Single-Cell Transcriptomics of a Human Kidney Allograft Biopsy Specimen Deﬁnes a Diverse Inﬂammatory Response

BASIC RESEARCH www.jasn.org

Haojia Wu,1 Andrew F. Malone,1 Erinn L. Donnelly,1 Yuhei Kirita,1 Kohei Uchimura,1 Sai M. Ramakrishnan,1 Joseph P. Gaut,2 and Benjamin D. Humphreys 1,3

1Division of Nephrology, Department of Medicine and Departments of 2Pathology and Immunology and 3Developmental Biology, Washington University in St. Louis School of Medicine, St. Louis, Missouri

ABSTRACT Background Single-cell genomics techniques are revolutionizing our ability to characterize complex tissues. By contrast, the techniques used to analyze renal biopsy specimens have changed little over several decades. We tested the hypothesis that single-cell RNA-sequencing can comprehensively describe cell types and states in a human kidney biopsy specimen.

Methods We generated 8746 single-cell transcriptomes from a healthy adult kidney and a single kidney BASIC RESEARCH transplant biopsy core by single-cell RNA-sequencing. Unsupervised clustering analysis of the biopsy specimen was performed to identify 16 distinct cell types, including all of the major immune cell types and most native kidney cell types, in this biopsy specimen, for which the histologic read was mixed rejection.

Results Monocytes formed two subclusters representing a nonclassical CD16+ group and a classic CD162 group expressing dendritic cell maturation markers. The presence of both monocyte cell subtypes was validated by staining of independent transplant biopsy specimens. Comparison of healthy kidney epithelial transcriptomes with biopsy specimen counterparts identified novel segment-specificproinflam- matory responses in rejection. Endothelial cells formed three distinct subclusters: resting cells and two activated endothelial cell groups. One activated endothelial cell group expressed Fc receptor pathway activation and Ig internalization genes, consistent with the pathologic diagnosis of antibody-mediated rejection. We mapped previously defined genes that associate with rejection outcomes to single cell types and generated a searchable online gene expression database. Conclusions We present the first step toward incorporation of single-cell transcriptomics into kidney biopsy specimen interpretation, describe a heterogeneous immune response in mixed rejection, and provide a searchable resource for the scientific community.

J Am Soc Nephrol 29: 2069–2080, 2018. doi: https://doi.org/10.1681/ASN.2018020125

The renal biopsy provides critical diagnostic and prognostic information, but interpretation rests Received February 5, 2018. Accepted May 5, 2018. on a limited number of techniques perfected de- H.W. and A.F.M. contributed equally to this work. cades ago. With few exceptions, there have been Published online ahead of print. Publication date available at no substantive changes in the way that biopsies www.jasn.org. are read in 25 years. Recent techniques for massively Correspondence: Dr. Benjamin D. Humphreys, Division of Ne- parallel single-cell RNA-sequencing (scRNA-seq) phrology, Washington University School of Medicine, 660 South have been developed, in which the expression of Euclid Avenue, CB 8129, St. Louis, MO 63110. Email: hum- thousands of genes in thousands of individual cells [email protected] can be measured rapidly, simultaneously, and quan- Copyright © 2018 by the American Society of Nephrology

J Am Soc Nephrol 29: 2069–2080, 2018 ISSN : 1046-6673/2908-2069 2069 BASIC RESEARCH www.jasn.org titatively. These approaches offer an unprecedented opportu- Significance Statement nity to define cell types and states comprehensively with molecular precision. Several international efforts, including the The renal biopsy provides critical diagnostic and prognostic in- Kidney Precision Medicine Project, are underway using these formation for clinicians when patients develop kidney disease. To- strategies and other tissue interrogation strategies to generate day, biopsies are read using a combination of light microscopy, electron microscopy, and indirect immunofluorescence with a lim- kidney cellular expression atlases in health and disease. Re- ited number of antibodies. These techniques were all perfected cently, the first comprehensive mouse scRNA-seq dataset was decades ago. More recently, new techniques in single-cell genomics published.1 have been transforming scientists’ ability to characterize cells. Hurdles limit application of scRNA-seq to human kidney Rather than measure expression of several genes at a time by im- fl biopsies.2 Tissue availability is unpredictable, requiring access muno uorescence, it is now possible to simultaneously measure fl expression of thousands of genes in thousands of single cells by to equipment, such as uorescence-activated sorting ma- single-cell RNA-sequencing (scRNA-seq). We show that compre- chines, with little notice. Additionally, the kidney biopsy con- hensive scRNA-seq of a single human kidney transplant biopsy is sists of a very small sample of about 1032 mm, but low-input feasible and that it allows the molecular characterization of this samples pose challenges in scRNA-seq, which relies on heterogeneous tissue. double-Poisson loading of cells and oligobearing beads, leading to loss of 95% of cellular input, for example, with the types, in many cases finding expression in unexpected cells. DropSeq approach. Despite these limitations, we hypothesized Our data prove the feasibility of applying comprehensive that scRNA-seq can be successfully applied to a single human scRNA-seq to human kidney biopsies and identify complex kidney biopsy and that doing so would reveal novel insights immune cell infiltrates and proinflammatory parenchymal about disease pathogenesis. Successful application of scRNA- responses. seq to human biopsies will enable evaluation of this powerful tool as an adjunct to traditional renal biopsy interpretation. We chose to focus on renal allograft biopsies, because a METHODS substantial body of data has already established bulk transcriptional profiling to be predictive of outcomes in transplantation. Clinical Sample These analyses, however, are limited, because population- The patient described in this study consented under the Wash- averaged measurements mask single-cell transcriptomic het- ington University Kidney Translational Research Core erogeneity, particularly with rare cell types. Moreover, rejection (WUKTRC) Institutional Review Board protocol (identifica- in all of its forms involves the complex interaction of many tion 201102312). The donor and recipient HLA haplotypes, the different immune and parenchymal cells. scRNA-seq is uniquely clinical biopsy Banff scores, and follow-up clinical outcome well suited to dissecting a complex tissue into multiple cellular data were received in a deidentified manner through the subpopulations, because it can be easily scaled to tens of WUKTRC. The healthy kidney tissue was from a discarded thousands of cells. Longitudinal clinicopathologic studies, human donor kidney where donor anonymity was preserved. such asthe DeKAF Study, and bulk transcriptionalanalysis using The donor was a 70-year-old white man with a serum creat- microarrays suggest that antibody-mediated immune injury is inine of 1.1 mg/dl. the most common driver of late allograft loss.3–5 Our understanding of the molecular and cellular mechanisms of rejection, Tissue Processing and Single-Cell Dissociation in particular, antibody-mediated rejection (ABMR), is poor. The renal biopsy was minced into small pieces with a razor Analyses of allograft biopsy transcriptional profiles by micro- blade and incubated at 37°C in freshly prepared dissociation array have classified patterns of gene expression that can define buffer containing 0.25% trypsin and 40 U/ml DNase I. Disso- the different clinical phenotypes of allograft rejection6,7 or ciated cells were harvested every 10 minutes by filtering the cell predict the development of IFTA using biopsy tissue.8,9 From suspension through a 70-mm cell strainer (pluriSelect) into these studies, a large number of genes have been described that 10% FBS buffer on ice. The residual biopsy tissue trapped on play a role in allograft rejection and failure. Whether resolution the cell strainer was dissociated once again with 1 ml dissoci- of these bulk transcriptional profiles to individual cell types ation buffer for 10 minutes and passed through the cell might improve our understanding of rejection or provide for strainer into the same FBS buffer from the first collection. better disease subclassification is unknown. We repeated this dissociation procedure three times until In this study, we present the first scRNA-seq analysis of a most of the tissue had been dissociated into single cells (total single human kidney allograft biopsy. We used unsupervised dissociation time was 30 minutes). Finally, cells were collected computational approaches to identify 16 different cell types by centrifugation at 4003g for 5 minutes, resuspended in in- and novel cell states within endothelium. Wecompared tubular Drops cell suspension buffer (9% Optiprep), and strained cell transcriptomes with healthy adult kidney and thereby, through a 40-mm cell mesh (pluriSelect) to further remove identified proinflammatory parenchymal responses in the re- cell clumps and large fragments. Cell viability was approxi- jecting kidney. Wemapped genes with expression that is known mately 90% for the biopsy used in this study as assessed by to associate with allograft pathologies to specific kidney cell Trypan Blue staining.

2070 Journal of the American Society of Nephrology J Am Soc Nephrol 29: 2069–2080, 2018 www.jasn.org BASIC RESEARCH

Single-Cell and Single-Nucleus InDrops and Bioinformatic Analysis Information is in Supplemental Material.

Data Availability All relevant data have been deposited in the Gene Expression Omnibus under accession numbers GSE109564 and GSE114156.

