic differentiation, Oil Red O staining (G1260; Solarbio, Beijing, China) was performed on days 7–14. The procedure included fixation, isopropanol rinsing, and Oil Red O staining, followed by observation under a fluorescent microscope.

Passage 5 (P5) MSCs were cultured in 24-well plates and osteogenic differentiation (A1007201; Gibco, Waltham, MA, USA) was induced when they reached 60–70% confluence. The culture medium was changed every 3 days. To confirm osteogenic differentiation, 0.2% Xylene Red staining (G1452; Solarbio, Beijing, China) was performed on days 14–21. The procedure included fixation, rinsing, and Alizarin Red S staining, followed by observation under a fluorescent microscope.

Single-cell RNA sequencing

Ten samples from P5 MSCs were isolated and cultured from adipose, amniotic membrane, chorionic villus, and umbilical cord tissues. Single cells from these samples were captured using a GEXSCOPE microfluidic chip according to the instructions provided by NewGen Biotech Co. (Dallas, TX, USA). Barcoded Beads were collected from the microarray, and the mRNAs captured by the Barcoded Beads were reverse-transcribed into cDNA and amplified. cDNA was fragmented and ligated to build a library for sequencing using the Illumina platform.

ScRNA-Seq and quality control of samples were performed using the Singleron Matrix platform. The results were annotated according to GENCODE (Version 28, https://www.gencodegenes.org/) using Star (Version: 2.5.2b). Further correction was performed using the distribution-based error correction (DBEC) algorithm, and all putative cells were distinguished from background noise using a second-order derivative analysis.

To further resolve the profiles of UCMSCs, we obtained eight samples from P0, P3, and P5 UCMSCs and P5L UCMSCs. The analysis procedure, from sample processing to cDNA library construction, was completed based on the instructions provided with the 10× Genomics platform. The quality of the library was evaluated using the Agilent Bioanalyzer and a High-Sensitivity DNA Kit (Agilent Technologies, Santa Clara, CA, USA). Libraries were sequenced in 2 × 150 bp double-end mode using the Illumina NovaSeq 6000 system (Oe Biotech, Shanghai, China). Cell Ranger v2.1.0 (https://support.10xgenomics.com/single-cell-gene-expression/software/pipelines/latest/what-is-cell-ranger) was used for cell demultiplexing, human genome (version GRCh38) matching, barcode processing, and unique molecular identifier (UMI) counting. STAR aligner was used for read mapping against the human genome and transcriptome.

Single-cell data processing

A power analysis for scRNA-seq was performed using the R package RNASeqPower (version: 1.40.0, https://bioconductor.org/packages/RNASeqPower/) with parameters set to effect = 0.5 and alpha = 0.05. RNASeqPower was 0.9994395. The Seurat R package (Version 4.0.3, https://satijalab.org/seurat) was used for further analyses of single-cell RNA-seq data. After the initial evaluation of Cell Ranger metrics, the data from adipose, amniotic, chorionic, and umbilical cord P5 cells under normoxic conditions were analyzed. Cells with fewer than 1,100 genes, more than 5,000 genes, or more than 15% mitochondrial genes were excluded, and 77,151 cells were retained for downstream bioinformatics analyses. For the datasets for P0, P3, P5 MSCs and P5 MSCs under hypoxic conditions, cells with fewer than 2,000 genes, more than 7,000 genes, or more than 15% mitochondrial genes were excluded, leaving 61,756 cells for downstream bioinformatics analyses. The sequencing reads for each gene were normalized to the total UMI in each cell to obtain the normalized UMI values via the “NormalizeData” function. In other words, the UMI for each gene was divided by the total UMI for the cell, multiplied by the median total UMI, and finally converted to the natural logarithmic form. The “ScaleData” function was used to scale and center the expression levels in the dataset for dimensionality reduction.

PCA downscaling and clustering analysis

To avoid batch effects between samples and experiments, the first 2,000 highly variable genes were evaluated using a canonical correlation analysis (CCA) implemented in Seurat. The aligned vectors were then used for downstream downscaling and clustering analyses. Subsequently, a principal component analysis (PCA) was performed using the “RunPCA” function. Since the cell cycle has a strong influence on properties of MSCs, the cell cycle score for each cell based on specific genes associated with the G2M and S phases was evaluated using the “CellCycleScoring” function, followed by the elimination of cell cycle effect using the “SCTransform” function. Then, the data were downscaled using the “RunUMAP” function and visualized using Uniform Manifold Approximation and Projection (UMAP). For P5 MSCs derived from adipose, amniotic, chorionic, and umbilical cord tissues, cell clustering was performed using the “FindClusters” function with a resolution of 0.8, and the first nine main components were used to define cell identity. For P0, P3, P5, and P5L generation UCMSCs, cell clustering was performed using the “FindClusters” function at a resolution of 0.4, and the first nine principal components were used to define cell identity. The Wilcoxon rank sum test was performed using the “FindAllMarkers” function to determine the marker genes for each cluster, and only genes with |avg_log₂FC| > 1, p_val < 0.05, and min.pct > 0.1 were considered markers.

