BASIC RESEARCH www.jasn.org

The Integrated RNA Landscape of Renal Preconditioning against Ischemia-Reperfusion Injury

Marc Johnsen,1 Torsten Kubacki,1 Assa Yeroslaviz ,2 Martin Richard Späth,1 Jannis Mörsdorf,1 Heike Göbel,3 Katrin Bohl,1,4 Michael Ignarski,1,4 Caroline Meharg,5 Bianca Habermann,6 Janine Altmüller,7 Andreas Beyer,3,8 Thomas Benzing,1,3,8 Bernhard Schermer,1,3,8 Volker Burst,1 and Roman-Ulrich Müller 1,3,8

Due to the number of contributing authors, the affiliations are listed at the end of this article.

ABSTRACT Background Although AKI lacks effective therapeutic approaches, preventive strategies using precondi- tioning protocols, including caloric restriction and hypoxic preconditioning, have been shown to prevent injury in animal models. A better understanding of the molecular mechanisms that underlie the enhanced resistance to AKI conferred by such approaches is needed to facilitate clinical use. We hypothesized that these preconditioning strategies use similar pathways to augment cellular stress resistance. Methods To identify genes and pathways shared by caloric restriction and hypoxic preconditioning, we used RNA-sequencing transcriptome profiling to compare the transcriptional response with both modes of preconditioning in mice before and after renal ischemia-reperfusion injury. Results The gene expression signatures induced by both preconditioning strategies involve distinct com- mon genes and pathways that overlap significantly with the transcriptional changes observed after ischemia-reperfusion injury. These changes primarily affect oxidation-reduction processes and have a major effect on mitochondrial processes. We found that 16 of the genes differentially regulated by both modes of preconditioning were strongly correlated with clinical outcome; most of these genes had not previously been directly linked to AKI. Conclusions This comparative analysis of the gene expression signatures in preconditioning strategies shows overlapping patterns in caloric restriction and hypoxic preconditioning, pointing toward common molecular mechanisms. Our analysis identified a limited set of target genes not previously known to be associated with AKI; further study of their potential to provide the basis for novel preventive strategies is warranted. To allow for optimal interactive usability of the data by the kidney research community, we provide an online interface for user-defined interrogation of the gene expression datasets (http://shiny. cecad.uni-koeln.de:3838/IRaP/).

JASN 31: 716–730, 2020. doi: https://doi.org/10.1681/ASN.2019050534

Received May 27, 2019. Accepted January 5, 2020. The incidence of AKI is steadily increasing, leading M.J., T.K., V.B., and R.-U.M. contributed equally to this work. to relevant morbidity and mortality and causing a growing economic burden to Western health care Published online ahead of print. Publication date available at systems.1 Despite extensive research, therapies for www.jasn.org. AKI are still lacking in clinical practice. In contrast, Correspondence: Dr. Roman-Ulrich Müller, Cologne Excellence preventive strategies using various so-called pre- Cluster on Cellular Stress Responses in Aging-Associated Dis- eases, Josef-Stelzmann-Street 26, 50931 Cologne, Germany, or conditioning protocols, including caloric restric- Dr. Volker Burst, Department II of Internal Medicine, University of tion (CR) and interventions activating hypoxia Cologne, Kerpener Street 62, 50937 Cologne, Germany. E-mail: signaling (ischemic preconditioning [IP] and hyp- [email protected] or [email protected] – oxic preconditioning [HP]),2 4 have been found to Copyright © 2020 by the American Society of Nephrology

716 ISSN : 1046-6673/3104-716 JASN 31: 716–730, 2020 www.jasn.org BASIC RESEARCH be extraordinarily effective in animal models. Although it has Significance Statement longbeenknownthatCR5 and activation of hypoxia signal- ing6,7 lead to an extension of healthy lifespan, more recent Preconditioning strategies, such as caloric restriction and hypoxic experiments demonstrated that a short-term application of preconditioning, show strongly protective effects in animal models CR or IP/HP induces robust protection of various organs, of AKI, and researchers hope exploration of these strategies might provide insights into translating these powerful interventions to the 8 9,10 9 including heart, kidney, brain, and liver. clinical setting. However, the molecular mechanisms underlying the So far, a comprehensive understanding of the molecular beneficial effects of short-term application of caloric restriction and mechanisms underlying the beneficial effect of short-term hypoxic preconditioning have remained elusive. The authors used CR and oxygen deficiency is still lacking. Strikingly, these RNA-sequencing transcriptome profiling to compare the tran- effects can be observed in a wide range of species from inver- scriptional response with both modes of preconditioning before and after renal ischemia-reperfusion injury, identifying genes and 10–13 tebrates to mammals, and the protective effect is not pathways commonly shared by the two strategies. A comparison of confined to distinct organs but rather affects whole organisms. these findings with genes dysregulated during AKI points to genes This strongly points to evolutionary conserved mechanisms involved in preconditioning-associated organ protection that can that are paramount for the cellular defense against various now be examined as potential therapeutic targets in AKI. injuries.5 The fact that CR and HP share their protective potential has day, respectively. No increase in mortality or morbidity was led us to hypothesize that they utilize common pathways. observed after HP. Here, we report our findings of a detailed comparative expres- sion analysis in kidneys of preconditioned mice. Renal Ischemia-Reperfusion Injury Model A warm renal ischemia-reperfusion injury (IRI) model was used as described elsewhere, with slight modifications.12 METHODS Briefly, after anesthesia with intraperitoneal application of ketamine/xylazine, the right kidney was removed and the Ethical Statement left renal pedicle was clamped for 40 minutes. Postsurgical All animal procedures were conducted in accordance with Eu- recovery was assessed on the basis of weight loss, activity (nest- ropean (European Union directive 86/609/EEC), national and ing, flight behavior, movement), and appearance (grooming, institutional guidelines and approved by local governmental tachypnoea, dehydration) on a daily basis. Mice were eutha- authorities (LANUV 84–02.04.2013.A158). nized at 4, 24, or 72 hours after ischemia. Animals Male C57BL6/J mice aged 8–12 weeks were housed under Analysis of Renal Function and Overall Health identical specific-pathogen-free conditions in group cages Performance (five animals per cage) at a relative humidity of 50%–60% Blood samples were collected via puncture of a buccal vein or fi and with a 12-hour light/dark rhythm. All mice received water by nal bleeding. Serum urea and creatinine levels were mea- ad libitum and all except the caloric restricted mice received sured on a Cobas C 702 and Creatinine Plus-Test version 2 food ad libitum.Foodwasobtainedfromssniff(Art. (both Roche Diagnostics, Mannheim, Germany). Mortality V1534–703; Soest, Germany). and a postoperative recovery score (see Supplemental Appendix 1) were assessed over a period of 72 hours after IRI. CR Average food intake was determined by daily weighing of re- Histopathology maining food pellets for a period of 2 weeks. For experiments, Acute tubular damage was evaluated in a blinded fashion by an CR was applied for 4 weeks and mice were fed 70% of observed experienced renal pathologist (H.G.) using sections stained average food consumption. Mice were weighed on a weekly with periodic acid–Schiff (five visual fields per section) and basis to monitor weight loss. Neither increased mortality nor categorized using the scoring system proposed by Tirapelli and morbidity were observed during CR. Goujon14,15 on the basis of the presence of vacuolization, ep- ithelial flattening, loss of brush border, loss of nuclei, and – HP necrosis. Results were graded 0 4 according to the affected For HP, mice were put in a sealed chamber with free access to area (1: 0%–25%, 2: 25%–50%, 3: 50%–75%, 4: 75%–100%). water and food on three consecutive days. To achieve a nor- mobaric, hypoxic environment, oxygen was gradually re- Terminal Deoxynucleotidyl Transferase–Mediated placed by nitrogen over a period of 20 minutes, yielding Digoxigenin-Deoxyuridine Nick-End Labeling Staining a final oxygen concentration of 8% (Oxygen sensor, Grei- DeadEnd Fluorometric TUNEL System (Promega) was used singer GMH 3690; GMH Meßtechnik GmbH, Remscheid, on formalin-fixed paraffinsections(2mm) according to the Germany). The animals were exposed to hypoxia for 2 hours manufacturer’s protocol. Pictures were taken with a Zeiss on the first day and 4 and 8 hours on the second and third Meta 710 confocal microscope for documentation.

JASN 31: 716–730, 2020 Gene Expression Signature of Renal Organoprotection 717 BASIC RESEARCH www.jasn.org mRNA and MicroRNA Sequencing mitoXplorer was used to analyze and visualize mitochon- RNeasy mini Kit (Qiagen, Hilden, Germany) was used to iso- drial expression dynamics.24 For visualization of the data, late total RNA from snap-frozen kidneys. After removal of access http://mitoxplorer.ibdm.univ-mrs.fr/, then go to ribosomal RNA using biotinylated target-specific oligos com- “analysis,” choose “Mouse” as organism and select project bined with Ribo-Zero ribosomal RNA removal beads, RNA “Kidney_Injury” to view the data. was fragmented into small pieces and copied into first-strand complementary DNA (cDNA) followed by second-strand Cell-Specific Marker Analysis cDNA synthesis. Products were purifiedandenrichedwith To gain more insight into cell type–specific changes we used a PCR to create the final cDNA library. After library validation published list of marker genes by Clark et al.25 The matrix of and quantification (2100 Bioanalyzer; Agilent), equimolar tpm values (generated with kallisto, version 0.43.1, and pro- amounts of five to six pooled libraries were quantified by using cessed with tximport, version 1.6.0, see also section “Pseudo- the Peqlab KAPA Library Quantification Kit and the Applied time analysis”) was Z-transformed for each gene separately Biosystems 7900HT Sequence Detection System and se- across all samples. The resulting Z-scores were averaged over quenced on a Hiseq2000 sequencer.16 all marker genes for each cell type for each group (type of MicroRNA (miRNA) libraries were prepared using the Il- preconditioning and time point). These mean Z-scores were lumina TruSeq Small RNA Sample Preparation Kit, as de- plotted as box-whisker plots for specific cell types and as a scribed elsewhere.17 bubble plot for all cell types. Mapping, trimming, and adapter removal are described in Supplemental Appendix 1. mRNA and miRNA expression Pseudotime Analysis were analyzed using DESeq2 (version 1.8.2)18 package of R Pseudotime analysis was performed as described before using (version 3.2.0) software (https://www.R-project.org/) the Monocle workflow in R26 (version 2.6.4). Here, a method (Supplemental Figure 1). Count data were normalized using that was originally created for single-cell data were applied to the size factor to estimate the effective library size.19 For du- our bulk RNA-seq dataset by treating each sample like a single plicated miRNAs the mean value was taken after calculating cell. Library tximport, version 1.6.0,27 was used to process the dispersion across all samples. Pairwise comparison of different kallisto, version 0.43.1,28 output files, resulting in tpm values conditions resulted in a list of differentially expressed RNAs for each gene. Genes were classified as being expressed with a applying a P value cut-off of ,0.05. P values were adjusted for tpm .0.1 in at least two samples. The differential expression multiple testing to reduce false discovery rate. analysis of Monocle was on the basis of the groups (type of preconditioning and time point of nephrectomy). Sample or- Data Accessibility and Data Sharing dering (“cell ordering” in the original Monocle workflow), For interactive online accessibility of RNA-sequencing (RNA- dimension reduction, and pseudotime trajectory calculation seq) data, a database was created with the “shiny” package in R was done according to the standard Monocle workflow. and is available at http://shiny.cecad.uni-koeln.de:3838/IRaP/. RNA-seq primary data can be found at https://www.ebi.ac.uk/ PCA arrayexpress/experiments/E-MTAB-7982. A standard PCA was done for all genes that monocle uses to reconstruct the pseudotime trajectories using the R-function Gene Ontology and Pathway Analyses prcomp. The matrix of tpms per gene was transformed into a Gene ontology (GO) and KEGG20 pathway analyses were done matrix of fold changes (each measurement versus the mean with the DAVID version 6.8 functional annotation tool over all samples). The loadings of the first and second princi- (https://david.ncifcrf.gov). An EASE score (modified Fisher pal component were visualized in a scatterplot. We defined exact P value) of ,0.05 was defined as significance threshold. upper, lower, and right outliers on the basis of a cut-off of The GO analyses of the outliers in the principal component 60.05, i.e.,defining the outer 2.13% of genes as outliers. For analysis (PCA) were done with the R package topGO, version these gene sets, GO and KEGG enrichments were calculated. 2.30.0,21 using the elim algorithm with Fisher exact test. All measured genes (transcripts per million [tpm] of .0inatleast Outcome Score and Outcome-Related Gene Clustering one sample) were used as the background. The KEGG analyses An outcome score on the basis of weight loss and serum urea of the outliers in the PCA were done with the R package clus- levels 24 and 72 hours after IRI was implemented rating from 1 terProfiler (version 3.6.0), using the function enrichKEGG.22 to 8 for each item, with 8 being the most severe damage or Signaling pathway impact analysis (SPIA) was done with the death (see Supplemental Appendix 1). As a caveat for this SPIA package for R, version 2.30.0,23 following the standard analysis, CR mice showed a lower weight and higher urea at workflow of the manual, using all measured genes (tpm of .0 baseline (Supplemental Figure 2, B and C), which was accoun- in at least one sample) as the background. From the output ted for by using fractional instead of absolute changes. For a tables we used observed total preturbation accumulation (tA) normalized baseline gene expression and to prevent bias and the False Discovery Rate adjusted global p-values (pGFdr) through expression outliers, the mean expression for each of the pathways, with a pGFdr,0.05 for plotting. gene was calculated. Spearman correlation was calculated

