seqFISH profiling of acute kidney injuryTo examine AKI outcomes after the initial phase of injury-invoked renal repair27, we subjected C57BL/6 J male mice (8–13 weeks of age) to a mild ischemia-reperfusion injury (IRI; serum creatinine levels at 48 h 0.3–1.7 mg/dL), waited four weeks and then collected and analyzed the kidneys. We selected 1300 target genes for seqFISH profiling based on key cell-type enriched markers from single-cell studies of normal and IRI kidney3,13,15 and a long-term study of injury-associated expression during the AKI to CKD transition in mice (27; Fig. 1A). We used Cellpose 2.042 and DAPI based segmentation to achieve maximal accuracy of our analysis and minimize segmentation errors. Integration of control (n = 3) and post AKI (n = 3) samples resulted in a data set comprising 245,171 single cells, with an average of 185 ± 153 (mean ± SD) total transcripts and 87 ± 53 individual genes detected per cell, with a strong concordance between samples (Supplementary Fig. 1 and Supplementary Fig 4). Following rigorous quality filtering, we retained 220,753 cells for subsequent analysis. By clustering on single-cell gene expression, we identified all major kidney cell types, highlighting the segmental cell diversity along the conjoined epithelial networks of the nephron and collecting system (Fig. 1B and Supplementary Data 1), as well as divergent vascular endothelial sub-sets, interstitial fibroblasts and distinct immune cell types (Fig. 1B, C). The cell types identified by seqFISH were concordant with cell types previously identified using single-cell RNA sequencing (scRNAseq) (Supplementary Fig. 2A, B). Further, their spatial localization conformed to well-documented kidney anatomy (Fig. 1C, D, F and Supplementary Figs. 3,  5). However, the seqFISH analysis captured a notably larger fraction of several kidney-resident cell types relative to that previously reported in scRNA-seq studies43,44, where the tissue dissociation procedures cause certain cell types to be over- or under-represented43,45,46. For instance, in past work, fibroblasts were reported to comprise ~ 4–10% and macrophages less than 5% in kidney single cell- and single nucleus-RNAseq datasets, whereas our data showed that fibroblasts comprise close to 20% and macrophages ~ 10% of all cells (Supplementary Fig. 2C). These results are consistent with the idea that detecting cell types in kidney tissue by seqFISH will be less biased compared with contemporary single cell-RNAseq protocols.Fig. 1: seqFISH reveals all major kidney cell types and their locations within the kidney, as well as compositional and spatial changes following AKI.A Experimental overview: mice were subject to IRI at day 0 to induce AKI. Control and AKI kidneys were collected on day 28, and seqFISH was performed on frozen kidney sections. B Umap depicting all cell types identified using seqFISH. Normal kidney-specific cell types are highlighted and their locations within the nephron are illustrated. DC – dendritic cells; Macroph – macrophages; DCT-CNT – distal convoluted and connecting tubule; Fib – fibroblasts; IC – intercalated cells; Injured PT – injured proximal tubule cells; LOH-TL-C – thin limb of the loop of Henle (comprising cells of the thin descending limb of the loop of Henle of cortical nephrons and the thin ascending limb of the loop of Henle of juxtamedullary nephrons); LOH-TL-JM – thin limb of the loop of Henle of juxtamedullary nephrons; PC – principal cells; Per – pericytes; Podo – podocytes; PT – proximal tubule; PTS1/2/3 – proximal tubule segment 1/2/3; T – T cells; TAL – thick ascending limb of the loop of Henle; Uro – urothelium; Vasc – endothelial cells. C Spatial location of all cell types in one representative control sample. Left – all cell types, right – 10 cell types out of the total 22 that were identified when plotted individually. These cell types span the cortex and medulla, showing that seqFISH analysis captures all major cell types in different areas of the kidney. D Representative field of view outlined in panel (C) showing the cell masks color-coded by cell type (top) and a zoomed-in image showing RNA expression of representative marker genes (bottom). The image is comprised of the sum of the background and DAPI image to illustrate the underlying tissue morphology. E Violin plot