GO enrichment analyses

Gene Ontology (GO) analyses of tagged genes were performed using Metascape (Version 3.5, http://metascape.org/) and visualized using the R package ggplot2 (Version 3.3.5, https://github.com/tidyverse/ggplot2).

Monocle 3 construction for a pseudo-time-series analysis

The pseudo-time-series analysis R package Monocle3 (Version 1.0.0, https://cole-trapnell-lab.github.io/monocle3/) was used to reconstruct the developmental trajectories of UCMSCs. Based on the routine process, the developmental start and end positions were selected using the “choose_graph_segments” function in combination with the timing diagram. Then, the “graph_test” function was used to identify genes associated with the development of MSCs. The “find_gene_modules” function was used to assign genes to modules, and the expression levels of the genes in these modules were evaluated. The “pheatmap” package was used to create a heat map of gene expression.

Identifying tissue-specific MSC subpopulations and expression profiles via scRNA-seq

MSC isolation methods and culture conditions were validated through the flow cytometry-based detection of surface markers as well as multilineage differentiation under in vitro induction conditions (Figs. S1–S4). To establish a comprehensive single-cell transcriptome atlas of human MSCs, we obtained MSCs at passage five from various tissue types from 10 healthy donors (Table S1), including adipose (two donors), chorionic villi (two donors), amniotic membrane (three donors), and umbilical cords (three donors). Subsequently, we performed scRNA-seq. After stringent quality control to eliminate low-quality cells with high mitochondrial gene signatures and doublets, our dataset comprised 77,151 cells. Our analysis, utilizing UMAP for dimensionality reduction, unveiled six cell clusters (Fig. 1A). This diversity of MSC subpopulations exhibited a tissue-dependent pattern, underscoring substantial transcriptomic heterogeneity among MSCs from distinct tissues (Figs. 1B and 1C). We observed differences in the proportions of subpopulation among MSCs derived from four different tissue sources (Fig. 2A). Additionally, we found that the differences in relatively subpopulation frequencies among different samples were relatively small, indicating that MSC samples from the same tissue source exhibited a degree of consistency in subpopulation distributions. However, AMMSC3 differed from AMMSC1 and AMMSC2. This observation suggests that there may be substantial differences in the number and composition of subpopulations of MSCs among different tissue sources, which is crucial for understanding the diversity and potential applications of MSCs. Furthermore, for MSCs from the same tissue source, subpopulation sizes and proportions were relatively consistent; however, there may still be variation among samples. To quantitatively support our observations, we performed a Chi-squared goodness of fit test to compare the subpopulation distributions among different groups. The results are presented in the Table S2, which includes statistical comparisons of the proportions across different MSC samples. The Chi-squared test results showed significant differences in the proportions of subpopulations across different tissue sources and samples. For example, the AMMSC3 sample shows significant differences in subpopulation proportions compared to AMMSC1 and AMMSC2, confirming our earlier visual observations. These differences could be caused by various factors, including individual variation, collection methods, and culture conditions, which will require further investigation in future studies. By applying the SCT method to remove cell cycle effects and conducting a PCA, we observed that the cell cycle phases were evenly distributed among cells from four different sources, and the samples also exhibited a uniform distribution (Figs. 1D and 1E). Furthermore, we did not detect a significant difference in cell cycle distribution among cells from different tissue sources. However, a noticeable difference in the cell cycle distribution existed among different subpopulations (Figs. 2C and 2D). Although we divided MSCs from four different tissue sources into six subpopulations, apart from the C5 subpopulation (which exhibited the upregulation of MEST, CDCA8, and KIF11), there were no significantly differentially expressed marker genes among subpopulations (Fig. 1F). This suggests that the subpopulations share similarities in functions or features. This similarity might reflect key shared biological characteristics. Nevertheless, in-depth functional analyses and feature descriptions are needed to gain a more comprehensive understanding of the biological roles and functions of each subpopulation.