718 JASN JASN 31: 716–730, 2020 www.jasn.org BASIC RESEARCH between normalized expression intensities and quantified after IRI (Figure 2, D–F), kidney architecture was markedly score values. The calculation was used to the complete dataset improved in HP and fully restored in CR kidneys after and a coefficient threshold value of rS.0.95 was set to define 72 hours. Nonpreconditioned animals still showed cast for- significantly correlated genes. Standard univariate linear re- mation and loss of nuclei at that time (Supplemental Figure 3). gression analyses were performed to estimate the association Attenuated cell death in preconditioned animals was con- of each gene with the outcome score. Because of the small firmed by terminal deoxynucleotidyl transferase–mediated sample size and a high degree of multicollinearity, we used a digoxigenin-deoxyuridine nick-end labeling (TUNEL) stain- mean-centered dataset for multivariate regression analysis but ing (Figure 2, G–I). included only two genes at a time. Gene Expression Quantitative Real-Time PCR We performed transcriptional analyses from preconditioned Quantitative PCR was performed using TaqMan Custom Ar- (HP, CR) and nonpreconditioned animals at three times: after rays on an ABI 7900 HT thermocycler (Applied Biosystems, preconditioning (0 hours, directly before IRI), and 4 and Life Technologies Cooperation, Carlsbad, CA). mRNA was 24 hours after IRI (Figure 3A). As shown in the PCA transcribed with a high-capacity reverse transcription kit (Ap- (Figure 3B), samples at 0 and 4 hours cluster closely together. plied Biosystems). Expression levels were normalized to At 24 hours after damage, samples are more spread out. Both housekeeping genes (Polr2a, Ubc, Gapdh, Rn18s-rs5)andcal- modes of preconditioning lead to profiles more similar to un- culated with the comparative threshold cycle method. For harmed animals, with CR animals showing the largest effect. primer identifiers see Supplemental Appendix 1. Analysis These results are confirmed using pseudotime trajectories29 was done with Expression Suite software from Thermo Fisher (Supplemental Figure 4). Scientific. We then examined the effect of IRI on the expression of recently published cell-specific markers25 as a surrogate pa- Statistical Analysis of Clinical and Laboratory rameter for cellular composition. This analysis showed a Parameters marked reduction of markers specific for the proximal tubule Statistical analysis was done with GraphPad Prism Software 24 hours after IRI in nonpreconditioned animals as compared version 6.0c. Results are presented as means6SD. For each with CR and HP (Figure 3C). Likewise, macrophage markers experiment, at least three biologic replicates were examined. increased markedly after IRI in nonpreconditioned animals To calculate differences between multiple groups we used (Figure 3D), whereas there were only minor changes seen in two-way ANOVAand a Tukey multiple comparisons test. Sig- CR and HP. For an overview of all cell-specific changes, see nificance of weight differences between CR and nonprecon- Supplemental Figure 5. ditioned animals was calculated with multiple t tests. Transcriptional Changes Induced by Preconditioning CR and HP induced the differential regulation of 3599 and 321 RESULTS genes, respectively, with a significant overlap of 230 genes dis- playing concordant transcriptional changes (i.e.,onlytwo Functional and Phenotypical Characterization genes were not regulated in the same direction) (Figure 4B, The experimental setup is depicted in Figure 1A. Both HP and see also Supplemental Table 1 or the online repository). These CR led to a significantly attenuated rise of creatinine and urea findings could be confirmed for 40 of 44 randomly chosen after 24 hours and a trend toward lower values after 72 hours genes by quantitative PCR (Supplemental Figure 6). When (Figure 1, B and C). Although none of the CR mice (n513) we compared the genes most strongly regulated by either CR and only two of 12 (17%) HP mice died within 72 hours after or HP, we found an overlap of six genes among the top ten IRI, eight of 14 (57%) nonpreconditioned animals died or had upregulated genes and one gene in the top ten downregulated to be euthanized according to previously defined criteria genes (Figure 4C, online repository). (Figure 1D). Preconditioned mice showed a markedly im- A comparison of HP and CR on the basis of separate GO proved general state of health after IRI, reaching significance term analyses revealed three overlapping biologic processes, in CR mice (Supplemental Figure 2D). “lipid metabolic processes,”“metabolic processes,” and A histology scoring system revealed statistically significant “oxidation-reduction processes,” that were enriched in re- differences in nonpreconditioned animals and HP mice sponse to both modes of preconditioning (Figure 4D, 24 hours after IRI, whereas kidneys in the CR group were Supplemental Table 2). KEGG pathway analyses showed “met- not different from unharmed organs (Figure 2B). This result abolic pathways,”“peroxisome,” and “glutathione metabo- was primarily driven by the extent of tubular necrosis lism” to be significant in both groups (Figure 4E, (Figure 2C). Although other common signs of AKI (i.e., brush Supplemental Figure 7, Supplemental Table 3). To allow for border loss, tubular cell flattening with prominent tubular a better insight into the actual effect on glutathione metabo- lumina, and focal formation of tubular casts) were observed lism and the peroxisome, we provide a visualization of the in all animals at 4 hours (Supplemental Figure 3) and 24 hours changes on the basis of KEGG maps in Supplemental

JASN 31: 716–730, 2020 Gene Expression Signature of Renal Organoprotection 719 BASIC RESEARCH www.jasn.org

A B 2.0 ** non-PC HP 1.5 CR = Ischemia-Reperfusion non-PC Injury (IRI) IRI - Removal right kidney - Histology - RNAseq 1.0 CR - Blood analysis = sacrifice with - Removal left kidney HP - Histology 0.5 - RNAseq creatinine mg/dl = blood analysis -28d -3 -2 -1 0h 4h 24h 72h 0.0 4h 24h 72h

C D 100 ** 800 non-PC ** *** HP 600 CR

400 50

urea mg/dl 200 Percent survival

0 0 4h 24h 72h 24h 72h non-PC HP CR

Figure 1. Experimental setup and preconditioning mediated protection against AKI. Experimental setup showing preconditioning modes and times of sample collection (blood, kidney tissue). For details see Methods section. (A) Overview of the study design il- lustrating experimental groups and timepoints of sampling. (B) Percentage survival after ischemic kidney injury and contralateral ne- phrectomy (B). (C) Creatinine and (D) urea values after IRI. n55 animals per group (P50.002 for nonpreconditioned animals versus CR; other comparisons are NS). *P,0.05; **P,0.01; *** P,0.001. non-PC, nonpreconditioned.

Figures 8 and 9. In addition, we performed a SPIA of genes NGAL],32 and Timp133;SupplementalFigure11B)among regulated by preconditioning resulting in seven significantly our top upregulated genes, as well as numerous other genes regulated pathways in response to CR (Figure 4F), of which previously described to be associated with early damage, in- “Alzheimer’s, Parkinson’s and Huntington’s disease,”“NAFLD cluding Jun, Fos, Btg2, Egr1, Zfp36, Klf4, Klf6,andCsrnp1 (non-alcoholic fatty liver disease),” and “Protein processing in (Supplemental Table 1). SPIA of the two damage timepoints endoplasmic reticulum” were also found via KEGG pathway in nonpreconditioned animals showed a transient pattern of analysis (Supplemental Table 3). The largest group of genes pathway activation with FOXO, C-type lectin receptor, and contributing to overrepresentation of neurodegenerative dis- IL-17 signaling pathway being part of an early response to ease terms and nonalcoholic fatty liver disease were genes en- IRI and NF-kB, PI3K-Akt, AGE-RAGE, and TNF-signaling coding for mitochondrial proteins (Supplemental Table 4). activated after 24 hours (Supplemental Figure 11C, HP did not show any significant terms in SPIA, most likely Supplemental Table 5). Of note, 203 of the 230 genes altered because of smaller overall gene expression changes induced by by preconditioning were also significantly regulated 24 hours HP. Apart from the mRNA analyses, we also sequenced after IRI in nonpreconditioned animals. Six of the seven sig- miRNAs to shed light on this potential additional layer of reg- nificant SPIA terms identified in CR animals before damage ulation. However, no miRNAs were regulated after HP.Related were also found significant in SPIA of nonpreconditioned an- information is provided in Supplemental Figure 10. imals 24 hours after IRI, with four of them being regulated in the opposite direction (Supplemental Figure 11D). Because Transcriptional Changes after IRI post-transcriptional mechanisms are likely to contribute to IRI induced major changes in gene expression with 8006 our gene expression patterns, we analyzed the expression of (4 hours) and 10,206 (24 hours) genes being differentially a set of genes known to affect mRNA stability and decay during regulated in nonpreconditioned animals (Supplemental stress.34 Seven stabilizing factors and nine decay factors were Figure 11, Supplemental Table 1, online repository). In line significantly dysregulated at 4 or 24 hours after damage, with with published data, Keratin 20 was the top upregulated gene only Rnpc1 and Mex3d reaching a log2 fold-change .1 at both postischemic timepoints.30 Furthermore, we found (Supplemental Table 6). transcription factors Atf3, Fosb, Maff, Dusp5, and well known Because a mere comparison of gene lists would not have markers of AKI (Havcr-1 [aka Kim1],31 Lcn-2 [encoding yielded conclusive results, a more global approach was chosen

720 JASN JASN 31: 716–730, 2020 www.jasn.org BASIC RESEARCH

A B C 15 ** ** 4 ***** 10 3 ** 2 5 quantity 1 damage score ø PAS baseline 0 0

baseline HP 24h CR 24h

non-PC 24h loss of nuclei tubular necrosis brush border loss tubular flattening baseline non-PC HP CR

non-PC CR HP D E F

# # # # #

PAS 24h # # #

G H I TUNEL 24h

Figure 2. Preconditioning reduces histologic signs of renal IRI. (A) Baseline kidney histology from nonpreconditioned animals. (B) Histologic damage score at baseline and 24 hours after IRI in the different groups. (C) Subitems of the damage score 24 hours after IRI. (D–F) Representative period acid–Schiff stains from baseline and postischemic kidneys 24 hours after IRI from nonpreconditioned controls, HP, and CR group (magnification 2003). *indicates tubular flattening; arrowheads indicate nuclear loss; arrows indicate tu- bular casts; and #indicates denuded tubuli with luminar debris. (G–I) Terminal deoxynucleotidyl transferase–mediated digoxigenin- deoxyuridine nick-end labeling assay from control, HP, and CR groups 24 hours after IRI, with cell death in green, (magnification 2003). to further analyze the interaction between preconditioning gene set. Figure 5C shows a histogram of all significant GO and IRI. Using the PCA shown in Figure 3B, we visualized terms (all gene sets and corresponding GO and KEGG analyses the loadings of the first and second principal component in are provided in Supplemental Table 7). a scatterplot to examine the degree to which specificgenes contribute (Figure 5, A and B). On the basis of the clustering Focused Analysis of Mitochondrial Processes of the experimental groups and the corresponding loadings, Because the KEGG, SPIA, and GO term analyses described we identified three gene sets of outliers that were assigned with above hinted, in line with the literature, toward mitochondria the following functional implications: “early damage,”“late being a key player in both damage35,36 and precondition- damage,” and “preconditioning-mediated protection.” Using ing,37,38 we used mitoXplorer24 to obtain a more detailed aGOanalysis,“early damage” was enriched for terms associ- view on mitochondrial processes. Preconditioning via CR af- ated with embryogenesis, inhibition of transcription, and the fected mitochondrial pathways much more than HP unfolded protein response. “Late damage” contained genes (Supplemental Figure 12). Both preconditioning interven- that can be associated with adaptation and repair. tions had a strong effect on mitochondrial reactive oxygen “Preconditioning-mediated protection” is primarily associ- species (ROS) defense (Figure 6A, see mitoXplorer); most reg- ated with metabolic processes. Interestingly, oxidation- ulated genes in mitochondrial ROS defense were identical in reduction processes were overrepresented in both the “late CR and HP (Figure 6A). IRI affected almost all mitochondrial damage” as well as the “preconditioning-mediated protection” processes, some of which were also induced by CR, with most