showing the normalized expression of marker genes for each cell type. The bar plot shows the relative abundance of each cell type within the control and AKI samples. F Spatial locations of PTS3, Vasc_2, Injured PT, fibroblasts, and macrophages within one representative control (top) and one AKI (bottom) sample.We next compared AKI with control samples to identify changes in the abundance and spatial locations of specific cell types. At the cortico-medullary boundary, which is known to be most sensitive to AKI47, the proximal tubule segment 3 (PTS3) and vascular endothelium (Vasc_2) were strongly reduced (Fig. 1E, F and Supplementary Figs. 2D, 5). Conversely, a marked increase was observed in injured proximal tubule cells, as identified by the expression of Vcam1 (13,15; Injured PT). In addition, fibroblast, macrophage, T cell, and dendritic cell populations were all elevated in proportion in AKI samples relative to control (Fig. 1E, F and Supplementary Fig. 2D,  5). In contrast to T cells and dendritic cells, the increase in fibroblasts and macrophages was associated with a cortical expansion in injured kidney samples (Fig. 1F and Supplementary Fig. 5); fibroblasts and macrophages predominated within the medulla of the uninjured kidney. In addition, we observed an over-representation of a Slc12a1+/Ptger3+ subset of thick ascending limb (TAL)-subtypes in the distal medullary loop-of-Henle (Fig. 1E48;). Thus, AKI induces global changes in both cell type composition and location.Distinct cellular microenvironments are specific for AKI and normal kidneysWe next sought to understand whether changes in cell populations also lead to the re-organization of the local kidney architecture. Specifically, we asked whether we could detect localized microenvironments (MEs) with distinct cell type compositions within the kidney and whether those environments changed following AKI. To this end, we calculated the frequencies of each cell type within a 30 µm radius of each of the cells in the kidney tissue. We reasoned that this radius is of physiological relevance as it represents ~ 2 cell layers around each cell, which is a distance scale that allows for short-range paracrine and juxtacrine signaling49. We then used the resulting cell-by-neighbor matrix to cluster individual cells using Leiden clustering (Fig. 2A and Supplementary Fig. 6). To determine the sensitivity of ME clustering to different radii, we varied the radii and repeated the neighboring calculation maintaining the same clustering parameters (Supplementary Fig. 6). Interestingly, this clustering approach detected MEs with similar compositions within a radius range of 10–50 µm (Supplementary Fig. 6A). However, smaller (10, 20 µm) and larger (50, 100 µm) radii resulted in multiple redundant MEs with highly similar cellular compositions (Supplementary Fig. 6B). In addition, larger radii produced multiple MEs that were sample specific reflecting slight differences in the orientation of sections, and potentially, biases caused by individual imaging Fields Of View (FOVs) (Supplementary Fig. 6C). In summary, decreasing or increasing the radius for neighborhood analysis from 30 µm does not improve the capture of biologically relevant cellular interactions and can introduce redundancy or bias.Fig. 2: Cellular microenvironments define the spatial architecture of Control and AKI kidneys.A Zoomed in image of one field of view in the AKI sample: The cellular composition within a 30um radius around each cell is calculated, and these compositions are then clustered to create 17 distinct MEs. B Cellular composition within each ME is calculated as the relative mean abundance of each cell type. Barplot showing the relative frequency of each ME within the Control and AKI samples. Gray asterisk represents enrichment in control and blue enrichment in AKI (*p < 0.05; **p < 0.01 of a one-sided t test). Source data are provided as a Source Data file. C Spatial locations of five MEs in one representative control and one AKI sample. D Spatial locations of all injured PT cells in one AKI sample plotted in color over the locations of all cells belonging to ME-5 plotted in gray. E Zoomed-in image of the box plot in D showing an overlay of cell masks on Vcam1 antibody and DAPI signal. The cell masks are colored according to the ME assignment of