To explore the effects of aging on WAT, we performed full-length snRNA-Seq analysis on abdominal subcutaneous WAT from 10 Older (≥ 65 years old) and 10 Younger (≤ 30 years old) individuals balanced for sex and BMI (Figure 1A). All individuals were metabolically healthy (Table 1, Supplementary Information 1). WAT biopsies were performed at Translational Research Institute (TRI) for the sole purpose of interrogating cell-specific aberrations with aging (see methods for processing details). The Older group had greater levels of systolic blood pressure, waist circumference, waist-to-hip-ratio, plasma cholesterol, very-low-density lipoprotein, triglycerides, thyroid stimulating hormone and aspartate transaminase vs. the Younger group (P < 0.05); however, all values were in the normal range. There were no differences in BMI, plasma glucose, plasma free fatty acids (FFAs), serum insulin, serum CRP, or adipose tissue insulin sensitivity measured by ADIPO-IR27,28 between the Older and Younger groups, indicative of comparable metabolic health, particularly for the WAT.
Aging human WAT single nuclei atlas
Our dataset included 25,736 nuclei composed of 12 different clusters (Figure 1B) with an average of 5304 genes detected per nuclei. Known cell markers were used for initial annotation (Figure 1C) revealing populations of stem cells, pre-adipocytes, adipocytes, vascular cells, macrophages and mast cells. By transcriptionally profiling the nuclei isolated from WAT, we determined cell composition of adipocytes and non-adipocyte cells using an array of markers that are not limited to pre-specific cell markers. The largest cell type proportion in WAT is adipocytes (~50%) followed by pre-ads (22-24%), vascular cells (~12%), stem cells (10-12%) and immune cells (3-6%) (Figure 1D). Each participant had every cell type quantified except for mast cells (Extended Figure 1A-B). The Older group in comparison to the Younger group had double the proportion of immune cells (Figure 1D) in agreement with previous findings13. There were no major cell compositional differences between males and females (Extended Figure 1C).
Unbiased annotation of cell types
We performed an unbiased annotation of the cell clusters by running DGE analysis on each cluster to identify uncommon cell type markers. The top 5 DEGs for each cluster are highlighted in Figure 2A (Known cell markers of each cluster [i.e. PDGFRA, CD38, LEP, PPARG, PECAM1, MRC1 and TPSAB1] are included as a reference for the main cell types. As anticipated, stem cells had the highest expression of DCN, LUM and COL1A2, vascular cells had the highest expression of VWF, macrophages had the highest expression of HLA-DRA and LIPA, and mast cells had the highest expression of KIT. Classical markers of adipocytes ADIPOQ and PLIN1 were upregulated to varying degrees in all five adipocyte clusters. Classical pre-adipocyte markers such as ZNF423, CD38 and DLK1 were not in top DEGs for each pre-adipocyte population; but rather, uncommon markers such as CTNNA2, RBFOX3, PTPRD and NRXN3 were highly expressed in our pre-adipocyte populations. Previous snRNA-Seq data sets in mouse and human WAT have identified pre-adipocyte populations within a subset of progenitor cells, usually marked by ICAM1 or CD3620,23,29–32; however, these previously identified pre-adipocytes still express known stem cells markers such as PDGFRA, COL4A1, COL1A120,29,31,32 which should diminish upon commitment to the adipogenic lineage33. Therefore, these previously identified pre-adipocyte populations may not be committed pre-adipocytes. In contrast, our pre-adipocyte populations are transcriptionally distinct from our stem cell pool and have reduced expression of stem cell markers (PDGFRA, LUM, DCN; Figure 2A).
