BASIC RESEARCH www.jasn.org

Proximal Tubule Translational Profiling during Kidney Fibrosis Reveals Proinflammatory and Long Noncoding RNA Expression Patterns with Sexual Dimorphism

Haojia Wu,1,2 Chun-Fu Lai,1,2,3 Monica Chang-Panesso,1,2 and Benjamin D. Humphreys 1,2,4

1Division of Nephrology, Departments of 2Medicine and 4Developmental Biology, Washington University in St. Louis School of Medicine, St. Louis, Missouri; and 3Renal Division, Department of Internal Medicine, National Taiwan University Hospital, Taipai, Taiwan

ABSTRACT Background Proximal tubule injury can initiate CKD, with progression rates that are approximately 50% faster in males versus females. The precise transcriptional changes in this nephron segment during fibrosis and potential differences between sexes remain undefined. Methods We generated mice with proximal tubule–specific expression of an L10a ribosomal subunit pro- tein fused with enhanced green fluorescent protein. We performed unilateral ureteral obstruction surgery BASIC RESEARCH on four male and three female mice to induce inflammation and fibrosis, collected proximal tubule–specific and bulk cortex mRNA at day 5 or 10, and sequenced samples to a depth of 30 million reads. We applied computational methods to identify sex-biased and shared molecular responses to fibrotic injury, including up- and downregulated long noncoding RNAs (lncRNAs) and transcriptional regulators, and used in situ hybridization to validate critical genes and pathways. Results We identified .17,000 genes in each proximal tubule group, including 145 G-protein–coupled receptors. More than 700 transcripts were differentially expressed in the proximal tubule of males versus females. The .4000 genes displaying altered expression during fibrosis were enriched for proinflam- matory and profibrotic pathways. Our identification of nearly 150 differentially expressed proximal tubule lncRNAs during fibrosis suggests they may have unanticipated regulatory roles. Network analysis prioritized proinflammatory and profibrotic transcription factors such as Irf1, Nfkb1, and Stat3 as drivers of fibrosis progression. Conclusions This comprehensive transcriptomic map of the proximal tubule revealed sexually dimorphic gene expression that may reflect sex-related disparities in CKD, proinflammatory gene modules, and previously unappreciated proximal tubule–specific bidirectional lncRNA regulation.

JASN 31: 23–38, 2020. doi: https://doi.org/10.1681/ASN.2019040337

The medical and economic burden of CKD remains high worldwide and new therapies are needed.1,2 Received April 2, 2019. Accepted August 1, 2019. fi The nal common pathway of all progressive H.W. and C.-F.L. contributed equally to this work. CKD is tubulointerstitial fibrosis, a process charac- fi Published online ahead of print. Publication date available at terized histologically by myo broblast expansion, www.jasn.org. extracellular matrix deposition, tubular atrophy, and peritubular capillary rarefaction. Activation Correspondence: Dr. Benjamin D. Humphreys, Division of Ne- – phrology, Department of Medicine, Washington University in of pericytes and mesenchymal stem cell like Saint Louis School of Medicine, 660 South Euclid Avenue, stroma into proliferative, matrix-secreting myofi- CampusBox8129,StLouis,MO63110.Email:humphreysbd@ broblasts represent key steps in pathologic fibrosis.3 wustl.edu Upstream stimuli for myofibroblast activation are Copyright © 2020 by the American Society of Nephrology

JASN 31: 23–38, 2020 ISSN : 1046-6673/3101-23 23 BASIC RESEARCH www.jasn.org less well understood but are of considerable interest because Significance Statement such pathways represent potential therapeutic targets. Increas- ing evidence implicates epithelial cells, in particular those of the Having a comprehensive transcriptional profile of the proximal tu- proximal tubule (PT), as fibrosis-initiating cells. More expres- bule in health and fibrosis would likely enhance understanding of fi sion quantitative trait loci for CKD are expressed in the PT than brosis and perhaps help explain why CKD progresses more quickly in males versus females. To obtain a more complete picture of gene 4 any other epithelial cell type. Both ischemic and toxic insults to expression in the proximal tubule, the authors performed deep the kidney primarily affect the PT and these insults—whether translational profiling of this segment in a mouse model of kidney genetic, ischemic, toxic, or otherwise—lead to adaptive and fibrosis. Their findings demonstrate substantial sex differences in maladaptive cellular responses including cell cycle arrest, met- transcripts expressed in proximal tubule cells of males versus fe- fi abolic reprogramming, and secretion of profibrotic and proin- males, and indicate that the proximal tubule drives brosis through inflammatory and profibrotic paracrine signaling. The study also fl fi 5–7 ammatory factors leading to myo broblast activation. identified 439 long noncoding RNAs expressed in the proximal Both PT dysfunction and the secreted factors that are a conse- tubule, 143 of which undergo differential regulation in fibrosis, quence of the response to epithelial injury represent logical suggesting that this type of RNA has unanticipated regulatory roles therapeutic targets to slow progression in CKD. kidney fibrosis. In part because they are unbiased, transcriptomic studies rep- resent a powerful approach to better understand kidney disease. protocols were conducted in accordance with the guidelines fi Bulk kidney transcriptional pro ling has revealed core gene ex- of the Institutional Animal Care and Use Committee of the pression signatures in diabetic kidney disease, AKI and transition Washington University in St. Louis (number 20180070). We to chronic disease, unilateral ureteral obstruction (UUO), and crossed homozygous Slc34a1-GFPCreERT2 against homozy- 8–11 new roles for B lymphocytes in dysfunctional kidney repair. gous R26-EGFP-L10a mice to generate compound heterozy- fi Cell-speci c transcriptomic studies are increasingly favored over gous Slc34a1-GFPCreERT2; R26-EGFP-L10a mice (hereafter bulk approaches. Fluorescence-activated cell sorting can gener- referred to as SLC34a1-eGFPL10a. Adult mice (8–12 weeks old) fi ate cell-speci c transcriptome data, but is limited by the require- were used for all experiments. Tamoxifen (Sigma-Aldrich, ment for surface antibodies and the cellular dissociation and St. Louis, MO) was dissolved in 3% (vol/vol) ethanol contain- FACS procedure itself, which induces transcriptional stress re- ing corn oil at a concentration of 10 mg/ml. To activate expres- sponses.12 Recently, single cell RNA–sequencing (scRNA-seq) sion of eGFP-L10a in PT, mice received two doses of 3 mg of approaches have been applied to understand mammalian kid- tamoxifen gavage over 3 days. This achieved .90% recombi- ney development and adult kidney in health and disease.13–16 nation in the cortical PTs.20 Ten days after the last dose of Although these are very powerful approaches, they still detect tamoxifen, mice underwent left UUO surgery, as previously only the more highly expressed genes in a cell and may miss low described.15 Mice were euthanized and UUO and contralateral to midlevel gene expression compared with cell-specific bulk kidney (CLK) were harvested at 5 or 10 days after the surgery. RNA sequencing (RNA-seq). Each group included three female and four male mice. The aim of our study was to generate a comprehensive and unbiased inventory of gene expression changes in the PT during fibrosis to serve as a scientific resource. We have previously vali- TRAP and RNA-Seq fi dated translating ribosome affinity purification (TRAP) as a pow- Puri cation of polysome-bound RNA from kidneys was per- 18,19 erful means to generate cell-specifictranscriptionalprofiles from formed according to our published reports and the orig- 21 podocytes and myofibroblasts in health and disease as well as the inal publication. RNA integrity was assessed using an RNA – transcriptional response of all nephronepithelia to acute injury and PicoChip (catalog number 5067 1513; Agilent Bioanalyzer). repair.17–19 We applied TRAP to the cortical PT in health, fibrosis, The Clontech SMARTer library kit (Takara Bio USA, Mountain and between sexes, and compared it to whole cortex mRNAs. We View, CA) was used for cDNA library construction. Seven report substantial differences in gene expression between male libraries were prepared in each group (three from female and female mice exist in healthy PT. We also define major proin- and four from male mice) and sequenced by HiSeq3000 to a flammatory gene modules upregulated in the PT after injury, depth of 30 million reads per sample. Reads were aligned to implicating this cell type as a key driver of fibrosis. Finally, we mm10 (Ensembl release 76) with STAR version 2.0.4b.22 report that over 143 long noncoding RNAs (lncRNAs) undergo Gene counts were derived from the number of uniquely both up- and downregulation in the PT, suggesting unappreciated aligned unambiguous reads by Subread:featureCount ver- cell-specific regulatory roles for lncRNAs in renal fibrosis. sion 1.4.5.23 Transcript counts were produced by Sailfish version 0.6.3. Sequencing performance was assessed by measuring the total number of aligned reads; total number METHODS of uniquely aligned reads, genes, and transcripts detected; ribosomal fraction known junction saturation; and read Animals distribution over known gene models with RSeQC version 2.3.24 Slc34a1-GFPCreERT2 and R26-EGFP-L10a mouse lines were Bioinformatic analyses were carried out as described in generated as described.19,20 All animal care and experimental Supplemental Methods.