MSCs from different tissue sources exhibit varying functional characteristics

To elucidate the functional differences among MSCs derived from four different tissue sources, we compared the relative frequencies of the subpopulations from the four different sources (Fig. 2B and Table S2). Additionally, we conducted a functional enrichment analysis of differentially expressed genes in the six subpopulations (Fig. 2E). Among the subpopulations, the most notable differences in cell proportions were observed in two clusters, C2 and C5. Based on the number of cells in each group within the subpopulation, cluster C2 was predominant, accounting for 60.38% of ADMSCs. Conversely, this subpopulation was much less prevalent in AMMSCs (12.59%) and UCMSCs (20.83%), with the lowest proportion in CMMSCs (6.19%). The gene expression patterns within cluster C2 were associated with blood vessel development and angiogenesis. This suggests that ADMSCs exhibit stronger functionality in terms of angiogenesis and blood vessel development when compared with that of cells from the other three sources. Furthermore, we identified a subpopulation, cluster C5, with variation in proportions among MSCs derived from four different tissue sources. Notably, the C5 subpopulation accounted for 85.39% of CMMSCs but was less prevalent in AMMSCs (9.29%) and UCMSCs (4.99%). Of note, this subpopulation was nearly absent in ADMSCs (0.33%). Gene expression patterns within cluster C5 were linked to oogenesis and the regulation of reproductive processes. This finding strongly suggests that CMMSCs possess more pronounced functional characteristics related to oogenesis and the regulation of reproductive processes, while ADMSCs may lack such characteristics.

Regarding other cell proportions, C0 (accounting for 36.7% of ADMSCs, 22.61% of AMMSCs, 18.06% of CMMSCs, and 22.63% of UCMSCs), C1 (accounting for 19.21% of ADMSCs, 27.36% of AMMSCs, 23.2% of CMMSCs, and 30.22% of UCMSC). Although the C0 and C1 subpopulations exhibited slightly different proportions across the four sources, these differences were relatively minor, and there was no significant variation in cell numbers. Cluster C0 was linked to cell growth and developmental processes, while cluster C1 displayed gene expression patterns associated with cell cycle regulation and mitosis. These results suggest that differences in regeneration and cell cycle regulation among MSCs from the four sources are not substantial.

Cluster C3 was prevalent in UCMSCs (41.92%) and AMMSCs (33.55%) and less common in ADMSCs (12.54%) and CMMSCs (11.99%). Cluster C3 exhibited gene expression patterns related to wound healing, skin development, and inflammatory responses. This implies that UCMSCSs and CMMSCs have important roles in wound repair and immune regulation. Cluster C4 was the least prevalent in ADMSCs (accounting for 8.78%). Cluster C4 demonstrated gene expression patterns associated with the regulation of osteoblast differentiation and neurogenesis. These results indicate that ADMSCs may have relatively weaker effects on osteoblast differentiation and neurogenesis than those of cells from the other three sources.

In summary, MSC subpopulations exhibit variation with respect to tissue sources, with ADMSCs associated with blood vessel development, CMMSCs associated with reproductive processes, and subtle differences in cell growth and cycle regulation. Umbilical cord and chorionic MSCs demonstrated roles in wound repair and immune regulation, while ADMSCs had relatively weak effects on osteoblast differentiation and neurogenesis compared with those of cells from other sources.

Passages and hypoxic culture umbilical cord MSC subpopulations and expression profiles via scRNA-seq

UCMSCs have become the most widely used cell sources in many clinical trials owing to their low immunogenicity, ease of acquisition, and minimal ethical controversies. To investigate the effects of in vitro culture conditions on subpopulations of UCMSCs, we performed a single-cell expression analyses. UCMSCs were obtained from four healthy donors (Table S1) and were evaluated at various passages and culture environments (UCMSCP0 = 2, UCMSCP3 = 2, UCMSCP5 = 2, and UCMSCP5L = 2) using the 10× Genomics platform. We identified six subpopulations through a UMAP clustering analysis (Fig. 3A). Through a PCA and UMAP clustering analysis, we discovered heterogeneity in UCMSCs across different passages and culture environments, with relatively minor individual differences between samples (Figs. 3B and 3C).

A more detailed examination indicated that the proportions of subpopulations among different samples from the same group were relatively consistent, despite variations in passages and hypoxic conditions (Fig. 4A). Nevertheless, there was variation in specific samples (e.g., between subpopulations S0 and S5 in UCMSCP3N1 and UCMSCP3N2). To quantitatively support our observations, we performed a Chi-squared goodness of fit test to compare the subpopulation distributions among different groups. The results are presented in the Table S3, which includes statistical comparisons of the proportions across different MSC samples. The Chi-squared test results show significant differences in the proportions of subpopulations across passages and hypoxic conditions. For example, between UCMSCP3N1 and UCMSCP3N2, there are significant differences in the proportions of subpopulations S0 to S5. These differences might be attributed to various factors, such as genetic background, necessitating further exploration.