each cell. The Vcam1 signal is shown in white over the DAPI gray signal. F Expression of top injury marker genes in injured PT cells within ME-5 and the Injured PT cells outside ME-5 in Control samples (left) and AKI (right). (*p < 0.05; **p < 0.01 of a paired two-sided t test). Source data are provided as a Source Data file. G Normalized expression of injury markers in Injured PT cells within ME-5, sorted by the eccentricity of each cell. Cells were sorted according to eccentricity values, and the gene expression was averaged using a moving window with a window size of 10% total number of cells. Expression values were normalized for each gene in the heatmap.Using a 30 µm radius for neighborhood analysis, we detected 17 MEs with a distinct cell type composition and spatial location (Fig. 2B, C and Supplementary Fig. 7). Eight MEs were not markedly different between individual control and AKI samples, five were enriched in the control kidneys and four were specific to AKI samples (Fig. 2B and Supplementary Fig. 7). To limit sampling bias, we focused our analysis on MEs that were equally enriched or depleted in all control and AKI replicates (indicated by an asterisk in Fig. 2B, gray – enriched in control and blue-enriched in AKI). In Fig. 2C, we show examples of MEs that are similar between control and AKI (ME-0), absent from AKI (ME-3), or enriched in AKI (ME-5, ME-15, ME-16). AKI-specific MEs were largely a combination of injured epithelium, fibroblasts, and immune cells in varying proportions, and were mostly depleted of the primary tissue cells. Interestingly, dendritic cells, macrophages, T cells, and fibroblasts were distributed across different MEs in the control samples, but concentrated within specific MEs in AKI samples (Supplementary Fig. 7B). Therefore, the emergence of these cells, whether by differentiation, proliferation, or recruitment from the blood, results in redistribution into distinct local environments upon AKI.Characterization of the pathogenic niche in AKIRecent studies in mouse and humans have drawn attention to the tissue micro-environment around injured proximal tubule cells, which have been associated with renal pathology13,15,28. Injured proximal tubule cells display the cell adhesion molecule Vcam1 and adopt a senescence-associated secretory phenotype (SASP)13,15,28,33. We identified a similar population of Vcam1-positive epithelial cells in our data set by unsupervised clustering (Cluster Injured PT, Fig. 1B, E). The majority of these injured epithelial cells reside within a single ME (ME-5, Fig. 2B, D). As a means of internal validation, we stained with anti-Vcam1 antibody concurrently with seqFISH analysis and found that Vcam1 protein is mainly present in cells within ME-5 (Fig. 2E).Figure 2F plots the mean expression of top injured PT marker genes in control (left) and AKI (right), comparing the expression of these genes in injured PT cells outside ME-5 to the expression of their counterparts within ME-5. Interestingly, while the expression of injury genes was elevated in injured PT cells in AKI samples compared to control, the expression within ME-5 is significantly higher than in cells outside ME-5 for both AKI and control samples. Injured PT cells have been detected in normal samples and are thought to be naturally occurring senescent cells13,15,17. Our analysis suggests that the cellular neighborhood of injured cells correlates with a more severe injury phenotype even within AKI and that cellular interactions within this neighborhood could be a driver of injury progression.Proximal tubule cells undergo a de-differentiation in response to AKI that manifests in several morphological changes, including a loss of the apical brush border, and a flattening of the epithelium, and likely with this, the nucleus of the cell50. To assess whether nuclear morphology might be an informative parameter, we correlated changes in the expression of injury genes to the eccentricity of the nucleus in Injured PT cell masks. An eccentricity value of 0 represents a perfect circle, and values between 0 and 1 represent an ellipse. We found that eccentricity analysis could sub-divide Injured PT cell expression profiles. With increasing eccentricity values (corresponding to a flattening of the nucleus), Vcam1 expression increased, while expression of other injury-related genes did not correlate with high eccentricity (Fig. 2G). Thus, morphological criteria can have a predictive value of the injury state of the cell. Taken together, our findings show that ME-5 represents an injured and likely pathogenic niche. The majority of injured PTs and injury-associated fibroblasts, vasculature, and immune cells reside within this niche, and morphological changes within the injured epithelium are associated with elevated expression of injury-response genes.