JASN 31: 716–730, 2020 Gene Expression Signature of Renal Organoprotection 721 BASIC RESEARCH www.jasn.org

A B HP 4h CR 4h 10 non-PC 4h non-PC 4h CR 4h HP 4h non-PC 4h HP 4h

5 non-PC 24h HP 24h non-PC 24h non-PC 0h non-PC = Ischemia-Reperfusion non-PC 0h IRI 0 non-PC 24h Injury (IRI) CR 0h - Removal right kidney non-PC 0h CR 0h

PC2 HP 0h CR 0h - Histology CR 24h - RNAseq non-PC 0h HP 0h CR 0h HP 24h CR - Blood analysis HP 0h CR 4h -5 CR 0h HP 24h = sacrifice with CR 4h HP 0h - Removal left kidney HP 0h - Histology HP 24h HP - RNAseq HP 4h -10 CR 24h = blood analysis non 0h CR 24h -28d -3 -2 -1 0h 4h 24h 72h non-PC 24h CR 24h non-PC 4h -15

-10 0 10 20 30 PC1

C Proximal tube D Macrophage

0.5 1.5

0.0 1.0

-0.5 0.5 Z-score Mean Z-score Mean

-1.0 0.0

-1.5 -0.5

h CR 0 CR 4h HP 0h HP 4h CR 0h CR 4h HP 0h HP 4h P 24h CR 24h HP 24h CR 24h H non-PC 0hnon-PC 4h non-PC 0non-PC 4h non-PC 24h non-PC 24h

Figure 3. PCA and cell-specific changes by preconditioning and damage. (A) Experimental setup overview. (B) PCA of all CR, HP, and nonpreconditioned (non-PC) animals at 0, 4, and 24 hours. (C and D) Analysis of expression of cell specific markers for proximal tubule (C) and macrophages (D) shown as Z-score on the basis of a recently published atlas of cell type–specificmarkers.25 The lower and upper hinges correspond to the first and third quartiles (the 25th and 75th percentiles). The middle line represents the median. of the genes affected being regulated in the same direction. change in urea, and death; see Supplemental Appendix 1). The Again, the most prominent changes were observed in ROS expression of 30 genes was strongly positively correlated with defense (upregulated: Gsta1, Gsta2, Mgst1;downregulated: outcome (rS.0.95) and the expression of four genes was Mpvl17L, Sod1, Prdx5)(Figure6,BandC)andtoalesser strongly negatively correlated with outcome (rS,20.95) extent in mitochondrial dynamics (Tcaim), metabolism of (Supplemental Figures 13 and 15). Of note, 16 of the 30 pos- lipids and lipoproteins (Nudt19, Hsd3b2), fatty acid degrada- itively correlated (but none of the negatively correlated) genes tion (Acadm, Acad9), and fatty acid biosynthesis (Acsm3). In had already been identified in our overlap analysis of genes contrast to this, CR led to downregulation of genes involved in regulated by both modes of preconditioning before IRI apoptosis, e.g., Bik and Dynll1 (Figure 6D), whereas IRI upre- (Figure 7A). At 24 hours after IRI, preconditioned animals gulated most genes in this process. uniformly showed less downregulation or stronger upregula- tion of these 16 outcome-correlated genes (Supplemental Correlation of the Baseline Expression of Single Genes Table 1, online repository). These effects were most prominent with Outcome in CR animals, which regarding five genes even showed an op- To further characterize the effect of transcriptional changes on posite regulation after IRI compared with nonpreconditioned clinical end points we correlated the transcriptome data of all animals (Odc1, Cmtm6, Gm7278, Ces2c, Slc39a11). Nhp2 was the 14 animals at baseline irrespective of the experimental group only one of these genes that showed upregulation in all with a predefined outcome score (calculated from weight loss, groups 24 hours after IRI (Figure 7B). Univariate regression

722 JASN JASN 31: 716–730, 2020 www.jasn.org BASIC RESEARCH

A B D GOTerms BP Overlap CR/HP 0h

lipid metabolic process non-PC IRI 23091 3369 metabolic process CR oxidation-reduction process HP 01234 enrichment factor -28d -3 -2 -1 0h HP0h CR 0h HP CR C E Top 10 upregulated genes after Top 10 upregulated genes after Caloric Restriction Hypoxic Preconditioning KEGG Pathways Overlap CR/HP 0h Gene name Log2FC Gene name Log2FC Metabolic pathways Slc7a12 3,47832243 Pcsk9 0,73939775 Kynu 2,16299815 Hao2 0,70298909 Glutathione metabolism Rdh16f2 1,94330539 Aqp4 0,67009173 Hao2 1,74300252 Slc7a12 0,66532859 Peroxisome Prlr 1,72105189 Prlr 0,58087969 Abcb1b 1,53769998 Cldn1 0,55187429 0246810 Abcb1a 1,53383682 Cyp2c44 0,53347737 enrichment factor Bhmt 1,50612378 Kynu 0,53232252 Aqp4 1,48188125 Palm3 0,52693101 HP CR Serpina1a1,38153707 Rdh16f2 0,52221903 F Top 10 downregulated genes after Top 10 downregulated genes after Thermogenesis Caloric Restriction Hypoxic Preconditioning Non-alcoholic fatty Gene name Log2FC Gene name Log2FC liver disease (NAFLD) Ctxn3 -2,6776447 9030619P08Rik -0,7612083 Parkinson disease -log10(pGFdr) Acsm3 -2,1044643 Slc22a7 -0,7293227 10.0 Gm15895 -2,0327822 Nudt19 -0,6583203 Huntington disease 7.5 Hsd11b1 -1,9192658 Reln -0,6450104 Protein processing in 5.0 Mki67 -1,9192143 Gm7278 -0,6108729 endoplasmic reticulum Cyp4a12a -1,918897 Srd5a2 -0,6067721 Alzheimer disease Slco1a1 -1,7597074 Odc1 -0,603844 Akr1c18 -1,7057526 Cndp2 -0,5948558 Viral myocarditis Lpl -1,6996584 Acsm3 -0,5491107 Kif20b -1,6700135 Vwa1 -0,5464013

Figure 4. HP and CR show common patterns of differential gene expression and signaling pathway modulation before IRI. (A) Overview. (B) Overlap of differentially regulated genes in the HP and CR group (P,0.001). (C) Top ten upregulated and downregulated genes in response to CR and HP. Genes in bold are common to both modes of preconditioning. (D and E) GO biologic process and KEGG pathway analysis were performed separately in HP and CR animals before IRI. A comparison of the significant results reveals three overlapping terms for biologic processes (D) and KEGG pathways (E) (adjusted P value ,0.05, false discovery rate ,0.05). (F) SPIA of CR animals (False Discovery Rate adjusted global P-values [pGFdr] ,0.05), y axis: pathway terms, x axis: observed total preturbation accumulation in the pathway (tA). FC, fold change; non-PC, nonpreconditioned animals. using standard linear models for each gene was applied. With contributions detected for the remaining ten genes. (see an estimated b-weight of .0.95, Tspan13, Myo5a, Cndp2, Supplemental Figure 14 for univariate analysis, multiple re- and Cyp7b1 appeared to be the four most important predic- gression models not shown). tors of outcome. Because inclusion of all 16 genes in a mul- tivariable regression analysis was not feasible owing to low sample size and, as expected from the retrieval process, DISCUSSION marked collinearity, we applied a standard linear model in- cluding each possible pair of genes at a time (i.e., all possible We identified a distinct set of genes and pathways that are permutations of gene pairs, n516315/2 pairs) to further linked to two preconditioning strategies, suggesting that there delineate the relative role of genes. By looking at the number might indeed be a common, conserved, molecular protective of significant (P,0.05) contributions of each individual mechanism. Metabolic as well as oxidation-reduction process- gene in the 15 analyses in which the respective gene was in- es were among the most prominent biologic processes in- cluded, again these four genes were retrieved with Myo5a volved, and mitochondria, peroxisomes, plasma membrane, reaching significance in 11, Cndp2 in 10, and Tspan13 and and the endoplasmic reticulum were revealed to be the major Cyp7b1 in seven of 15 analyses. There were no significant cellular components. Many of these findings are in line with

JASN 31: 716–730, 2020 Gene Expression Signature of Renal Organoprotection 723 BASIC RESEARCH www.jasn.org

A 10

5

0

-5 C -log10P platelet-derived growth factor receptor signaling pathway -10 negative regulation of transcription from RNA PC2 (18.8% explained var.) PC2 (18.8% explained polymerase II promoter cellular zinc ion homeostasis -15 I face morphogenesis

-100 102030 post-embryonic development PC1 (50.9% explained var.) endoplasmic reticulum unfolded protein response CR0h HP0h nonPC0h membrane raft assembly CR-IR24h HP-IR24h nonPC-IR24h cellular response to acid chemical

CR-IR4h HP-IR4h nonPC-IR4h II angiogenesis

regulation of receptor-mediated endocytosis B oxidation-reduction process Fosb positive regulation of cell-substrate adhesion 0.2 Atf3 oxidation-reduction process Vasn Ppp1r15a Csrnp1 Slc20a1 Chka Abl2 steroid metabolic process Tiparp Jun Dnajb4 Dusp14 Tgif1 III Gm10720 Ddit3 Cdcp1 Mafk Bach1 Emp1 Nek6 alcohol metabolic process 0.1 Mt2 Plin2 Dusp1 Cry2 Jmjd1c Ccnl1 Slc39a14 Pdlim7 Sfn Olfr1033 Ankrd33b Osgin1 Dnajb9 Lrrfip2AC109138.2 Mt1 Serpine1 organic hydroxy compound metabolic process Sgk1 S100a10 Fam107b Eps8 Plekho1 Col4a1 Nedd4l Sqstm1 Pakap Zbtb38 Ucp2 Eif4ebp1 Sgms2 Slc16a1 Gm45551 Tuft1 Arid5b Rtn4 Stat3 Col4a2 Tagln Vat1 Mvp Anxa2 Txnip Clu H3f3b Nrip1 Bcar1 Rras2 Col18a1 Fabp4 0.0 Acat3 Arg2 Akr1b8 Ctnnal1 Riok1 Arpc1b Flot1 Hacd1S100a11 PC2 Steap1 Spp1 Sparc Cstb Sh3bgrl3 Col3a1 Plp2 G6pc Lgals1 2200002D01Rik Tpm2 Pea15a Sh3tc2 Mgst1 Capg Gsta1 Gm3776 Kynu Npl Fdps Trf Jpt1 Tuba1b Sprr1a Hp Apoc3 Cyp24a1 Aldh1a1 Aldh1a7 -0.1 Gc

Gm10639

-0.2 Ugt1a2 -0.05 0.00 0.05 0.10 0.15 PC1

Figure 5. Genes influencing PCA and GO analysis for biologic processes. (A) Biplot (PCA plus loadings) of all CR, HP, and non-PC samples pre- and post-IRI. (B) Loading plot showing genes with the most influence on the principal components. Cut-offs were set to 0.05 to receive three gene sets (I: upper outliers, 50 genes; II: right outliers, 58 genes; III: lower outliers, 15 genes) corresponding to the sample clusters of early damage, late damage, and precondition-mediated protection. (C) Histogram showing–log10 of P values of GO of biologic processes of the three gene sets (I, II, and III) identified via loading plot (Padjusted ,0.05, minimum three genes per term). non-PC, nonpreconditioned animals. the existing literature.2,38–43 However, the comparative ap- kidneys after reperfusion46 among our top genes 4 hours after proach to CR and HP in its interaction with IRI revealed sev- IRI showing the validity of our approach. Pathway analysis eral interesting candidates that have not been in the focus of using SPIA showed activation of FoxO signaling, a key regu- AKI research so far and warrant further investigation. lator of apoptosis, cell cycle progression, glycolysis, stress re- Using markers for specific nephron segments, we con- sistance, longevity, and immune cells.47–49 The importance of firmed the loss of proximal tubular cells and the accumulation immune responses after damage is further highlighted in our of immune cells as the major cellular sequelae after IRI. Be- dataset through activation of the IL-17 signaling and C-type cause proximal tubule cells are primary targets to damage in lectin receptor signaling pathways. The fundamental role of AKI and account for around 50% of the total kidney sub- IL-17 signaling in the pathogenesis and protection from AKI stance,25 this approach outlines an important tool for future has been shown in murine experiments50,51 and already avail- studies using similar datasets in whole-kidney samples. able therapies for spondyloarthritis and psoriasis that block As expected, IRI led to profound changes in the transcrip- IL-17 signaling52,53 make it an interesting target in humans to tional profile similar to findings of other investigators.44,45 We be evaluated in clinical studies. C-type lectin receptor signaling— found numerous genes, which had been described to be in- activated via damage-associated molecular patterns in the duced early after IRI in mice30 as well as in human donor early phase after ischemia-reperfusion—has been shown as