Additionally, after applying the SCT method to remove cell cycle effects and conducting a PCA, we compared the distribution of cell cycle phases within subpopulations among samples. We observed that the variation in cell cycle phases among the UCMSCP0, UCMSCP3, UCMSCP5, and UCMSCP5L groups was relatively minor. The differences in cell cycle phases between various subpopulations were more pronounced (Figs. 3D and 3E). A detailed analysis of relative frequencies of cell cycle phases revealed no significant differences in the cell cycle distribution among UCMSCP0, UCMSCP3, and UCMSCP5. However, in UCMSCP5L, we observed an increased proportion of G1 phase cells and a decreased proportion of G2M phase cells. This suggests that the low oxygen environment may slow cell cycle progression (Fig. 4C). Nevertheless, distinct differences in the cell cycle distribution were evident among subpopulations (Fig. 4D). These results suggest that these subpopulations have distinct characteristics in terms of cell cycle control and regulation, which could potentially impact their functions and functional specificity. Further research on the cell cycle characteristics of these subpopulations improve our understanding of their biological roles. Furthermore, we analyzed relatively specific marker genes of cluster S0–S5 (Fig. 3F). The absence of significant marker genes for these subpopulations implies that they exhibit shared functions or features. For a more comprehensive grasp of the biological roles and functions of subpopulation, in-depth functional analyses and feature descriptions are warranted.

Variation in the proportional composition of umbilical cord MSC subpopulations across passages P0, P3, P5, and low-oxygen culture conditions

To elucidate the functional differences among UCMSCs derived from various passages and oxygen environments, we compared cell proportions and conducted a functional enrichment analysis of differentially expressed genes among the six subpopulations (Figs. 4B, 4E and Table S3). The cell proportions in different subpopulations showed variation within UCMSCP0, with a higher proportion of cluster S5 (36%) and lower proportions of cluster S1 (13.52%) and S4 (7.85%). Differentially expressed genes in cluster S5 were enriched for functions related to DNA replication, while genes in cluster S1 showed enrichment for functions related to vascular development, follicular development, and reproductive system development. Differentially expressed genes in cluster S4 exhibited enrichment for vascular development and regeneration. These findings suggest that UCMSCP0 cells have a strong capacity for proliferation; however, their potential for vascular and reproductive functional regeneration may be relatively weak. In UCMSCP3, the proportions of various subpopulations were relatively consistent. In UCMSCP5, the S5 subpopulation was most abundant (41.33%), while the S4 subpopulation was less frequent (14%). This distribution suggests that in P5, cells were disproportionately involved in DNA replication, indicating a faster rate of cell proliferation. However, this may also indicate a relatively weaker capacity for vascular regeneration. Additionally, when analyzing the proportions of subpopulations in UCMSCP5L, we detected variation. The S4 subpopulation accounted for 56.71% of cells, followed by the S1 subpopulation (39.54%), suggesting that UCMSCP5L from P5 under low oxygen conditions contributes to vascular regeneration and reproductive function. Interestingly, the S5 subpopulation was notably scarce (0.29%), and the S2 subpopulation only accounted for 10.68% of cells. These results imply that a limited number of cells in P5 under low oxygen conditions are involved in DNA replication and chromosome separation, possibly due to a slow rate of cell cycle progression induced by the low oxygen environment. UCMSCP5 may have a greater role in reproductive functions when compared with those of UCMSCP0 and UCMSCP3. Moreover, a low-oxygen culture environment might promote vascular regeneration while inhibiting cell cycle progression.

To further confirm the specialized functions of the S1 and S4 subpopulations in clinical applications related to vascular and reproductive function, we performed a pseudotime trajectory analysis of UCMSCs across different passages and oxygen conditions. This analysis confirmed that the S1 and S4 subpopulations follow unique differentiation trajectories (Figs. 5A and 5B). Importantly, these trajectories appeared to be minimally influenced by the cell cycle and displayed subpopulation-specific and tissue-specific characteristics (Figs. 5C and 5D).

Based on correlations between pseudotime and gene expression results, we categorized 14,837 genes into 20 gene modules, revealing distinct expression patterns in the S0–S5 subpopulations. Notably, the S1 subpopulation exhibited the specific upregulation of module 17 (Fig. 5E). A functional enrichment analysis of the genes in module 17 indicated their association with reproductive system development (Fig. 5F). This further corroborates the enrichment of reproduction-related functions within the S1 subpopulation in UCMSCP5 and UCMSCP5L. Similarly, a functional enrichment of genes within modules 3 and 8, specifically upregulated in the S4 subpopulation, indicated roles in vascular development (Figs. 5E and 5F). This provides additional evidence of the enhanced vascular regeneration functions of the S4 subpopulation, particularly in UCMSCP5L. Additionally, we observed that subpopulations S0, S2, S3, and S5 were characterized by the upregulation of genes in modules 12, 16, 10, and 11. A functional enrichment analysis of these gene modules (see Fig. S5) revealed enrichment for similar functions to those of subpopulations S0, S2, S3, and S5 (Fig. 4E).