Aging modulates the effects of ischemic injury upon mesenchymal cells within the renal interstitium and microvasculature

Abstract The renal mesenchyme contains heterogeneous cells, including interstitial fibroblasts and pericytes, with key roles in wound healing. Although healing is impaired in aged kidneys, the effect of age and injury on the mesenchyme remains poorly understood. We characterized renal mesenchymal cell heterogeneity in young vs old animals and after ischemia‐reperfusion‐injury (IRI) using multiplex immunolabeling and single cell transcriptomics. Expression patterns of perivascular cell markers (α‐SMA, CD146, NG2, PDGFR‐α, and PDGFR‐β) correlated with their interstitial location. PDGFR‐α and PDGFR‐β co‐expression labeled renal myofibroblasts more efficiently than the current standard marker α‐SMA, and CD146 was a superior murine renal pericyte marker. Three renal mesenchymal subtypes; pericytes, fibroblasts, and myofibroblasts, were recapitulated with data from two independently performed single cell transcriptomic analyzes of murine kidneys, the first dataset an aging cohort and the second dataset injured kidneys following IRI. Mesenchymal cells segregated into subtypes with distinct patterns of expression with aging and following injury. Baseline uninjured old kidneys resembled post‐ischemic young kidneys, with this phenotype further exaggerated following IRI. These studies demonstrate that age modulates renal perivascular/interstitial cell marker expression and transcriptome at baseline and in response to injury and provide tools for the histological and transcriptomic analysis of renal mesenchymal cells, paving the way for more accurate classification of renal mesenchymal cell heterogeneity and identification of age‐specific pathways and targets.


| INTRODUCTION
Acute kidney injury (AKI) occurs in approximately 1 in 5 hospital admissions, 1 and can leave patients with varying degrees of renal fibrosis-an important contributor to the transition to chronic kidney disease (CKD). In addition, the likelihood of a fibrotic outcome post AKI and of progression to CKD increases with age. [2][3][4][5][6][7] It is therefore crucial to understand the mechanism of fibrosis following renal injury and the age-associated factors that drive this.
Bilateral ischemia reperfusion injury (bIRI) is a common model of AKI, 8 but the long-term effects on the aged kidney cannot be easily assessed in this model as aged mice tolerate such severe injury poorly leading to a high mortality. 2,9 In contrast, unilateral ischemia reperfusion injury (uIRI) 8 transiently interrupts blood flow to one kidney but leaves the contralateral kidney unaffected, thus enabling the study of more unilaterally severe injury than the models of bIRI or uIRI with contralateral nephrectomy. 10 Importantly, this facilitates the observation of long-term fibrotic end-points, increasing the relevance to patients. 11 Although AKI results in widespread tubular cell injury, the effect upon the interstitium is key for overall kidney outcome: rarefaction of peritubular capillaries with subsequent tissue hypoxia exacerbates the loss of nephrons, 12 while a progressive fibrotic phenotype in the interstitium is the hallmark of CKD. 13 These processes are influenced by mesenchymal perivascular/interstitial cells which support the vasculature 14 and contribute substantially to myofibroblast generation and expansion. 15 However there has been little research into how age affects perivascular/ interstitial progenitor cells, 16,17 and crucially no research to date into the effect of age on the perivascular/interstitial cell response to injury.
There are two broad types of mesenchymal cells present in the interstitium. "Pericytes" enwrap the microvascular endothelium and are embedded in the capillary basement membrane. 18,19 "Interstitial fibroblasts" are embedded in and structurally maintain the collagenous extracellular matrix (ECM) of the interstitium. [20][21][22][23] These fibroblasts are necessarily in close proximity to the capillaries, but are less intimately associated than pericytes. There are likely many subpopulations of pericytes and interstitial fibroblasts, such as the "perivascular fibroblasts" that reside in the collagenous matrix around larger vessels. 24 In the text below the term "interstitial cell" refers to all mesenchymal cells in the interstitial compartment.
Their heterogeneity in expression is not well characterized, and is indicative of an underlying functional heterogeneity. 17 Kidney pericyte functional heterogeneity has been demonstrated previously: Gli1 + pericytes/perivascular fibroblasts give rise to the majority of myofibroblasts following kidney injury, 25 and there is a pericyte subset that produce renin. 26 CD146 is reportedly a ubiquitous human pericyte marker, [27][28][29] but is poorly characterized in murine kidney. Thus, more detailed characterization of the interstitium is required given the central role of mesenchymal cells in renal injury and recovery.
This work characterizes the interstitial distribution of common perivascular cell markers in the young and aged murine kidney and tests the effects of injury and age on mesenchymal cell phenotypes. This is achieved by multiplexed immunolabeling, which reveals the relative spatial distribution of multiple surface markers within the complex renal architecture, in combination with single cell transcriptomic technology, which provides high-dimensional information on gene expression within individual cells and allows unbiased clustering into transcriptionally distinct subpopulations. In this work these two technologies independently identify similar subpopulations within the renal interstitium and provide insight into their functional properties through their anatomical localization, their reaction to age and ischemia, and through their gene expression profile.