To validate our cluster annotations in the pre-adipocyte populations, we compared gene expression profiles to previously published literature. We recently published full-length scRNA-Seq on the stromal vascular fraction (SVF) of abdominal subcutaneous WAT and transcriptionally profiled the commitment of pre-adipocytes from pluripotent stem cells (Whytock et al., 2022). We integrated our current dataset (annotated as BA) with the scRNA-Seq SVF and ran pairwise correlation analysis of the gene expression profiles between the identified cell clusters. Gene expression profiles from our defined cell populations correlated strongly with each other from both data sets, including our pre-adipocyte populations (black dashed box, Extended Figure 2A). Recently, Emont et al.,20 published a single cell atlas of WAT from various mouse and human fat pads and depots, respectively. We filtered the Emont data set to only include human abdominal subcutaneous WAT and compared gene expression profiles of major cell clusters against our data set. Gene expression profiles from major cell populations highly correlated with each other between the two data sets (Extended Figure 2B). Interestingly, our pre-adipocyte populations correlated most strongly with the lymphatic endothelial cell (EC) cluster in the Emont data set (black dashed box, Extended Figure 2B). All three of our pre-adipocyte clusters did not express upregulated endothelial cell makers (PECAM1, VWF, CDH5) or the specific lymphatic EC marker LYVE1. We are therefore confident that our pre-adipocyte populations are not lymphatic ECs and are indeed pre-adipocytes based on expression of expected markers (CD38, ZNF423 and DLK1)34–36. Both Emont et al.20 and a recent meta-analysis of human scRNA-Seq data 25 had more defined immune cell populations (NK/T/B cells) in comparison to our data set, which is likely driven by the manner in which we procure our abdominal subcutaneous WAT samples. Ours24 are the only human single cell/nuclei RNA-Seq data generated from subcutaneous WAT obtained via lipoaspiration of the superficial subcutaneous depot rather than during surgery, the latter of which likely induces tissue inflammation.
Pre-Adipocytes
The top five DEGs for each of the three pre-adipocyte clusters (Figure 2A) have established roles in cell development including cell-to-cell adhesion, proliferation and differentiation which have primarily been shown in synapse/neuronal cells37–49. It is therefore plausible that these novel genes play important yet understated roles in pre-adipocyte proliferation, commitment, and differentiation, particularly in humans. CTNNA2 was recently identified as a gene with a single nucleotide polymorphism in polyadenylation signals upregulated in mouse lines for fatness50. PRPTD has been identified in pre-adipocyte populations, but there is conflicting evidence on whether it contributes to or downregulates adipogenesis51,52. NELL1 inhibits adipogenesis in 3T3-L1 and adipose-derived stem cell lines53. CDH4 is among differentially expressed genes in pre-adipocytes compared to bone-resident progenitor cells and has been identified as negative regulator of UCP154.
To characterize these pre-adipocyte populations further, transcription factor (TF) enrichment analysis was performed on the top DEGs (logfc >0.5) from all pre-adipocytes using ChEA3 55 (Supplementary Table 2). Top ranking TFs include: 1) SATB2 which interacts with and increases transcriptional activity of pre-adipocyte marker ZNF42356; 2) ZNF365 which increases in expression during differentiation in pre-adipocytes57; and 3) CTCF a multifactorial protein required during adipogenesis58. Therefore, our pre-adipocyte populations express genes that can be transcriptionally regulated by known TFs during pre-adipocyte differentiation.
Cadherin-Associated Protein Alpha 2 (CTNNA2) was among one of the highest expressed novel pre-adipocyte genes and was ubiquitously expressed across the three pre-adipocyte populations. We validated this particular pre-adipocyte marker at the protein level using immunofluorescence staining of SVF cells collected from abdominal subcutaneous WAT (Figure 2B). We co-stained CTNNA2 with known pre-adipocyte marker CD3836. CTNNA2 protein was present on the majority of SVF cells but to varying degrees. CTNNA2+ cells with the highest CTNNA2 abundance were also CD38+ positive cells, confirming cells with high CTNNA2 expression are pre-adipocytes (Figure 2B).
DGE and pathway analysis between the three pre-adipocyte populations (Supplemental Table 3) revelated Pre-Ad 1 had an upregulation of genes encoding ribosomal proteins (RPS6, RPL19) indicative of greater cell proliferation59; whereas, Pre-Ad2 and 3 had an upregulation of gens related to synaptic and cell-to-cell signaling that was more pronounced in Pre-Ad 3.There were no proportional differences among the pre-adipocyte populations between the Older and Younger groups (Figure 2C). Mitochondrial capacity is essential for preadipocyte proliferation and differentiation60 and can be blunted in WAT of metabolically unhealthy individuals61,62. Differences in mitochondrial capacity are more pronounced in pre-adipocytes compared to in vitro differentiated adipocytes63, as putative differences between groups can be diminished following induction with an adipogenic cocktail. Using high-resolution respirometry, we observed no differences between the Older and Younger groups in basal, leak nor maximal (uncoupled) respiration (electron transport system, ETS) of cultured pre-adipocytes (Figure 2D), suggesting that mitochondrial capacity in WAT is not attenuated with aging.