24 JASN JASN 31: 23–38, 2020 www.jasn.org BASIC RESEARCH

In Situ Hybridization Real-Time PCR For each gene of interest, 400- to 800-bp DNA fragments were cDNAwas synthesized by reverse transcription from extracted synthesized by PCR of cDNA with a 59 SP6 promoter and a RNA by iScript cDNA Synthesis Kit (Bio-Rad Laboratories, 39 T7 promoter. Primer sequences are in Supplemental Meth- Hercules, CA). Expression of target genes was determined ods. In situ hybridization (ISH) was performed in RNase-free by real-time quantitative PCR using the iTaq SYBR Green conditions according to a published protocol9 with modifica- Supermix and the iQ5 Multicolor Real-Time PCR Detection tions. Briefly, 15 mm frozen tissue sections were fixed in 4% System (Bio-Rad). Glyceraldehyde-3-phosphate dehydroge- paraformaldehyde/PBS overnight at 4°C, permeabilized with nase was used as the internal control. Proteinase K (10 mg/ml) for 20 minutes, postfixed, and incu- bated in acetylation solution (0.375% acetic anhydride) for Data Availability 10 minutes. Slides were then incubated with digoxigenin- RNA-seq data for all samples have been deposited in the Gene labeled sense or antisense riboprobes (0.5 mg/ml riboprobes Expression Omnibus (GSE125015). To increase rigor and re- in hybridization buffer of 50% formamide, 50 mg/ml yeast producibility, the raw sequencing data as well as analysis files transfer RNA, 1% SDS, 50 mg/ml heparin, and 53 SSC) at 68°C generated as part of this study were also uploaded into the (Re) overnight. Hybridization was followed by stringency washes, Building a Kidney consortium database and are fully accessible blocking (2% blocking reagent in PBS; Roche), and incubation at https://doi.org/10.25548/16-E08W.25 with anti–digoxigenin-alkaline phosphatase antibody (catalog number 11093274910, 1:4000; Roche) overnight at 4°C. After washing, sections were treated with NTMT solution (100 mM RESULTS sodium chloride, 100 mM Tris pH 9.5, 50 mM magnesium chloride, 0.1% Tween 20, 2 mM tetramisole) for 10 minutes. Translational Profiling of the PT during Fibrosis The final reaction was developed by adding chromogenic We generated Slc34a1-eGFPL10a mice, performed UUO sur- substrate (BM Purple, catalog number 11442074001; Roche) gery, and euthanized mice at day 5 or 10 (Figure 1, A and B). for 1–7 days until sufficient staining intensity was reached eGFP-L10a expression was exclusively localized to the PT in or signals started to appear in the negative control sections a typical ribosomal nucleolar pattern (Figure 1C). At day 10 incubated by sense riboprobes. Sections were fixed by 4% of UUO, Acta2 (aSMA) expression was dramatically induced paraformaldehyde/PBS again and then mounted by mounting in the interstitium as expected (Figure 1D). RNA quality from medium (ProLong gold; ThermoFisher). Images were captured TRAP was excellent among all groups (average RNA integrity with a 203 objective on a Zeiss Axio Scan Z1. scoreofmorethannine,SupplementalFigure1A).Wese- quenced each sample to a depth of 30 million reads, enabling Immunofluorescence Staining detection of .17,000 genes per sample (Supplemental Table 1 After fixation in 4% paraformaldehyde/PBS for 90 minutes, and Supplemental Figure 1A). This is more than twice the gene kidney specimens were immersed in 30% sucrose/PBS overnight detectionatlessthanhalfthesequencing depth compared at 4°C, embedded in Tissue-Tek optimal cutting temperature with a recent report of bulk RNA-seq of microdissected compound (Miles Inc., Elkhart, IN) in liquid nitrogen, and PT which could detect approximately 7300 transcripts in stored at 280°C until cryostat sectioning. the S1 and S2 segments (Supplemental Figure 1B and C).26 Cryopreserved sections (6 mm) were subsequently washed There was a strong correlation between the biologic replicates in PBS for 10 minutes, permeabilized by 0.25% Triton X-100/PBS in each group (Supplemental Figure 1D), suggesting high for 10 minutes, and blocked with 5% BSA for 60 minutes reproducibility across experiments. at room temperature. Sections were incubated with chicken Multidimensional scaling (MDS) analysis did not distin- anti-GFP antibodies (catalog number GFP-1020, 1:300; Aves, guish between sex or disease time points (day 5 UUO and Tigard, OR) at 4°C overnight. On the next day, the sections day 10 UUO), whereas robust differences were detected be- were washed twice in 0.1% Tween 20/PBS for 5 minutes each tweenthesampletypes(PTortotalcortex)anddiseasestate and then incubated with Alexa Flour 488–conjugated donkey (CLKandUUO;Figure1E).Asshownbythescatterplotof anti-chicken antibody (catalog number 703-545-155, 1:200; the whole translatome (Figure 1F) and quantitative PCR of Jackson ImmunoResearch) and Cy3-conjugated anti–mouse- selected genes (Figure 1G), PT samples were appropriately aSMA antibody (catalog number C6198, 1:400; Sigma-Aldrich) enriched for PT markers as expected (Slc34a1, Slc5a2,and for 1 hour at room temperature. After three washes in 0.1% Slc2a2) but not other kidney-type markers (e.g., Nphs1 and Tween 20/PBS for 5 minutes each, sections were colabeled Synpo for podocytes, Aqp2 and Aqp4 for collecting duct, with 4’,6-diamidino-2-phenylindole (1:1000; Invitrogen) Col1a1 and Acta2 for myofibroblasts, Slc12a1 and Umod for and mounted with ProLong Gold. The primary antibodies loop of Henle, and Pecam1 and Emcn for endothelial cells). were omitted in sections as negative controls. Images were Compared with total cortex, 119 transporters and ion channels captured and processed using a confocal microscope were significantly enriched in the PT samples (false discovery (Eclipse Ti, Nikon). Sections were examined in a blinded rate [FDR] ,0.05, Supplemental Table 2), consistent with the fashion. physiologic role of the PT in the kidney.

JASN 31: 23–38, 2020 Proximal Tubule Transcriptomics 25 BASIC RESEARCH www.jasn.org

A exon2 B Slc 34a1 eGFP CreERT2 SV 40pA CLK D5UUO D10UUO

loxP loxP Rosa 26 eGFP-L10a 3-pA pA CAGGS

Slc34a1GCE-eGFPL10a

anti-GFP Ab

beads CD GFP-L 10a CLK CLK D10UUO Bound mRNA: PT Unbound mRNA: Total

42 Samples

Male Female GFP/ aSMA / DAPI

PT Total PT Total GFP epifluorescence

CLK D5 D10 CLK D5 D10 CLK D5 D10 CLK D5 D10 UUO UUO UUO UUO UUO UUO UUO UUO

n=4 per group n=3 per group

EFMulti-dimensional scaling (MDS)

Disease 2 CLK.PT Slc34a1 10 CLK.Total Slc5a2 1 D5UUO.PT Slc2a2 Umod D5UUO.Total 5 0 D10UUO.PT

Dim2 Slc12a1 Synpo Aqp2 D10UUO.Total Nphs1 Aqp4 Acta2 -1 0 Gender Bound (PT) moderated log2CPM Col1a1 Emcn Female Pecam1

-2 Male

-2 0 2 0510 Dim1 Unbound (cortex) moderated log2CPM

G Slc34a1 Slc5a2 Slc2a2 Nphs1 2.0 5 5 3

1.5 4 4

3 2 3 1.0

Fold change 2 2 1 0.5 1 1 0

Umod Aqp2 Aqp4 2.0 4 Group

0.9 CLK.Cortex 3 1.5 CLK.PT UUO.Cortex 2 0.6 1.0 UUO.PT Fold change 1 0.3 0.5

0 0.0 0.0

Figure 1. Successful translational profiling of PT using the SLC34a1-eGFPL10a line. (A) Schematic illustrating the experimental work flow. (B) Trichrome stain showing collagen deposition in the kidney interstitium (male mice) after day 5 (D5UUO) and 10 (D10UUO) UUO. (C) Perinuclear and nucleolar expression of eGFP in PT (male mice), a pattern consistent with ribosomal location. Scale bar, 50 mm. (D) Immunofluorescence staining shows a strong induction of interstitial aSMA in the UUO kidney (male mice) and no

26 JASN JASN 31: 23–38, 2020 www.jasn.org BASIC RESEARCH

33 TRAP Reveals Sexually Dimorphic t1/2 and increasing TNF-a proinflammatory signaling. Finally, Gene Expression in PT RhoC encodes Ras homolog family member C, a small GTPase MDS analysis on the PTsamples showed no difference between that regulates cell motility, cell division, and protein trafficking. day 5 and 10, so we combined those time points. The combined Although RhoC has never been described in the kidney, Rho- PTsamples were separated based on both healthy versus disease kinase inhibitors are known to be antifibrotic.34 We observed and male versus female (Figure 2A). A total of 1013 genes were strong downregulation of genes expressed in differentiated ep- differentially expressed between male and female in healthy ithelia. Slc22a8, also known as organic anion transporter 3, is PT (FDR,0.05), and the number of differentially expressed strongly expressed in cortical PT, but after UUO it was almost genes (DEGs) was linearly correlated with the cutoff set for undetectable (Figure 3B). 2 fold change (R =0.97; Supplemental Figure 2A). When log2 We next quantified the added specificity that PT TRAP fold change was set to .1 and less than 21forup-anddown- provides compared with profiling of whole cortex. We com- regulated genes, we detected 748 genes were differentially pared the DEGs identified in the PT to those identified in the expressed between sexes (Supplemental Figure 2B). The vast whole cortex. Of the DEGs from fibrotic PT, 22.7% were not majority of these sex-specific genes (approximately 95%) were differentiallyexpressedinwholecortex(Figure3C,Supplemental located in autosomes rather than sex chromosomes (Figure 2B). Figure 3C). By contrast, 26.3% of the DEGs from whole cortex In addition, 64.8% of these genes were also differentially did not change in PT (Figure 3D, Supplemental Figure 3C). Thus expressed during UUO (Figure 2C, Supplemental Table 2). PT-specificprofiling by TRAP is both more sensitive and more We validated two of these genes by ISH (male, Cndp2; female, specific than bulk cortex profiling, despite the fact that PT Hao2) (Figure 2D) and additionally by reanalysis of published makes up the majority of cortical cell mass. scRNA-seq data sets (female, Kynu and Rdh16; male, Ces1f, We then compared the disease genes identified in this study Cyp4b1, Slc22a30,andCyp2e1; Figure 2E, Supplemental to the disease signature identified in our prior TRAP data using Table 3).15,16,27 These expression differences may reflect some a Six2-eGFPL10a line. Although this comparison was limited of the known sex-related disparities in CKD epidemiology28 because these two data sets were generated from different Cre and progression.29 driver lines (Slc34a1-CreERt2 versus Six2-Cre), different animal models (UUO versus ischemia reperfusion injury [IRI]), and PT Gene Signatures of Fibrosis different profiling platforms (RNA-seq versus microarray); There were .4000 genes with altered PT expression after ob- we found 64.9% of disease DEGs identified in the Six2 data struction (Figure 3A). About 80% of the DEGs were shared set (IRI versus sham) were detected by this study (UUO versus between day 5 and 10 UUO versus CLK (Supplemental Figure control) (Supplemental Figure 4A). We then performed gene 3A), suggesting most of the transcriptional changes occurring ontology (GO) analysis separately on the upregulated DEGs in this model of fibrosis happen by day 5, although we cannot unique to Six2 TRAP, unique to Slc34a1 TRAP, and shared by exclude the contribution of changes in RNA stability. We both (Supplemental Figure 4B). Interestingly, the top enriched observed a very similar trend when comparing day 5 versus terms for the Six2 DEGs, Slc34a1 DEGs, and shared DEGs were day 10 whole cortex (Supplemental Figure 3B). apoptosis, inflammation, and fibrosis, respectively (Supple- We validated expression of four upregulated genes (Cstb, mental Figure 4C). These results are consistent with the tubular S100a10, Carhsp1,andRhoC)infibrotic PT by ISH (Figure disease phenotypes we expect for IRI and UUO. 3B). None of these genes have known roles in the kidney. CstB encodes cystatin B, a protease inhibitor of lysosomal cathep- PT Gene Signatures for Epithelial-to-Mesenchymal sins with antiapoptotic and anti-inflammatory actions in the Transition, Fatty Acid Metabolism, Hedgehog brain and immune system.30 S100a10 encodes a plasminogen Signaling, and G protein–coupled receptors receptor that binds to plasminogen and facilitates its activa- Our comprehensive data set can be used to assess the role of tion to plasmin, a protease that regulates matrix remodeling. the PT in fibrogenesis. Tubular epithelia were once thought Intriguingly, S100a10 has been shown to promote mesenchy- to directly contribute to the interstitial myofibroblast pool in mal transition of epithelial cells in cancer models.31 Carhsp1 fibrosis through a process of “full epithelial-to-mesenchymal encodes calcium-regulated heat stable protein 1 containing a transition” (full EMT). Lineage tracing data have disproven cold-shock domain and two RNA-binding motifs.32 It binds to that hypothesis, although epithelia do take on a mesenchymal the 39 untranslated region of TNF-a mRNA, doubling transcript phenotype within the tubule.35,36 We selected candidate

expression in eGFP-positive proximal tubular cells. (E) Sample clustering in MDS plot. Scale bar, 50 mm. (F) Robust enrichment of PT-specific genes through scatterplot of normalized expression values in bound RNA (PT) versus unbound RNA (whole cortex). (G) Validation of cell type specificity of the SLC34a1-eGFPL10a line using quantitative PCR. PT markers, Slc34a1, Slc5a2 (Sglt2)andSlc2a2 (Glut2); podocyte markers, Nphs1 and Synpo; loop of Henle markers, Slc12a1 and Umod; collecting duct markers, Aqp2 and Aqp4; endothelial markers, Emcn and Pecam1;myofibroblast markers, Col1a1 and Acta2. Statistical analysis was performed using one-way ANOVA to compare data among groups (four male mice per group). Ab, antibody; CPM, count per million; DAPI, 49,6-diamidino-2-phenylindole.