Clcf1 – Crlf1 interactions between injured PT and fibroblasts shape the injured nicheWe sought to identify signaling between Injured PT and the other cell types present within ME-5 (fibroblasts, macrophages, T cells, DCs, Vasc_1, few normal PT cells), that would be potential drivers organizing the injured niche. We first identified predicted signaling initiated by the Injured PTs using Nichenet51 and found that Clcf1 – Crlf1 and Jag1-Notch3 signaling were upregulated between Injured PTs and fibroblasts within the AKI samples (Fig. 3A and Supplementary Fig. 9A). Upregulation of Crlf1 and Clcf1 in kidney injury and disease has been indicated in several prior studies27,52,53,54, and Crlf1 upregulation has been reported in the fibrotic lung55. Single-cell RNAseq data analyzing IRI in the mouse kidney has predicted a Clcf1-Crlf1 signaling axis between injured PT cells and fibroblasts15. Clcf1 encodes a member of the Il-6 cytokine family, which is thought to engage Crlf1 as a chaperone in signaling through the receptor Cntfr56,57. While Clcf1 was specifically expressed within injured PTs (Fig. 3A and Supplementary Figs. 9A, 10), Crlf1 was predominantly expressed in a distinct subset of ME-5 associated fibroblasts (Fig. 3A and Supplementary Figs. 9A, 11).Fig. 3: Injured PT are interacting with fibroblasts in a spatially dependent manner.A Mean expression of two top ligands and their receptors identified by NicheNet calculated for the cell types comprising ME-5. B Illustration showing hypothesized cellular behaviors under distance-dependent and distance-independent signals. Cell A expresses a ligand while the expression of the receptor on cell B can be constant over the distance between Cell A and Cell B (distant independent) or changing over distance (distant dependent). C Fraction of Fibroblasts expressing Crlf1 and Notch3 over distance from the nearest Injured PT cell. Cells were divided into 20um distance bins and the fraction of expressing fibroblasts was calculated out of all fibroblasts in each bin. D Spatial expression of Clcf1 on Injured PT (gold) and Crlf1 (blue) on fibroblasts. E Zoomed-in image on the inset in (D) showing Clcf1 on Injured PT and Crlf1 on fibroblasts. F Zoomed-in image showing mRNA dots of the corresponding genes. G Antibody-staining of the Injured PT marker Vcam1 coupled with RNAscope evaluation of Clcf1, Crlf1, and Cntfr, the receptor for the Clcf1-Crfl1 complex. Staining was done on a control sample. The image is of one representative control sample out of two. A larger kidney area from this experiment is presented in Supplementary Fig. 13. H Same as in (G), measured on AKI sample. The image is of one representative AKI sample out of three. A larger kidney area from this experiment is presented in Supplementary Fig. 13.Our analysis also identified upregulation of Cxcr6 in T cells and Itgax in DCs within AKI samples suggestive of Injured PT signaling to T cells and to dendritic cells through Cxcl16 – Cxcr6 and Icam1 – Itgax signaling axes, respectively, recruiting immune cells within AKI (Supplementary Fig. 9A). In the context of liver fibrosis, Cxcr6-dependent recruitment of CD8 T cells and NKT cells is known to contribute to disease progression58,59,60 and a similar role has been suggested for a Cxcl16 – Cxcr6 axis in kidney fibrosis61,62,63. Macrophages in AKI also express Cxcl16, consistent with a contribution to T-cell recruitment (Supplementary Fig. 9A). Csf1 – Csf1r signaling axis between injured PTs and macrophages was also upregulated within AKI samples (Supplementary Fig. 9A), in line with previous studies that link injured epithelial Csf1 production to macrophage-mediated recovery following AKI64. Analysis of two published scRNAseq and snRNAseq datasets of dissociated human and mouse kidney samples showed that the elevated expression of