Cell sub-clustering
We resolved adipocyte and pre-adipocyte heterogeneity in our initial clustering analyses. To resolve our remaining cell types further, we performed additional sub-clustering on our macrophage, stem cell and vascular populations. Macrophages sub-clustered into two distinct clusters (Supplementary Figure 1A). Macrophage_0 had markers of resident macrophages (MRC1, PDGFC); whereas Macrophage_1 had markers of lipid-associated macrophages (LIPA, TREM2, Supplementary Figure 1B)64. The increased proportion of immune cells in the Older group was due to a greater proportion of resident type macrophages (Macrophage_0) compared to the Younger group (Supplementary Figure 1C). There were no proportional differences in Macrophage_1 between the Older and Younger groups (Supplementary Figure 1C).
Stem cells sub-clustered into three stem cell populations (Supplementary Figure 2A). Stem_0 had upregulation of pre-adipocyte markers compared to other stem cell populations (LSAMP, PDE4D) indicative of committing to the adipogenic lineage. Stem_1 was defined by nascent stem cell markers (DCN, LUM, COL1A1) and Stem_2 had upregulation of genes associated with lipid metabolism (PLIN4, G0S2; Supplementary Figure 2B). There was a trend towards Stem_0 having a higher proportion in the Older group when expressed as a proportion of total nuclei (Supplementary Figure 2C). There were, however, no differences in the proportions of Stem_1 or Stem_2 between the Older and Younger groups.
Vascular cells subclustered into four cell populations (Supplementary Figure 3A). Vascular_0 had upregulation of lipid metabolism genes (CIDEC, G0S2, PLIN4) defining a fatty-acid handling endothelial phenotype as previously described21,24. Vascular_1 had markers of pre-ads (DPP10, RBFOX1, PTRPD) indicative of being a vascular originating progenitor cell65,66. Vascular_2 had upregulation of classical markers of endothelial cells (PECAM1, VWF). Vascular_3 had markers of pericytes (PDGFRB, NOTCH3; Supplementary Figure 3B). There were also no differences in proportion of the each vascular subcluster between Older and Younger groups (Supplementary Figure 3C).
Adipocyte Heterogeneity
We identified five distinct adipocyte populations and sought to identify characteristics of each cluster. We performed DGE analysis comparing each adipocyte cluster to the remaining adipocyte clusters (Supplementary Table 4). We characterized Adipocyte 1 as “Nascent” adipocytes due to an upregulation of genes observed in pre-adipocyte populations (PTRPD, DPP10, NRXN3) and lower expression of classical adipocyte markers (ADIPOQ and LEP; Figure 3A). Adipocyte 2 had an upregulation of genes associated with adipocyte growth (PPARG, MALAT1 and NEAT1)67,68in addition to insulin signaling genes (INSR and IRS1) and PDE3B which suppresses lipolysis69 and was therefore classified as “Immature” adipocytes (Figure 3A). Adipocyte 3 had upregulation of lipid synthesis genes (GPAM, SCD, DGAT), lipolytic genes (PNPLA2) and the classic adipokine ADIPOQ (Figure 3A). These features are characteristic of a “Mature” adipocyte carrying out routine adipocyte functions. Adipocyte 4 had upregulation of oxidative (ATP5F1B, ATP5MF and COX5A) and glutathione peroxidase (GPX1,3-4) genes but not typical browning or beiging markers and was therefore classified as “Oxidative”. Adipocyte 5 was defined as a “Remodeling” adipocyte due to upregulation of lysosomal proteases, cathepsins (CTSD and CTSS), complement factor (C3)and chemokines (CXCL12 and CXCL14). There were comparable adipocyte proportions between the Older and Younger groups (Figure 3B). When divided by sex, males tended to have a higher proportion of Adipocyte 5 (Remodeling) and a lower proportion of Adipocyte 4 (Oxidative) compared to females (Extended Figure 3A).