JASN 31: 23–38, 2020 Proximal Tubule Transcriptomics 27 BASIC RESEARCH www.jasn.org

ACFemale CLK Female UUO (D5+D10) Male Female Male CLK Male UUO (D5+D10) 2 1 UUO up 0 Cyp2e1 1 -1 -2

PT male genes Cyp4b1

Dim2 0

UUO down Ces1f

-1

-2 -1 0 123 Cndp2 Dim1

B 387 genes 361 genes (4.9% sex-linked) (4.1% sex-linked) Slc22a30 UUO up

Y-linked Hao2

X-linked

PT female genes Kynu

Autosome UUO down

Rdh16

-5 05

XX Enriched Log2 Fold Difference XY Enriched CLK CLK D5UUO D5UUO D10UUO D10UUO

DEMale Female

Kynu Rdh16 Gender 60 Female(n=3) 30 50 Male(n=10) 20 40 scRNAseq_study

Cndp2 30 Park et al, Science 2018 10 Schaum et al, Nature 2018 20 Wu et al, JASN 2018 0

Ces1f Cyp4b1 SIc22a30 Cyp2e1 80 60 80 60 75 60 40 40 50 40 Hao2 20 20 25 20 % PT expressing the sexually dimorphic genes 0 0

Figure 2. TRAP reveals sexually dimorphic gene expression in PT. (A) MDS plot shows separation of the PT TRAP samples by disease and sex. (B) Chromosomal distribution of sexually dimorphic genes (FDR,0.05, log fold change .1 or less than 21). (C) Heatmap visualizing the sex-specific genes that are differentially expressed in UUO kidney. (D) Sex-specific expression pattern of Cndp2 and Hao2 validated by RNA ISH. Red boxes indicate region enlarged. (E) Reanalysis of published scRNA-seq data sets confirms the specific expression of Kynu and Rdh16 in female PT, and Ces1f, Cyp4b1, Slc22a30,andCyp2e1 in male PT. Statistical analysis was performed using the Welch t test to compare data between sexes. D5, day 5; D10, day 10.

28 JASN JASN 31: 23–38, 2020 www.jasn.org BASIC RESEARCH

ABD5UUO vs CLK D10UUO vs CLK D10UUO vs D5UUO 15 1,895 2,795 15 1,729 3,175 15 2 55 CLK UUO

10 10 10 q value (FDR) Cstb 5 5 5 10 -Log

0 0 0

-5 0 5 -5 0 5 -2.5 0.0 2.5 5.0 Log2 fold change C PT Cortex CLK D5UUO D10UUO CLK D5UUO D10UUO 1 S100a10

0

-1 Carhsp1

D PT Cortex CLK D5UUO D10UUO CLK D5UUO D10UUO 2 RhoC 1 0 -1 Slc22a8

Figure 3. Validation of PT-specific injury signature in fibrosis. (A) Volcano plots revealing widespread gene expression changes in PT beginning at UUO day 5 (D5UUO). Longer kidney obstruction time (day 10 [D10UUO]) altered the PT transcriptome only minimally. (B) RNA ISH of select PT injury responsive genes identified from TRAP analysis. Representative images were obtained from female kidneys. Red boxes indicate region enlarged. (C) Heatmap illustrating genes whose expression are altered in fibrotic PT but not detected in whole cortex. (D) Heatmap of expression profiles in all samples for DEGs that are changed during UUO in whole cortex but not in the PT. FDR,0.05, log fold change .1orlessthan21.

EMT markers, including Snai1, Snai2, Twist,andvarious in whole cortex (Figure 4A, Supplemental Figure 5A), matrix metalloproteases and collagens and visualized their suggesting a trivial contribution of tubular EMT to kidney expression in PT versus whole cortex. UUO significantly fibrosis. Concordant with this result, reanalysis our UUO altered the expression of these EMT markers in both PT and scRNA-seq data revealed that fewer than 3% of PT cells ex- whole cortex (Supplemental Table 2), but the expression level pressed the EMT markers (Acta2, Col1a1, Col3a1, Fn1,and of these genes in the UUO kidney was much lower in PT than Mmp2) in the UUO kidney compared with the healthy control

JASN 31: 23–38, 2020 Proximal Tubule Transcriptomics 29 BASIC RESEARCH www.jasn.org

A Male Female

PT Total PT Total

CLK D5 D10 CLK D5 D10 CLK D5 D10 CLK D5 D10 UUO UUO UUO UUO UUO UUO UUO UUO 4 Acta2 Pdgfrb Snai1 0 Snai2 Ctgf Col1a1 4 Fn1 S100a4 Col3a1 Cd248 Des Twist1 Mmp2 Col1a2 Col4a1 Col4a2 Col5a1 Col5a2 Mmp9

B Fatty Acid Metabolism (GSEA: UUO vs CLK) C CLK D5UUO D10UUO 0.00 Gcdh Pcbd1 -0.05 Enrichment score = -0.42 -0.10 Bphl Nominal p-value = 0.0 -0.15 Gstz1 -0.20 FDR q-vale = 0.0 Aldh9a1 -0.25 Acaa2 -0.30 Suclg1 -0.35 -0.40 Mdh1

Enrichment score (ES) Fmo1 Hadh Sucla2 Gpd1 Dlst UUO correlated CLK correlated Retsat Enrichment profile Hits Grhpr Crat Pdhb Echs1 Shh expression (RPKM) Acsl1 D Acadm Sdhc 0.3 Aco2 Ppara Acss1 Type 0.2 Idh3b Female PT Hmgcl Female Total Hibch Male PT Aadat 0.1 Male Total Cpox Idh3g Expression (RPKM) Sms 0.0 Acads Acat2 CLK D5 UUO D10 UUO Mlycd Ech1 Ehhadh Setd8 E Ihh expression (RPKM) Hsd17b7 Fh1 5 Slc22a5 Sdha Hao2 4 Type Reep6 Vnn1 Female PT 3 Eci1 Female Total Pdha1 Male PT Male Total Acox1 2 Mgll

Expression (RPKM) Odc1 Uros 1 Car4 Inmt CLK D5 UUO D10 UUO Fabp1

Figure 4. Evaluation of epithelial-to-mesenchymal transition, fatty acid metabolism, and hedgehog ligand gene expression signatures. (A) Heatmap showing the expression of EMT markers in whole kidney but not PT. (B) Gene set enrichment analysis (GSEA) confirms defective fatty acid metabolism in PT during UUO. The total height of the curve indicates the extent of enrichment, with the normalized enrichment score (ES), P value, and FDR value indicated. (C) Heatmap of leading edge genes in the fatty acid metabolism gene set. (D) Expression of Ihh in PT by reads per kilobase of transcript per million mapped reads (RPKM) value. (E) Expression of Shh in PT by RPKM value. D5UUO, 5 days after UUO; D10UUO, 10 days after UUO.

30 JASN JASN 31: 23–38, 2020 www.jasn.org BASIC RESEARCH

A B Cytokine Signaling in Immune system 19.3 22.3

Innate Immune System 17.6 18.7 -1 1PT cytokines Total receptors -1 1 Cytokine-cytokine receptor interaction 10.2 15.7 Ccl12 Ccr1 TNF signaling pathway 13.4 12.4 Ccl2 Ccr2 Signaling by Interleukins 11.3 11.9 Ccl5 Neutrophil degranulation 9.2 9.4 Sdc1 Ccl7 Chemokine signaling pathway 4.7 9.2 Cx3cr1 Ccl8 Platelet activation, signaling and aggregation 6.9 8.8 Csf1r Cx3cl1 Interleukin-10 signaling 6.9 8.2 Pdgfra ll34 Interleukin-4 and 13 signaling 6.8 7.8 Pdgfrb Pdgfa Inflammation mediated by chemokine signaling 5.7 7.6 Eng Pdgfb Complement and coagulation cascades 7.3 6.3 Tgfbr1 Pdgfd NF-kappa B signaling pathway 5.9 7.3 Tgfb2 Tgfbr2 PI3K-Akt signaling pathway 5.9 6.7 Leukocyte transendothelial migration 3.5 6

Genes encoding secreted soluble factors 2.7 5.9 CLK(F) CLK(M) CLK(F) CLK(M)

Chemokine receptors bind chemokines 3.2 5.8 D5UUO(F) D5UUO(M) D5UUO(F) D5UUO(M) D10UUO(F) D10UUO(M) D10UUO(F) Toll-Like Receptors Cascades 5.3 4.7 D10UUO(M) IL-17 signaling pathway 5 2.9 C Immunoregulatory interactions 4.6 4.2 MAPK signaling pathway 4.6 3.3 Interferon gamma signaling 2.6 4.4 IL12-mediated signaling events 2.3 4.4 Interferon alpha/beta signaling 3.6 4.2 Adaptive Immune System 3.8 3.9 CD40/CD40L signaling 3.2 2.4 Th1 and Th2 cell differentiation 1.8 3.1 Fc gamma R-mediated phagocytosis 2.4 2.2 Regulation of TLR by endogenous ligand 1.7 2.2 Cell-Cell communication 2 1.4

D5UUO D10UUO

Adj. P value (-log10): >8 3-8 <3

DE-log10 (Adj.P value) 024 D5UUO PT TRAP D10UUO PT TRAP TNF signaling pathway 1.00 1.00 0.69 0.60 NF-kappa B signaling pathway Cell type

Interleukin-10 signaling 0.75 0.75 PT(S1+S2) PT(S3) IL-17 signaling pathway Inflammatory PT NOD-like receptor signaling pathway 0.50 0.50 Cycling PT

Monocyte and its Surface Molecules Frequency 0.13 0.13 Adhesion and Diapedesis of Lymphocytes 0.25 0.15 0.25 0.13 0.12 Cytokine Signaling in Immue system 0.05 Cells and Molecules involved in local acute inflammatory response 0.00 0.00

Figure 5. Astrongproinflammatory gene expression signature is detected in PT during kidney fibrosis. (A) Pathway analysis using the DEG lists from PT (day 5 UUO [D5UUO] versus CLK and day 10 UUO [D10UUO] versus CLK). (B) Ligand-receptor analysis illustrating the interaction between PT and other cell types in the cortex. Heatmap showing the expression of cytokines/growth factors in PT (left) and their cognate receptors in the whole cortex (right). Straight lines connect the ligand-receptor pairs. (C) Reclustering of proximal tubular cells from day 14 UUO scRNA-seq data set identifies five PT subtypes including a subcluster expressing inflammatory markers. (D) Pathway analysis using the marker gene list from the inflammatory PT cluster. (E) Deconvolution of the subtype composition in the PT TRAP samples using the molecular signature identified from scRNA-seq. Adj., adjusted; F, female; inflamm., inflammatory; M, male; Th1/-2, T helper cell 1/2; TLR, Toll-like receptor.