the top ligands identified here, including Clcf1 and Csf1 as well as Cxcl16, Icam1, and Tgfb2, was conserved across species (Supplementary Fig. 9B).We hypothesized that spatially localized ligand-receptor interactions (distance dependent; Fig. 3B) will contribute specifically to the structuring of the microenvironment around and via injured PTs as opposed to signaling with no spatial preference (distance independent; Fig. 3B). To assess the spatial localization of the signals, we quantified the fraction of receptor-expressing target cells as a function of the distance of the target cells from the nearest Injured PT, regardless of ME assignment. Figure 3C plots the fraction of Crlf1 + (left) and Notch3 + (right) fibroblasts over distance from injured PTs. While there is a relatively constant fraction of Notch3 + fibroblasts at varying distances from injured PT cells, the fraction of Crlf1 + fibroblasts decreases as the distance from the injured cells increases. Similar to Notch3, little to no spatial preference was detected for other receptors associated with Injured PTs signaling to fibroblasts, vasculature, and different immune cell types (Supplementary Fig. 12). To validate the predicted expression and spatial distribution of the Clcf1-Crlf1 pairings, we combined anti-Vcam1 immunodetection with serial RNA-FISH on control and AKI samples. Consistent with seqFISH analysis, Crlf1 + cells were specifically concentrated adjacent to Clcf1 + Vcam1 + Injured PTs (Fig. 3G, H and Supplementary Fig. 13). Contrary to the specific localization of Crlf1 around injured cells, Acta2, a well-documented marker of inflammatory myofibroblasts15, exhibited heterogeneous expression (Supplementary Fig. 14). Clcf1, the predicted target of Crfl1 activity, signals through Cntfr. Cntfr expression colocalized with Crlf1 in peri-epithelial myofibroblasts but was not limited to this population (Fig. 3G, H).Taken together, our findings implicate injured PT signals in the remodeling of vasculature, recruitment of immune cells, and reshaping of the fibroblast population within the AKI samples. We show that while several signaling events are upregulated within AKI samples, only Clcf1-Crlf1 interactions are highly spatially localized. Signaling which is upregulated but does not show spatial localization could present global changes following AKI or past events leading to the formation of the injured niche. Our combined spatial data and validation suggest that Crlf1 is a specific identifier of fibrotic processes closely coupled to Injured PTs and that Clcf1 – Clcf1 interaction with fibroblast cells is a constitutive defining feature of the injured niche.Fibroblasts show distinct expression patterns between different MEsTo determine whether gene expression changes could be identified more precisely in different MEs, we analyzed fibroblasts within the four AKI-specific MEs (ME-5, ME-9, ME-15, and ME-16). Both the composition of cell types and location in the tissue were distinct for each ME (Fig. 4A – left, middle). ME-5 localized to the cortex and corticomedullary boundary, ME-15 extended from the corticomedullary boundary to the medulla, ME-9 was concentrated around TAL_2 cells in the medulla, and ME-16, comprising predominantly immune cell types, was scattered throughout the kidney (Fig. 4A and Supplementary Fig. 8).Fig. 4: Within different MEs, Fibroblasts have distinct gene expression patterns, with a subset of Fibroblasts representing acute response to injury.A Left – coordinates plot showing the composition of four MEs enriched in AKI samples. Middle-spatial locations of each of the MEs in one AKI sample. Right – expression of marker genes of each ME in all fibroblasts in the sample. B Differentially expressed genes within fibroblasts between the different MEs. The heatmap shows the mean expression of each of the genes, calculated for fibroblasts within each ME as well as for all fibroblasts in control and AKI. The depicted MEs represent four enriched in AKI (ME-5,15,9,16) and five enriched in control (ME-1,3,7,8,12). C Top – sub-clusters of Fibroblasts based on gene expression presented in Umap space. Clusters were calculated using Leiden clustering. Bottom – same ME and marker genes as in (A), presented on Umap space. D Mean expression score for the DE genes shown in panel (B) calculated