RESULTS

A clinically indicated kidney allograft biopsy was performed on a 21-year-old white man with ESRD secondary to hemolytic uremic syndrome. He received a deceased donor kidney transplant 2 years prior (HLA mismatch 1A, 2B, 1DR) with thymoglobulin induction (Supplemental Table 1). At the time of the biopsy, his creatinine was 6.1 mg/dl, and it had increased from a base- line of 1.2 mg/dl 4 months prior. Mainte- nance immunosuppression was tacrolimus (trough levels 4–8 ng/ml), mycophenolic acid, and prednisone. He failed to get lab- oratory testing in the 4 months before biopsy, raising suspicion of medication noncompliance. The histologic diagnosis was acute T cell–mediated rejection with plasma cells graded Banff 1B and acute C4d-negative ABMR (Figure 1A, Supplemental Table 2). Donor-speciﬁc antibodies (DSAs) were positive at the time of biopsy (Supple- mental Table 3). Treatment consisted of high-dose corticosteroids (750 mg methyl- prednisone), thymoglobulin (3.5 mg/kg), IVIG (1 g/kg), and rituximab (200 mg). Allograftfunctiondidnotimprove,and he returned to dialysis 2 months later.

acid–Schiff staining revealed diffuse inflammatory infiltrates. (C) tSNE plot of cell clusters identified on the basis of the expression of highly variable genes. (D) Heat map of all cells clustered by recursive hierarchical clustering and Louvain–Jaccard clustering (Seurat) showing selected marker genes for each population. CD, collecting duct; EC, en- Figure 1. Comprehensive single-cell RNA-sequencing of an allograft biopsy reveals dothelial cell; LOH (AL), loop of Henle, ascend- diverse kidney and immune cell types. (A) Four 16-mm biopsy cores were subjected to ing limb; LOH (DL), loop of Henle, distal limb; cell dissociation. Cell yield varied between 44,000 and 75,000 cells, and viability was Mono, monocyte; PT, proximal tubule; tSNE, t- 88%–96%. (B) Biopsy A14 was used for single-cell RNA-sequencing, and periodic Distributed Stochastic Neighbor Embedding.

J Am Soc Nephrol 29: 2069–2080, 2018 Kidney Biopsy Single-Cell RNA-Sequencing 2071 BASIC RESEARCH www.jasn.org scRNA-seq Identifies Kidney Allograft Biopsy Cell proinflammatory, nonclassic monocytes.17 Of note, CD16++ Types cells are strongly associated with allograft rejection.18,19 In preliminary experiments, we used a 30-minute 0.25% trypsin Monocyte 2 seems to be a classic or intermediate monocyte cell dissociation procedure to compare the cell yield and viability population (Figure 2A, Supplemental Table 4). Interestingly, from four human kidney biopsies. Biopsy length ranged from ABCA1, which mediates sterol efflux in activated dendritic 10 to 15 mm, cell yield ranged from 44,000 to 75,000 cells, and cells (DCs),20,21 was uniquely detected in monocyte 1 (Figure viability was between 88% and 96% (Figure 1A). We next assessed 2D), likely indicating a phenotypic transition toward a DC fate the ability of two common droplet-based techniques, DropSeq10 by monocytes in allograft rejection. In support of this inter- and inDrops,11 to generate single-cell transcriptomes from hu- pretation, direct comparison of monocyte 1 with the mon- man kidney (Supplemental Figure 1). We consistently failed to o_CD16+_C1qa cluster from the PBMC dataset revealed a generate libraries using DropSeq lysis buffer and beads, but we clear separation of monocyte 1 and the PBMC monocytes, were successful using inDrops reagents (Supplemental Figure 1B). although they were highly correlated on the basis of averaged Of note, the DropSeq lysis buffer includes the detergent expression (Figure 2C). Although both cells clusters expressed sarcosyl, whereas the inDrops lysis buffer includes NP40, typical monocyte markers (Figure 2D), monocyte 1 from the suggesting the possibility that incomplete cell lysis might ex- biopsy dataset highly expressed two receptors (SDC3 and plain our failure to generate libraries using DropSeq. ABCA1) (Figure 2D) and a panel of DC maturation markers, We enzymatically dissociated one 16G biopsy core (Figure 1B) including APOE,22 PDE3A, IGKC,23 LGMN,andiCD83,24 sug- into a single-cell suspension, performed droplet-based microfluidic gesting differentiation into DC in situ (Figure 2D). cell separation using InDrops, generated libraries, and sequenced We identified unique marker genes for each monocyte clus- the cells to a read depth of approximately 50,000 reads per cell. A ter (Figure 2E) and performed immunohistochemistry on in- total of 4487 cells passed our quality filters. We detected, on dependent transplant biopsies, with histologic diagnoses of no average, 1481 transcripts from 827 different genes per cell. Un- disease, mixed rejection, or ABMR. There was sparse intersti- supervised clustering analysis using Seurat identified 16 distinct tial staining for both the monocyte 1 marker (FCGR3A or cell clusters (Figure 1C). A similar number of unique clusters CD16) and the monocyte 2 marker (FCN1) in biopsies with were identified using an independent clustering approach12 no disease. By contrast, there was strong staining for both (Supplemental Figure 2). We confirmed that there was no batch monocyte subsets in mixed rejection, with lesser infiltration effect by projecting the cells from different tubes onto the same in pure ABMR (Figure 2F, Supplemental Figure 5). Costaining tSNE (Supplemental Figure 3). by immunofluorescence analysis confirmed that the monocyte We annotated clusters on the basis of anchor gene expression subtypes are separate populations (Figure 2G). The presence and the literature. We independently validated the clusters by of these monocyte subsets in all six independent biopsies with computing Pearson correlation on averaged expression profiles mixed rejection or ABMR validates the use of scRNA-seq to derived from a recently generated mouse P1 kidney scRNA- identify novel cell types associated with kidney rejection. seq dataset and a PBMC dataset from a healthy donor (http:// Ligand-receptor analysis revealed expression of 14 receptors support.10xgenomics.com/single-cell/datasets) (Figure 2A, (excluding collagens) for which we could detect expression of Supplemental Figure 4A).13 We could, therefore, confi- their cognate ligands (Figure 2B). These ligands were detected dently annotate four tubular cell types (38.0%), three leukocyte in all cell types, emphasizing the integration of signals between populations (21.4%), four types of lymphocytes (19.8%), three multiple kidney and leukocyte cell types. Pericytes, fibroblasts, stromal cell types (8.2%), endothelial cells (ECs; 11.5%), and an and myofibroblasts expressed the chemoki CXCL12,which actively proliferating cell population (1.1%) (Supplemental promotes lymphocyte and monocyte chemotaxis through its Figure 4B). The expression of unique marker genes in each cognate receptor CXCR4, which itself is expressed in T cells, cell population confirmed the robustness of our cell type classi- monocytes, and mast cells in our dataset. Further evidence for fication approach (Figure 1D). Full lists of differentially parenchymal cell regulation of the inflammatory response expressed genes in each cluster are provided in Supplemental comes from the observation that collecting duct epithelia ex- Table 4. A searchable database for this dataset, including gene press KITLG, also known as stem cell factor, which binds the expression projected onto the tSNE diagram, is available at transmembrane receptor encoded by KIT and is expressed on http://humphreyslab.com/SingleCell/. mast cells in the biopsy (Figure 2B).25 Stem cell factor promotes mast cell migration, adhesion, proliferation, and sur- Comparative Analysis Reveals Monocyte Cell-State vival. Mast cells transcripts correlate strongly with allograft Transition and Cell-Cell Interactions between Many biopsy fibrosis.26 These results suggest the unexpected hy- Cell Types pothesis that collecting duct epithelia actively coordinate Monocyte infiltration is quantitatively associated with kidney dys- mast cell infiltration during rejection. Consistent with an im- function during allograft rejection.14,15 By mapping our dataset portant role for mast cells in kidney injury, a recent study to a published single-cell PBMC dataset,16 we identified two dis- showed that mast cell ablation in the early phases of renal tinct monocyte clusters (Figure 2A). Monocyte 1 strongly ex- injury is sufficient to reduce subsequent fibrosis by decreasing pressed FCGR3A (CD16) and was most similar to CD16-positive, the inflammatory response.27,28

2072 Journal of the American Society of Nephrology J Am Soc Nephrol 29: 2069–2080, 2018 www.jasn.org BASIC RESEARCH

Figure 2. Annotation of leukocyte subsets. (A) Heat map indicating Pearson correlations of the averaged transcriptional proﬁles between human PBMCs and leukocyte clusters from the kidney biopsy. (B) Ligand-receptor pair expression according to cell type. Ligands are indicated in the left panel, and receptors are indicated in the right panel. Straight lines indicate ligand-receptor pairs. (C) The biopsy monocyte 1 cluster is most similar to CD16+ peripheral blood monocytes on the basis of average expression; however, these cells form different clusters on the basis of tSNE, indicating that they are different cell types. (D) The monocyte 1 cluster is differen- tiating toward an acquired dendritic cell phenotype on the basis of expression of marker genes. (E) Violin plot showing that FCGR3A (CD16) distinguishes monocyte 1 from monocyte 2, whereas FCN1 is expressed in monocyte 2 but not monocyte 1. MSR1 is expressed in both clusters. (F) Immunohistochemistry for FCGR3A or FCN1 on normal, mixed rejected, or pure antibody-mediated rejection (ABMR) transplant kidney biopsies. Distinct monocyte 1 and 2 cell types can be seen. Upper and lower panels are serial sections. Scale bar, 50 mm. (G) Immunoﬂuorescence analysis of mixed rejection. FCN1+ cells (arrowheads) and FCGR3A+ cells (arrows) are separate cell types. CD, collecting duct; DAPI, 49,6-diamidino-2-phenylindole; EC, endothelial cell; LOH (AL), loop of Henle, ascending limb; LOH (DL), loop of Henle, distal limb; Mono, monocyte; PT, proximal tubule. Scale bar, 10 mm.