| Animals and surgery
Interstitial cell quantification in the cortex and inner stripe was performed on male FVB mice from the National Institute of Aging colony (Charles River, Boston, USA) that were either young (3-5 months) or old (18 months

Significance statement
The mesenchymal cell compartment plays a key role in kidney disease, but the varied cell types within are poorly defined and the effect of aging on mesenchymal cells is incompletely understood. Here, for the first time the authors perform histological analysis of common mesenchymal markers with accompanying transcriptomic profiling on young and old mice following unilateral ischemia reperfusion injury. This results in a more refined understanding of mesenchymal marker expression as they align with cell subtypes within the mesenchymal compartment. Age associated changes in mesenchymal populations are identified, furthering our understanding of the differences in injury response that occur with age.

| Human tissues
Human kidney tissue was collected with prior written informed consent. Ethical approval for the use of human tissues in research was obtained from the South East Scotland Research Ethics Committee.
Tissue was obtained from uninephrectomy operations that were performed following the detection of renal carcinoma. Plugs were taken at maximum distance from neoplasms from the cortex and outer medulla regions. These were fixed in formalin and paraffin embedded, before processing for immunofluorescent staining as described.

| Histopathology
Histopathology was performed as previously described. 3 Kidney halves were fixed overnight in methacarn before paraffin embedding. Five micrometer sections were stained using hematoxylin and eosin, or picrosirius red. For acute tubular necrosis (ATN) scoring at day one, 4-8 random fields in the outer stripe at 20× magnification were acquired from H&E stained sections. Images were blinded, randomized, and the proportion of tubules with evidence of tubular death was quantified. Fibrosis quantification was performed on eight random fields in the outer stripe of picrosirius red stained sections. % red-positive area was quantified using ImageJ software.

| Immunofluorescence staining
All immunofluorescence staining was performed on methacarn-fixed paraffin embedded slides as previously described. 3 The choice of endothelial cell marker (CD31 or CD34) in each case depended on technical considerations such as species of antibody and brightness of labeling. Full details of protocols in Supplementary Methods.

| Cell quantification
Positive cell nuclei were putatively identified automatically via colocalization of antigen signal and DAPI using ImageJ software, followed by manual verification. Only nuclei in the interstitium or on the outer surface of capillaries were deemed positive. Areas of outer stripe were digitally extracted from scans of stained sections. These were either analyzed in their entirety or had 6-8 high power fields digitally extracted from random locations. Total area quantified for CD146 and PDGFR-β dual-labeling was 0.89 mm 2 per mouse, for α-SMA and NG2 was 0.17 mm 2 , and for PDGFR-α and -β was 0.09 mm 2 ; cell numbers were normalized by total area analyzed.
Quantification in the cortex, and outer and inner stripe analyzed 0.35 mm 2 per region per mouse; cell numbers were normalized by total cross-sectional area of vasculature analyzed, as determined by CD34 labeling.