724 JASN JASN 31: 716–730, 2020 www.jasn.org BASIC RESEARCH

A ROS Defence C ROS Defence D Apoptosis 2 Txnrd1 Aifm3 Mgst1 Bak1 Gsta1 Tmem173 Gsta2 Bbc3 Gsta1 Mpv17l Bcl2l1 Gsta2 Gsr Mcl1 Pmaip1 1 Mgst1 Ogg1 Gstp1 Bik

log2 fold change Gstp2 Bax Gsta2 Msrb3 Dynll1 Mgst1 Gsta1 Txn1 Tspo Gclc Bcl2l13 Tldc2 Clic4 0 Prdx1 Casp8 Ctsb Gclc Gclm Txnrd2 Apaf1 Mpv17l Nudt2 Bnip1 Gpx4 Bcl2l11 Prdx3 Pacrg -1 Gclc Mpv17l Gss Bnip3 Msra Ghitm Romo1 Aifm1 Mpv17 Trap1 Park7 Alkbh7 Glrx2 Bnip3l -2 Ngb Bcl2l2 CR 0h HP 0h Oxr1 Bad Txn2 Bok Sod2 Aifm2 B ROS Defence Cat Htra2 Mgst1 Gpx1 Casp3 Msrb2 Mapk3 Gsta1 Prdx5 Armc10 2 Gsta2 Txnrd1 Sod1 Xiap Gfer Triap1 Txnrd1 Gsta1 Gsta2 Gsr CR 0h HP 0h Mapk8 non-PC 4h Diablo non-PC 24h 1 log2 fold change Mgst1 Mapk1 Parl Gsr Prelid1 Gpx1 Tldc2 Casp9 0 Sod1 Bid Sod2 Mtch1 Apopt1 Mpv17l Anp32a Sod1 Bcl2 -1 Casp2 Tldc2 Gpx1 Csnk2a1 1.6 Mtch2 Sod2 0.8 0.0 Casp7 -2 –0.8 Fam162a –1.6 Letmd1 Rps6kb1 Mpv17l

non-PC 4 h non-PC 24h CR 0h HP 0h non-PC 4h non-PC 24h

Figure 6. Mitochondrial dynamics in preconditioning and IRI. (A and B) Scatterplots of all genes in mitochondrial ROS defense in CR 0hourandHP0hour.x axis: experimental group, y axis: fold-change compared with non-PC animals (A) and non-PC animals 4 and 24 hours after damage fold-change compared with non-PC 0 hour (B). (C and D) Heatmap of relevant genes in mitochondrial ROS defense (C) and apoptosis in CR 0 hour HP 0 hour and non-PC animals 4 and 24 hours after damage (D). non-PC, nonpreconditioned animals. a target to alleviate IRI.54–56 Further, C-type lectin receptor candidates to be studied in more detail in respect of their signaling may also be the reason for the activation of NF-kB functional involvement in future studies. One example is the signaling observed 24 hours after IRI in our data.57 Targeting acyl-CoA synthetase Acsm3, which is involved in the first step this proinflammatory pathway injury reduces tubular injury, of fatty acid metabolism and was downregulated by both pre- apoptosis, necrosis, and accumulation of inflammatory conditioning treatments. Acsm3 has been associated with cells.58 In general, pathways triggering inflammatory immune fasting-induced renal organoprotection before,13 but had responses were prominent in our pathway analyses of post- not been linked to HP. Peroxisomal Hao2, one of the top up- ischemic changes. regulated genes after both modes of preconditioning in our A detailed comparative analysis of two modes of precondi- study, is involved in a reaction that ultimately leads to the tioning and their transcriptional consequences had not been production of hydrogen peroxide. Fatty acid metabolism/ performed before. Regarding the overlap between CR and b-oxidation and redox activity in peroxisomes as well as mi- HP on a single-gene level, we found a row of interesting tochondria play central roles in the promotion of ischemic

JASN 31: 716–730, 2020 Gene Expression Signature of Renal Organoprotection 725 BASIC RESEARCH www.jasn.org

A B log2 FC -6 -4 -2 0 2

Kap Kap Cndp2 Cndp2 Odc1 Odc1 Cd36 Cd36 Akr1c14 Akr1c14 Cmtm6 Cmtm6 Gm7278 Gm7278 Ces2c Ces2c Cyp7b1 Cyp7b1 Myo5a Myo5a Nhp2 Nhp2 Slc39a11 Slc39a11 Tmem237 Tmem237 Slco1a4 Slco1a4 Tc2n Tc2n Tspan13 Tspan13

non-PC 24 h

HP 1 HP HP 2 HP HP 3 HP

CR 2 CR CR 1 CR CR 4 CR

CR 3 CR HP 24 h

Control 2 Control Control 1 Control Control 3 Control -2 -1 0 1 2 CR 24 h

Figure 7. A distinct subset of genes predicts the individual animals’ outcome. (A) Heatmap for 16 outcome-correlated genes that were also significantly regulated in response to CR as well as HP at baseline: gene expression of 10 animals at baseline (nonpreconditioned

Control 1–4, and preconditioned HP 1–3 and CR 1–3). (B) Regulation (log2 fold-change) of outcome-correlated genes 24 hours after IRI in nonpreconditioned (black bars) and preconditioned (HP blue, CR orange) animals. damage to the kidney.40,44 Kynureninase (Kynu), a commonly same genes to be differentially regulated after IRI, confirming upregulated gene in response to preconditioning, is a hydro- their implication in AKI. Apart from altered ROS handling 1 lase necessary for de novo synthesis of NAD from that was common to CR and HP, CR showed changes in terms 1 L-tryptophan.59,60 NAD is an important cofactor for genome of glycolysis, lipid metabolism, and pyruvate metabolism, all 1 stability, stress tolerance, and metabolism.61 Enhancing NAD of which are known to be important processes during IRI, e.g., availability, e.g., via supplementation, has been shown to im- as part of an ischemia-induced metabolic switch, which is prove renal function after AKI through augmented mitochon- predominant in the oxygen-sensitive proximal tubule.36 Fur- drial stress resistance.39,60,62–64 Interestingly, Olenchock et al.65 thermore, the strong CR-associated downregulation of genes showed that a-ketoglutarate–induced secretion of kynurenic involved in apoptosis at baseline may protect the kidney from acid mediates protective effects of remote IP, emphasizing the the upregulation of the same genes that is observed upon IRI central role of L-tryptophan metabolism to strategies that in- and may be one reason for the strong effect of CR on cell death. crease cellular stress resistance. Other genes from the list of the This insight could allow for more targeted interventions aim- most strongly regulated overlapping candidates (Slc7a12, ing at a modulation of cell death in the prevention and treat- Rdh16f2, Prlr) had not been associated with AKI or precondi- ment of AKI. tioning before. Aqp4 has been implicated to fulfill a protective To obtain a more detailed view of the interplay between IRI function in ischemic brain damage,66,67 but a contribution to and preconditioning, we calculated the contribution of single preconditioning in AKI has not been described. genes to variation using PCA loadings. Undamaged samples A comparative analysis of preconditioning-associated alter- (0 hour) and samples at 4 hours formed tight clusters, and ations in transcripts involved in mitochondrial processes re- samples at 24 hours showed a separation between precondi- vealed key genes associated with ROS handling that are targets tioned and nonpreconditioned animals. These clusters al- of both HP and CR (Gsta1, Gsta2, Mgst1, Mpv17l). Soluble and lowed us to identify the key genes contributing to early and membrane-bound glutathione transferases like Gsta and late damage–associated patterns as well as preconditioning- Mgst1 detoxify harmful products of membrane-lipid peroxi- mediated protection after IRI (lower outliers). Although “early dation caused by ROS, and thereby protect mitochondria and damage” (upper outliers) was associated with processes of reg- cells against oxidative damage.68 Of note, we could also see the ulation of transcription, “late damage” (right outliers) was

726 JASN JASN 31: 716–730, 2020 www.jasn.org BASIC RESEARCH linked to organization and repair, which is largely confirma- not driven by differences between the groups alone because tory of previous studies.30,45 The lower outliers, which are the association was also detected within the distinct groups likely to be connected to “preconditioning-mediated protec- (Supplemental Figure 13). Regression analysis suggested tion” were associated with oxidation-reduction and metabolic Cndp2, Tspan13, Myo5a,andCyp7b1 to be among the most processes. Here, apart from Kynu, most other genes were re- relevant predictors; however, these findings will require a con- lated to metabolic pathways and processes. Examples are firmation by a focused analysis of these candidates using larger Apoc3, a key regulator of triglyceride metabolism,69 and numbers of animals. Notwithstanding, these 16 genes do not Fdps, a key enzyme in isoprenoid biosynthesis.70 Both hapto- pertain to the well established “usual suspects” with regard to globin (HP), a well known circulating acute phase protein with AKI and protection from IRI, and thus certainly warrant a antioxidant function,71 and Transferrin (TF), necessary for more detailed examination in further studies. iron handling in our lower outlier group, have been shown A comprehensive discussion of all identifiedgenesisbe- to ameliorate toxic effects of iron and confer organ protection yond the scope of this paper. However, these data along with in the kidney and other organs.72–77 Interestingly, two of the the online repository providing researchers with detailed 15 “preconditioning-mediated protection” associated genes information (http://shiny.cecad.uni-koeln.de:3838/IRaP/) (Cyp24a1, Gc) are involved in vitamin D metabolism. may serve as a source for future studies using respective 78 CYP24A1 is a key regulator of 1,25(OH)2D3 catabolism genetically modified animal models. Importantly, the versa- and the vitamin D binding protein (GC), which plays a role tile usability of such data are also highlighted by the fact that in actin scavenging, for example after cell damage. Disruption the protective potential of preconditioning is not limited to and clumping of actin filaments has been observed in arteries IRI. A recent study published by our group, in which we an arterioles in the kidney after IRI.79 Of note, regulation of focused on differential regulation of the proteome, could actin cytoskeleton was among the top regulated pathways show the enormous potential in the setting of cisplatin- 4 hours after damage in nonpreconditioned animals, under- induced AKI.88 lining its significance in tissue damage. GC also exerts im- In conclusion, we demonstrate in this study for the first munomodulatory functions like macrophage modulation, time that common organ-protective pathways, biologic enhancing complement factor 5 (C5)–mediated signaling processes, and gene sets are activated by CR as well as HP. and the binding of endotoxins.80,81 Ugt1a2,alsobelonging Our data indicate that possible shared mechanisms between to the cluster of genes associated with “preconditioning- these two preconditioning strategies include strong regula- mediated protection,” encodes a glucuronosyltransferase tion of mitochondrial energy balance and cellular radical ho- that is highly expressed in the kidney and belongs to a family meostasis. Strikingly, the expression of a limited set of the of proteins involved in elimination of steroids, heme metab- genes induced by preconditioning was strongly associated olites, environmental toxins, and drugs, e.g. glucuronidation with the clinical outcome of individual animals after IRI. of bilirubin.82 Neuraminic acid pyruvate-lyase (NPL) is nec- These genes have so far not been linked to AKI and may offer essary to form ManNAc, the metabolites of which ultimately the opportunity to develop novel strategies for both the iden- enter the hexosamine biosynthesis pathway.83 Increased rates tification of patients at risk and the prevention of AKI. Our ofproteinO-GlcNAcylationhavebeenidentified to be an online repository will empower other researchers in the field early response to cellular stress and to be protective in models to exploit the full power of our datasets for the design of future of IRI to the heart, brain, and kidney.84–87 studies. In a final step, we hypothesized that the state of an organ at baseline could predefine its resistance to damage and that this is reflected by the transcriptome. Hence, we correlated gene expression with clinical outcome in individual animals irre- ACKNOWLEDGMENTS spective of the pretreatment (i.e., nonpreconditioned, HP or CR). Although a qualitative association between expression of Dr. Müller and Dr. Burst designed the study. Dr. Johnsen, Dr. Kubacki, some genes and clinical outcome has to be expected owing to Dr. Späth, Dr. Altmüller, and Mr. Mörsdorf performed the experi- methodological reasons, we reckoned that a quantitative cor- ments. Dr. Yeroslaviz, Dr, Habermann, Dr. Beyer, Dr. Meharg, and relation between expression levels and a predefined combined Dr. Bohl analyzed the RNA-sequencing data. Dr. Göbel analyzed end point could robustly reveal clinically meaningful gene the histopathological images. Dr. Bohl programmed the shiny candidates. A total of 34 genes were identified that showed app. Dr. Johnsen and Dr. Kubacki prepared the figures. Dr. Kubacki, an exquisitely strong correlation with the clinical outcome Dr. Johnsen, Dr. Burst, and Dr. Müller drafted the manuscript. score and 16 of these genes were also identified in the overlap Dr. Benzing and Dr. Schermer revised the manuscript critically for of genes that were common to both modes of preconditioning. intellectual content. DifferencesatbaselineinureaandbodyweightintheCR We thank Martyna Brütting and Ruth Herzog for excellent tech- group are a caveat regarding these analyses, which we accoun- nical assistance. Special thanks to the Bioinformatics Core Facility at ted for by analyzing fractional instead of absolute changes in Cologne Excellence Cluster on Cellular Stress Responses in Aging- these parameters. Importantly, the observed correlation was Associated Diseases (CECAD) for help with the analysis of the data.