Activation of Epithelial, Endothelial, and Stromal Cells sequenced 4259 nuclei to a similar depth as the biopsy and in Allograft Rejection identiﬁed six distinct epithelial cell clusters, including podo- We next compared epithelial transcriptomes from the biopsy cytes, proximal tubule, loop of Henle, distal tubule, principal with their healthy counterparts. Multiple attempts at scRNA- cells, and intercalated cells (Figure 3, A–C). The absence of seq of healthy nephrectomy tissue failed to generate libraries; stromal or leukocyte populations presumably reﬂects either however, we were successful in generating adult human kid- dissociation bias and/or a cell frequency below our limit of ney single-nucleus RNA-sequencing (RNA-seq) data. We detection.

J Am Soc Nephrol 29: 2069–2080, 2018 Kidney Biopsy Single-Cell RNA-Sequencing 2073 BASIC RESEARCH www.jasn.org

Comparison of scRNA-seq and single- nucleus RNA-seq datasets has been shown to be valid after normalizing to reduce method-specificdifferences.29 Using a computational approach allowing for integrated analysis of multiple datasets across techniques,30 we compared proximal tubule, loop of Henle, and collecting duct cells from our biopsy dataset with the same cell types from our healthy adult kidney single-nucleus RNA-seq dataset. Unsu- pervised analysis of combined cells from all six clusters (three healthy and three rejection) by tSNE resulted in the expected three cell types, indicating that rejection did not fundamentally alter tubular cell identity (Figure 3D, Supplemental Figure 6). Con- sistent with this interpretation, projecting thesourceofcellsontothistSNEcon- firmed that healthy proximal tubule was overlaid with rejecting proximal tubule as was loop of Henle and collecting duct (Figure 3E). Despite the overall similarity between healthy and rejecting proximal tubule, loop of Henle, and collecting duct, differential gene analysis revealed downregula- tion of terminal differentiation markers and upregulation of proinflammatory genes in biopsy epithelia (Figure 3F). The proximal tubule in particular expressed proinflammatory cytokine genes, such as CXCL14, which induces monocyte chemotaxis, and IL32, which induces expression of TNFa by macrophages (Figure 3F). ITGB6, a gene previously shown to be expressed in distal tubule of the diseased kidney transplant,31 was specifically induced in the loop of Henle in our biopsy dataset. Within collecting duct epithelia, the top statistically significant canonical signature Figure 3. Comparison of epithelia from single-cell RNA-sequencing of healthy adult by GSEA that distinguished the biopsy col- kidney with transplant biopsy reveals activated and proinflammatory cell states. (A) lecting duct from the healthy collecting Unsupervised clustering identified six distinct cell types in human adult kidney. These duct was the TGF-b/BMP signaling signa- types include three tubular cell types (proximal tubule [PT], loop of Henle (LH), and ture (Supplemental Figure 7). Collectively, distal tubule (DT), two collecting duct (CD) cell populations principal cells (PC) and intercalated cells (IC), and one podocyte population (P). (B) The heat map showed that putative molecular signature marks the identity of each cluster. (C) The violin plot further confirmed the clean expression of well known cell type–specificmarkersin kidney. (G) Ordering of healthy and activated each cell population, which makes it suitable for use in benchmarked comparison PT cells along pseudotime using Monocle. (H) analysis. (D) tSNE analysis of PT, loop of Henle (LOH), and CD cells from the allograft Selected transcription factors that are upre- biopsy and healthy adult human kidney cluster together indicating that cell identity is gulated during PT activation. LOH-bx, loop of maintained despite allograft inflammation. (E) Collecting duct cells from the biopsy Henle from the biopsy; LOH-h, loop of Henle (CD-bx) coproject by tSNE onto the collecting duct cluster from healthy kidney (CD-h). from healthy kidney; PT-bx, proximal tubule The same is true of PT and LOH. (F) A dot plot comparing expression of terminal from the biopsy; PT-h, proximal tubule from differentiation or inflammatory genes in epithelial cells from the biopsy or healthy healthy kidney.

2074 Journal of the American Society of Nephrology J Am Soc Nephrol 29: 2069–2080, 2018 www.jasn.org BASIC RESEARCH these results highlight the degree to which epithelial damage predominant EC cluster was in the resting state, because it during rejection amplifies proinflammatory and profibrotic expressed EC markers (PECAM1, VWF,andENG)butno responses. transcripts previously associated with ABMR or angiogenesis To further investigate proximal tubule injury responses, we (Figure 4C). We compared average gene expression of the performed cell trajectory analysis by reconstituting the healthy three clusters with healthy human brain and pancreas ECs, and biopsy proximal tubule clusters by pseudotime analysis. and the resting cluster correlated best (Figure 4D). A smaller We found a smooth transition between healthy and activated subcluster expressed the angiogenic program,33 with several of proximal tubule cells (Figure 3G and F). We next asked what these genes previously described in association with ABMR genes might be regulating proximal tubule activation by plot- (Figure 4, C and E)34,35 A third EC cluster expressed markers ting differentially expressed transcription factors across pseu- of endoplasmic stress (XBP1) and the cold shock gene RBM3, dotime. Among others, we observed strong upregulation of suggesting cell stress and activation. This cluster was also Sox9, which has recently been identified as a master regulator actively transcribing Ig, such as has been observed in vitro of proximal tubule repair after injury, validating our (Figure 4C).36 The src family kinases have been shown to be approach.32 activated on ingestion of IgG. Consistent with this, FYN is Donor ECs are the primary targets of the recipients’ hu- highly expressed in this EC subcluster.37 Thus, these ECs moral immune response in ABMR. Reclustering of the ECs maybe internalizing bound IgG, consistent with DSA-EC revealed three separate cell states in rejection (Figure 4, A and binding in ABMR. Studies have shown that this process occurs B). Differential gene expression analysis suggested that the in EC via Fc receptor–mediated phagocytosis.38 In line with

Figure 4. Analysis of endothelial cell (EC) subsets. (A) tSNE plot of endothelial subclusters from the kidney biopsy. (B) Heat map showing selected marker genes for the three EC subclusters. (C) Violin plots showing conserved expression of endothelial markers across all three subclusters. The genes SEMA3G, IGFBP3, SERPINE2,andAQP1 indicate activation of an angiogenic program. A separate EC subcluster expresses Igs as well as markers of endoplasmic reticulum stress (XBP1) and cold shock (RBM3). (D) Pearson correlation of all three EC subsets shows best correlation between the resting EC cluster and healthy brain or pancreas ECs. (E) Markers of antibody-mediated rejection are detected in the angiogenic cluster.

J Am Soc Nephrol 29: 2069–2080, 2018 Kidney Biopsy Single-Cell RNA-Sequencing 2075 BASIC RESEARCH www.jasn.org this notion, our GO analysis data showed that the top GO express genes involved in antigen presentation (CD86) and T terms for this cluster were all related to phagocytosis (Supple- cell regulation (indoleamine 2,3-dioxygenase, IDO1), thus be- mental Figure 8). having like a naïve B cell. Plasma cell clusters highly express Ig We identified three separate stromal clusters (Supplemental genes; however, cluster 1 was k-chain restricted, suggesting Table 4). Cluster 1 highly expresses RSG5, a pericyte marker,39 that these cells are monoclonal. Plasma cell cluster 2 was not as well as CACNA1C, which is expressed in a pericyte pattern. light-chain restricted, and it highly expressed both k-and Stromal cluster 2 expresses MoxD1, a gene upregulated in early l-gene types (Supplemental Figure 9B). These clusters were kidney transplant biopsies and associated with subsequent fi- labeled plasma cells on the basis of the correlation with a brosis.40 This cluster likely represents a fibroblast cell type. PBMC reference dataset; however, these cells could also rep- The third stromal cluster uniquely expressed COL8A1 and resent mature B cells. COL12A1, two genes that we have previously shown to be strongly upregulated in kidney myofibroblasts during fibrotic Reassessing the Molecular Signature for Disease injury, suggesting a myofibroblast identity for this cluster.41 Classification at the Single-Cell Level The interstitial expression patterns for marker genes from We next assigned expression of genes previously associated each stromal cluster are shown in Figure 5. with allograft pathologies to single-cell types.42 When we mapped putative EC-associated transcripts, we found that Antigen Expression and Antibody Production most of these genes were actually not specifically expressed Because alloantigens play a central role in ABMR, we profiled in ECs. Some of them were even uniquely expressed in non- the expression of HLA transcripts. Appropriately, class 2HLA EC types, such as ENPEP (PT), ANGPT1 (fibroblast), MGP transcripts were predominantly expressed in the antigen- (myofibroblast), and FOSB (mast cells) (Figure 6A). Over one presenting cell clusters (Supplemental Figure 9A). Seurat anal- half of the transcripts were expressed in multiple cell types. ysis of our biopsy revealed two plasma cell clusters and one B Whether disease outcome correlates best with expression of cell cluster. The B cell cluster expressed few Ig genes but did these genes in one cell type and not others will require further