(Supplemental Figure 5B), and no mesenchymal transition state enrichment analysis using DEGs revealed the majority of fatty was identified in the PT subtypes of the day 14 UUO kidney. acid metabolism–related genes were enriched in the CLK group We then tested if our TRAP data supported defective fatty (Figure 4B), with 83.3% of the genes in this GO term (60 of 72) acid metabolism in PT during UUO, because this has recently significantly downregulated in the tubular fraction of the UUO been shown to contribute to fibrosis progression.5,37 Gene set kidney (Figure 4C).

JASN 31: 23–38, 2020 Proximal Tubule Transcriptomics 31 BASIC RESEARCH www.jasn.org

A All IncRNA (439) Other genes Log2FC = 1 or -1 B Up IncRNA (79) Down IncRNA (64) Log2FC = 0 Multi-dimensional scaling (MDS) using IncRNA

8 PT UUO vs CLK

0.5 Group 4 PT.CLK Total.CLK PT.D5UUO

0 Dim1 0.0 Total.D5UUO PT.D10UUO Log2 fold change Total.D10UUO -4 -0.5

0 5 10 15 -1.0 -0.5 0.0 0.5 1.0 Average expression (log2RPKM) Dim1

C Control UUO

anti-sense probe sense probe anti-sense probe sense probe Snhg18 Gm20513

D Snhg18 Gm20513

Irx5 Pknox2 Alox5ap Dlg4 Grk5 Macc1 Rnf150 Fcgr2b Kcnj15 Nr2f1 Ptgfrn Cx3cr1 Wdr66 Fgd3 Xaf1 Isoc2a Lpl Trim12c 2 CLK (M) 0 CLK (F) -2 D5UUO (M)

D5UUO (F)

D10UUO (M)

D10UUO (F)

Figure 6. Many lncRNAs are detected and undergo differential regulation in PT during fibrosis. (A) MA plot displaying the differentially expressed lncRNAs in PT (UUO versus CLK). (B) MDS plot for the PT TRAP samples according to lncRNA expression profiles alone. (C) ISH validates two PT lncRNAs selected from the DEG list (UUO versus CLK). Data were obtained from male kidneys. Red boxes indicate region enlarged. (D) Network approach to show the protein coding mRNAs that might be targeted by Snhg18 and Gm20513.Nodes are lncRNAs (circle) and their target mRNAs (square). Edge sizes are determined by the energy (absolute log scale) required for

32 JASN JASN 31: 23–38, 2020 www.jasn.org BASIC RESEARCH

There is controversy within the literature regarding whether transitional state based on the expression of unique signatures PT upregulates Indian hedgehog (Ihh)38 or sonic hedgehog identified from unsupervised clustering analysis (Figure 5C). (Shh) after injury. We clearly identified upregulation of PT Very similar pathways were identified in this proinflammatory Ihh,38 whereas Shh was expressed at almost undetectable levels PT cluster (Figure 5D) compared with the pathways we have at baseline and subsequently underwent downregulation after observed to be enriched in the UUO TRAP (Figure 5A). To UUO, despite reports to the contrary (Figure 4, D and E).39,40 validate the existence of these PT subtypes in our TRAP G protein–coupled receptors (GPCRs) constitute the larg- samples, we applied the BSEQ-sc algorithm to estimate cell est family for approved drugs, with a third of the 364 human heterogeneity within each TRAP group from the UUO kidneys, endo-GPCRs serving as targets for approved drugs.41 We ex- based on the PT-subtype signature identified from scRNA-seq amined GPCR expression in PT and identified 132 GPCRs analysis of the day 14 UUO kidney. This revealed a consistent expressed at baseline. Of these GPCRs, 91 underwent up- or fraction of these subpopulations in the PT from scRNA-seq and downregulation during fibrosis, and 13 additional GPCRs TRAP (Figure 5E, Supplemental Figure 6B). were expressed de novo during fibrosis (Supplemental Table 2). Of these 13 de novo upregulated GPCRs, four of them send Identification of Differentially Expressed lncRNAs proinflammatory signals in response to their ligands, including during Fibrosis the formyl peptide receptor Fpr2,thechemokinereceptor lncRNAs are a very large (up to 30,000 different lncRNAs in Cxcr3, the free fatty acid receptor Ffar2, and the sphingosine- humans) and diverse class of transcribed RNAs of longer than 1-phosphate receptor S1pr4. Intriguingly, Ffar2 also regulates 200 nucleotides. They do not encode known proteins, and energy metabolism as does another of these upregulated recep- regulate gene expression through diverse mechanisms. tors, the chemokine-like receptor 1 Cmklr1, suggesting possible lncRNA expression is regulated developmentally, in a tissue- mechanisms underlying the metabolic switch in PT during and cell-specific fashion as well as in disease. In a bulk kidney fibrosis (Figure 4C). transcriptomic survey, Arvaniti et al.45 found that UUO al- tered the expression of 126 lncRNA genes, 77 of which were PT Acquires a Proinflammatory Phenotype intergenic lncRNAs. We could reproduce 63.6% of these inter- during Fibrosis genic lncRNAs when we compared the mRNA expression in Previous studies implicate immune cell infiltrationinthepro- whole cortex between UUO and control using the same sig- gression of kidney fibrosis in the mouse UUO model.42–44 nificance threshold (FDR,0.05). But we also identified an Consistent with these studies, analysis of the total cortex sam- additional 183 differentially expressed lncRNAs not identified ples detected a variety of markers for various immune cell by Arvaniti and colleagues (Supplemental Figure 7A). Because types (B cells, CD4+ T cells, CD8+ T cells, CD14+ monocytes, recent studies have reported that lncRNAs can be pulled down dendritic cells, CD16+ monocytes, megakaryocytes, and natural by TRAP,46,47 we examined lncRNA expression in PT. A total killer cells) that were all upregulated during UUO (Supplemental of 439 lncRNAs were identified in our PT data set (Figure 6A). Figure 6C). We hypothesize proximal tubular cells play an Differential gene analysis revealed 79 upregulated and important role in recruiting the immune cells to the fibrotic 64 downregulated lncRNAs in PT during UUO compared kidney by secretion of cytokines and chemokines. To test this with control (Figure 6A). Strikingly, MDS analysis using hypothesis, we performed pathway enrichment analysis on the the lncRNAs grouped the samples in the same way as when upregulated DEGs identified from UUO versus CLK using the using the whole translatomes (Figures 1E and 6B), suggest- ToppFun Suite. This highlighted differential regulation of two ing the ribosome-bound lncRNAs are expressed in a highly key pathways in PT during fibrosis: immune response and cell-specific fashion both in health and fibrosis. inflammation (Figure 5A). Receptor-ligand analysis re- RNA ISH analysis of Snhg18 (average reads per kilobase vealed very robust proinflammatory (Ccl2, -5, -7, -8, -12, of transcript per million mapped reads, 87.1) and Gm20513 IL34)andprofibrotic (Pdgfa, Pdgfb, Pdgfd, Tgfb)ligandup- (average reads per kilobase of transcript per million mapped regulation in PT (Figure 5B), with their cognate receptors reads, 1.0), two representative lncRNAs with high expression expressed in whole kidney, implicating PTas a central driver and low expression, respectively, confirmed the expected of the inflammatory and fibrotic cascade. UUO-induced expression patterns (Figure 6C, Supplemental Integrative analysis of our previously published scRNA-seq Figure 7B). Because a major role of ribosome-bound lncRNA of PT in UUO15 confirmed the strong induction of Il34 and is regulation of mRNA translation,48 we next determined the Pdgfd during kidney fibrosis (Supplemental Figure 6A). potential target mRNAs that were regulated by Snhg18 and Furthermore, a separate cluster in the day 14 UUO data set Gm20513. Using a computation prediction tool (http:// was classified as a PTsubtype that acquired a proinflammatory rtools.cbrc.jp/LncRRIsearch/), we were able to rank the target

lncRNA-mRNA interaction (LncRRI tools). Heatmap showing the target genes that are differentially expressed in PT during UUO. D5UUO,

5 days after UUO; D10UUO, 10 days after UUO; F, female; log2FC, log2 fold change; M, male; RPKM; reads per kilobase of transcript per million mapped reads.

JASN 31: 23–38, 2020 Proximal Tubule Transcriptomics 33 BASIC RESEARCH www.jasn.org

A Group B PT.CLK PT.D5UUO PT.D10UUO Healthy PT Injured PT

Egr1 3 2 Runx1 Stat3 Smad2 Ets1 Fosl2 2

Dim2 0 Myc Creb5 lrf1 Fosl1 1 -2 Klf4 Sox9 Nfkb1 Jun Gender Maff 0 Female Male Klf6 Ets2 Relb Nfkb2 Jund -1 2 Klf2 Atf3 Junb Rela -2

Dim2 0 Hlf Hoxa9 Rxra Tef -2 Hnf4a -3 Etv5 Nr1h3 Hnf1a Ovol1 Pseudotime Hes6 Hoxa7 Nme2 0204060 Zfp871 Ncor1 Rcor1 Ppara Klf15 2 Klf9 Thrb Creb3l2 Lhx1 0 Tbx2 Dim2 Esrra Mllt10 Tmf1 -2 TFs up TFs down -20 -10 0 10 20 Dim1 D 25 TFs C TFs_down TFs_up 20 Stat3 Egr1 Nfkb1 Runx1 Sox9 15 Nfkb2 Myc Creb5 lrf1 Fos Junb Bach1 Ets1 10 Jun KIf4 Prrx1 Atf3 Relb Tcf12 Fosl2 Expression (RPKM) 5 Nfkb2 Nfix Nfkb1 Maff Zeb1 Klf6 Fosl1 0 Stat3 Smad2 Jund 0 25 50 75 100 Klf2 Rela Healthy to diseased PT (Pseudotime) Irf1 Ets2 Rcor1 Hes6 Thrb E Motif Names Total # of DE genes with Top20 DE genes whose TSS+/- Zfp871 model (NES) motif instances 10kbp with motif instances Klf9 Lhx1 transfac_pro_M07221 Ovol1 2.0 Nme2 Cxcl5, Madcam1, Runx1, Vcam1, Klf15 Cd44, ll1m, Mmp20, Ccdc33, 1.0 Nfkb1 motif Nr1h3 bits 527 Cxcl2, Traf1, Tmem154, Ngf, Tmf1 (3.79) Fam167a, Ccl2, Kcnip4, Cxcl1, Hoxa7 0.0 5’ 3’ Etv5 1 2 3 4 5 6 7 8 9 Ubd, Cxcl3, Pik3r5, Creb5 Tef 10 11 Esrra homer_ACAGGATGTGGT_ETS:RUNX MIlt10 2.0 Rxra Ccdc33, Traf1, Havcr1, Ngf, Hnf4a Pik3r5, Oit3, Anxa2, Adamts5, 1.0 Runx1 motif Ncor1 bits 178 Flna, Bcl2a1b, Psd4, Arhgap25, Hoxa9 (3.36) Hnf1a 0.0 Lmo1, Lif, Cd276, Rnd1, Gpr68, 3’ 5’ 1 2 3 4 5 6 7 8 9