JASN 31: 716–730, 2020 Gene Expression Signature of Renal Organoprotection 727 BASIC RESEARCH www.jasn.org

DISCLOSURES Supplemental Table 2. GO term analysis. Supplemental Table 3. KEGG pathway analysis. None. Supplemental Table 4. Gene lists of SPIA from CR 0 hour. Supplemental Table 5. SPIA. Supplemental Table 6. Regulation of AUMD RNA degrading FUNDING proteins after damage. Supplemental Table 7. Analysis from PCA with loadings. Dr. Späth was supported by the Koeln Fortune Program/Faculty of Medi- cine, University of Cologne. Dr. Müller was supported by the Nachwuchs- gruppen.NRW program of the Ministry of Science North Rhine Westfalia. REFERENCES Furthermore, this study received funding from the German Research Foun- dation (MU3629/2-1 to Dr. Müller) and the German Federal Ministry of Education and Research (FKZ0315893A Systems Biology of Ageing Cologne 1. Rewa O, Bagshaw SM: Acute kidney injury-epidemiology, outcomes – Initiative). Dr. Müller, Dr. Benzing, and Dr. Schermer received additional and economics. Nat Rev Nephrol 10: 193 207, 2014 support from the German Research Foundation (grants BE2212 to Dr. Benzing, 2. Lempiäinen J, Finckenberg P, Mervaala EE, Sankari S, Levijoki J, SCHE1562/6 to Dr. Schermer, and KFO329 to Dr. Benzing, Dr. Schermer, Mervaala EM: Caloric restriction ameliorates kidney ischaemia/re- a and Dr. Müller). Dr. Beyer reports grants from German Federal Ministry of perfusion injury through PGC-1 -eNOS pathway and enhanced auto- – Education and Research, during the conduct of the study. phagy. Acta Physiol (Oxf) 208: 410 421, 2013 3. Vesnina ZV, Lishmanov YB, Alexandrova EA, Nesterov EA: Evaluation of nephroprotective efficacy of hypoxic preconditioning in patients undergoing coronary artery bypass surgery. Cardiorenal Med 6: SUPPLEMENTAL MATERIAL 328–336, 2016 4. Kapitsinou PP, Haase VH: Molecular mechanisms of ischemic pre- This article contains the following supplemental material online at conditioning in the kidney. Am J Physiol Renal Physiol 309: F821–F834, http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2019050534/-/ 2015 5. Fontana L, Meyer TE, Klein S, Holloszy JO: Long-term calorie restriction DCSupplemental. is highly effective in reducing the risk for atherosclerosis in humans. Supplemental Appendix 1. Supplemental methods. Proc Natl Acad Sci U S A 101: 6659–6663, 2004 Supplemental Figure 1. Raw count of mapped reads for total RNA 6. Vigne P, Frelin C: Plasticity of the responses to chronic hypoxia and samples. dietary restriction in an aged organism: Evidence from the Drosophila – Supplemental Figure 2. Experimental setup and baseline charac- model. Exp Gerontol 42: 1162 1166, 2007 7. Ristow M, Schmeisser K: Mitohormesis: Promoting health and lifespan teristics of animals. by increased levels of reactive oxygen species (ROS). Dose Response Supplemental Figure 3. Damage after IRI is strongly ameliorated 12: 288–341, 2014 in preconditioned animals. 8. Murry CE, Jennings RB, Reimer KA: Preconditioning with ischemia: A Supplemental Figure 4. Pseudotime analysis. delay of lethal cell injury in ischemic myocardium. Circulation 74: – Supplemental Figure 5. Cell-specific changes in preconditioned 1124 1136, 1986 9. Morris KC, Lin HW, Thompson JW, Perez-Pinzon MA: Pathways for is- and nonpreconditioned animals before and after IRI. chemic cytoprotection: Role of sirtuins in caloric restriction, resveratrol, Supplemental Figure 6. Quantitative PCR validation for RNA-seq and ischemic preconditioning. JCerebBloodFlowMetab31: 1003–1019, results of selected genes that were commonly regulated in response to 2011 CR as well as HP. 10. Fontana L, Partridge L: Promoting health and longevity through diet: – Supplemental Figure 7. Fifty one genes from overlapping KEGG From model organisms to humans. Cell 161: 106 118, 2015 11. Müller RU, Fabretti F, Zank S, Burst V, Benzing T, Schermer B: The von pathways. Hippel Lindau tumor suppressor limits longevity. J Am Soc Nephrol 20: Supplemental Figure 8. Genes in peroxisome pathway regulated in 2513–2517, 2009 response to CR and HP. 12. Johnsen M, Späth MR, Denzel MS, Göbel H, Kubacki T, Hoyer KJ, Supplemental Figure 9. Genes in glutathione pathway regulated in et al.: Oral supplementation of glucosamine fails to alleviate acute response to CR and HP. kidney injury in renal ischemia-reperfusion damage. PLoS One 11: e0161315, 2016 Supplemental Figure 10. Differential regulation of miRNAs and 13. Mitchell JR, Verweij M, Brand K, van de Ven M, Goemaere N, van den fi biological signi cance of regulation in response to CR and 24 hours Engel S, et al.: Short-term dietary restriction and fasting precondition after IRI in nonpreconditioned animals. against ischemia reperfusion injury in mice. Aging Cell 9: 40–53, 2010 Supplemental Figure 11. IRI-induced gene expression patterns 14. Tirapelli LF, Barione DF, Trazzi BF, Tirapelli DP, Novas PC, Silva CS, and pathways at 4 and 24 hours. et al.: Comparison of two models for evaluation histopathology of ex- perimental renal ischemia. Transplant Proc 41: 4083–4087, 2009 Supplemental Figure 12. Interactome overview of mitochondrial 15. Goujon JM, Hauet T, Menet E, Levillain P, Babin P, Carretier M: His- processes (by mitoXplorer). tological evaluation of proximal tubule cell injury in isolated perfused Supplemental Figure 13. Heatmap for outcome correlated genes. pig kidneys exposed to cold ischemia. JSurgRes82: 228– 233, 1999 Supplemental Figure 14. Univariate linear regression analysis of 16. Boucas J, Fritz C, Schmitt A, Riabinska A, Thelen L, Peifer M, et al.: fi outcome correlated genes. Label-free protein-RNA interactome analysis identi es khsrp signaling downstream of the p38/Mk2 kinase complex as a critical modulator of Supplemental Figure 15. Spearman blots for outcome regulated cell cycle progression. PLoS One 10: e0125745, 2015 genes that were enriched after CR as well as HP. 17. Awazawa M, Gabel P, Tsaousidou E, Nolte H, Krüger M, Schmitz J, Supplemental Table 1. Differentially regulated genes. et al.: A microRNA screen reveals that elevated hepatic ectodysplasin A

728 JASN JASN 31: 716–730, 2020 www.jasn.org BASIC RESEARCH

expression contributes to obesity-induced insulin resistance in skeletal 39. Katsyuba E, Mottis A, Zietak M, De Franco F, van der Velpen V, Gariani 1 muscle. Nat Med 23: 1466–1473, 2017 K, et al.: De novo NAD synthesis enhances mitochondrial function and 18. Love MI, Huber W, Anders S: Moderated estimation of fold change and improves health. Nature 563: 354–359, 2018 dispersion for RNA-seq data with DESeq2. Genome Biol 15: 550, 2014 40. Lismont C, Nordgren M, Van Veldhoven PP, Fransen M: Redox in- 19. Anders S, Huber W: Differential expression analysis for sequence count terplay between mitochondria and peroxisomes. Front Cell Dev Biol 3: data. Genome Biol 11: R106, 2010 35, 2015 20. Ogata H, Goto S, Sato K, Fujibuchi W, Bono H, Kanehisa M: KEGG: 41. Zhao J, Li L, Pei Z, Li C, Wei H, Zhang B, et al.: Peroxisome proliferator Kyoto encyclopedia of genes and genomes. Nucleic Acids Res 27: activated receptor (PPAR)-g co-activator 1-a and hypoxia induced 29–34, 1999 factor-1a mediate neuro- and vascular protection by hypoxic pre- 21. Alexa A, Rahnenfuhrer J: topGO: Enrichment analysis for gene ontol- conditioning in vitro. Brain Res 1447: 1–8, 2012 ogy. 2010 42. Young JM, Nelson JW, Cheng J, Zhang W, Mader S, Davis CM, et al.: 22. Yu G, Wang LG, Han Y, He QY: clusterProfiler: An R package for com- Peroxisomal biogenesis in ischemic brain. Antioxid Redox Signal 22: paring biological themes among gene clusters. OMICS 16: 284–287, 109–120, 2015 2012 43. Silachev DN, Plotnikov EY, Pevzner IB, Zorova LD, Babenko VA, Zorov 23. Tarca AL, Draghici S, Khatri P, Hassan SS, Mittal P, Kim JS, et al.: A novel SD, et al.: The mitochondrion as a key regulator of ischaemic tolerance signaling pathway impact analysis. Bioinformatics 25: 75–82, 2009 and injury. Heart Lung Circ 23: 897–904, 2014 24. Yim A, Koti P, Bonnard A, Marchiano F, Dürrbaum M, Garcia-Perez C, 44. Kumar S, Liu J, McMahon AP: Defining the acute kidney injury and et al.: mitoXplorer, a visual data mining platform to systematically an- repair transcriptome. Semin Nephrol 34: 404–417, 2014 alyze and visualize mitochondrial expression dynamics and mutations. 45. Correa-Costa M, Azevedo H, Amano MT, Gonçalves GM, Hyane MI, Nucleic Acids Res 48: 605–632, 2020 Cenedeze MA, et al.: Transcriptome analysis of renal ischemia/re- 25. Clark JZ, Chen L, Chou CL, Jung HJ, Lee JW, Knepper MA: Repre- perfusion injury and its modulation by ischemic pre-conditioning or sentation and relative abundance of cell-type selective markers in hemin treatment. PLoS One 7: e49569, 2012 whole-kidney RNA-Seq data. Kidney Int 95: 787–796, 2019 46. Damman J, Bloks VW, Daha MR, van der Most PJ, Sanjabi B, van der 26. Trapnell C, Cacchiarelli D, Grimsby J, Pokharel P, Li S, Morse M, Vlies P, et al.: Hypoxia and complement-and-coagulation pathways et al.: The dynamics and regulators of cell fate decisions are revealed in the deceased organ donor as the major target for intervention to by pseudotemporal ordering of single cells. Nat Biotechnol 32: improve renal allograft outcome. Transplantation 99: 1293–1300, 381–386, 2014 2015 27. Soneson C, Love MI, Robinson MD: Differential analyses for RNA-seq: 47. Link W: Introduction to FOXO biology. Methods Mol Biol 1890: 1–9, 2019 Transcript-level estimates improve gene-level inferences. F1000 Res 4: 48. Tissenbaum HA: DAF-16: FOXO in the context of C. elegans. Curr Top 1521, 2015 Dev Biol 127: 1–21, 2018 28. Bray NL, Pimentel H, Melsted P, Pachter L: Near-optimal probabilistic 49. Cabrera-Ortega AA, Feinberg D, Liang Y, Rossa C Jr., Graves DT: The RNA-seq quantification. Nat Biotechnol 34: 525–527, 2016 role of forkhead box 1 (FOXO1) in the immune system: Dendritic cells, 29. Cippà PE, Sun B, Liu J, Chen L, Naesens M, McMahon AP: Transcrip- T cells, B cells, and hematopoietic stem cells. Crit Rev Immunol 37: tional trajectories of human kidney injury progression. JCI Insight 3: 1–13, 2017 123151, 2018 50. Li L, Huang L, Vergis AL, Ye H, Bajwa A, Narayan V, et al.: IL-17 pro- 30. Liu J, Kumar S, Dolzhenko E, Alvarado GF, Guo J, Lu C, et al.: Molecular duced by neutrophils regulates IFN-gamma-mediated neutrophil mi- characterization of the transition from acute to chronic kidney injury gration in mouse kidney ischemia-reperfusion injury. J Clin Invest 120: following ischemia/reperfusion. JCI Insight 2: 94716, 2017 331–342, 2010 31. Han WK, Bailly V, Abichandani R, Thadhani R, Bonventre JV: Kidney 51.LeeJW,BaeE,KwonSH,YuMY,ChaRH,LeeH,etal.:Transcrip- injury molecule-1 (KIM-1): A novel biomarker for human renal proximal tional modulation of the T helper 17/interleukin 17 axis ameliorates tubule injury. Kidney Int 62: 237–244, 2002 renal ischemia-reperfusion injury. Nephrol Dial Transplant 34: 32. Paragas N, Qiu A, Zhang Q, Samstein B, Deng SX, Schmidt-Ott KM, 1481–1498, 2019 et al.: The Ngal reporter mouse detects the response of the kidney to 52. Balato A, Scala E, Balato N, Caiazzo G, Di Caprio R, Monfrecola G, et al.: injury in real time. Nat Med 17: 216–222, 2011 Biologics that inhibit the Th17 pathway and related cytokines to treat 33. Eddy AA, Kim H, López-Guisa J, Oda T, Soloway PD: Interstitial fibrosis inflammatory disorders. Expert Opin Biol Ther 17: 1363–1374, 2017 in mice with overload proteinuria: Deficiency of TIMP-1 is not pro- 53. Dubash S, Bridgewood C, McGonagle D, Marzo-Ortega H: The advent tective. Kidney Int 58: 618–628, 2000 of IL-17A blockade in ankylosing spondylitis: Secukinumab, ixekizumab 34. von Roretz C, Di Marco S, Mazroui R, Gallouzi IE: Turnover of AU-rich- and beyond. Expert Rev Clin Immunol 15: 123–134, 2019 containing mRNAs during stress: A matter of survival. Wiley Interdiscip 54. Hoving JC, Wilson GJ, Brown GD: Signalling C-type lectin receptors, Rev RNA 2: 336–347, 2011 microbial recognition and immunity. Cell Microbiol 16: 185–194, 2014 35. Zhao Y, Ding C, Xue W, Ding X, Zheng J, Gao Y, et al.: Genome-wide 55. Fan Q, Tao R, Zhang H, Xie H, Lu L, Wang T, et al.: Dectin-1 contributes DNA methylation analysis in renal ischemia reperfusion injury. Gene to myocardial ischemia/reperfusion injury by regulating macrophage 610: 32–43, 2017 polarization and neutrophil infiltration. Circulation 139: 663–678, 2019 36. Lan R, Geng H, Singha PK, Saikumar P, Bottinger EP, Weinberg JM, 56. Shushimita S, van der Pol P, W F de Bruin R, N M Ijzermans J, van Kooten Venkatachalam MA: Mitochondrial pathology and glycolytic shift dur- C, Dor FJ: Mannan-binding lectin is involved in the protection against ing proximal tubule atrophy after ischemic AKI. JAmSocNephrol27: renal ischemia/reperfusion injury by dietary restriction. PLoS One 10: 3356–3367, 2016 e0137795, 2015 37. Martinez-Lopez N, Tarabra E, Toledo M, Garcia-Macia M, Sahu S, 57. Kingeter LM, Lin X: C-type lectin receptor-induced NF-kB activation in Coletto L, et al.: System-wide benefits of intermeal fasting by auto- innate immune and inflammatory responses. Cell Mol Immunol 9: phagy. Cell Metab 26: 856–871.e5, 2017 105–112, 2012 38. Jankauskas SS, Pevzner IB, Andrianova NV, Zorova LD, Popkov VA, 58. Markó L, Vigolo E, Hinze C, Park J-K, Roël G, Balogh A, et al.: Tubular Silachev DN, et al.: The age-associated loss of ischemic precondition- epithelial NF-kB activity regulates ischemic AKI. J Am Soc Nephrol 27: ing in the kidney is accompanied by mitochondrial dysfunction, in- 2658–2669, 2016 1 creased protein acetylation and decreased autophagy. Sci Rep 7: 59. Katsyuba E, Auwerx J: Modulating NAD metabolism, from bench to 44430, 2017 bedside. EMBO J 36: 2670–2683, 2017