| Statistical analysis
Data are presented as mean ± SD, or in the case of ratios geometric mean ± 95% CI. For single comparisons in Figure S7B  A substantial population of PGDFR-β + CD146 − NG2 − α-SMA − cells were present. They were distributed in the interstitial ECM and did not show the same intimate association with vessels as CD146 + PDGFR-β − and CD146 + PDGFR-β + pericytes (Figures 1A-C and 2A, and S1-S3) indicating that PDGFR-β + cells lacking CD146 expression are interstitial fibroblasts. NG2 expression was occasionally observed on these interstitial fibroblasts (PDGFR- Based on these observations, labeling for CD146 and PDGFR-β alongside an endothelial cell marker, such as CD31, effectively identifies pericytes and distinguishes them from interstitial fibroblasts in the baseline murine kidney. In endothelial marker negative cells, CD146 + (±PDGFR-β + ) staining identifies pericytes, while a CD146 − PDGFR-β + phenotype identifies interstitial fibroblasts (summarized in Figure S5).

| Reductions in pericyte and fibroblast numbers in the renal interstitium occur with age
Reductions in vascular area and pericyte numbers with age have been reported. 16 To determine the effects of age on perivascular/interstitial cell numbers, kidney sections from young (3-5 months) and old (18 months) mice were triple-labeled for PDGFR-β, CD146, and the endothelial cell marker CD34 43,44 (Figure 2A,B) and the relative abun- and CD146 + PDGFR-β + (CD34 − ) interstitial cells were quantified in the cortex, outer stripe, and inner stripe regions ( Figure 2C-F).
CD146 + PDGFR-β + pericytes were a minority subset of each population, namely 27-33% of the CD146 + population and 7-28% of the PDGFR-β + positive population ( Figure S6). This indicates that the major portion of cells in the total PDGFR-β + interstitial population are not pericytes but rather interstitial fibroblasts. This analysis revealed a general increase in CD146 − PDGFR-β + interstitial cells (ie, interstitial fibroblasts) from cortex to inner stripe ( Figure 2E). The number of CD146 − PDGFR-β + interstitial fibroblasts in the inner stripe decreased from young to aged kidneys ( Figure 2E). There was also a decrease in CD146 + PDGFR-β + and CD146 + PDGFR-β − cells (ie, pericytes) in the cortex and inner stripe, but not the outer stripe ( Figure 2D,F). These results indicate that there is a loss of pericyte coverage on renal T A B L E 1 Properties of perivascular cell surface markers and their previous use in murine renal studies Interstitial surface marker Other names Description Use in murine renal studies

PDGFR-α CD140a
Receptor for PDGF-A, -B, and -C. 31 Linked to a fibrotic interstitial cell phenotype in studies of muscle. 32,33 Studies of glomerular and interstitial fibrosis 34

PDGFR-β CD140b
Receptor for PDGF-A, -B, and -D. 31 Commonly used pericyte marker in multiple tissues including kidney. Involved in pericyte recruitment during angiogenesis/ vasculogenesis. Brain studies suggest that PDGFR-β dependent binding to CD146 in pericyte progenitors facilitates their coverage of endothelial cells. 35 Common pericyte marker. Studies of glomerular and interstitial fibrosis 34

NG2
Chondroitin sulfate proteoglycan 4 Proposed roles in detecting extracellular matrix components and relaying signals to the cytoskeleton. 40 Necessary for full pericyte coverage of retinal vessels. 41 In human it is specifically not expressed on venular pericytes. 27 Expression lost during pericyte quiescence and regained upon stimulation (eg, following injury). 24 Common pericyte marker α-SMA α-actin-2 Role in cell contraction. Presence on stromal cells indicates a collagen producing myofibroblast. 42   Log ratios significantly different from one by one sample t test indicated with asterisks: *P < .05; **P < .01. Bars show geometric mean ± 95% CI. N = 4-10 per group. Con., contralateral kidney; IRI, ischemic kidney vessels with age in the cortex and inner stripe, along with a loss of interstitial fibroblasts in the inner stripe.