within fibroblast populations in a time course snRNAseq data following injury in a mouse model of AKI. The scores were calculated for individual fibroblasts using Seurat and averaged for each mouse. The data presented in the heatmap shows the average of all mice in each condition. The minimum values were subtracted for each score such that minimum score = 0. N = 3 mice per time point. E The same calculation as in (D) was performed on a dataset of samples collected from patients who have undergone AKI (n = 11), exhibit CKD (n = 13), and are compared to normal kidney biopsies (n = 14). F Vcam1 antibody-staining with Clcf1, Crlf1 and Npr3 expression measured using RNAscope on control sample. The image is of one representative control sample out of two. G Same as in (F) on the AKI sample. The image is of one representative AKI sample out of three.The top enriched fibroblast gene sets for each ME, shown in Fig. 4A, B, revealed specific ME-associated gene expression (see Supplementary Data 1 for a full list of differentially expressed genes). Figure 4A shows that the expression of the top ME-associated genes is highly restricted to the location of each ME in the physical space. Interestingly, we found that many of these genes were not detected when clustering fibroblasts based on gene expression alone without considering spatial ME information (Fig. 4C). Out of the four ME-specific fibroblast maker genes Crlf1, Actg2, Igfbp5 and Iigp1, only Igfbp5 was enriched in the gene expression-based cluster (cluster 3 in Fig. 4C), while the other genes did not show cluster-specific expression. These data underscore the importance of considering the spatial environment of fibroblasts to identify genes with a potential functional significance.We found that Crlf1, Npr3 (which encodes the natriuretic peptide receptor 3), and Timp1 (which encodes TIMP metallopeptidase inhibitor 1) were markedly enriched in the ME-5 fibroblasts (Fig. 4A, B). In previous work, Npr3, a known blood pressure regulator, reduced disease severity upon inhibition in a model of cardiac fibrosis65, whereas upregulation of Timp1 was reported to increase kidney scarring66,67. Here, we validated the co-expression of Crlf1 and Npr3 in association with injured PTs using immunostaining and serial RNA-FISH (Fig. 4F, G). In ME-15, which comprises fibroblasts and macrophages, Actg2, encoding a member of the smooth muscle actin family, was the top gene identifier of ME-15 fibroblasts (Fig. 4A). Actg2 has not previously been associated with kidney disease, although a recent study identified Actg2 as a candidate in long term kidney impairment following acute decompensated heart failure68. ME-15 fibroblasts also elevate the expression of Apoe and C1qa, key inflammatory genes highly expressed by macrophages, suggesting both macrophages and fibroblasts contribute to the inflammatory environment. Fibroblasts in ME-16 were distinguished by the expression of Ligp1, which encodes the interferon-induced GTPase1, and the expression of Bcl3, an inhibitor of apoptosis, and the caspase Casp4. In contrast to ME-5, -15, and -16 fibroblasts, ME-9 fibroblast marker genes were not highly expressed within the total AKI fibroblast population, and several genes were higher in control fibroblasts. One of the most strongly enriched fibroblast genes in this ME was the insulin growth factor binding protein (IGFBP), encoded by the Igfbp5 gene. IGFBPs have been linked to CKD progression69 and disease progression in a mouse model of diabetic kidney disease70.Spatially dependent stromal gene expression in the mouse and human kidneyGiven the highly distinct gene signatures of fibroblasts across the different MEs, we next asked if the expression of the identified ME-specific fibroblast marker genes changes along the AKI-to-CKD transition. To answer this question, we analyzed published mouse snRNA-seq15 collected at five-time points post IRI-invoked AKI (4 h, 12 h, 2 days, 14 days, and 6 weeks; Fig. 4D). We calculated a ME score for fibroblasts, defined as the combined expression of all the ME-specific genes for each of the AKI-enriched MEs (ME-5, ME-15, ME-9, and ME-16). We also calculated scores for fibroblast marker genes in control and AKI samples regardless of ME assignment (Control-Fib and AKI-fib). Figure 4D presents the average score for all the