JASN 31: 716–730, 2020 Gene Expression Signature of Renal Organoprotection 729 BASIC RESEARCH www.jasn.org

60. Shi H, Enriquez A, Rapadas M, Martin EMMA, Wang R, Moreau J, et al.: whereas iron-free Tf confers protection against brain ischemia by NAD deficiency, congenital malformations, and niacin supplementa- modifying blood Tf saturation and subsequent neuronal damage. Re- tion. NEnglJMed377: 544–552, 2017 dox Biol 15: 143–158, 2018 61. Cantó C, Menzies KJ, Auwerx J: NAD(1) metabolism and the control of 75. de Vries B, Walter SJ, von Bonsdorff L, Wolfs TG, van Heurn LW, energy homeostasis: A balancing act between mitochondria and the Parkkinen J, et al.: Reduction of circulating redox-active iron by apo- nucleus. Cell Metab 22: 31–53, 2015 transferrin protects against renal ischemia-reperfusion injury. Trans- 62. Guan Y, Wang SR, Huang XZ, Xie QH, Xu YY, Shang D, et al.: Nicotin- plantation 77: 669–675, 2004 1 amide mononucleotide, an NAD precursor, rescues age-associated 76. Li Y, Zhou Y, Zhang D, Wu WY, Kang X, Wu Q, et al.: Hypobaric susceptibility to AKI in a sirtuin 1-dependent manner. JAmSoc hypoxia regulates iron metabolism in rats. J Cell Biochem 120: Nephrol 28: 2337–2352, 2017 14076–14087, 2019 63. Tran MT, Zsengeller ZK, Berg AH, Khankin EV, Bhasin MK, Kim W, et al.: 77. Hato T, Winfree S, Kalakeche R, Dube S, Kumar R, Yoshimoto M, et al.: PGC1a drives NAD biosynthesis linking oxidative metabolism to renal The macrophage mediates the renoprotective effects of endotoxin protection. Nature 531: 528–532, 2016 preconditioning. J Am Soc Nephrol 26: 1347–1362, 2015 64. Poyan Mehr A, Tran MT, Ralto KM, Leaf DE, Washco V, Messmer J, 78. Veldurthy V, Wei R, Campbell M, Lupicki K, Dhawan P, Christakos S: 25- 1 et al.: De novo NAD biosynthetic impairment in acute kidney injury in Hydroxyvitamin D3 24-hydroxylase: A key regulator of 1,25(OH)2D3 humans. Nat Med 24: 1351–1359, 2018 catabolism and calcium homeostasis. Vitam Horm 100: 137–150, 2016 65. Olenchock BA, Moslehi J, Baik AH, Davidson SM, Williams J, Gibson 79. Kwon O, Phillips CL, Molitoris BA: Ischemia induces alterations in actin WJ, et al.: EGLN1 inhibition and rerouting of a-Ketoglutarate suffice for filaments in renal vascular smooth muscle cells. Am J Physiol Renal remote ischemic protection. Cell 164: 884–895, 2016 Physiol 282: F1012–F1019, 2002 66. Hirt L, Ternon B, Price M, Mastour N, Brunet JF, Badaut J: Protective 80. Meier U, Gressner O, Lammert F, Gressner AM: Gc-globulin: Roles in role of early aquaporin 4 induction against postischemic edema for- response to injury. Clin Chem 52: 1247–1253, 2006 mation. J Cereb Blood Flow Metab 29: 423–433, 2009 81. Bikle DD, Schwartz J: Vitamin D binding protein, total and free vitamin 67. Monai H, Wang X, Yahagi K, Lou N, Mestre H, Xu Q, et al.: Adrenergic D levels in different physiological and pathophysiological conditions. receptor antagonism induces neuroprotection and facilitates recovery Front Endocrinol (Lausanne) 10: 317, 2019 from acute ischemic stroke. Proc Natl Acad Sci U S A 116: 82. Heydel JM, Garnier P, Faure P, Artur Y: Ciprofibrate regulation of rat 11010–11019, 2019 hepatic bilirubin glucuronidation and UDP-glucuronosyltransferases 68. Aniya Y, Imaizumi N: Mitochondrial glutathione transferases involving a expression. Eur J Drug Metab Pharmacokinet 37: 233–240, 2012 new function for membrane permeability transition pore regulation. 83. Wen XY, Tarailo-Graovac M, Brand-Arzamendi K, Willems A, Rakic B, Drug Metab Rev 43: 292–299, 2011 Huijben K, et al.: Sialic acid catabolism by N-acetylneuraminate 69. Taskinen MR, Packard CJ, Borén J: Emerging evidence that ApoC-III pyruvate lyase is essential for muscle function. JCI Insight 3: inhibitors provide novel options to reduce the residual CVD. Curr 122373, 2018 Atheroscler Rep 21: 27, 2019 84. Zachara NE, Hart GW: Cell signaling, the essential role of O-GlcNAc! 70. Kim HY, Kim DK, Bae SH, Gwak H, Jeon JH, Kim JK, et al.: Farnesyl Biochim Biophys Acta 1761: 599–617, 2006 diphosphate synthase is important for the maintenance of glioblastoma 85. Jensen RV, Zachara NE, Nielsen PH, Kimose HH, Kristiansen SB, Bøtker stemness. Exp Mol Med 50: 137, 2018 HE: Impact of O-GlcNAc on cardioprotection by remote ischaemic 71. Bertaggia E, Scabia G, Dalise S, Lo Verso F, Santini F, Vitti P, et al.: preconditioning in non-diabetic and diabetic patients. Cardiovasc Res Haptoglobin is required to prevent oxidative stress and muscle atro- 97: 369–378, 2013 phy. PLoS One 9: e100745, 2014 86. Gu JH, Shi J, Dai CL, Ge JB, Zhao Y, Chen Y, et al.: O-GlcNAcylation 72. Belcher JD, Chen C, Nguyen J, Abdulla F, Zhang P, Nguyen H, et al.: reduces ischemia-reperfusion-induced brain injury. Sci Rep 7: Haptoglobin and hemopexin inhibit vaso-occlusion and inflammation 10686, 2017 in murine sickle cell disease: Role of heme oxygenase-1 induction. PLoS 87. Hu J, Wang Y, Zhao S, Chen J, Jin S, Jia P, et al.: Remote ischemic One 13: e0196455, 2018 preconditioning ameliorates acute kidney injury due to contrast ex- 73. Baek JH, D’Agnillo F, Vallelian F, Pereira CP, Williams MC, Jia Y, et al.: posure in rats through augmented O-GlcNAcylation. Oxid Med Cell Hemoglobin-driven pathophysiology is an in vivo consequence of the Longev 2018: 4895913, 2018 red blood cell storage lesion that can be attenuated in guinea pigs by 88. Späth MR, Bartram MP, Palacio-Escat N, Hoyer KJR, Debes C, Demir F, haptoglobin therapy. JClinInvest122: 1444–1458, 2012 et al.: The proteome microenvironment determines the protective ef- 74. DeGregorio-Rocasolano N, Martí-Sistac O, Ponce J, Castelló-Ruiz M, fect of preconditioning in cisplatin-induced acute kidney injury. Kidney Millán M, Guirao V, et al.: Iron-loaded transferrin (Tf) is detrimental Int 95: 333–349, 2019

AFFILIATIONS

1Department II of Internal Medicine and Center for Molecular Medicine, University of Cologne, Faculty of Medicine and University Hospital Cologne, Cologne, Germany; 2Max Planck Institute of Biochemistry, Martinsried, Germany; 3Institute for Pathology, Diagnostic and Experimental Nephropathology Unit, 4Cologne Excellence Cluster on Cellular Stress Responses in Aging-Associated Diseases, 7Cologne Center for Genomics, and 8Systems Biology of Ageing Cologne, University of Cologne, Cologne, Germany; 5Institute for Global Food Security, Queen’s University Belfast, Belfast, Northern Ireland, United Kingdom; and 6Development Biology Institute of Marseille, Aix- Marseille University, CNRS, Marseille, France

730 JASN JASN 31: 716–730, 2020 Supplementary material

Title: The integrated signature of renal preconditioning against ischemia reperfusion injury

Running title: Gene expression signature of renal organoprotection

Authors: Marc Johnsen, Torsten Kubacki, Assa Yeroslaviz, Martin Späth, Jannis Mörsdorf, Heike Göbel, Katrin Bohl, Michael Ignarski, Bianca Habermann, Janine Altmüller, Andreas Beyer, Thomas Benzing, Bernhard Schermer, Volker Burst and Roman-Ulrich Müller