Figure 5. Marker gene localization to independent stromal cell types. The expression of known and new cell type discriminating genes. Violin plots are shown adjacent to immunohistochemistry, which is from the Human Protein Atlas (https://www.proteinatlas.org/). Examples of pericyte-, fibroblast-, and myofibroblast-specific marker genes are shown. In addition, DCN and COL6A1 are expressed in both fibroblast and myofibroblast populations, whereas SPARC, BGN, and COL6A2 are expressed in all three stromal cell types.

2076 Journal of the American Society of Nephrology J Am Soc Nephrol 29: 2069–2080, 2018 www.jasn.org BASIC RESEARCH

Figure 6. Mapping cell type expression of genes previously associated with kidney allograft pathology. (A) Expression of 35 endothelial cell (EC)–associated genes identified in bulk transcriptional profiling assigned to individual biopsy cell types (likelihood ratio test). A heat map was used to visualize the z score–normalized average gene expression of the candidate genes for each cell cluster identified from the biopsy dataset. A majority of EC-associated transcripts are not expressed in ECs. (B) A similar analysis performed on transcripts associated with T cell–mediated rejection reveals that nearly all of them are expressed in leukocytes and that the majority are expressed in T cells. (C) A similar analysis on transcripts associated with antibody-mediated rejection reveals that most of the genes were expressed in EC clusters. CD, collecting duct; LOH (AL), loop of Henle, ascending limb; LOH (DL), loop of Henle, distal limb; Mono, monocyte; PT, proximal tubule. investigation. By contrast, the 30 genes with expression that is the 103 Chromium system are much better suited to analyze associated with T cell–mediated rejection are predominantly single human kidney biopsy. expressed in immune cell types, with the majority expressed Our study is limited to a single biopsy, and therefore, results in T cells and almost no expression in kidney parenchyma cannot be generalized. However, our findings do reveal several (Figure 6B).43 Finally, most of the transcripts reported to be novel insights into human allograft rejection. Anti-HLA anti- enriched in ABMR biopsies35 were mainly detected in the EC bodies are known to play important roles in rejection, with cluster in our biopsy dataset (Figure 6C), substantiating the much attention focused on EC responses to antibody binding. critical role of endothelium in the pathogenesis of ABMR.44 Previous studies suggest that antibody binding causes a number of EC responses, including cell proliferation,45–48 induction of secondary factors (FGFR and VEGF),49,50 and DISCUSSION leukocyte recruitment (including vWF and P-selectin expression),51–53 as well as cell survival responses. Our data show two This study establishes the feasibility of using scRNA-seq to distinct pathologic EC responses in mixed rejection: a resting comprehensively measure gene expression from thousands state and two ABMR response states consisting of an angio- of cells from a single human kidney biopsy. We were able to genic state or an Ig phagocytosis state. The resting-state cells identify 16 separate cell types from just under 5000 single-cell probably elicit the initial humoral response, because they ex- libraries. We are confident that future improvements of our press donor-specific HLA and leukocyte adhesion molecules basic workflow will enable the generation of ten times as many (PECAM1). Evaluating the EC responses to different anti- single-cell libraries from a single human biopsy, which should bodies in the setting of ABMR will help us better understand allow for much finer cell type separation and deeper transcript why DSAs confer variable degrees of pathogenicity. Therapeu- detection. Our experience suggests that standard DropSeq ap- tically, the B cell-plasma cell lineage is an attractive target for proaches are poorly suited for this application, however, po- therapy, because inhibition of alloantibody production will tentially due to incomplete cell lysis as well as inefficient likely improve outcomes in ABMR. To this end, we identified cell capture rates. By contrast, InDrops, as we used here, or B cells that are likely naïve, polyclonal plasma cells (or mature

J Am Soc Nephrol 29: 2069–2080, 2018 Kidney Biopsy Single-Cell RNA-Sequencing 2077 BASIC RESEARCH www.jasn.org

B cells), and a monoclonal plasma cell cluster representing the ACKNOWLEDGMENTS spectrum of antibody-producing cells. Wehaveadditionally provided the first single-nucleus RNA- This work was supported by National Institutes of Health/National seq dataset from healthy adult human kidney. The reasons for Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) our failure to generate scRNA-seq libraries from healthy adult grant DK107374 (to B.D.H.), grant 173970 (to B.D.H.) from the kidney compared with our success with a diseased transplant Chan Zuckerberg Initiative, NIDDK Diabetic Complications Con- biopsy are unclear. We speculate that the robust inflammation sortium (www.diacomp.org) grants DK076169 (to B.D.H.) and present in the transplant biopsy made the tissue more amena- DK115255 (to B.D.H.), and The Keller Family Fund at the Founda- ble to dissociation. However, this points to a major current tion for Barnes-Jewish Hospital (B.D.H.). challenge and limitation in the application of single-cell technologies to human kidney—the difficulty of successful dissociation without RNA degradation and the associated problem DISCLOSURES of dissociation bias. For example, we could not identify podo- None. cytes in the transplant biopsy, whereas they were present in the healthy kidney single-nucleus dataset, although similar numbers of cells/nuclei were analyzed. However, the healthy REFERENCES human kidney dataset contained only epithelial cell types without stroma, endothelium, or leukocytes. These striking 1. Park J, Shrestha R, Qiu C, Kondo A, Huang S, Werth M, et al.: Single-cell differences indicate that dissociation methods selectively en- transcriptomics of the mouse kidney reveals potential cellular targets of rich for certain cell types, while losing others—either because kidney disease [published online ahead of print April 5, 2018]. Science doi:10.1126/science.aar2131 they die during dissociation or because they remain in clumps. 2. Malone AF, Wu H, Humphreys BD: Bringing renal biopsy interpretation Clearly, future work is needed to define the optimal cell/ into the molecular age with single-cell RNA sequencing. Semin nucleus dissociation approach for human kidney. It is worth Nephrol 38: 31–39, 2018 considering that nuclear dissociation can be carried out en- 3. Sellarés J, de Freitas DG, Mengel M, Reeve J, Einecke G, Sis B, et al.: tirely on ice—which eliminates the artefactual transcriptional Understanding the causes of kidney transplant failure: The dominant role of antibody-mediated rejection and nonadherence. Am J Trans- stress responses that are known to arise on incubation of kid- plant 12: 388–399, 2012 13 ney tissue with proteases at 37°C. 4. El-Zoghby ZM, Stegall MD, Lager DJ, Kremers WK, Amer H, Gloor JM, An additional important consideration is that information et al.: Identifying specific causes of kidney allograft loss. Am J Trans- concerning cell positioning is lost with scRNA-seq. Such infor- plant 9: 527–535, 2009 mation may be critical to properly interpret results. For 5. Gaston RS, Cecka JM, Kasiske BL, Fieberg AM, Leduc R, Cosio FC, example, a ligand-receptor analysis between podocytes and in- et al.: Evidence for antibody-mediated injury as a major determinant of late kidney allograft failure. Transplantation 90: 68–74, 2010 tercalated cells does not make sense biologically, because they are 6. Halloran PF, Pereira AB, Chang J, Matas A, Picton M, De Freitas D, et al.: positioned so far apart, whereas the same comparison between Potential impact of microarray diagnosis of T cell-mediated rejection in principle cells and intercalated cells is biologically meaningful. kidney transplants: The INTERCOM study. Am J Transplant 13: 2352– New experimental and computational tools are in development 2363, 2013 to address the loss of spatial context with scRNA-seq. One ap- 7. Halloran PF, Pereira AB, Chang J, Matas A, Picton M, De Freitas D, et al.: Microarray diagnosis of antibody-mediated rejection in kidney trans- proach, fluorescent in situ sequencing,54 allows for genome-wide fi fi — plant biopsies: An international prospective study (INTERCOM). Am J expression pro ling in xed tissues preserving tissue architec- Transplant 13: 2865–2874, 2013 ture for RNA localization studies. In addition, computational 8. O’Connell PJ, Zhang W, Menon MC, Yi Z, Schröppel B, Gallon L, et al.: approaches can integrate scRNA-seq data with in situ hybridiza- Biopsy transcriptome expression profiling to identify kidney transplants tion RNA expression patterns, allowing inference of spatial ex- at risk of chronic injury: A multicentre, prospective study. Lancet 388: – pression patterns by virtual in situ hybridization.55 983 993, 2016 9. Modena BD, Kurian SM, Gaber LW, Waalen J, Su AI, Gelbart T, et al.: Despite current challenges, we envision a time in the future Gene expression in biopsies of acute rejection and interstitial fibrosis/ when kidney biopsies are routinely sent for scRNA-seq as part tubular atrophy reveals highly shared mechanisms that correlate with of a molecular diagnostic workup. scRNA-seq can unmask the worse long-term outcomes. Am J Transplant 16: 1982–1998, 2016 transcriptional heterogeneity that is hidden in population- 10. Macosko EZ, Basu A, Satija R, Nemesh J, Shekhar K, Goldman M, et al.: fi averaged measurements. We illustrate the power of this Highly parallel genome-wide expression pro ling of individual cells using nanoliter droplets. Cell 161: 1202–1214, 2015 approach by analyzing paracrine signaling pathways between 11. Klein AM, Mazutis L, Akartuna I, Tallapragada N, Veres A, Li V, et al.: infiltrating leukocytes and kidney parenchyma and by map- Droplet barcoding for single-cell transcriptomics applied to embryonic ping predictive gene expression to single-cell types. We expect stem cells. Cell 161: 1187–1201, 2015 that application of scRNA-seq technologies to human native 12. Kiselev VY, Kirschner K, Schaub MT, Andrews T, Yiu A, Chandra T, et al.: and allograft biopsies will ultimately generate clinically useful SC3: Consensus clustering of single-cell RNA-seq data. Nat Methods 14: 483–486, 2017 datasets to improve diagnostic and prognostic accuracy, enable 13. Adam M, Potter AS, Potter SS: Psychrophilic proteases dramatically disease subphenotyping, and accelerate adoption of molecular reduce single-cell RNA-seq artifacts: A molecular atlas of kidney de- biopsy interpretation. velopment. Development 144: 3625–3632, 2017