Creb3l2 10 11 12 ll2rb, Bcl2a1a, Bcl2a1d Hlf Tbx2 dbcorrdb_IRF1_ENCSR000EGU_1_m1 Ppara 2.0 Creb5, Gsdmd, Cd40, Ccl5, Psmb8, Arhgap25, Marcksl1, Fn1, -100 0 100 200 1.0 bits Irf1 motif 87 Dock10, Atf3, Birc3, E/f4, TF ranking (Mogrify score) (3.17) Ccdc109b, Uba7, Cxcl10, Gli3, 0.0 5’ 3’ Ube2l6, Ankrd6, Dusp10, Socs1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Figure 7. Identification of transcription factors and regulatory regions driving disease progression in kidney fibrosis. (A) Sample state ordering in reduced dimension space using the Monocle algorithm. Samples are color coded by actual time point, sex, and pseu- dotime. (B) Top differentially expressed transcription factors along the pseudotime from healthy to fibrotic PT. (C) Transcription factor (TF) ranking using the influence score as determined by the Mogrify algorithm. Negative score indicates downregulated transcription factors. (D) Visualization of the reads per kilobase of transcript per million mapped reads (RPKM) values for the selected transcription

34 JASN JASN 31: 23–38, 2020 www.jasn.org BASIC RESEARCH mRNAs that interacted with Snhg18 and Gm20513 based on and 87 DEGs, respectively, were for Nfkb1, Runx1,and their known expression in PT, although it should be empha- Irf1 (Figure 7E, Supplemental Table 5). These three motifs sized that these tools can be inaccurate and should be viewed were also identified by the Mogrify algorithm, providing as a means to generate hypotheses. The expression of some of independent confirmation for their potential importance these target genes were also altered by UUO (Figure 6D), sug- in regulating fibrosis. gesting but not proving the possibility that translation of these genes might be regulated by lncRNA during fibrosis. Detailed annotations for putative target genes for Snhg18 and Gm20513 DISCUSSION revealed that some of these genes were related to proinflam- matory responses such as Fcgr2b, Cx3cr1, Lpl,andAlox5ap In this study, we tested the hypothesis that TRAP could (Supplemental Table 4). Lpl was also involved in fatty acid me- provide enhanced sensitivity and specificity for PT gene tabolism; Fcgr2b and Lpl were also members of the profibrotic expression detection to better define the role of this cell GO terms including extracellular region and extracellular space type in renal fibrosis. The results demonstrate that TRAP (Supplemental Table 4). outperformsbulkaswellasmicrodissectionapproaches. Whole-cortex profiling will miss about a quarter of the Data-Driven Analysis Identifies Key Transcription DEGs in PT, while also identifying another quarter of genes Factors that Drive Disease Progression in Kidney that are differentially expressed but are not expressed in PT. Fibrosis Thus TRAP improves both sensitivity and specificity, even We next sought to identify the key drivers controlling disease for the most abundant kidney cell type. Although laser cap- progression in kidney fibrosis. We implemented a data-driven ture microdissection or tubule segment microdissection approach, Monocle,49 to order the TRAP samples in pseudo- certainly improves specificity, when we compared our data time based on global gene expression changes rather than time set to those generated by such approaches, TRAP identified point. In this analysis, samples were not ordered according to at least twice as many total genes at half the sequencing time point or sex, revealing heterogeneity between samples depth. fromthesametimepoints(Figure7A).Wecouldvalidate Although scRNA-seq is a powerful and increasingly adopted the pseudotemporal ordering of the global gene expression transcriptomic technology in kidney studies, bulk cell-specific by examining expression of known fibrosis regulators on the approaches like TRAP offer complementary advantages in cer- pseudotime trajectory (Stat350 and Nfkb1;51 Supplemental tain applications. Our PT molecular map identified about Figure 8, A and B). We then performed a differential gene 17,000 genes expressed in PT, whereas scRNA-seq has more test across pseudotime and highlighted the top transcription limited gene detection sensitivity. Regulatory genes may be factors derived from the DEG list (Figure 7B). We then used a expressed at lower levels, so sensitive gene detection may be regulatory network approach, the Mogrify algorithm (http:// required to measure those transcripts. Our ability to define www.mogrify.net),52,53 to rank these transcription factors 145 GPCRs illustrates this point. Overall these results sug- according to importance. This implicated a number of pre- gest TRAP can be a good option when (1) an appropriate Cre viously unrecognized transcription factors such as Maff, Klf6, driver line is available to activate eGFP-L10a expression in and Creb5 as well as a variety of transcription factors known the desired cell types; (2) cell-specificbutnotsingle-cell to play roles in kidney injury, such as Stat3,50 Nfkb2,51 Irf1,54 transcriptomics is desired; (3) the translatome is of more Egr1,55 and Jund56 (Figure 7C). Notably, the top transcrip- interest than the transcriptome; and (4) a simple, relatively tional regulators control inflammation (e.g., Stat3, Nfkb1, inexpensive method that does not require any specialized Nfkb2,andIrf1), whereas transcription factors that regulate equipment is optimal. EMT did not increase in expression (Bach1, Prrx1, Tcf12, Nfix, This molecular map of PT in fibrosis also provided impor- and Zeb1) as predicted by the Mogrify algorithm (Figure 7D). tant biologic insights concerning kidney fibrosis. A finding of In addition, we identified 150 NF-kB target genes that were central importance from our study is that the PTactively drives markedly upregulated in PTafter UUO (Supplemental Table 2). fibrosis through activation of proinflammatory and profibrotic Finally, we performed motif enrichment analysis in the pathways. This is consistent with a model whereby nephron 610-kbp region surrounding the transcription start site of epithelia are the first cells to sense and react to damage, whether the DEGs (see Methods) to identify candidate transcription from albuminuria, genetic, toxic, or immune causes. These factors that might be driving differential expression. The findings are consistent with Beckerman et al.57 whose tran- three binding motifs significantly enriched in 527, 178, scriptome analysis of microdissected tubulointerstitium also

factors that are related to inflammation (solid lines) and EMT (dashed lines). (E) Motif analysis on the list of upregulated genes identified in PT after UUO. Three representative motifs that are significantly enriched within 10 kbp upstream and downstream of the transcription start site (TSS) for the DEGs are shown. D5UUO, 5 days after UUO; D10UUO, 10 days after UUO; DE, differentially expressed; NES, normalized enrichment score.

JASN 31: 23–38, 2020 Proximal Tubule Transcriptomics 35 BASIC RESEARCH www.jasn.org identified strong proinflammatory gene modules that corre- fees from Celgene, personal fees from Chinook Therapeutics, personal lated with estimated glomerular filtration loss. fees from Genetech, grants from Biogen Idec, grants from Janssen, and Several large observational studies have shown differing other from Chinook Therapeutics, outside the submitted work. rates of progression in CKD between men and women, but the molecular explanation for this epidemiologic observation FUNDING remains unknown. Recently, using data from the Chronic Renal fi Insuf ciency Cohort Study, Ricardo and colleagues reported This work was supported by National Institutes of Health, National that women in this cohort had both a lower risk of CKD pro- Institute of Diabetes and Digestive and Kidney Diseases grants DK103740 gression, ESKD, and death, compared with men.29 Although and DK107374 (to Dr. Humphreys). there may be many potential reasons for this disparity, our finding that a substantial number of genes exhibit sexually SUPPLEMENTAL MATERIAL dimorphic expression both at baseline and during fibrosis suggests gene expression differences may be at least partly This article contains the following supplemental material online at responsible. The McMahon group (A. Ransick, N. O. Lindström, http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2019040337/-/ J. Liu, Z. Qin, J.-J. Guo, G. F. Alvarado, et al., unpublished ob- DCSupplemental. servations) has recently also reported that the PT, among all Supplemental Figure 1. TRAP sample quality control. nephron segments, displays a large degree of sexually dimorphic Supplemental Figure 2. Sexually dimorphic genes in proximal gene expression, confirming our results. tubule. Finally, a recent estimate suggested .98% of transcribed Supplemental Figure 3. Disease signature for kidney fibrosis. RNAs are not translated to protein in eukaryotic cells and Supplemental Figure 4. Comparison of the DEG identified from most of these noncoding transcripts are classified as lncRNA.58 this study and the published Six2-TRAP dataset. It is increasingly clear that lncRNA may participate in the path- Supplemental Figure 5. EMT marker expression in TRAP-seq and ogenesis of many kidney diseases, through the mechanisms scRNA-seq. of chromatin modification, transcriptional regulation, post- Supplemental Figure 6. Proximal tubule pro-inflammatory gene transcriptional control, or by functioning as decoy or molecu- expression and subtype composition revealed by scRNA-seq. lar sponges.59–61 Many lncRNAs have recently been reported Supplemental Figure 7. lincRNA expression in UUO kidneys. to contain functional small open reading frames that may Supplemental Figure 8. Sample ordering validationby transcription be translated into peptides.47,62,63 We identified many lncRNAs factors known to be upregulated in proximal tubule during kidney differentially expressed during renal fibrosis by TRAP sequencing fibrosis. (Figure 6, A–E, Supplemental Table 2), presumably indicating that Supplemental Table 1. Sequencing characteristics. these lncRNAs are bound by mRNA undergoing translation con- Supplemental Table 2. TRAP gene expression summary. sistent with other recent analyses.46,47 It should be noted that Supplemental Table 3. PTsexually dimorphic gene expression from functional lncRNAs may not require a poly-A tail, and whether scRNA-seq dataset. we isolated such lncRNAs is not clear from our analysis. Supplemental Table 4. Annotations of putative target genes for In conclusion, this article provides an unbiased and deep de- lncRNAs Snhg18 and Gm20513. scription of PT gene expression in health, during fibrosis and Supplemental Table 5. Enriched PT binding motifs during fibrosis. across sexes.OurstudyhighlightsthepoweroftheTRAPapproach Supplemental Methods. to provide high-quality, cell-specific mRNA and reinforces the concept that the PT is a critical, active driver of cortical fibrosis through regulation of proinflammatory and profibrotic pathways. REFERENCES

1. Glassock RJ, Warnock DG, Delanaye P: The global burden of chronic ACKNOWLEDGMENTS kidney disease: Estimates, variability and pitfalls. Nat Rev Nephrol 13: 104–114, 2017 Dr. Chang-Panesso designed and carried out experiments and ana- 2. United States Renal Data System: 2017 USRDS annual data report: lyzed results. Dr. Humphreys designed experiments, analyzed results, Epidemiology of kidney disease in the United States, Bethesda, MD, National Institutes of Health, National Institute of Diabetes and Di- and wrote the manuscript. Dr. Lai and Dr. Wu designed and carried gestive and Kidney Diseases, 2017 out experiments, analyzed results, and contributed to writing of 3. Humphreys BD: Mechanisms of renal fibrosis. Annu Rev Physiol 80: the manuscript. 309–326, 2018 4. Wu H, Uchimura K, Donnelly EL, Kirita Y, Morris SA, Humphreys BD: Comparative analysis and refinement of human PSC-derived kidney organoid differentiation with single-cell transcriptomics. Cell Stem Cell DISCLOSURES 23: 869–881.e868, 2018 5. Kang HM, Ahn SH, Choi P, Ko YA, Han SH, Chinga F, et al.: Defective Dr. Humphreys reports personal fees from Merck, personal fees from fatty acid oxidation in renal tubular epithelial cells has a key role in Janssen, personal fees from Medimmune, personal fees from Roche, personal kidney fibrosis development. Nat Med 21: 37–46, 2015