The adipocyte continuum
We assessed whether the different adipocyte populations were end-fate phenotypes with cell-specific biological functions or rather a continuum of biological processes through time and the impact of age. We performed pseudotime analysis on the adipocyte populations after re-clustering (Figure 3C), which revealed a clear one-dimensional trajectory initiating at Adipocyte 1 (Nascent) and terminating at Adipocyte 5 (Remodeling) (Figure 3D). The pseudotime visually displayed a similar trajectory when split into Older and Younger groups (Extended Figure 3B). The Older group had a lower proportion of nuclei in the first pseudotemporal bin (<5 pseudotime score; 30% vs. 34%) and higher proportion of nuclei in the last psuedotemporal bin (>20 pseudotime score; 39% vs. 35%) compared to the Younger group (Extended Figure 3C), indicating a shift towards later stages of the adipocyte continuum in the Older group. Unsupervised hierarchical gene clustering identified four modules of genes (Supplementary Table 5) whose patterns changed throughout the pseudotime (Figure 3E, Supplementary Figure 4A), with similar patterns observed when split by age (Extended Figure 3D, Supplementary Figure 4B-C). The module analysis revealed that adipocyte evolution initiates with early development, followed by maturation and lipid regulation and ends with remodeling (Figure 3E, Supplementary Figure 4). We therefore conclude that the observed adipocyte heterogeneity (i.e. five distinct adipocyte populations) is due to different developmental stages throughout time rather than end-fate phenotypes with cell-specific functions. Importantly, the biological processes of the adipocyte continuum that explain adipocyte heterogeneity are largely retained with aging; however, the Older group has more adipocytes in the later stages of remodeling, and the Younger group has more adipocytes in the early developmental stage. While the composition and trajectory of the modules were similar between groups (Supplementary Figure 4), some of the key temporally-regulated genes from each module showed disparity along the pseudotime with aging (Figure 3F, Supplementary Figure 5). Lipid homeostasis genes (e.g. PPARG, SRSF6, SCD, RBP4) had elevated expression in the Younger group, whereas remodeling genes (e.g. CLU, CFD) were elevated in the Older group (Figure 3F, Supplementary Figure 5). Taken together, our data indicate that the Younger group have a greater potential to develop adipocytes with more metabolically active properties, whereas the Older group have more adipocytes actively remodeling.
Adipocyte Size
Differences in adipocyte size were determined by histology70 with an average of 158 adipocytes measured per participant. Average adipocyte diameter (μm) was not statistically different between the Older and Younger groups (Figure 3G). Distribution frequency analyses revealed that the Older group displayed a greater frequency of larger adipocytes (>90 μm) and a reduced frequency of smaller adipocytes (<70 μm) compared to the Younger group (Figure 3H, Extended Figure 3E).
Adipokines
We quantified key adipokines (adiponectin, leptin and resistin) in the plasma and conditioned media (CM) of WAT explants (Extended Figure 3F). There was a trend towards leptin concentrations being higher in the CM from the Younger Group compared to the Older Group (P = 0.09, Extended Figure 3F) and no differences between groups for CM adiponectin and resistin. Plasma resistin concentrations were higher in the Older Group compared to the Younger group (P < 0.05) with no differences plasma adiponectin and leptin Extended Figure 3F). Adipokine concentrations in plasma and CM had expected correlations with adipocyte size and BMI (Supplementary Figure 6).
Pre-adipocyte differentiation
We next sought to track pre-adipocyte differentiation in vivo using a pseudotime trajectory. We re-clustered the pre-adipocyte populations with Adipocytes 1 (Nascent) and 2 (Immature) which are in the earlier stages of adipocyte development (Figure 4A). The pseudotime trajectory initiates in Pre-ad 1, branches off to Pre-ad 2 or 3, with Pre-ad 3 then converting towards Adipocyte 1 and Adipocyte 2 (Figure 4B). Similar pseudotime trajectories and similar proportions of nuclei in each pseudotemporal bin were observed between Older and Younger groups (Extended Figure 4A-B). Unsupervised hierarchical gene clustering identified four modules of genes (Supplementary Table 6) whose patterns changed throughout the pre-adipocyte differentiation pseudotime (Figure 4C, Supplementary Figure 7A), with similar patterns observed when split by age (Extended Figure 4C, Supplementary Figure 7B-C). The module analysis revelated that pre-adipocyte differentiation initiates with adipocyte development, transitions with adipocyte maturation and stabilization and has a fluctuation of inflammatory responses throughout (Figure 4C, Supplementary Figure 7). To validate our pre-adipocyte differentiation pseudotime, we plotted known markers of pre-adipocyte differentiation (CEPBA, CEBPB & PPARG; Figure 4D). As expected, early regulators of pre-adipocyte differentiation CEBPA and CEBPB increased during the early phases of the pseudotime (Figure 4D), while PPARG continues to increase throughout the pseudotime (Figure 4D). Both CEBPB and PPARG had a blunted response in the Older group (Figure 4D). Key genes from Module 1-3 showed little disparity with aging when transitioning along the pseudotime (Supplementary Figure 8). Due to module 4 expressing genes that align with commitment of pre-adipocytes to the adipogenic lineage, we mapped these novel markers along the pseudotime (Figure 4D). Known committed pre-adipocyte marker ZNF423 was used as reference. Several of these genesdisplayed a similar pattern to ZNF423 exhibiting a rapid increase at pseudotime score 7-8 corresponding to pre-adipocyte commitment (Figure 4D). Interestingly, the expression of the genes declined as the pseudotime transitioned to a score of 10-15 which corresponds to a Nascent adipocyte being formed and where a rapid increase in PPARG is observed (Figure 4D). This decline was dampened in the Younger group which retained elevated expression of several of these genes (IL1RAPL1, IL1RAPL2, PTPRD, NRXN3 & DPP10) as the pseudotime progressed (Figure 4D). We propose that previously assigned neuron/synapse genes play an essential role in forming committed pre-adipocytes and early adipocyte development and the elevated expression in the Younger group indicates a more poised state for differentiation during early development.