2078 Journal of the American Society of Nephrology J Am Soc Nephrol 29: 2069–2080, 2018 www.jasn.org BASIC RESEARCH

14. Mannon RB: Macrophages: Contributors to allograft dysfunction, re- 33. Reinders ME, Rabelink TJ, Briscoe DM: Angiogenesis and endothelial pair, or innocent bystanders? Curr Opin Organ Transplant 17: 20–25, cell repair in renal disease and allograft rejection. JAmSocNephrol17: 2012 932–942, 2006 15. Girlanda R, Kleiner DE, Duan Z, Ford EA, Wright EC, Mannon RB, et al.: 34. Sis B, Jhangri GS, Bunnag S, Allanach K, Kaplan B, Halloran PF: Monocyte infiltration and kidney allograft dysfunction during acute Endothelial gene expression in kidney transplants with alloantibody rejection. Am J Transplant 8: 600–607, 2008 indicates antibody-mediated damage despite lack of C4d staining. Am 16. Zheng GX, Terry JM, Belgrader P, Ryvkin P, Bent ZW, Wilson R, et al.: J Transplant 9: 2312–2323, 2009 Massively parallel digital transcriptional profiling of single cells. Nat 35. Sellarés J, Reeve J, Loupy A, Mengel M, Sis B, Skene A, et al.: Molecular Commun 8: 14049, 2017 diagnosis of antibody-mediated rejection in human kidney transplants. 17. Ziegler-Heitbrock L, Ancuta P, Crowe S, Dalod M, Grau V, Hart DN, Am J Transplant 13: 971–983, 2013 et al.: Nomenclature of monocytes and dendritic cells in blood. Blood 36. Zhao Y, Liu Y, Chen Z, Korteweg C, Gu J: Immunoglobulin g (IgG) ex- 116: e74–e80, 2010 pression in human umbilical cord endothelial cells. J Histochem Cy- 18. van den Bosch TPP, Hilbrands LB, Kraaijeveld R, Litjens NHR, tochem 59: 474–488, 2011 Rezaee F, Nieboer D, et al.: Pretransplant numbers of CD16+ 37. Majeed M, Caveggion E, Lowell CA, Berton G: Role of Src kinases and monocytes as a novel biomarker to predict acute rejection after Syk in Fcgamma receptor-mediated phagocytosis and phagosome- kidney transplantation: A pilot study. Am J Transplant 17: 2659– lysosome fusion. J Leukoc Biol 70: 801–811, 2001 2667, 2017 38. Goebl NA, Babbey CM, Datta-Mannan A, Witcher DR, Wroblewski 19. Vereyken EJ, Kraaij MD, Baan CC, Rezaee F, Weimar W, Wood KJ, VJ, Dunn KW: Neonatal Fc receptor mediates internalization of Fc in et al.: A shift towards pro-inflammatory CD16+ monocyte subsets with transfected human endothelial cells. Mol Biol Cell 19: 5490–5505, preserved cytokine production potential after kidney transplantation. 2008 PLoS One 8: e70152, 2013 39.HeL,VanlandewijckM,RaschpergerE,AndaloussiMäeM,JungB, 20. Westerterp M, Gautier EL, Ganda A, Molusky MM, Wang W, Fotakis P, Lebouvier T, et al.: Analysis of the brain mural cell transcriptome. Sci et al.: Cholesterol accumulation in dendritic cells links the in- Rep 6: 35108, 2016 flammasome to acquired immunity. Cell Metab 25: 1294–1304.e6, 40. Wei C, Menon M, Zhang W, Woytovich C, Philippe N, He J, Murphy B, 2017 Wei C: Genomic Analysis of Gene Expression in Human Kidney 21. Castella B, Kopecka J, Sciancalepore P, Mandili G, Foglietta M, Mitro Transplants Identifies Moxd1 as a Key Driver for Renal Fibrosis N, et al.: The ATP-binding cassette transporter A1 regulates phos- [abstract]. Am J Transplant 2017 17(suppl 3). Available at: http:// phoantigen release and Vg9Vd2 T cell activation by dendritic cells. Nat atcmeetingabstracts.com/abstract/genomic-analysis-of-gene- Commun 8: 15663, 2017 expression-in-human-kidney-transplants-identifies-moxd1-as-a-key- 22. Stephens TA, Nikoopour E, Rider BJ, Leon-Ponte M, Chau TA, driver-for-renal-fibrosis/. Accessed May 21, 2018 Mikolajczak S, et al.: Dendritic cell differentiation induced by a self- 41. Grgic I, Krautzberger AM, Hofmeister A, Lalli M, DiRocco DP, Fleig SV, peptide derived from apolipoprotein E. JImmunol181: 6859–6871, et al.: Translational profiles of medullary myofibroblasts during kidney 2008 fibrosis. J Am Soc Nephrol 25: 1979–1990, 2014 23. Worah K, Mathan TSM, Vu Manh TP, Keerthikumar S, Schreibelt G, Tel 42. Einecke G, Sis B, Reeve J, Mengel M, Campbell PM, Hidalgo LG, et al.: J, et al.: Proteomics of human dendritic cell subsets reveals subset- Antibody-mediated microcirculation injury is the major cause of late specific surface markers and differential inflammasome function. Cell kidney transplant failure. Am J Transplant 9: 2520–2531, 2009 Rep 16: 2953–2966, 2016 43. Venner JM, Famulski KS, Badr D, Hidalgo LG, Chang J, Halloran PF: 24. Lechmann M, Berchtold S, Hauber J, Steinkasserer A: CD83 on den- Molecular landscape of T cell-mediated rejection in human kidney dritic cells: More than just a marker for maturation. Trends Immunol 23: transplants: Prominence of CTLA4 and PD ligands. Am J Transplant 14: 273–275, 2002 2565–2576, 2014 25. Okayama Y, Kawakami T: Development, migration, and survival of mast 44. Wang S, Zhang C, Wang J, Yang C, Xu M, Rong R, et al.: Endothelial cells. Immunol Res 34: 97–115, 2006 cells in antibody-mediated rejection of kidney transplantation: Patho- 26. Mengel M, Reeve J, Bunnag S, Einecke G, Sis B, Mueller T, et al.: genesis mechanisms and therapeutic implications. J Immunol Res Molecular correlates of scarring in kidney transplants: The emergence 2017: 8746303, 2017 of mast cell transcripts. Am J Transplant 9: 169–178, 2009 45. Coupel S, Leboeuf F, Boulday G, Soulillou JP, Charreau B: RhoA 27. Danelli L, Madjene LC, Madera-Salcedo I, Gautier G, Pacreau E, Ben activation mediates phosphatidylinositol 3-kinase-dependent pro- Mkaddem S, et al.: Early phase mast cell activation determines the liferation of human vascular endothelial cells: An alloimmune chronic outcome of renal ischemia-reperfusion injury. JImmunol198: mechanism of chronic allograft nephropathy. JAmSocNephrol15: 2374–2382, 2017 2429–2439, 2004 28. Holdsworth SR, Summers SA: Role of mast cells in progressive renal 46. Jin YP, Korin Y, Zhang X, Jindra PT, Rozengurt E, Reed EF: RNA in- diseases. JAmSocNephrol19: 2254–2261, 2008 terference elucidates the role of focal adhesion kinase in HLA class I- 29. Lake BB, Codeluppi S, Yung YC, Gao D, Chun J, Kharchenko PV, et al.: mediated focal adhesion complex formation and proliferation in human A comparative strategy for single-nucleus and single-cell tran- endothelial cells. J Immunol 178: 7911–7922, 2007 scriptomes confirms accuracy in predicted cell-type expression from 47. Jin YP, Jindra PT, Gong KW, Lepin EJ, Reed EF: Anti-HLA class I anti- nuclear RNA. Sci Rep 7: 6031, 2017 bodies activate endothelial cells and promote chronic rejection. 30. Butler A, Hoffman P, Smibert P, Papalexi E, Satija R: Integrating single- Transplantation 79[Suppl]: S19–S21, 2005 cell transcriptomic data across different conditions, technologies, and 48. Jindra PT, Jin YP, Rozengurt E, Reed EF: HLA class I antibody-mediated species. Nat Biotechnol 36: 411–420, 2018 endothelial cell proliferation via the mTOR pathway. J Immunol 180: 31. Trevillian P, Paul H, Millar E, Hibberd A, Agrez MV: Alpha(v)beta(6) in- 2357–2366, 2008 tegrin expression in diseased and transplanted kidneys. Kidney Int 66: 49. Bieri M, Oroszlan M, Farkas A, Ligeti N, Bieri J, Mohacsi P: Anti-HLA I 1423 –1433, 2004 antibodies induce VEGF production by endothelial cells, which in- 32. Kumar S, Liu J, Pang P, Krautzberger AM, Reginensi A, Akiyama H, creases proliferation and paracellular permeability. Int J Biochem Cell et al.: Sox9 activation highlights a cellular pathway of renal repair in Biol 41: 2422–2430, 2009 the acutely injured mammalian kidney. Cell Rep 12: 1325–1338, 50. Bian H, Harris PE, Reed EF: Ligation of HLA class I molecules on 2015 smooth muscle cells with anti-HLA antibodies induces tyrosine