SUPPLEMENTARY MATERIAL ...... 1

SUPPLEMENTARY METHODS ...... 3 Total RNA and miRNA mapping, trimming and adapter removal ...... 3 Outcome correlation score ...... 3 Primer IDs for qPCR assays ...... 4 Recovery score after IRI ...... 5 SUPPLEMENTARY FIGURE LEGENDS ...... 6 Supplementary Figure 1 Raw count of mapped reads for total RNA samples ...... 6 Supplementary Figure 2 Experimental setup and baseline characteristics of animals ...... 6 Supplementary Figure 3 Damage after IRI is strongly ameliorated in preconditioned animals ...... 7 Supplementary Figure 4 Pseudotimeanalysis ...... 7 Supplementary Figure 5: Cell specific changes in precondtioned and non-preconditioned animals before and after IRI ...... 7 Supplementary Figure 6 qPCR validation for RNAseq results of selected genes that were commonly regulated in response to CR as well as HP ...... 7 Supplementary Figure 7 51 genes from overlapping KEGG Pathways ...... 8 Supplementary Figure 8 Genes in Peroxisome pathway regulated in response to CR and HP ..... 8 Supplementary Figure 9 Genes in Glutathione pathway regulated in response to CR and HP ..... 8 Supplementary Figure 10 Differential regulation of miRNAs and biological significance of regulation in response to CR and 24 hours after IRI in non-preconditioned ...... 8 Supplementary Figure 11 IRI-induced gene expression patterns and pathways at 4h and 24h .... 9 Supplementary Figure 12 Interactome overview of mitochondrial processes (by mitoXplorer) ...... 9 Supplementary Figure 13 Heatmap for outcome correlated genes ...... 10 Supplementary Figure 14 Univariate linear regression analysis of outcome correlated genes .... 10

1 Supplementary Figure 15A-P Spearman blots for outcome regulated genes that were enriched after CR as well as HP ...... 10 SUPPLEMENTARY TABLE LEGENDS ...... 10 Supplementary Table 1 Differentially regulated genes ...... 10 Supplementary Table 2 GO term analysis ...... 10 Supplementary Table 3 KEGG pathway analysis ...... 10 Supplementary Table 4 Gene lists of SPIA from CR0h ...... 11 Supplementary Table 5 SPIA ...... 11 Supplementary Table 6 Regulation of AUMD RNA degrading proteins after damage ...... 11 Supplementary Table 7 Analysis from PCA with loadings ...... 11

REFERENCES FOR SUPPLEMENTARY MATERIAL ...... 12

SUPPLEMENTARY FIGURES ...... 13

2 Supplementary methods

Total RNA and miRNA mapping, trimming and adapter removal Raw mRNA data quality was analyzed using the fastqc software (0.10.1). mRNA trimming and adapter removal were done with the fastx-Toolkit (0.0.14) and cutadapt (1.5) 1 before mapping to the mouse genome from Ensembl (GRCm38) using STAR aligner (2.4.2a) 2. Mapped reads were run through featureCount (1.4.6-p4) 3. miRNA adapter removal was done using cutadapt 1, mapping and quantification using the mirDeep2 software package 4. Products were evaluated for suitability on a thermodynamic scale with bowtie aligner 5 and their structure was using Randfold 6 to classify as possible miRNAs.

Outcome correlation score Mean of serum urea on day 1 and 3 after IRI and the weight from day 2 post IRI divided by the weight at baseline before IRI were rated from 1 to 8 and summed up to a maximum score of 16.

mg/dl score % baseline weight score

<50 1 105-110 1

50-99 2 104-100 2 100-149 3 95-99 3 150-199 4 90-94 4 200-249 5 85-89 5 250-299 6 80-84 6 300-349 7 75-79 7

349 / deceased 8 70-74 8

3 Primer IDs for qPCR assays For qPCR Applied Biosystems Customized TaqMan Arrays were utilized.

Acsm3-Mm00489774_m1 Target Ubc-Mm01201237_m1 Control Slc22a7-Mm00460672_m1 Target Tmprss9-Mm01177636_m1 Target Tspan13-Mm00481226_m1 Target Reln-Mm00465200_m1 Target Cyp7b1-Mm00484157_m1 Target Ace-Mm00802048_m1 Target Folh1-Mm00489655_m1 Target Ctxn3-Mm01718023_m1 Target 18S-Hs99999901_s1 Control Cndp2-Mm01261959_mH Target Gusb-Mm01197698_m1 Control Pde6a-Mm00476664_m1 Target Mpv17l-Mm00473833_m1 Target Hsd17b11-Mm00504410_m1 Target Als2cr11-Mm00613608_m1 Target Slc16a13-Mm00655437_m1 Target Nudt19-Mm00473613_m1 Target Hmox1-Mm00516005_m1 Target Aqp4-Mm00802131_m1 Target Ldhd-Mm00459138_g1 Target Pdk3-Mm00455220_m1 Target Ces2c-Mm01250994_m1 Target 9030619P08Rik-Mm01205874_m1 Target Mogat1-Mm00503358_m1 Target Apoe-Mm01307193_g1 Target Amacr-Mm00507717_m1 Target Slc18a1-Mm00461868_m1 Target Acadm-Mm01323360_g1 Target Entpd4-Mm00491888_m1 Target Acy3-Mm00503587_g1 Target Hao2-Mm00469507_m1 Target

4 Cyp4a12a-Mm00514494_m1 Target Kif20b-Mm01205010_m1 Target Kynu-Mm00551012_m1 Target Slc7a12-Mm00499866_m1 Target Prlr-Mm00599957_m1 Target Alas1-Mm01235914_m1 Target Gapdh-Mm99999915_g1 Control Tnfrsf21-Mm00446361_m1 Target Wnt5b-Mm01183986_m1 Target Hbb-b1;Beta-s;Hbb-b2-Mm03646870_gH Target Alas2-Mm00802083_m1 Target Grem2-Mm00501909_m1 Target Hdc-Mm00456104_m1 Target Polr2a-Mm00839493_m1 Control Il34-Mm01243248_m1 Target

Recovery score after IRI

The Score was put together based on the “Belastungskataloge zur Bewertung von Tierversuchen” of the „Forum Tierversuche in der Forschung“. http://www.dfg.de/download/pdf/dfg_magazin/forschungspolitik/tierschutz2015/dialogf orum_tierversuche/belastungskataloge_zur_bewertung_von_tierversuchen.pdf

Weight loss compared to baseline (day 0) before surgery 0 points No weight loss 1 point < 10% 2 points 10-20% 3 points > 20%

Activity nest building, flight reflex, fur cleansing, movement in cage 0 points normal 1 point slightly reduced 2 points severly reduced 3 points no activity

5

Appearance 0 points normal 1 point shaggy fur 2 points + nose bump, elevated respiratory rate, signs of slight dehydration 3 points + discharge from eyes and / or nose, standing skin folds

Total Score ≥ 7 points sacrifice of animal in anaesthesia 5-7 points close observation and consultation with experimental supervisor 0-4 points continue experiment according to protocol

Supplementary Figure / Table Legends

Supplementary Figure 1 Raw count of mapped reads for total RNA samples Raw count of mapped reads with gene rank on x axis and raw counts on y axis for non- preconditioned animals (CTRL) at baseline – 0 hours (A), 4 (B) and 24 hours (C) after IRI, caloric restricted animals (CR) at baseline (D) and 24 hours after IRI (E), as well as hypoxic preconditioned animals (HP) at baseline (F) and 24 hours after IRI (G).

Supplementary Figure 2 Experimental setup and baseline characteristics of animals Baseline creatinine (A) and urea (B) values for non-preconditioned animals and animals preconditioned with calorie restriction (CR). Weight curves for animals from A+B (C). Recovery score 24h and 72h after IR (n=9 in each group) D). Lower scores correspond to better recovery. (p=0.0023 at 24 h after IRI, p<0.0001 at 72 h after IRI) Caloric restriction, CR; hypoxic preconditioning, HP; ischemic reperfusion injury, IRI. N.S. not significant, ** p<0.01, **** p<0.0001

6 Supplementary Figure 3 Damage after IRI is strongly ameliorated in preconditioned animals Representative PAS stainings from post ischemic kidneys 4h and 72h after IRI from non-preconditioned controls, HP and CR animals. * tubular flattening, ➤ nuclear loss, ⬈ tubular casts, # denuded tubuli with luminar debris. TUNEL assay from controls, HP and CR group 24h after IRI showing cell death in green (D/H/L). Caloric restriction, CR; hypoxic preconditioning, HP; ischemic reperfusion injury, IRI.

Supplementary Figure 4 Pseudotimeanalysis Pseudotime analysis including all samples (CR, HP, non-PC) before and after IRI. Sample ordering (“cell ordering” in the original Monocle workflow) in a reduced dimensional space as determined by the Monocle algorithm CR; hypoxic preconditioning, HP; non-preconditioned animals, non-PC; ischemic reperfusion injury, IRI

Supplementary Figure 5: Cell specific changes in precondtioned and non- preconditioned animals before and after IRI Shown are the cell specific changes of different kidney and immune cells according to the expression of their marker genes calculated as Z-score. Z-score is showing the number of standard deviations from the mean expression of each cell specific marker set across all conditions. CR; hypoxic preconditioning, HP; non-preconditioned animals, non-PC; ischemic reperfusion injury, IRI

Supplementary Figure 6 qPCR validation for RNAseq results of selected genes that were commonly regulated in response to CR as well as HP Validation of RNAseq expression changes with qPCR TaqMan arrays of 44 genes that were commonly regulated in RNAseq in response to preconditioning with CR as well as HP. Log2FC expression changes from RNAseq (dark blue CR / red HP) and qPCR (light blue CR / orange HP). Genes are sorted according to Log2FC in RNAseq for CR animals starting with the most downregulated gene in the upper panel to the most upregulated gene in the lower panel to the right.

7 Supplementary Figure 7 51 genes from overlapping KEGG Pathways KEGG pathway analysis was performed separately in HP and CR animals before IRI. Shown are the genes contributing to each of the 3 overlapping KEGG pathways for CR and HP before IRI (padj <0.05, FDR<0.05).

Supplementary Figure 8 Genes in Peroxisome pathway regulated in response to CR and HP Depiction of Peroxisome KEGG pathway with significantly regulated genes in response to CR (A) and HP (B).

Supplementary Figure 9 Genes in Glutathione pathway regulated in response to CR and HP Depiction of Glutathione KEGG pathway with significantly regulated genes in response to CR (A) and HP (B).

Supplementary Figure 10 Differential regulation of miRNAs and biological significance of regulation in response to CR and 24 hours after IRI in non- preconditioned Regulation of miRNAs in CR animals compared to non-preconditioned animals at baseline (A) and 24 hours after IRI (B). Upregulated miRNAs depicted as red bars, downregulated miRNAs as green bars. Percentage of downregulated target mRNAs for significantly up- and downregulated miRNAs at baseline (C) and 24 hours after IRI (D) compared to the whole gene set (black bar). Targets that show a significantly different percentage of downregulation compared to the whole gene set (black bar chart) are marked with an asterix (*). Mean log2foldchange for the whole gene set (black bar) and for targets of significantly regulated miRNAs at baseline (E) and 24 hours after IRI (F). Significantly regulated miRNAs in response to CR or IRI Since posttranscriptional regulators are expected to add another layer of complexity to the regulation of gene networks, we performed small RNA sequencing in addition to standard RNAseq. Four microRNAs were significantly regulated in response to CR. No microRNAs were significantly regulated in response to HP. 24 hours after IRI, 91 microRNAs were differentially regulated in non-preconditioned animals. For a

8 complete list of regulated microRNAs see Table 5. All microRNAs regulated in response to CR were also regulated 24 hours after IRI (A+B). mir-802-5p was upregulated and mir-22-3p downregulated in both situations. To evaluate their functional relevance, we compared the differences between all significantly regulated mRNAs with the significantly regulated predicted miRNA targets in the same condition. If functionally relevant, inverse regulation of a miRNA compared to its targets would be expected as miRNAs are thought to silence their targets. From the 4 miRNAs regulated by CR, such an association was revealed only for miR-22-3p (miR-22-3p downregulated, targets upregulated, mean log2FC (.007) higher than for the regulated mRNAs in the whole sample (p=.007)). Consistently, only 2 downregulated miRNAs showed a pattern of inversely regulated targets 24 hours after IRI: miR-29-3p and miR- 182-5p, which both had a higher percentage of upregulated targets and also a higher mean log2FC than regulated genes in the whole sample (C-F).