| Initial injury is not significantly worse in aged vs young mice following severe unilateral ischemia/ reperfusion
Following 25 minutes of warm uIRI young and old mice were culled at one-and 28 days post-ischemia. Subsequent investigations unless otherwise indicated focused on the outer stripe of the outer medulla, as this is the region most susceptible to tubular cell injury following ischemia and relatively little injury was observed in the cortex and inner stripe. ATN scoring at day one post-IRI indicated that a significant injury was inflicted ( Figure S7A,B). However, there was no difference in ATN score between the ischemic kidneys of old and young mice ( Figure S7B), indicating that the degree of initial tubular injury was equivalent between ages.

| Old animals have increased interstitial fibrosis compared with young both before and after unilateral ischemia reperfusion injury
To test whether old animals have a greater fibrotic response following a similar initial ischemic injury, young and old animals were culled at 28 days post-IRI and fibrosis was quantified in the outer stripe by picrosirius red staining ( Figure S7C-E). There was significantly more fibrosis in the old kidneys of the contralateral and ischemic groups ( Figure S7D,E). There was no obvious change in fibrosis in contralateral kidneys between days one and 28 ( Figure S8).
CD146 − PDGFR-β + interstitial fibroblasts were significantly increased in ischemic kidneys of young and old animals at day 28 compared with contralateral kidneys and day-one ischemic kidneys ( Figure 3G-I). No difference was detected between ages ( Figure 3G).
In young ischemic kidneys strong CD146 labeling was observed around denuded tubules (identified morphologically and by lack of nuclei) at day 1 ( Figure 3C). In old animals CD146 was not so obviously activated and localized ( Figure 3D). CD146 + pericytes may be separated into subtypes with or without PDGFR-β expression.
CD146 + PDGFR-β − pericyte numbers were unchanged in ischemic compared with contralateral kidneys in both age groups ( Figure S9B, C), however young animals exhibited a marked increase in CD146 + PDGFR-β + pericytes at day one in ischemic kidneys in contrast to old ( Figure 3J,K). CD146 + PDGFR-β + pericyte numbers in ischemic kidneys were equivalent to contralateral levels at both ages by day 28 ( Figure 3L). These data demonstrate a transient increase in a CD146 + PDGFR-β + pericyte subtype early in the wound healing response of young animals that is absent in old animals.
At 28 days post-IRI, NG2 + α-SMA +/− cells were significantly more abundant in old ischemic kidneys than at baseline and were significantly higher in number than in corresponding young kidneys ( Figure 4D,E), suggesting a more active pathological phenotype in old kidneys following ischemic injury. Supporting this idea, there was an increase in NG2 expression in the inner medulla of old kidneys at day 28 post-IRI that was absent in young kidneys ( Figure S12), indicating that activation of NG2 expression was more widespread in old kidneys.
Old ischemic kidneys had significantly more α-SMA + myofibroblasts than young ( Figure 4F) suggesting a more active fibrotic phenotype. Furthermore, in day-28 post-IRI ischemic kidneys NG2 + α-SMA +/− cell numbers correlated more closely with fibrosis area than α-SMA + myofibroblast numbers ( Figure S10D-F). Previous work showed that not all α-SMA + cells express collagen, 24 and α-SMA expression on other renal cell types such as macrophages. 46 We thus asked whether more specific myofibroblast markers exist.

PDGFR-α was observed mainly on interstitial fibroblasts, whereas
PDGFR-α − -β + cells only occurred rarely and in perivascular locations ( Figure 5A,B). This suggests that PDGFR-α + -β +/− cells are interstitial fibroblasts, whereas PDGFR-β + cells that lack PDGFR-α expression are pericytes. PDGFR-α + -β + cells increased significantly with ischemia at both ages and correlated more strongly with fibrosis than α-SMA + cells ( Figures 5G,H; S10D,I), suggesting they may together label fibrosisproducing myofibroblasts. There was substantial, but not absolute, colocalization of PDGFR-α and α-SMA in injured kidneys ( Figure S11E). In contralateral kidneys the PDGFR-α − -β + pericyte population tended to be lower in old than young animals and was significantly lower following ischemia, consistent with a loss of pericytes with age ( Figure 5E,H).
In contralateral kidneys the PDGFR-α + -β − population was significantly higher in old than young ( Figure 5F,H). Interestingly, young day-28 ischemic kidneys had also acquired a population of PDGFR-α + -β − cells, suggesting PDGFR-α + -β − cells are a subtype of interstitial fibroblasts arising in response to pathological stimuli ( Figure 5F,H). This suggests a persistent activated state of interstitial cells 28 days post-IRI and in aged kidney at baseline.  Figure 6D, Table S2).  post IRI we found Pdgfra and Pdgfrb co-localized only in the high collagen-expressing myofibroblast population, whereas Acta2 was most highly expressed in the pericyte/vSMC population ( Figure 7D-F).