samples in each condition, normalized such that the minimum score for each gene set = 0.As expected, following AKI, the control-fibroblast score is reduced immediately and remains relatively low until 6 weeks post AKI, while the AKI-fibroblast score is low in control samples, but increases and peaks at day 2 after AKI, remaining elevated at 6 weeks. This timeframe suggests that the injury induces fibroblast differentiation and recruitment, as captured in the AKI-fibroblast score, and that AKI-induced fibroblasts persist 6 weeks after injury. We observed divergent patterns in the ME-specific fibroblast scores, where ME-5 and ME-15 scores were similar to the score of AKI-fibroblast, peaking at day 2 post AKI, with ME-15 score remaining high also at later time points. However, the ME-9 score was significantly delayed and peaked at 14 days post-AKI, suggesting a later appearance for this fibroblast population following AKI. Although ME-16 genes were highly specific to AKI samples in our data (Fig. 4B), the ME-16 score was low throughout the time course. As this population of fibroblasts is relatively small, the signal may be diluted within other populations within the sequencing data.We repeated this analysis with a scRNA-seq data set from human patients with AKI, CKD, or normal kidneys as control28(Fig. 4E). Similarly to the mouse data, we find that here, too, ME-5 and ME-15 scores are high in AKI relative to control and CKD. ME-16 scores were elevated in both AKI and CKD, although at lower levels. Assuming that the sequencing data is capturing cells from different sites of the tissue, our analysis suggests that the abundance of the different fibroblast subsets detected in each condition reflect changes along the AKI-to-CKD transition and that these populations are common to the mouse and human kidney. Thus, beyond spatial mapping, seqFISH analysis provides additional molecular granularity as to how cell populations acquire distinct properties during a biological process.Spatial distribution of immune cells correlates with inflammatory potentialNext, we focused on immune cells as drivers of inflammatory fibrosis. To identify T cell subtypes present in our data, we mapped T cell populations onto a reference data set of mouse T cells71, using SEURAT. We identified known T cell populations, including CD4 and Tregs (high Cd83 and Ctla4 expression), naïve, effector (high Cxcr3 expression and expression of cytotoxic genes such as Nkg7), and exhausted (no Cxcr3 expression but expression of the inhibitory molecule Pdcd1) CD8 T cells (Fig. 5A). When we quantified the fraction of each T cell subtype within MEs − 5, − 15 and − 16 in the AKI samples, we found a difference in the fraction of CD4 T cells amongst all T-cells: ~ 40% in ME-5, 50% in ME-15 and 55% in ME-16 (Fig. 5B). We noticed a similar trend for Tregs, although, as expected, Tregs represented a smaller subset of the T cell population (Fig. 5B). The inverse trend was apparent for effector CD8 T cells, and no clear trend was evident for naïve and exhausted CD8 cells (Fig. 5B).Fig. 5: T cell subtypes preferentially reside into different MEs, correlating with their functionality.A All T cells in our dataset were mapped to a tumor-infiltrating lymphocyte dataset. The T cells were mapped to six known cell populations – helper T cells (CD4T), cytotoxic T cells (CD8T Eff_mem), naive and exhausted CD8 (CD8_Naive, CD8_exhauseted), and regulatory T cells (Treg). The violin plot shows marker gene expression for each of the T cell subtypes in our data. B Fraction of T cells of each subtype in AKI enriched MEs (5, 15, 16) for n = 3 AKI mice. Each dot represents one mouse. The fraction is the total fraction of each subtype out of all T cells within the specified ME. The boxplot middle line represents the median value, and box boundaries show the 25th and 75th percentiles. Source data are provided as a Source Data file. C Average expression of three M1 (blue) and four M2(red) markers within Macrophages in the AKI enriched MEs. D Fractions of Mrc + (left) and Ccr7 + (right) Macrophages within the AKI enriched MEs boxplots showing the distribution for n = 3 AKI mice as in (B). Dots represent individual mice. The boxplot middle line represents the median value, and box boundaries show the 25th and 75th percentiles. Source data are provided