J Am Soc Nephrol 29: 2069–2080, 2018 Kidney Biopsy Single-Cell RNA-Sequencing 2079 BASIC RESEARCH www.jasn.org

phosphorylation, ﬁbroblast growth factor receptor expression and cell expression proﬁling in intact cells and tissues. Nat Protoc 10: 442–458, proliferation. Int Immunol 10: 1315–1323, 1998 2015 51. Lowenstein CJ, Morrell CN, Yamakuchi M: Regulation of Weibel- 55. Karaiskos N, Wahle P, Alles J, Boltengagen A, Ayoub S, Kipar C, et al.: Palade body exocytosis. Trends Cardiovasc Med 15: 302–308, The Drosophila embryo at single-cell transcriptome resolution. Science 2005 358: 194–199, 2017 52. Yamakuchi M, Kirkiles-Smith NC, Ferlito M, Cameron SJ, Bao C, Fox-Talbot K, et al.: Antibody to human leukocyte antigen triggers endothelial exocytosis. Proc Natl Acad Sci U S A 104: 1301–1306, 2007 See related perspective, “Single-Cell Sequencing the Glomerulus, Unraveling 53. Fishbein MC, Kobashigawa J: Biopsy-negative cardiac transplant re- the Molecular Programs of Glomerular Filtration, One Cell at a Time,” on pages – jection: Etiology, diagnosis, and therapy. Curr Opin Cardiol 19: 166 2036–2038. 169, 2004 54. Lee JH, Daugharthy ER, Scheiman J, Kalhor R, Ferrante TC, Terry R, This article contains supplemental material online at http://jasn.asnjournals. et al.: Fluorescent in situ sequencing (FISSEQ) of RNA for gene org/lookup/suppl/doi:10.1681/ASN.2018020125/-/DCSupplemental.

2080 Journal of the American Society of Nephrology J Am Soc Nephrol 29: 2069–2080, 2018 Supplemental Information

Single Cell Transcriptomics of a Human Kidney Allograft Biopsy Defines a Diverse Inflammatory Response

Haojia Wu*, Andrew F. Malone*, Erinn Donnelly, Yuhei Kirita, Kohei Uchimura, Sai M. Ramakrishnan, Joseph Gaut and Benjamin D. Humphreys

* These authors contributed equally

Inventory of Supplemental Information

Figures S1 – S9

Supplementary Methods

Supplemetary Tables S1 – S4

Supplemental References

Figure S1. Bioanalyzer traces of library preparations resulting from the two microfluidic-based single cell techniques, Dropseq and inDrops. (A) cDNA trace for a library prepared using the Dropseq method. This method was unsuccessful in making quality libraries from a human biopsy tissue core sample. (B) cDNA trace for a library prepared using the inDrops method. Using this method we were able to create a quality cDNA library from a human biopsy tissue core.

Figure S2. Consensus clustering of our single cell RNA-seq data using SC3. The SC3 method is another unsupervised method used to cluster cells. This method is highly accurate and robust and combines multiple clustering approaches into one. Using SC3 we found that our initial clustering using Seurat was replicated using SC3.

Figure S3. Minimal batch effects observed between sequencing platforms and library preparations. Cells processed using the inDrops method on one biopsy sample were collected sequentially into 4 tubes. Libraries prepared from tubes 1 to 3 were sequenced using the Illumina HiSeq 2500 platform and the library prepared from tube 4 was sequenced using the Illumina NextSeq platform. Projecting cells into one tSNE map shows that all clusters from each tube overlap.

Figure S4. Comparison of cell-type specific gene expression from a P1 mouse kidney and a human PBMC RNAseq dataset. (A) Heatmap comparing cell-type specific gene expression from a P1 mouse kidney RNAseq dataset (y-axis) with our cell-type specific gene expression (x-axis). The results confirm our cluster annotations. Color key denotes the Pearson correlation score. (B) Proportion of cell types within our biopsy sample.

Figure S5. Low power views of independent transplant biopsies stained for monocyte#1 and monocyte#2 subsets. FCGR3A identifies the monocyte#1 subset, and staining was absent from transplant biopsies with a histologic diagnosis of no rejection, but strong staining was present in biopsies with a histologic diagnosis of mixed rejection, and intermediate staining in ABMR biopsies. The monocyte#2 marker FCN1 exhibited a very similar pattern. The top and bottom slides were from serial sections. Scale bar 100 µm.

Figure S6. Confirmation of proximal tubule, loop of Henle and collecting duct cell clusters. Clustering analysis of epithelial cell types from the combined biopsy and normal human samples show proximal tubular cell markers CUBN and LRP2 in the proximal tubule cluster, UMOD and SLC12A1 in the loop of Henle cluster and AQP2 and AQP3 in the collecting duct cluster.

Figure S7. Upregulated TGF-β/BMP signaling signature in collecting duct cells from biopsy. (A) GSEA enrichment plot of the TGF-β receptor signaling pathway, one of the top pathways upregulated in biopsy CD when compared with the healthy CD. (B) Expression of selected TGF-β/BMP pathway genes were mapped to healthy vs. biopsy epithelial cell types. Unexpectedly, a very strong TGF-β/BMP signature was observed in collecting duct epithelia from the biopsy kidney sample (CD-bx) compared to healthy collecting duct (CD-h). Receptor-ligand pairs were identified from human the Database of Ligand−Receptor Partners (DLRP), IUPHAR and Human Plasma Membrane Receptome (HPMR).

Figure S8. GO analysis on the EC subgroup that expressed immunoglobulins. The results were obtained from the ToppGene suite using differentially expressed genes in the activated EC subcluster (p value<0.01) as the input gene list. Note the enrichment for immunoglobulin binding and phagocytosis, suggesting antibody- mediated activation of endothelial cells in this biopsy with antibody-mediated rejection.

Normalized UMI CountNormalized UMI

Figure S9. Distribution of HLA and light chain transcript expression across cell types. (A) The class I HLA transcripts are expressed equally across all cell types. The Class II HLA transcripts are predominantly expressed in the professional antigen presenting cells (APCs), monocyte cluster 1 and 2 and B cells. The class II transcripts corresponding to the donor specific antibodies circulating at the highest concentration (anti-HLA-DRB1/DRA and anti- HLA-DRB5/DRA) are expressed highly in the APCs. (B) Light chain transcript expression in lymphocytes and plasma cells. Plasma cell cluster number 1 is expressing only kappa light chains (IGKC) and is thus light chain restricted. This suggests that this cluster represents a single clone. Plasma cell cluster number 2 expresses kappa and lambda light chain constant regions (IGKC and IGLC) and is therefore not light chain restricted is thus expresses polyclonal.