| DISCUSSION
The renal mesenchyme has vast importance in disease and consists of a complex assortment of cells with disparate roles and biologies. 17 Despite this there is a lack of understanding, and awareness, of renal mesenchymal heterogeneity that inevitably has a detrimental impact on the design and interpretation of studies. It is therefore imperative to thoroughly characterize this cell compartment and provide tools for its effective study. Here we have identified mesenchymal subtypes both transcriptomically and histologically, developed methods to distinguish these subtypes histologically, and determined subtype localization at baseline and with injury and aging.
Using scRNAseq, we have identified three major subtypes of renal mesenchymal cells, namely pericytes/vSMCs, fibroblasts, and myofibroblasts. The same populations were identified in two independently performed sequencing experiments which encompass physiological aging (TMS dataset) 30 and the effect of injury (IRI dataset).
Importantly these studies were performed without pre-selection based on a transgene or surface marker, allowing us to avoid selection bias due to incomplete marker coverage (a problem highlighted by our histology studies) or variations in transgene expression. Our analysis represents an improvement on previous unbiased approaches, where only a generic "fibroblast" population was identified in mouse wholekidney scRNAseq, and "pericyte/vSMC" and "interstitial" populations were identified in human single nucleus RNA sequencing. 47,48 Our ability to identify myofibroblast populations was likely due to our inclusion of aged and injured kidney tissue.
Many previous studies in kidney have used PDGFR-β as a marker of "pericytes." 14,15,17,24,43,[49][50][51][52] The data presented here demonstrate that PDGFR-β, used alone, 14,[49][50][51][52] is an inappropriate pericyte marker in murine kidney because the majority of its labeling identifies interstitial fibroblasts. In our experience CD146 is the most specific marker for distinguishing pericytes from other renal mesenchymal cells and is also highly sensitive, as is the case in other human and mouse organs. 27,35,53 We propose a combination of CD146, PDGFR-β and endothelial cell labeling as the most accurate method of identifying and distinguishing pericytes and interstitial fibroblasts in histological sections of murine kidney. Under this model, within the non-endothelial population pericytes are CD146 + (±PDGFR-β labeling), and interstitial fibroblasts are CD146 − PDGFR-β + (see Figure S5). In the future, prospective pericyte markers identified through our scRNAseq analysis should be tested, such as Purkinje cell protein 4-like protein 1 (Pcp4l1) which was in the top differentially expressed genes for pericytes in both the IRI and TMS datasets.
We have also identified a combination of PDGFR-α and PDGFR-β labeling as a potentially superior method of identifying myofibroblasts.
This is through corroboration of histological data, in which this population increased substantially with ischemia ( Figure 5G) and correlated strongly with fibrosis extent (Figure S10I), and scRNAseq data in which only the collagen-expressing myofibroblast population expressed both the Pdgfra and Pdgfrb transcripts ( Figures 6C,E and 7C-E). In line with this PDGFR-α is a marker of fibrotic populations and progenitors in other tissues such as muscle, skin, and multiple visceral organs. 25,32,[54][55][56] It is often in association with PDGFR-β 25,32,56 which is itself reported a myofibroblast progenitor marker. 57 Furthermore, a contemporaneous study recently published also identified PDGFR-α + PDGFR-β + cells as the major ECM producers in human and mouse renal fibrosis. 58 These two proteins are therefore emerging as reliable fibrotic population markers across multiple organ systems.
Other prospective myofibroblast markers may be identified from scRNAseq data. Of the top differentially expressed genes for myofibroblasts three, Abca8a, Fndc1, and Sult5a1, have probable intracellular expression (www.uniprot.org) and thus represent markers to investigate in future.