as a Source Data file. E Spatial distribution of the same genes as in (D). Left – ME-15 is indicated and Mrc+ cells are colored in red. Right – ME-16 is indicated and Ccr7+ cells are colored in blue. F. Spatial locations of the T cell subtypes within one AKI sample. ME-16 is depicted in color. G Vcam1 antibody staining, as well as the markers Cd4, Cd8, and the DC marker Clec9a, are detected with RNAscope on a control sample. The image is of one representative control sample out of two. A larger kidney area from this experiment is presented in Supplementary Fig. 15. H Same as in G for an AKI sample. The image is of one representative AKI sample out of three. A larger kidney area from this experiment is presented in Supplementary Fig. 15.Since macrophages represent a large fraction of the immune cells in AKI-specific MEs, we measured marker genes linked to differential activation states of M1 and M2 macrophages. In general, M1 macrophages promote disruptive, disease-associated inflammatory responses, whereas M2 macrophages promote constructive, inflammatory-associated tissue repair72. We found that the average expression of the M2-related genes Mrc1, Cd163 and Arg1 was increased in macrophages within the ME-5 and ME-15 groupings, while M1-related genes Cxcl10, Ccr7, Cd40 and Cd86, were expressed at higher levels on the average in ME-16 macrophages (Fig. 5C). In agreement with these data, Mrc1+ macrophages were more prevalent in ME-5 and ME-15, and Ccr7 + macrophages were more prevalent in ME-16 (Fig. 5D, E).Tertiary Lymphoid Structures (TLS), increasingly recognized as contributors to chronic inflammation22, are organized lymphoid aggregates that form ectopically in response to a disturbance in tissue homeostasis and then act as localized hubs that enable signal exchange between immune cells, which in turn promotes the development of adaptive immunity within a tissue. Unlike secondary lymphoid structures (such as the lymph nodes and the spleen), TLS is unencapsulated, exposing cells within to multiple signals from the environment. In the context of autoimmune and chronic inflammation, the presence of TLS is correlated with severe disease, and in the context of chronic kidney disease, with more severe inflammation and fibrosis20. Mechanistically, the emergence of TLS in a long-term model of the AKI to CKD transition was linked to the maturation of B cell-directed autoimmunity against the target kidney tissue27,73.We found that the cellular composition and appearance of ME-16 show characteristics of immune aggregates reminiscent of nascent TLS. ME-16 is composed of T cells, DCs, and macrophages with a small number of fibroblasts (Fig. 4A). ME-16 clusters also appear scattered without a preference for the cortex or the medulla (Figs. 2C,4A). When overlaid on the locations of ME-16 (Fig. 5F), the spatial distribution of CD4 and CD8 T cells were dispersed within and outside of ME-16, with CD4 T cells clearly aggregated in ME-16. Thus, ME-16 has the expected properties of early-forming immune aggregates in which CD4 T cells and Tregs are providing instructive signals to CD8 and other immune cells. Further, employing RNA-FISH detection of Cd4, Cd8a, and the DC marker Clec9, we detect early aggregation of DCs, CD4, and CD8 T cells 4 weeks following AKI (Fig. 5G, H and Supplementary Fig. 15). CD8 T cells trend towards cortical prevalence and CD4 towards medullary enrichment. Moreover, elevated Cxcr6 expression in CD8 T cells (Supplementary Fig. 16) suggests interaction with Cxcl16-expressing Injured PTs (Supplementary Fig. 9A).Although the clear enrichment of T cells and DCs within ME-16 is consistent with localized antigen presentation, we did not detect any B cells, in line with kidney profiling data showing a later engagement of the B cell response26,73. Because M1 macrophages are preferentially enriched within ME-16, we propose that ME-16 represents an early lymphoid aggregate that propagates inflammation. In contrast, the presence of M2 macrophages in ME-15 and ME-5 suggests that these MEs are undergoing a combination of inflammatory and fibrotic processes where the inflammatory response is sequestered and replaced by fibrotic processes74.
Spatial transcriptomics defines injury specific microenvironments and cellular interactions in kidney regeneration and disease