Supplementary Methods InDrops single cell RNA-seq InDrops was performed as described1. In brief, cells were diluted into 60,000 cells/mL in 9% Optiprep buffer. Single cell encapsulation was carried out using an inDrops instrument and microfluidic chip manufactured by 1CellBio. In total, four tubes of cells with about 1,000 cells/tube were collected. Library preparation was performed according to the protocol provided by the manufacturer. Of note, we used low PCR cycles (8-9 cycles) to amplify the libraries from this sample as our previous trials with higher PCR cycles (12-13 cycles) resulted in an overamplification of the libraries where averaged transcripts/cell detected was less than 300. Libraries were sequenced by HiSeq 2500 and NextSeq with a sequencing depth of 50K mapped reads/cell.

InDrops single nucleus RNA-seq Nuclei were isolated with Nuclei EZ Lysis buffer (Sigma #NUC-101) supplemented with protease inhibitor (Roche #5892791001) and RNase inhibitor (Promega #N2615, Life Technologies #AM2696). Samples were cut into <2 mm pieces and homogenized using a Dounce homogenizer (Kimble Chase #885302-0002) in 2ml of ice-cold Nuclei EZ Lysis buffer and incubated on ice for 5 min with an additional 2ml of lysis buffer. The homogenate was filtered through a 40-µm cell strainer (pluriSelect #43-50040-51) and then centrifuged at 500 x for 5 min at 4 ºC. The pellet was resuspended and washed with 4 ml of the buffer and incubated on ice for 5 min. After another centrifugation, the pellet was resuspended with Nuclei Suspension Buffer (1x PBS, 0.07% BSA, 0.1% RNase inhibitor), filtered through a 20- µm cell strainer (pluriSelect 43-50020-50) and counted. RNA from single nucleus was encapsulated, barcoded and reversed transcribed using an InDrop microfluidics system (1CellBio). The library was sequenced in HiSeq2500 with custom primers.

InDrops data preprocessing We used a recently developed inDrops computational pipeline, dropEst,1 to process the single cell and single nuclei InDrops data. In brief, cell barcodes and UMIs from the library were extracted from read1 by the dropTag program and added to the names of the transcript reads, resulting in a new fastq file for read alignment. We used STAR (version 2.5.3a) to map the high quality reads to the human genome (GRCh38). Only reads that were uniquely mapped to the genome (~70% of the total reads) were used for UMIs counts. We next ran the dropEst program to estimate the accurate molecular counts, which generated a UMI count matrix for each gene in each cell.

Unsupervised clustering and cell type identification UMI count matrices from four tubes were combined and loaded into the R package Seurat. For normalization, the DGE matrix was scaled by total UMI counts, multiplied by 10,000 and transformed to log space. Only genes found to be expressed in >10 cells were retained. Cells with a relatively high percentage of UMIs mapped to mitochondrial genes (>=0.3) were discarded. Moreover, cells with fewer than 300 or more than 4,000 detected genes were omitted, resulting in 4,487 cells. We also regressed out the variants arising from library size and percentage of mitochondrial genes using the function RegressOut in R package Seurat. The highly variable genes were identified using the function MeanVarPlot with the parameters: x.low.cutoff = 0.0125, x.high.cutoff = 6 and y.cutoff = 1, resulting in an output of 2,404 highly variable genes. The expression level of highly variable genes in the cells was scaled and centered along each gene, and was conducted to principal component analysis. We then assessed the number of PCs to be included in downstream analysis by (1) plotting the cumulative standard deviations accounted for each PC using the function PCElbowPlot in Seurat to identify the ‘knee’ point at a PC number after which successive PCs explain diminishing degrees of variance, and (2) by exploring primary sources of heterogeneity in the datasets using the PCHeatmap function in Seurat. Based on these two methods, we selected first 20 PCs for two-dimensional t-distributed stochastic neighbor embedding (tSNE), implemented by the Seurat software with the default parameters. Based on the tSNE map, sixteen clusters were identified using the function FindCluster in Seurat with the resolution parameter set to 0.6. Alternatively, we used SC32 to validate the clusters identified by Seurat with cluster number set to 14 (ks=14). We found that the cells clustered by Seurat as a cell type were also grouped together by SC3. Differentially expressed genes that were expressed at least in 25% cells within the cluster and with a fold change more than 0.25 (log scale) were considered to be marker genes. In total, 2,837 marker genes were identified for the clusters in the biopsy dataset. A heatmap of selected marker gene expression across clusters was plotted using a Python plotting library Matplotlib. We applied the same unsupervised clustering analysis on the single nucleus dataset. First, we generated the digital expression matrix using the same dropEst pipeline that utilizes both exonic and intronic reads. After filtering low quality nuclei, 4,259 nuclei with > 400 genes expressed were imported into Seurat for clustering analysis. In total, we identified six kidney cell types in the single nucleus dataset, including PT, LOH, intercalated cells (IC), principal cells (PC), DCT and podocytes.

Comparison of immune cell types to a published PBMC single cell dataset Cell-type-specific expression patterns of the cell clusters identified in our dataset were compared to signatures previously defined in a PBMC dataset by calculating the pairwise Pearson correlations coefficients between each pair of cell types for the same set of genes. First, a precomputed Seurat object containing cell cluster information for 33K human PBMCs was downloaded from the Satija lab (http://satijalab.org/seurat/get_started.html). Only genes detected in both our dataset and the PBMC dataset were used for downstream correlation analysis. Second, pearson correlation was computed between the cell clusters in our dataset and the cell clusters identified in the PBMC dataset, using the previously defined cell-type annotations and normalized average gene expression values for each cell type. Data was shown by pheatmap R package.

To compare monocyte transcriptomes between the biopsy and PBMC datasets, we extracted the expression profiles for the cluster we annotated as monocyte#1 and the corresponding monocyte cluster (mono_CD16+_C1qa) in the PBMC dataset. We then clustered the cells after removing cell-cell variations driven by the number of detected molecules and mitochondrial gene contents. Cells in tSNE were colored by the original dataset where they are extracted, or the cell cluster where they are assigned based on unsupervised clustering analysis. To compare the marker gene expression in the monocyte from different sources, we extracted the scaled UMI expression values for the selected genes that are known as monocyte markers, receptors and DC differentiation markers. Average expression level of these selected genes was visualized as the violin plot using ggplot2 package in R.

Correlation analysis of the kidney cell types from allograft biopsy and P1 mouse kidney To assess the similarity between biopsy cell types and embryonic kidney, we re-analyzed a recently published Dropseq dataset from P1 kidney (GSE94333).3 We used the Seurat clustering parameters described by the authors and reproduced the same cell types from the datasets. We calculated the Pearson correlation based on the expression patterns of highly variable genes between cell populations within the mouse embryonic kidney dataset against the cell types identified in our biopsy dataset. Correlation matrix were visualized by R package pheatmap. Color keys represent the range of the coefficients of determination (r2) in this analysis.

Ligand-receptor interaction analysis To study ligand-receptor interactions across the cell types identified from the transplant biopsy, we used a human ligand–receptor list comprising 2,557 ligand–receptor pairs curated by the Database of Ligand−Receptor Partners (DLRP), IUPHAR and Human Plasma Membrane Receptome (HPMR).4, 5 We selected the receptors that were only differentially expressed in each cell type from the biopsy dataset. To determine the ligand-receptor pairs to plot on the heatmap, we required (i) the receptors are uniquely expressed in each cell type (q-val<0.05 and logFC>0.6); (ii) Each receptor should have at least one corresponding ligand to pair with. We used heatmap.2 function from gplots package to visualize the ligand- receptor pairs in each cell type.

Comparison of healthy kidney and allograft kidney To compare the transcriptional difference across each tubular cell type in healthy and disease state, we extracted the expression profiles for PT, LOH, and PC (CD) from the healthy kidney single nucleus dataset and the biopsy single cell dataset. We then performed integrated analysis on both datasets using a recently developed computational strategy6 (implemented in Seurat v2.0). First, we selected the union of the top 2,000 genes with the highest dispersion from both datasets for a canonical correlation analysis (CCA) to identify common sources of variation between the two datasets. Then we aligned the CCA subspaces using the first 15 dimensions of the CCA. After CCA alignment, we performed clustering analysis on the nuclei and cells with the resolution parameter set to 0.6. We visualized the cells by their original identity or by their cluster identity classified by this integrated analysis. Differential gene analysis was performed on the nuclei and cells from different datasets but grouped in the same cluster by the alignment analysis. Differential genes were visualized using DotPlot function in Seurat.