Aging is one of the largest risk factors for renal disease, 7 yet the effect of age on the mesenchymal cell response to injury has not to our knowledge been investigated beyond 12 months in mice. 59 Following IRI, the response of old kidney mesenchyme differed to young in several ways. First, increased fibrosis with age was detected in line with previous ischemia studies. 2,3,59 Second, transient increases in pericyte numbers and CD146 labeling intensity seen in young animals at day 1 post-IRI were not observed in old. Third, at day 28 there was substantially more ischemia-linked subtypes in old animals, especially NG2 + populations. Indeed, analysis of mesenchymal subtypes suggests that the aged mesenchyme at baseline is in a "post-ischemic" state. For example, NG2 + α-SMA − , NG2 + α-SMA + , and PDGFR-α + -β − cells increased with ischemia in both age groups, and while they were difficult to detect in young contralateral kidneys, in old contralateral kidneys they were reasonably abundant. Taken together, this points to a situation in which the old kidney has a subdued acute response to injury, chiefly in the pericyte population, but an extended or nonresolving long-term reaction within fibroblast and myofibroblast populations. Non-resolving fibrosis is well documented with aging, 60,61 and previous studies have shown an impaired proliferative response in kidney in the acute phase of injury, although none have targeted pericytes specifically. 2,9,62,63 It is possible that age related senescence is a factor in both these observations 2,64 ; however, further testing is required to ascertain whether and how these differences impact the progression of renal disease.
Closer analysis of scRNAseq gene expression data leads to some interesting findings. First, expression of Acta2 (α-SMA), the classical myofibroblast marker, was more substantial in the pericyte/vSMC than in myofibroblast population, as observed elsewhere. [65][66][67][68][69][70] It has been shown that only 75% of α-SMA + cells express collagen, 24 and removal of significant numbers of α-SMA + cells by inhibiting TGF-β signaling had only a small effect on fibrosis. 46  also noticed an induction of NG2 with disease. 24 Finally, renal mesenchymal populations with "stem-like" or "MSC-like" properties, such as multilineage differentiation potential in vitro and ability to integrate into the kidney when delivered in an injury setting, have been identified in multiple studies using various combinations of surface marker proteins Sca-1, c-Kit, CD24, CD29, CD44, CD73, CD90, and CD105. 84,[88][89][90][91][92][93] We have found that transcription of these surface markers is distributed across all of our scRNAseq populations (not shown). Given that MSCs can be generated in vitro from both pericytes and adventitial fibroblasts, and that these MSC-like populations are often selected by digests and plating of whole kidneys, it is likely that they have a mixed origin within the kidney.
In future, renal mesenchymal heterogeneity can be probed more deeply by increasing the yield of cells used in transcriptomic analysis.
Besides improved cell dissociation protocols, fluorescent fate mapping using pan-mesenchymal gene drivers such as FoxD1 15 or myelin protein zero 94 should label the majority of mesenchyme and provide an efficient method of cell sorting. Alternatively, subpopulations can be interrogated individually using markers such as those used here. This approach has been recently performed with PDGFR-α and PDGFR-β, revealing multiple subpopulations within this compartment. 58 Much work is still required to investigate the effect of age on renal mesenchyme, and whether and how differences such as those identified here impact on the progression and treatment of disease.

| CONCLUSION
Despite the heterogeneity in perivascular marker expression in the renal interstitium, there is correlation between marker expression and anatomical location, and clear injury-and age-linked differences have been identified. Further research is required to elucidate the full extent and functional implications of renal interstitial cell heterogeneity, and the intriguing questions raised in this study of marker heterogeneity should stimulate such investigations.

CONFLICT OF INTEREST
The authors declared no potential conflicts of interest.