36 JASN JASN 31: 23–38, 2020 www.jasn.org BASIC RESEARCH

6. Zhou HL, Zhang R, Anand P, Stomberski CT, Qian Z, Hausladen A, et al.: 28. Carrero JJ, Hecking M, Chesnaye NC, Jager KJ: Sex and gender dis- Metabolic reprogramming by the S-nitroso-CoA reductase system parities in the epidemiology and outcomes of chronic kidney disease. protects against kidney injury. Nature 565: 96–100, 2019 Nat Rev Nephrol 14: 151–164, 2018 7. Canaud G, Brooks CR, Kishi S, Taguchi K, Nishimura K, Magassa S, 29. Ricardo AC, Yang W, Sha D, Appel LJ, Chen J, Krousel-Wood M, et al.:

et al.: Cyclin G1 and TASCC regulate kidney epithelial cell G2-M arrest CRIC Investigators: Sex-related disparities in CKD progression. JAm and fibrotic maladaptive repair. Sci Transl Med 11: eaav4754, 2019 Soc Nephrol 30: 137–146, 2019 8. Woroniecka KI, Park AS, Mohtat D, Thomas DB, Pullman JM, Susztak K: 30. Kopitar-Jerala N: The role of stefin B in neuro-inflammation. Front Cell Transcriptome analysis of human diabetic kidney disease. Diabetes 60: Neurosci 9: 458, 2015 2354–2369, 2011 31. Bydoun M, Sterea A, Weaver ICG, Bharadwaj AD, Waisman DM: A 9. Liu J, Kumar S, Dolzhenko E, Alvarado GF, Guo J, Lu C, et al.: Molecular novel mechanism of plasminogen activation in epithelial and mesen- characterization of the transition from acute to chronic kidney injury chymal cells. Sci Rep 8: 14091, 2018 following ischemia/reperfusion. JCI Insight 2: 94716, 2017 32. Nastasi T, Scaturro M, Bellafiore M, Raimondi L, Beccari S, Cestelli A, 10. Cippà PE, Sun B, Liu J, Chen L, Naesens M, McMahon AP: Transcrip- et al.: PIPPin is a brain-specific protein that contains a cold-shock do- tional trajectories of human kidney injury progression. JCI Insight 3: main and binds specifically to H1 degrees and H3.3 mRNAs. JBiol 123151, 2018 Chem 274: 24087–24093, 1999 11. Cippà PE, Liu J, Sun B, Kumar S, Naesens M, McMahon AP: A late B 33. Pfeiffer JR, McAvoy BL, Fecteau RE, Deleault KM, Brooks SA: CARHSP1 lymphocyte action in dysfunctional tissue repair following kidney injury is required for effective tumor necrosis factor alpha mRNA stabilization and transplantation. Nat Commun 10: 1157, 2019 and localizes to processing bodies and exosomes. Mol Cell Biol 31: 12. Adam M, Potter AS, Potter SS: Psychrophilic proteases dramatically 277–286, 2011 reduce single-cell RNA-seq artifacts: A molecular atlas of kidney de- 34. Nagatoya K, Moriyama T, Kawada N, Takeji M, Oseto S, Murozono T, velopment. Development 144: 3625–3632, 2017 et al.: Y-27632 prevents tubulointerstitial fibrosis in mouse kidneys with 13. Lindström NO, Guo J, Kim AD, Tran T, Guo Q, De Sena Brandine G, unilateral ureteral obstruction. Kidney Int 61: 1684–1695, 2002 et al.: Conserved and divergent features of mesenchymal progenitor 35. Humphreys BD, Lin SL, Kobayashi A, Hudson TE, Nowlin BT, cell types within the cortical nephrogenic niche of the human and Bonventre JV, et al.: Fate tracing reveals the pericyte and not epi- mouse kidney. JAmSocNephrol29: 806–824, 2018 thelial origin of myofibroblasts in kidney fibrosis. Am J Pathol 176: 14. Menon R, Otto EA, Kokoruda A, Zhou J, Zhang Z, Yoon E, et al.: Single- 85–97, 2010 cell analysis of progenitor cell dynamics and lineage specification in the 36. Kramann R, DiRocco DP, Humphreys BD: Understanding the origin, human fetal kidney. Development 145: dev164038, 2018 activation and regulation of matrix-producing myofibroblasts for 15. Wu H, Kirita Y, Donnelly EL, Humphreys BD: Advantages of single- treatment of fibrotic disease. J Pathol 231: 273–289, 2013 nucleus over single-cell RNA sequencing of adult kidney: Rare cell 37. Chung KW, Lee EK, Lee MK, Oh GT, Yu BP, Chung HY: Impairment of types and novel cell states revealed in fibrosis. JAmSocNephrol30: PPARa and the fatty acid oxidation pathway aggravates renal fibrosis 23–32, 2019 during aging. J Am Soc Nephrol 29: 1223–1237, 2018 16. Park J, Shrestha R, Qiu C, Kondo A, Huang S, Werth M, et al.: Single-cell 38. Fabian SL, Penchev RR, St-Jacques B, Rao AN, Sipilä P, West KA, et al.: transcriptomics of the mouse kidney reveals potential cellular targets of Hedgehog-Gli pathway activation during kidney fibrosis. Am J Pathol kidney disease. Science 360: 758–763, 2018 180: 1441–1453, 2012 17. Grgic I, Hofmeister AF, Genovese G, Bernhardy AJ, Sun H, Maarouf OH, 39. Zhou D, Li Y, Zhou L, Tan RJ, Xiao L, Liang M, et al.: Sonic hedgehog is a et al.: Discovery of new glomerular disease-relevant genes by trans- novel tubule-derived growth factor for interstitial fibroblasts after kid- lational profiling of podocytes in vivo. Kidney Int 86: 1116–1129, 2014 ney injury. J Am Soc Nephrol 25: 2187–2200, 2014 18. Grgic I, Krautzberger AM, Hofmeister A, Lalli M, DiRocco DP, Fleig SV, 40. Ding H, Zhou D, Hao S, Zhou L, He W, Nie J, et al.: Sonic hedgehog et al.: Translational profiles of medullary myofibroblasts during kidney signaling mediates epithelial-mesenchymal communication and pro- fibrosis. J Am Soc Nephrol 25: 1979–1990, 2014 motes renal fibrosis. JAmSocNephrol23: 801–813, 2012 19. Liu J, Krautzberger AM, Sui SH, Hofmann OM, Chen Y, Baetscher M, 41. Sriram K, Insel PA: G protein-coupled receptors as targets for approved et al.: Cell-specific translational profiling in acute kidney injury [pub- drugs: How many targets and how many drugs? Mol Pharmacol 93: lished correction appears in JClinInvest124: 2288, 2014]. J Clin Invest 251–258, 2018 124: 1242–1254, 2014 42. Vielhauer V, Allam R, Lindenmeyer MT, Cohen CD, Draganovici D, 20. Kusaba T, Lalli M, Kramann R, Kobayashi A, Humphreys BD: Differen- Mandelbaum J, et al.: Efficient renal recruitment of macrophages and tiated kidney epithelial cells repair injured proximal tubule. Proc Natl T cells in mice lacking the duffy antigen/receptor for chemokines. Am J Acad Sci U S A 111: 1527–1532, 2014 Pathol 175: 119–131, 2009 21. Heiman M, Schaefer A, Gong S, Peterson JD, Day M, Ramsey KE, et al.: 43. Tapmeier TT, Fearn A, Brown K, Chowdhury P, Sacks SH, Sheerin NS, A translational profiling approach for the molecular characterization of et al.: Pivotal role of CD4+ T cells in renal fibrosis following ureteric CNS cell types. Cell 135: 738–748, 2008 obstruction. Kidney Int 78: 351–362, 2010 22. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, et al.: STAR: 44. Han H, Zhu J, Wang Y, Zhu Z, Chen Y, Lu L, et al.: Renal recruitment of B Ultrafast universal RNA-seq aligner. Bioinformatics 29: 15–21, 2013 lymphocytes exacerbates tubulointerstitial fibrosis by promoting 23. Liao Y, Smyth GK, Shi W: FeatureCounts: An efficient general purpose monocyte mobilization and infiltration after unilateral ureteral ob- program for assigning sequence reads to genomic features. Bioinformatics struction. J Pathol 241: 80–90, 2017 30: 923–930, 2014 45. Arvaniti E, Moulos P, Vakrakou A, Chatziantoniou C, Chadjichristos C, 24. Wang L, Wang S, Li W: RSeQC: Quality control of RNA-seq experi- Kavvadas P, et al.: Whole-transcriptome analysis of UUO mouse model ments. Bioinformatics 28: 2184–2185, 2012 of renal fibrosis reveals new molecular players in kidney diseases. 25. Humphreys BD: (Re)Building a Kidney Consortium, https://doi.org/ Sci Rep 6: 26235, 2016 10.25548/16-E08W. 2019 46.ZhouP,ZhangY,MaQ,GuF,DayDS,HeA,etal.:Interrogating 26. Lee JW, Chou CL, Knepper MA: Deep sequencing in microdissected translational efficiency and lineage-specific transcriptomes using ribo- renal tubules identifies nephron segment-specific transcriptomes. JAm some affinity purification. Proc Natl Acad Sci U S A 110: 15395–15400, Soc Nephrol 26: 2669–2677, 2015 2013 27. Tabula Muris Consortium: Single-cell transcriptomics of 20 mouse 47. Bazin J, Baerenfaller K, Gosai SJ, Gregory BD, Crespi M, Bailey-Serres organs creates a Tabula Muris. Nature 562: 367–372, 2018 J: Global analysis of ribosome-associated noncoding RNAs unveils new