Pseudotemporal ordering of PT single cells and TF analysis We used Monocle27 to draw a minimal spanning tree connecting the PT cells collected from healthy and diseased kidneys. As input into Monocle2, we selected the highly variable genes for cell ordering as described in the Monocle2 tutorial (http://cole-trapnell- lab.github.io/monocle-release/docs_mobile/). We then reduced the data space to two dimensions using the reduceDimension function with ‘DDRTree’ method and ordered the cells using the orderCells function in Monocle2. Individual cells were color-coded based on the kidneys where they were collected. To identify the transcription factors whose expression are dynamically changing across the trajectory from healthy PT to diseased PT, we first identified the genes that were differentially expressed across pseudotime using the differentialGeneTest function in Monocle2 with the fullModelFormulaStr parameter set to ‘Psudeotime’. To identify the TFs, we crossed the differential genes to the human TF list downloaded from AnimalTFDB (http://bioinfo.life.hust.edu.cn/AnimalTFDB/). We selected some of the important TFs that have been reported by the literature as key regulators of proximal tubule injury and repair and visualized them by R package ggplot2.

Analysis of heterogeneity in the EC population We performed unsupervised clustering analysis on the endothelial cells extracted from the biopsy dataset using Seurat with similar parameters as described above. Top markers in each subcluster were visualized with the DoHeatmap function in Seurat. We used a stacked violin plot to show the expression of endothelial markers, angiogenic markers, and immunoglobulin genes across the three EC subgroups. Marker genes were selected based on literature search, and were differentially expressed among the EC subclusters. GO analysis was performed by uploading the differential genes from the EC subclusters to ToppGene Suite (http://toppgene.cchmc.org). Top 4 GO terms from categories of biological process, molecular function and cellular component were visualized using ggplot2 package.

Re-assessing disease-associated gene lists from the public datasets To map the disease-associated genes to single cell, we selected the top genes from three public datasets where the authors identified EC enriched transcripts8, TCMR enriched genes,9 and ABMR enriched markers10 by comparing the microarray data from the biopsies with different rejection patterns. Expression values of these genes in our single cell dataset were normalized, z-scored and visualized the heatmap.2 in gplots R package.

DropSeq single-cell RNA sequencing Dropseq was performed on the dissociated biopsy cells as previously described.11 Briefly, biopsy was dissociated into single cell using the same methods as being used in inDrops. Single cell suspension was visually inspected under a microscope, counted by hemocytometer (INCYTO C-chip) and resuspended in PBS + 0.01% BSA. Single cells were coencapsulated in droplets with barcoded beads purchased from Chemgenes (MACOSKO- 2011-10). cDNA libraries were constructed according to an online Dropseq protocol (McCarroll’s lab: http://mccarrolllab.com/dropseq/). cDNA quality was determined by BioAnalyzer (Agilent) with a high sensitivity DNA chip.

GSEA pathway analysis GSEA (http://software.broadinstitute.org/gsea/index.jsp) was used to estimate the enriched pathways in the collecting duct subgroups from biopsy and healthy kidney. We used the normalized UMI count matrix generated by Seurat (NormalizeData and ScaleData) as an expression dataset, and a gene set containing all human pathways downloaded from Bader Lab (http://download.baderlab.org/EM_Genesets/current_release/Human/Entrezgene/) as gene set file input to GSEA. We used 1,000 gene label permutations and identified significantly enriched pathways defined by adjusted p value < 0.05 between the collecting duct cells from biopsy and healthy kidney.

Supplementary Tables

Table S1

HLA DSAs at biopsy Mean fluorescence intensity DRB5*01:01 mfi 12143 (Immunodominant DSA) DRB1*15:01 mfi 2354 DQA1*03:02/DQB1*03:03 mfi 1750 DQA1*03:01/DQB1*03:03 mfi 1660

Table S2

Banff t i v ptc g cg ci ct mm ah cv C4d criteria score 3 3 0 3 2 0 0 0 0 1 0 0

Table S3

HLA A B C Bw4 Bw6 DRB1 DRB3/4/5 DQB1 DPB1 Recipient 24/11 40(61)/51 01/03(10) + + 07/14 B4*01:03N 03(9)/05 B3*02

Donor 03:01/ 07:02/ - + 15:01/ DRB4*01:03N 03:03/ 04:01 24:02 07:61 07:01 DRB5*01 06:02

Tables S1-3: Clinical Features of Study Biopsy. (1) Haplotype matching of donor a recipient at class I and class II loci. (2) Full Banff criteria scoring for the study biopsy showed histologic features consistent with Banff 1B T-cell mediated rejection and acute C4d-negative antibody mediated rejection. (3) Donor specific antibody (DSA) found in the patients serum at the time of biopsy. mfi- mean fluorescent intensity, t = tubulitis, i = interstitial lymphocyte infiltration, v = intimal arteritis, ptc = peritubular capillaritis, g = glomerulitis, cg = transplant glomerulopathy, ci = interstitial fibrosis, ct = tubular atrophy, mm = mesangial matrix expansion, ah = arteriolar hyalinosis, cv = vascular intimal fibrosis.

Supplementary Table S4. Link to data set .xls

Supplementary References

1. Klein, AM, Mazutis, L, Akartuna, I, Tallapragada, N, Veres, A, Li, V, Peshkin, L, Weitz, DA, Kirschner, MW: Droplet barcoding for single-cell transcriptomics applied to embryonic stem cells. Cell, 161: 1187-1201, 2015. 2. Kiselev, VY, Kirschner, K, Schaub, MT, Andrews, T, Yiu, A, Chandra, T, Natarajan, KN, Reik, W, Barahona, M, Green, AR, Hemberg, M: SC3: consensus clustering of single- cell RNA-seq data. Nat Methods, 14: 483-486, 2017. 3. Adam, M, Potter, AS, Potter, SS: Psychrophilic proteases dramatically reduce single-cell RNA-seq artifacts: a molecular atlas of kidney development. Development, 144: 3625-3632, 2017. 4. Ramilowski, JA, Goldberg, T, Harshbarger, J, Kloppmann, E, Lizio, M, Satagopam, VP, Itoh, M, Kawaji, H, Carninci, P, Rost, B, Forrest, AR: A draft network of ligand- receptor-mediated multicellular signalling in human. Nat Commun, 6: 7866, 2015. 5. Hrvatin, S, Hochbaum, DR, Nagy, MA, Cicconet, M, Robertson, K, Cheadle, L, Zilionis, R, Ratner, A, Borges-Monroy, R, Klein, AM, Sabatini, BL, Greenberg, ME: Single-cell analysis of experience-dependent transcriptomic states in the mouse visual cortex. Nat Neurosci, 2017. 6. Butler, A, Satija, R: Integrated analysis of single cell transcriptomic data across conditions, technologies, and species. bioRxiv, 2017. 7. Qiu, X, Mao, Q, Tang, Y, Wang, L, Chawla, R, Pliner, HA, Trapnell, C: Reversed graph embedding resolves complex single-cell trajectories. Nat Methods, 14: 979-982, 2017. 8. Einecke, G, Sis, B, Reeve, J, Mengel, M, Campbell, PM, Hidalgo, LG, Kaplan, B, Halloran, PF: Antibody-mediated microcirculation injury is the major cause of late kidney transplant failure. Am J Transplant, 9: 2520-2531, 2009. 9. Venner, JM, Famulski, KS, Badr, D, Hidalgo, LG, Chang, J, Halloran, PF: Molecular landscape of T cell-mediated rejection in human kidney transplants: prominence of CTLA4 and PD ligands. Am J Transplant, 14: 2565-2576, 2014. 10. Sellares, J, Reeve, J, Loupy, A, Mengel, M, Sis, B, Skene, A, de Freitas, DG, Kreepala, C, Hidalgo, LG, Famulski, KS, Halloran, PF: Molecular diagnosis of antibody- mediated rejection in human kidney transplants. Am J Transplant, 13: 971-983, 2013. 11. Macosko, EZ, Basu, A, Satija, R, Nemesh, J, Shekhar, K, Goldman, M, Tirosh, I, Bialas, AR, Kamitaki, N, Martersteck, EM, Trombetta, JJ, Weitz, DA, Sanes, JR, Shalek, AK, Regev, A, McCarroll, SA: Highly Parallel Genome-wide Expression Profiling of Individual Cells Using Nanoliter Droplets. Cell, 161: 1202-1214, 2015.

SIGNIFICANCE STATEMENT

The renal biopsy provides critical diagnostic and prognostic information for clinicians when patients develop kidney disease. Today, biopsies are read using a combination of light microscopy, electron microscopy, and indirect immunoﬂuorescence with a limited number of antibodies. These techniques were all perfected decades ago. More recently, new techniques in single-cell genomics have been transforming scientists’ ability to characterize cells. Rather than measure expression of several genes at a time by immunoﬂuorescence, it is now possible to simultaneously measure expression of thousands of genes in thousands of single cells by single-cell RNA-sequencing (scRNA-seq). We show that comprehensive scRNA-seq of a single human kidney transplant biopsy is feasible and that it allows the molecular characterization of this heterogeneous tissue.