To further validate a subset of these novel pre-adipocyte markers, we tracked the expression levels during differentiation in vitro. Cultured pre-adipocytes derived from digested human subcutaneous abdominal WAT were differentiated using an established adipogenic cocktail71 and cells were harvested for RT-qPCR at time-points Day 0, Day 0 + 4hrs, Day 2, Day 4, Day 9 and Day 12 (Extended Figure 4D). To match the in vivo pseudotime trajectory, Days were converted to hours and gene expression patterns were modelled throughout the differentiation period. As anticipated, known adipogenesis markers CEBPA and PPARG increased with differentiation (Figure 4E). Novel pre-adipocyte markers from Module 4, CDH4, SHANK2 and IL1RAPL1 had their highest expression levels at time point 0 before the differentiation cocktail was added and when they would still be classified as pre-adipocytes. This aligns with a pseudotime score 7-8 corresponding to pre-adipocyte commitment in vivo. CDH4, SHANK2 and IL1RAPL1 expression rapidly decreased upon addition of the differentiation cocktail coinciding with the cells rapidly transitioning to adipocytes, with no discernable differences between the Older and Younger groups. Our data confirm that these are pre-adipocyte markers are temporally regulated during pre-adipocyte differentiation. We posit, however, that nuances in gene expression levels between groups are better detected using an in vivo pseudotime approach rather than artificial induction in vitro.
Cell-Type transcriptional differences between Older and Younger groups
Random Forest Classification
Random forest classification72 was performed on each cluster to determine the genes most important in classifying an Older and Younger cell for that specific cluster (Supplementary Table 7). The top genes based on a Mean Decrease Gini above 3 are shown in Figure 5A. Out of the 12 separate clusters, the five Adipocyte clusters, Pre-adipocyte 1, Vascular cells and Stem cells showed the highest accuracy (> 0.8) of random forest classification (Figure 5B, Extended Figure 5A), meaning that cells from Older and Younger individuals can be classified more accurately if selected at random from these clusters. Among the top genes contributing to this high accuracy that were also upregulated in Older individuals were DPT and CXCL14. DPT regulates extracellular matrix (ECM) formation and is upregulated in WAT in obesity-induced T2D73. Treatment of isolated adipocytes with DPT increases gene expression of ECM regulators (COL6A3, ELN, MMP9, TNMD) and inflammatory markers (IL6, IL8, TNF)73. CXCL14 is notable for its chemotactic properties74 but has recently been identified as a mediator of macrophage communication in brown adipose tissue75, as well as a regulator of insulin-mediated glucose uptake in 3T3-L1 adipocytes76. Together, these genes indicate an enhanced immune response in cells from the Older group. Among the top genes contributing to high accuracy that were also upregulated in Younger individuals were genes related to lipid metabolism (SCD, FASN, G0S2, ACSL1 and ADH1B) and adipogenesis (PPARG), indicative of active routine regulatory processes typical in WAT. Macrophages and Mast Cells had the lowest accuracy scores suggesting lower transcriptional diversity between Younger and Older groups for these cell types (Extended Figure 5A).