JASN 31: 23–38, 2020 Proximal Tubule Transcriptomics 37 BASIC RESEARCH www.jasn.org

modes of translational regulation. Proc Natl Acad Sci U S A 114: 55. Ho LC, Sung JM, Shen YT, Jheng HF, Chen SH, Tsai PJ, et al.: Egr-1 E10018–E10027, 2017 deficiency protects from renal inflammation and fibrosis. J Mol Med 48. Pircher A, Gebetsberger J, Polacek N: Ribosome-associated ncRNAs: (Berl) 94: 933–942, 2016 An emerging class of translation regulators. RNA Biol 11: 1335–1339, 56. Pillebout E, Weitzman JB, Burtin M, Martino C, Federici P, Yaniv M, 2014 et al.: JunD protects against chronic kidney disease by regulating 49. Qiu X, Mao Q, Tang Y, Wang L, Chawla R, Pliner HA, et al.: Reversed paracrine mitogens. J Clin Invest 112: 843–852, 2003 graph embedding resolves complex single-cell trajectories. Nat 57. Beckerman P, Qiu C, Park J, Ledo N, Ko YA, Park AD, et al.: Human Methods 14: 979–982, 2017 kidney tubule-specific gene expression based dissection of chronic 50. Bienaimé F, Muorah M, Yammine L, Burtin M, Nguyen C, Baron W, et al.: kidney disease traits. EBioMedicine 24: 267–276, 2017 Stat3 controls tubulointerstitial communication during CKD. JAmSoc 58. Mattick JS: The central role of RNA in human development and cog- Nephrol 27: 3690–3705, 2016 nition. FEBS Lett 585: 1600–1616, 2011 51. Sanz AB, Sanchez-Niño MD, Ramos AM, Moreno JA, Santamaria B, 59. Zhou Q, Huang XR, Yu J, Yu X, Lan HY: Long noncoding RNA arid2-IR Ruiz-Ortega M, et al.: NF-kappaB in renal inflammation. JAmSoc is a novel therapeutic target for renal inflammation. Mol Ther 23: 1034– Nephrol 21: 1254–1262, 2010 1043, 2015 52. Rackham OJ, Firas J, Fang H, Oates ME, Holmes ML, Knaupp AS, et al.: 60. Lorenzen JM, Thum T: Long noncoding RNAs in kidney and cardio- FANTOM Consortium: A predictive computational framework for direct vascular diseases. Nat Rev Nephrol 12: 360–373, 2016 reprogramming between human cell types. Nat Genet 48: 331–335, 61. Lai CF, Chen YT, Gu J, Nerbonne JM, Lin CH, Yang KC: Circulating long 2016 noncoding RNA DKFZP434I0714 predicts adverse cardiovascular outcomes 53. Zhu Q, Fisher SA, Dueck H, Middleton S, Khaladkar M, Kim J: PIVOT: in patients with end-stage renal disease. Int J Cardiol 277: 212–219, 2019 Platform for interactive analysis and visualization of transcriptomics 62. Chekulaeva M, Rajewsky N: Roles of long noncoding RNAs and circular data. BMC Bioinformatics 19: 6, 2018 RNAs in translation. Cold Spring Harb Perspect Biol 11: a032680, 2018 54. Wang Y, John R, Chen J, Richardson JA, Shelton JM, Bennett M, et al.: 63. Li LJ, Leng RX, Fan YG, Pan HF, Ye DQ: Translation of noncoding RNAs: IRF-1 promotes inflammation early after ischemic acute kidney injury. Focus on lncRNAs, pri-miRNAs, and circRNAs. Exp Cell Res 361: 1–8, JAmSocNephrol20: 1544–1555, 2009 2017

38 JASN JASN 31: 23–38, 2020 Supplemental Information

Proximal tubule translational profiling during kidney fibrosis reveals pro- inflammatory and lncRNA expression patterns with sexual dimorphism

Haojia Wu, Chun-Fu Lai, Monica Chang-Panesso and Benjamin D. Humphreys

Supplementary Figures Supplementary Figure S1. TRAP sample quality control. Supplementary Figure S2. Sexually dimorphic genes in proximal tubule. Supplementary Figure S3. Disease signature for kidney fibrosis. Supplementary Figure S4. Comparison of the DEG identified from this study and the published Six2-TRAP dataset Supplementary Figure S5. EMT marker expression in TRAP-seq and scRNA-seq. Supplementary Figure S6. Proximal tubule pro-inflammatory gene expression and subtype composition revealed by scRNA-seq. Supplementary Figure S7. lincRNA expression in UUO kidneys. Supplementary Figure S8. Sample ordering validation by transcription factors known to be upregulated in proximal tubule during kidney fibrosis.

Supplementary Methods In Situ Hybridization Bioinformatic analysis 1. Differential expression analysis 2. Gene-set enrichment and pathway analysis 3. Comparative analysis of the differential genes from TRAP-seq, microdissection RNA-seq and TRAP microarray 4. scRNA-seq re-analysis 5. Cytokine-receptor interaction analysis 6. Long noncoding RNA and G protein-coupled receptor (GPCR) analysis 7. Transporters, ion channels and NFκB target gene analysis. 8. Reconstruction of disease progression trajectory with Monocle 9. Transcription factor and motif enrichment analysis

Supplementary References

A. B. Number Integruty RNA RNA

Control D5 D10 Control D5 D10

Proximal tubule Cortex

C. D.

Supplementary Figure S1. TRAP sample quality control. A. RNA integrity as determined by BioAnalyzer. N=7 for each group. B. Read depth and total gene detection in Lee et al1 compared with this study. C. Venn diagram revealing the overlapping genes detected in this study and in Lee et al.1 D.Heatmap showing the correlation between each two samples. The scale bar represents the Pearson correlation coefficient. PTEC, proximal tubule epithelial cells.

A. B.

Supplementary Figure S2. Sexually dimorphic genes in proximal tubule. A. Linear relationship of log fold change cutoffs and the number of detected DEG. B. Volcano plot visualizing the sexually dimorphic genes that have been validated by ISH and the published scRNA-seq datasets.

Supplementary Figure S3. Disease signature for kidney fibrosis. A. Venn diagram showing ~80% of the DE genes identified in fibrotic PT are overlapped at day 5 and day 10 UUO. B. Very few differentially expressed genes are identified in the whole cortex from day 10 UUO and day 5 UUO kidneys. C. Venn diagram summarizing the overlap of the DE genes across pairwise comparisons.

Supplementary Figure S4. Comparison of the DEG identified from this study and the published Six2-TRAP dataset A. Venn diagram showing ~65% of the DEGs identified in this study are overlapped with DEGs reported by Liu et al2 using a Six2 TRAP line. B. Venn diagram showing the upregulated genes unique in each study or shared by both datasets. C. GO enrichment analysis on the DEGs unique to each dataset, or shared by both datasets.

Supplementary Figure S5. EMT marker expression in TRAP-seq and scRNA-seq. A. Expression of the EMT markers (RPKM value) are much higher in whole cortex than in PT. B. scRNA-seq3 revealed that less than 3% PT cells from the day 14 UUO kidney are expressing the EMT markers.

Supplementary Figure S6. Proximal tubule pro-inflammatory gene expression and subtype composition revealed by scRNA-seq. A. Feature plot confirms the upregulation of Il34 and Pdgfd at day 14 UUO. B. Fraction of PT subtypes in the day 14 UUO kidney. C. Expression of immune cell markers in total cortex during UUO. Left: heatmap showing the marker expression in a publicly available PBMC dataset (https://satijalab.org/seurat/). Right: heatmap showing the expression of the same set of immune cell markers in our total cortex.

Supplementary Figure S7. lincRNA expression in UUO kidneys. A. Venn diagram to compare the DE lincRNAs identified from this study (whole cortex) with a published dataset (whole kidney).4 B. Box plot showing the RPKM value for the selected lincRNAs Snhg18 and Gm20513 in proximal tubule.

Supplementary Figure S8. Sample ordering validation by transcription factors known to be upregulated in proximal tubule during kidney fibrosis. The accuracy of sample ordering can be validated by the expression of the known TFs such as Stat3 (A) and Nfkb1 (B).

Supplementary Methods In Situ Hybridization Primers were as follows: Cndp2, SP6: 5’- CATTTAGGTGACACTATAGCTCCCTGCATGGGATCGAAG – 3’; T7: 5’ – TAATACGACTCACTATAGGGTTCTCATTCTGCGAGTGGGC – 3’; Hao2, SP6; 5’ - CATTTAGGTGACACTATAG GGCAGACTTTAAGGCACAAGC – 3’; T7: 5’ – TAATACGACTCACTATAGGGCTCTCCAGGGGATCGGAGAT – 3’; Cstb, SP6; 5’ – CATTTAGGTGACACTATAGGCCAGGTTTTTCTAGGGTCCA – 3’; T7; 5’ – TAATACGACTCACTATAGGGCAAAGGAGCCCCGAATCAGA – 3’; S100a10, SP6; 5’ - CATTTAGGTGACACTATAG GAGTGCTCATGGAACGGGAG – 3’; T7; 5’ - TAATACGACTCACTATAGGG TTGAGGGCAATGGGATGCAAA – 3’; Carhsp1, SP6; 5’ – CATTTAGGTGACACTATAGCTGGAGGAGTAGGACGTGTCG – 3’; T7: 5’ – TAATACGACTCACTATAGGGGGTCTCATGCTTGGTCCCTG – 3’; RhoC: 5’ – CATTTAGGTGACACTATAGATCGAAGTGGATGGCAAGCA – 3’; T7: 5’ – TAATACGACTCACTATAGGGCTTGGGGCTGGGAAACTCAT – 3’; Slc22a8, SP6; 5’ – CATTTAGGTGACACTATAGCTGGCTACAGTTGTCCGTGT – 3’; T7: 5’ – TAATACGACTCACTATAGGGGACAGGCATCCCTTCCCAAA – 3’; Snhg18, SP6; 5’ – CATTTAGGTGACACTATAGTCGCCTGGAAGAGACCTTCT – 3’; T7: 5’ – TAATACGACTCACTATAGGGTAAAACCGAAGCAGCACCGA – 3’; Gm20513, SP6: 5’ – CATTTAGGTGACACTATAGGAGGAGCAACCTTCCCTGGT – 3’; T7: 5’ – TAATACGACTCACTATAGGGATGTTCCCCAAATCAACTGCAT – 3’. In vitro transcription was then performed to generate digoxigenin (DIG)-labeled sense and antisense riboprobes DIG labeling mixture (cat #11277073910, Roche) by SP6 and T7 RNA polymerases, respectively.

Bioinformatic analysis 1. Differential expression analysis Genes with fewer than 0.3 RPKM (Reads Per Kilobase of transcript, per Million mapped reads) in one sixth of the samples were excluded from further analysis. The filtered gene-level counts were imported into the R/Bioconductor package edgeR.5 Raw counts were normalized using the TMM (Trimmed Mean of M-values) technique. P values were reported using a negative binomial test in edgeR, followed by a Benjamini–Hochberg multiple testing correction to derive FDR. Only differentially expressed genes with FDR < 0.05 and absolute log2 fold change >1 were considered for downstream analysis. Performance of the samples was assessed with a Pearson correlation matrix and multi-dimensional scaling (MDS) plots. Unless otherwise noted, we used R packages heatmap.2, VennDiagram and ggplot2 to generate the heatmaps, venn diagrams and all other plots in this study.

2. Gene-set enrichment and pathway analysis GSEA (http://software.broadinstitute.org/gsea/index.jsp) was used to estimate the enriched gene ontology (GO) terms from the differentially expressed (DE) genes. We used the DE gene list ranked by log2 fold change as a ranked list, and a gene sets database downloaded from Bader Lab (http://download.baderlab.org/EM_Genesets/) that contained all mouse GO terms as gene set file input to GSEA. We then used 1,000 gene label permutations to identify the significantly enriched gene-sets as defined by adjusted p value < 0.05. Pathway analysis results were obtained from the ToppGene suite6 (https://toppgene.cchmc.org) using the DE genes derived from pairwise comparison as input.

3. Comparative analysis of the differential genes from TRAP-seq, microdissection RNA-seq and TRAP microarray To compare the differential gene list produced from this study to the genes from the public cell-type specific profiling datasets, we re-analyzed a study reported by Lee et al1 (GSE56743; http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE56743) where each tubular segment was profiled using microdissection RNA-seq, and a study published by us using TRAP and microarray to profile the Six2 cells in the ischemia-reperfusion kidney disease model2 (GSE52004; https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE52004). To minimize the bias arising from different analytical pipelines and parameters chosen, we used the same pipeline and cutoffs to generate the differential gene lists. For the microdissection RNA-seq dataset, we removed the lowly expressed genes using the same cutoff reported in this study (i.e. RPKM>0.3 in at least 1/6 samples), followed by the edgeR package to produce the differential gene list. For the Six2-TRAP microarray data, since edgeR was not a suitable tool to analyze microarray data due to the underlying statistical model (negative binomial) but limma can analyze both microarray and RNA-seq data (if the normalized counts were transformed by voom),7we therefore re-analyzed our TRAP data using limma. In this analysis, 95% of differential genes generated by edgeR and limma were overlapped, confirming the similar performance of both pipelines on our TRAP data. We then used the same cutoffs to select the DEGs (log fold change >1 and FDR <0.05) for the comparison.

4. scRNA-seq re-analysis To validate genes with sexually dimorphic expression, we re-analyzed three independent scRNA-seq datasets generated by three different laboratories.3, 8, 9 The proximal tubule population was extracted from these datasets and the percentage of cells expressing the sexually dimorphic genes were compared between male and female. In total, we obtained PT single cell transcriptomes from 3 female and 10 male mice (Park et al8: 6 male; Wu et al3: 2 male + 2 female; and Schaum et al9: 2 male + 1 female). Selected genes were reported from the gene list that passed the significance threshold (p<0.05, student t-test). To confirm the existence of a proinflammatory PT state, we re-clustered the PT population obtained from a D14UUO snRNA-seq data using the Seurat package. PT subtype identities were annotated based the marker genes identified from each subtype. Pathway analysis was carried out on the proinflammatory PT subtype using the ToppGene suite mentioned above. We then applied the single cell deconvolution algorithm BSeq-sc10 to estimate the proportion of each PT subtype identified from scRNA-seq in the D5 and D10 UUO PT bulk RNA-seq data. The marker genes for each PT subtype and the RPKM normalized gene expression matrix from TRAP were used as input according to the tutorial from BSeq-sc package (https://shenorrlab.github.io/bseqsc/vignettes/bseq-sc.html). To obtain the marker gene list for each immune cell types, we downloaded the precomputed Seurat R object from PBMC scRNA- seq in Satija’s lab (https://satijalab.org/seurat/). We then used FindAllMarker function in Seurat to produce the marker gene list.

5. Cytokine-receptor interaction analysis To study ligand-receptor interactions between PT and other cell types within the UUO kidney, we used a 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).11, 12 To determine the ligand-receptor pairs to plot on the heatmap, we required that (i) the ligands were significantly upregulated in the PT fraction and receptors were only enriched in the total cortex fraction (FDR<0.05 and |logFC|>1); (ii) Each receptor should have at least one corresponding ligand. To highlight paracrine interactions, we further reduced the gene list using the cytokine-receptor pairs documented in the iTALK package.13

6. Long noncoding RNA and G protein-coupled receptor (GPCR) analysis We extracted lncRNAs from the DE gene list based on the categories in Ensembl NCBIM38 annotations. lncRNA-mRNA interaction was predicted by a publicly available database LncRRIdb (http://rtools.cbrc.jp/LncRRIsearch/). We used network visualization software Cytoscape (version 3.4.0)14 to visualize lncRNA-mRNA interactions. To compare the lncRNAs identified from this PT-TRAP line to those reported by a bulk transcriptomic study (14), we set the same cutoffs to select the differentially expressed lncRNAs (|logFC|>1 and FDR <0.05). The number of overlapped and non-overlapped lncRNAs were visualized by venn diagram. To catalog the GPCRs expression in PT, we obtained a mouse GPCR gene list curated by IUPHAR/BPS Guide to PHARMACOLOGY database (http://www.guidetopharmacology.org/GRAC/GPCRListForward?class=A), and crossed it to the whole gene list detected in our TRAP dataset. Differential gene test was performed on the detected GPCRs (UUO vs CLK), and all differentially expressed GPCRs were reported if they passed the statistical threshold (|logFC|>1 and FDR <0.05).

7. Transporters, ion channels and NFκB target gene analysis. To categorize the genes that were enriched in PT, we obtained a complete mouse list of transporters and ion channels from the PHARMACOLOGY database (http://www.guidetopharmacology.org/targets.jsp). This gene list was crossed to the DEG list produced from the comparison of PT and total cortex. Using a similar approach, we generated the data for the NFκB target genes by crossing the complete NFκB target gene list curated by Thomas Gilmore’s lab (http://www.bu.edu/nf-kb/gene-resources/target-genes/) to the DEG list from this study (PT UUO versus PT CLK).

8. Reconstruction of disease progression trajectory with Monocle We used Monocle215 to draw a minimal spanning tree connecting the PT samples from CLK, D5UUO and D10UUO kidneys. As input into Monocle2, we selected the highly variable genes for sample ordering as described in the Monocle2 tutorial (http://cole-trapnell- lab.github.io/monocle-release/). We then reduced the data space to two dimensions using the reduceDimension function with ‘DDRTree’ method and ordered the samples using the orderCells function in Monocle2. Individual sample were color-coded based on the kidneys where they were collected. To identify the genes whose expression were dynamically changing across the trajectory from healthy to fibrotic kidneys, we performed differentially expression test on the samples across pseudotime using the differentialGeneTest function in Monocle2 with the fullModelFormulaStr parameter set to ‘Psudeotime’. The output gene list was used for the downstream TF analysis.

9. Transcription factor and motif enrichment analysis To identify the TFs, we cross referenced differentially expressed genes obtained from Monocle to the mouse TF list downloaded from AnimalTFDB16 (http://bioinfo.life.hust.edu.cn/AnimalTFDB/). We selected important TFs that have been reported by the literature as key regulators of proximal tubule injury and visualized them by heatmap or line chart. We then used a mogrify algorithm to score the TFs. The top 25 upregulated and downregulated TFs were selected and visualized by bar chart. To identify the TF binding motif enriched in the 10kbp regions of transcription start site (TSS) for the DE genes, we implemented an R package RcisTarget17 on the differential gene list with the following steps: 1) A precomputed mouse motif database with scores for each gene-motif pair was downloaded from Aerts Lab (https://resources.aertslab.org/cistarget/). In this analysis, we used a motif database named mm9-tss-centered-10kb-7species.mc9nr.feather which ranked the motifs enriched in 10kbp upstream and 10kbp downstream of the TSS. 2) A motif annotation database (motifAnnotations_mgi) was used to associate the motif to transcription factors. 3) Over-representation of each motif on the gene-set is estimated by computing the Area Under the Curve (AUC) for each pair of motif-geneset using the calcAUC function provided by the RcisTarget package. 4) The significantly enriched motifs with normalized enrichment score greater than 3 were reported and annotated to transcription factors using the RcisTarget functions addMotifAnnotation and addSignificantGenes.

1. Lee, JW, Chou, CL, Knepper, MA: Deep Sequencing in Microdissected Renal Tubules Identifies Nephron Segment-Specific Transcriptomes. J Am Soc Nephrol, 26: 2669- 2677, 2015. 2. Liu, J, Krautzberger, AM, Sui, SH, Hofmann, OM, Chen, Y, Baetscher, M, Grgic, I, Kumar, S, Humphreys, BD, Hide, WA, McMahon, AP: Cell-specific translational profiling in acute kidney injury. J Clin Invest, 124: 1242-1254, 2014. 3. Wu, H, Kirita, Y, Donnelly, EL, Humphreys, BD: Advantages of Single-Nucleus over Single- Cell RNA Sequencing of Adult Kidney: Rare Cell Types and Novel Cell States Revealed in Fibrosis. J Am Soc Nephrol, 30: 23-32, 2019. 4. Arvaniti, E, Moulos, P, Vakrakou, A, Chatziantoniou, C, Chadjichristos, C, Kavvadas, P, Charonis, A, Politis, PK: Whole-transcriptome analysis of UUO mouse model of renal fibrosis reveals new molecular players in kidney diseases. Sci Rep, 6: 26235, 2016. 5. Robinson, MD, McCarthy, DJ, Smyth, GK: edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics, 26: 139-140, 2010. 6. Chen, J, Bardes, EE, Aronow, BJ, Jegga, AG: ToppGene Suite for gene list enrichment analysis and candidate gene prioritization. Nucleic Acids Res, 37: W305-311, 2009. 7. Ritchie, ME, Phipson, B, Wu, D, Hu, Y, Law, CW, Shi, W, Smyth, GK: limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res, 43: e47, 2015. 8. Park, J, Shrestha, R, Qiu, C, Kondo, A, Huang, S, Werth, M, Li, M, Barasch, J, Susztak, K: Single-cell transcriptomics of the mouse kidney reveals potential cellular targets of kidney disease. Science, 360: 758-763, 2018. 9. Tabula Muris, C, Overall, c, Logistical, c, Organ, c, processing, Library, p, sequencing, Computational data, a, Cell type, a, Writing, g, Supplemental text writing, g, Principal, i: Single-cell transcriptomics of 20 mouse organs creates a Tabula Muris. Nature, 562: 367-372, 2018. 10. Baron, M, Veres, A, Wolock, SL, Faust, AL, Gaujoux, R, Vetere, A, Ryu, JH, Wagner, BK, Shen-Orr, SS, Klein, AM, Melton, DA, Yanai, I: A Single-Cell Transcriptomic Map of the Human and Mouse Pancreas Reveals Inter- and Intra-cell Population Structure. Cell Syst, 3: 346-360 e344, 2016. 11. 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. 12. Wu, H, Malone, AF, Donnelly, EL, Kirita, Y, Uchimura, K, Ramakrishnan, SM, Gaut, JP, Humphreys, BD: Single-Cell Transcriptomics of a Human Kidney Allograft Biopsy Specimen Defines a Diverse Inflammatory Response. J Am Soc Nephrol, 29: 2069- 2080, 2018. 13. Wang, Y, Wang, R, Zhang, S, Song, S, Jiang, C, Han, G, Wang, M, Ajani, J, Futreal, A, Wang, L: iTALK: an R Package to Characterize and Illustrate Intercellular Communication. bioRxiv: 507871, 2019. 14. Shannon, P, Markiel, A, Ozier, O, Baliga, NS, Wang, JT, Ramage, D, Amin, N, Schwikowski, B, Ideker, T: Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res, 13: 2498-2504, 2003. 15. 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. 16. Hu, H, Miao, YR, Jia, LH, Yu, QY, Zhang, Q, Guo, AY: AnimalTFDB 3.0: a comprehensive resource for annotation and prediction of animal transcription factors. Nucleic Acids Res, 47: D33-D38, 2019. 17. Aibar, S, Gonzalez-Blas, CB, Moerman, T, Huynh-Thu, VA, Imrichova, H, Hulselmans, G, Rambow, F, Marine, JC, Geurts, P, Aerts, J, van den Oord, J, Atak, ZK, Wouters, J, Aerts, S: SCENIC: single-cell regulatory network inference and clustering. Nat Methods, 14: 1083-1086, 2017.