Differential gene expression
We performed DGE analysis comparing Older and Younger groups for each cell cluster (Supplementary Table 8). The Older group had the highest number of upregulated genes among the cell clusters (Figure 5C) except for Adipocyte 1 (Nascent adipocyte). Pathway analysis revealed the main pathways upregulated in the Older group were related to innate immune response and complement activation and were observed to the greatest extent in in Pre-Ad 1, Adipocyte 1 (Nascent adipocyte), Adipocyte 4 (Oxidative adipocyte) and Adipocyte 5 (Remodeling adipocyte) (Figure 5D). Genes contributing to immune response and complement activation pathways are shown in Figure 5E. In addition to bona fide immune cells, adipocytes themselves are now recognized as key innate immune cells within adipose tissue77. The data suggests that in the Older group, these pre-adipocyte and adipocyte populations are initiating an innate immune response via alternative complement pathway activation. As part of the innate immune response, these cells undergo remodeling of ECM (ADAMTS5, VIM, HTRA1, MFAP4, SPP1, TIMP1), phospholipid membrane (PLD3), lipids (APOD, APOE,) actin cytoskeleton (ACTG1, GSN, LSP1, MYO1C) and carbohydrate binding structure (LGALS1, RNASE1) that is in part mediated by lysosomal activity (LSAM1, LAPTMS) and ROS regulation (ROMO1). In the Younger group, the main pathways upregulated were related to neurogenesis and synapse organization in Adipocyte 1 (Extended Figure 5B). These genes have previously been attributed to the development of neurons; however, given their upregulation in Adipocyte 1 (Nascent) and in agreement with our pre-adipocyte differentiation pseudotime, we again reason that these genes are early regulators of pre-adipocyte differentiation.
Senescence
Cellular senescence is caused by accumulation of DNA damage78 and other cellular stressors79–81, leading to enlarged cells that are resistant to regulated cell death 82 and secrete chemokines, cytokines, growth factors, and matrix metalloproteinases 83 that can negatively impact the function of other cells in the microenvironment11,84. Senescence typically occurs in proliferating cell types but has also been found in terminally differentiated cells85,86 including adipocytes9,87. Senescent cells activate innate immunity to assist in the removal of nearby dying cells10,88. Given that the Older group has larger (ie, hypertrophic) adipocytes, upregulated innate immunity specifically in adipocyte populations and increased proportions of adipocytes in the end-stage of the adipocyte pseudotime, we reasoned that the Older group would have increased levels of senescence, particularly among their adipocytes. p16 (CDKN2A) and p21 (CDKN1A) are canonical markers of senescence but are unreliable in detecting senescence at the transcriptional level due to very low expression. To measure senescence profiles in our snRNA-Seq data set, we performed GSEA on each nucleus using the recently published SenMayo gene set which identifies the transcriptional profile of senescent cells with a combination of 125 genes89 (Extended Figure 6A). In line with Saul et al.89, we determined that the top 10% of cells could be classified as those with a senescent profile. The Older group had a significantly greater proportion of cells with a senescent profile in Adipocyte clusters 2, 5 and Pre-Adipocyte 1 and a trend towards a greater proportion of cells with a senescent profile in Adipocyte clusters 1 and 4 (Figure 6A, Extended Figure 6B). It was notable from our data set, that the only chemokine or cytokine upregulated in clusters that have an increased proportion of cells with a senescent profile was CXCL14 (Figure 6B). This gene was also included as high importance in distinguishing between Older and Younger individuals in our prior random forest classification (Figure 5A).
To confirm Older individuals had a greater proportion of cells with a senescent profile, we performed immunohistochemistry staining for the classic senescent marker p16. Abundance of p16+ cells was quantified as the percentage of 1) total fields (20x magnification) analyzed containing p16+ cells9 and 2) p16+ adipocytes. The Older group had significantly more fields containing p16+ cells than the Younger group (Figure 6C-D). The percentage of fields containing p16+ adipocytes was also greater among Older individuals (Figure 6C), suggesting the increased senescence profile in the Older group is driven by adipocytes and aligns with the snRNA-seq data (Figure 6A-B).
Given that CXCL14 may mediate macrophage infiltration90 and that our snRNA-seq data show a greater number of resident macrophages in the Older group, we quantified macrophage content via immunohistochemistry using a pan-macrophage marker CD6891. The numbers of macrophages quantified with CD68 per frame and per adipocyte were not statistically significant between the Older and Younger groups (Figure 6E). Three participants had crown-like structures of macrophages (Figure 6F), which form around dying or damaged adipocytes92. All of these individuals were male and two of them were from the Older group. Obesity is associated with macrophage infiltration91. In agreement, we noted a significant and positive correlation between a marker of abdominal obesity, waist-hip-ratio (WHR), and the number of macrophages per adipocytes (Extended Figure 6C). WHR was greater in the Older group (Table 1),and this was driven specifically by Older males (Extended Figure 6D).We therefore reason that macrophage infiltration occurs in Older males with higher WHR that can lead to crown-like structures developing around damaged adipocytes, perhaps as a means to maintain healthy WAT.