Supplementary Figure 11 IRI-induced gene expression patterns and pathways at 4h and 24h Overlap of differentially regulated genes in non-preconditioned animals 4h and 24h after IRI (A) (p<0.001). Top ten up- and downregulated genes by IRI after 4h and 24h (B). SPIA analysis showing the Top 20 signalling pathways after 4h and 24h after IRI in non-PC animals (C) (pGfdr <0.05), y-axis: pathway terms, x-axis: observed total preturbation accumulation in the pathway (tA). Heatmap of SPIA pathways of preconditioned animals before damage and non-preconditioned animals 4h and 24 hours after IRI (D). CR; hypoxic preconditioning, HP; non-preconditioned animals, non- PC; ischemic reperfusion injury, IRI

Supplementary Figure 12 Interactome overview of mitochondrial processes (by mitoXplorer) Interactome view is the visualisation method for all mitochondrial processes of a single dataset. Shown are CR 0h vs non-PC 0h (A), HP 0h vs non-PC 0h (B), non-PC-IR4h vs non-PC 0h (C) and non-PC-IR24h vs non-PC 0h (D). Red - down-regulated, Blue - up-regulated, size of the ball represents higher/lower log2FC (p-value is not considered at this visualisation).

9 Supplementary Figure 13 Heatmap for outcome correlated genes Heatmap of 34 genes whose regulation at baseline was highly correlated with individual animal’s outcome after IRI (r Spearman >0.95 and <-0.95). Expression of 30 genes is positively correlated and of 4 genes is negatively regulated with outcome.

Supplementary Figure 14 Univariate linear regression analysis of outcome correlated genes Univariate linear regression analyses for the 16 genes, that were highly correlated with the damage score and that were also identified in the overlap between CR and HP. Regression coefficient B is provided as the change of damage score per change of transcripts by 100 copies.

Supplementary Figure 15A-P Spearman blots for outcome regulated genes that were enriched after CR as well as HP

Raw read count for genes with rS > 0.95 and rs < -0.95 blotted against outcome score.

Supplementary Table Legends

Supplementary Table 1 Differentially regulated genes Shown are the differentially regulated genes of CR, HP and non-preconditionend animals at different timepoints (see legend in the table)

Supplementary Table 2 GO term analysis Shown are GO BP, CC and MF terms of CR, HP and non-preconditioned animals at different timepoints (see legend in the table).

Supplementary Table 3 KEGG pathway analysis Shown are the KEGG pathway terms of CR, HP and non-preconditioned animals at different timepoints (see legend in the table).

10 Supplementary Table 4 Gene lists of SPIA from CR0h Shown are the genes contributing to the 7 significant pathways from the Signaling pathway impact analysis (SPIA) in CR0h animals (see legend in the table). Supplementary Table 5 SPIA Signalling pathway impact analysis of CR0h, HP0h and non-preconditioned animals 4h and 24h after ischemia-reperfusion injury (see legend in the table).

Supplementary Table 6 Regulation of AUMD RNA degrading proteins after damage Selected RNA stabilizing and destabilizing adenine/uridine-rich element binding proteins (AUBPs) and their differential expression (Log2FC) in response to IRI induced damage (adapted from von Roretz et al. 2011) (see legend in the table).

Supplementary Table 7 Analysis from PCA with loadings Shown are the genes from the 3 gene sets (upper, right and lower outliers) from the PCA analysis (->Fig.5) and the GO term analysis (BP, CC, MF) as well as KEGG pathway analysis (see legend in the table).

11 References for Supplementary Material

1. Martin M: Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet.journal 17: pp. 10–12, 2011

2. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, Batut P, Chaisson M, Gingeras TR: STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29: 15–21, 2013

3. Liao Y, Smyth GK, Shi W: featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics 30: 923–930, 2014

4. Friedländer MR, Mackowiak SD, Li N, Chen W, Rajewsky N: miRDeep2 accurately identifies known and hundreds of novel microRNA genes in seven animal clades. Nucleic Acids Research 40: 37–52, 2012

5. Langmead B, Trapnell C, Pop M, Salzberg SL: Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biology 10: R25, 2009

6. Bonnet E, Wuyts J, Rouzé P, Van de Peer Y: Evidence that microRNA precursors, unlike other non-coding RNAs, have lower folding free energies than random sequences. Bioinformatics 20: 2911–2917, 2004

12 A B C

D E

F G A B

0.3 150 N.S. ****

0.2 100

0.1 urea mg/dl 50 creatinine mg/dl

0.0 0 CR CR non-PC non-PC C D 30 10 ** ****

8 25 6 gram 4 20 recovery score 2

15 0

HP CR HP CR day -7 day -28 day -21 day -14 baseline non-PC non-PC non-PC CR 24 hours after IR 72 hours after IR 4 hours 72 hours

# * * *

* *

*

* * CR4 controls HP

* Pseudotime 0 10 20 30

nonPC-IR4h CR-IR4h CR-IR4h HP-IR4h HP-IR4h 10 HP-IR4h nonPC-IR4h nonPC-IR4h

5

nonPC-IR24h HP-IR24h nonPC-IR24h

Component 2 0 nonPC-IR24h CR0h HP-IR24h CR0h HP-IR24h 2 CR0h CR-IR24h 1 CR0h -5 CR-IR24h CR-IR4h CR-IR24h HP0h nonPC0h HP0h nonPC0h HP0h HP0h nonPC0hnonPC0h -20 -10 0 10 Component 1 Podocyte Mesangial Cells Macula Densa Granular cell of aerent arteriole Proximal Long Descending Limb Short Loop Descending Limb Thin Ascending Limb Thick Ascending Limb Connecting Tubule Distal Convoluted Tubule Collecting Duct Principal Cells Z-score Endothelial Cell Intercalated A Cells -1.2 Intercalated B Cells -0.8 Inner Medullary Collecting Duct -0.4 Transitional Epithelium 0.4 Fibroblast Interstitial Cells 0.8 Pericyte 1.2 Smooth Muscle Cell Erythrocyte Monocyte Macrophage Dendritic Cell Basophil Eoisinophil B-lymphocytes T-lymphocyte Mast Cell Plasma Cell Polymorphonuclear Leukocyte Megakaryocyte Neuronal Cell nonPC0h nonPC-IR4h nonPC-IR24h CR0h CR-IR4h CR-IR24h HP0h HP-IR4h HP-IR24h 0 log2 FC

-2

-4

Il34 Reln Ldhd Ace Ctxn3 Kif20b Cndp2 Gusb Pde6a Acsm3 Nudt19 Mpv17l Cyp7b1 Mogat1 Slc22a7 Tmprss9Tspan13**Hmox1 Cyp4a12a Hsd17b11 Slc16a13

9030619P08Rik CR4 RNAseq HP RNAseq CR4 qPCR HP qPCR

4

2 log2 FC

0

-2

Prlr Pdk3 Acy3 Alas1 Alas2 Folh1 **Hdc Aqp4 Hao2 Kynu Amacr Ces2c Acadm Entpd4 Wnt5b Hbb-b1 **Apoe Slc18a1 Slc7a12 ++Grem2 **Als2cr11 ++Tnfrsf21

CR4 RNAseq HP RNAseq CR4 qPCR HP qPCR

Metabolic Acadm, Acat1, Acox1, Acsm3, Agps, Aldh3a2,Amacr, Anpep, Bdh1, Bhmt, Chpt1, Pathways Cndp2, Cox15, Cox5a, Cyp2c23,Cyp2e1,Degs2, Dhrs4, Folh1, Galns, Galnt1, Gclc, Gda, Ggt1, Glb1, Gusb, Hao2, Hsd11b1, Kynu, Me1, Ndufa1, Ndufb8, Odc1, Pank3, Papss1, Ppt2, Tpk1, Treh Glutathione Abcd3, Acox1, Agps, Amacr, Ddo, Dhrs4, Hao2, Mpv17l Metabolism Nudt19, Phyh, Sod1, Sod2 Peroxisome Anpep, Gclc, Ggt1, Gpx3, Gsta1, Gsta2, Mgst1, Odc1

A Caloric restriction B Hypoxic preconditioning A Caloric restriction B Hypoxic preconditioning A 1.0 B 4

0.5 2

0.0 0 log2 FC miRNA -0.5 log2FC miRNA -2

-1.0 -4 miR-6238 miR-6538 miR-22-3p miR-3470b

C miR-802-5p miR-30a-5p D miR-21a-5p miR-383-5p miR-29a-3p miR-182-5p miR-29b-3p miR-6937-5p 100 100

80 80 * * * *

60 60

40 40

20 20 % of down regulated mRNAs % of down regulated mRNAs 0 0 mRNA mRNA miR-6238 miR-6538 miR-29-3p miR-22-3p miR-3470b miR-21a-5p miR-182-5p E miR-802-5p miR-30a-5p F miR-6937-5p miR-383-5p.2 miR-383-5p.1 0.050 0.3

0.2 0.025 0.1

0.000 0.0

-0.1 mean log2FC mean log2FC -0.025 -0.2 mRNAs

-0.050 -0.3 miR-6238 miR-6538 miR-29-3p miR-21a-5p miR-182-5p miR-6937-5p miR-383-5p.2 miR-383-5p.1 mRNAs miR-22-3p miR-3470b miR-802-5p miR-30a-5p A C

B

D

Fig S9 a caloric restriction (CR) b hypoxia (HP)

Prot. stability & degrad.

Transcription Transcription Transmembrane (nucl.) Prot. stability FE-S cluster (nucl.) transport Nucleotide & degrad. biosynth. metab. Metab. of FE-S cluster vitamins & TCA cycle biosynth. co-factors Pyruvate FA-metab. metab. Metab. of vitamins & co-factors mito-Carrier FA-biosynth. & ROS defense elongation Amino acid FA-biosynth. & Amino acid metabolism elongation FA-metab. metabolism Glycolysis Ca2+ signaling & ROS defense transport Apoptosis mito- Lipid & lipoprot. Signaling metab. Pyruvate Nucleotide metab. metab. Lipoic acid Transmembrane metab. transport Translation OXPHOS OXPHOS Translation Lipoic acid Apoptosis metab. Translation mito-Carrier FA degrad. & (mt) beta-oxid. Cardiolipin Import & biosynth. Ca2+ signaling & Lipid & lipoprot. sorting transport metab. mito- PP Cardiolipin Signaling Pathway biosynth. Heme Bile acid Replication & Glycolysis biosynth. mito-Dynamics Import & FA degrad. & mito-Dynamics Replication & sorting synth. Folate & Pterin transcription Ubiquinone UPRmt transcription beta-oxid. metab. biosynth. OXPHOS Fructose Translation Fructose (mt) metab. Nitrogen Nitrogen metab. PP Unk. Bile acid metab. (mt) Mitophagy UPRmt Unk. Pathway TCA cycle Mitophagy Heme OXPHOS synth. metab. Folate & Pterin biosynth. (mt) Ubiquinone metab. biosynth.

c 4hrs after IR vs non-preconditioned d 24hrs after IR vs non-preconditioned

Folate & Pterin Nitrogen FA-biosynth. & metab. elongation Lipid & lipoprot. metab. metab. Metab. of Mitophagy FA degrad. & TCA cycle vitamins & Transcription beta-oxid. co-factors (nucl.) Pyruvate metab. Folate & Pterin Ca2+ signaling & metab. transport Import & mito-Carrier sorting Amino acid FE-S cluster FE-S cluster biosynth. biosynth. metabolism mito-Carrier Translation Translation (mt) FA-biosynth. & elongation Prot. stability Bile acid Translation Glycolysis & degrad. Transcription Import & synth. (mt) (nucl.) sorting Lipoic acid metab. Ca2+ signaling & FA degrad. & transport FA-metab. beta-oxid. Translation OXPHOS Cardiolipin Amino acid OXPHOS Replication & metabolism Lipoic ROS defense biosynth. transcription acid Transmembrane metab. transport Heme Fructose Apoptosis biosynth. metab. mito-Dynamics PP Cardiolipin mito- Pathway Transmembrane biosynth. Signaling mito-Dynamics Glycolysis transport Lipid & lipoprot. ROS defense Ubiquinone Ubiquinone metab. Pyruvate UPRmt biosynth. Nitrogen Replication & biosynth. Apoptosis metab. metab. transcription Fructose TCA cycle Mitophagy Metab. of FA-metab. metab. Bile acid vitamins & OXPHOS Unk. OXPHOS PP synth. co-factors (mt) Prot. stability (mt) Pathway Heme & degrad. Unk. UPRmt Nucleotide mito- Nucleotide biosynth. metab. Signaling metab. Density 0 0.1 0.2 0.3 0.4 - 2 and DensityPlot - Row Z-Score Color Key 1 0 1 2

CR 3

CR 2 Outcome Corre

CR 1 l CR 4 ation Score

Control 3

HP 3

Control 1

Control 2

HP 1

HP 2 Ugt2b38 Cd36 Lgals1 Tmem237 Slc39a11 Nhp2 Dnah3 Pla2g2c Vwc2 Ankrd55 Kif2c Gm9115 Rtp3 Ifi206 Asns Hspb11 Gm15895 Cntnap5a Slc7a2 Tspan13 Ralgds Tsga10 Tc2n Slco1a4 Myo5a Cyp7b1 Ces2c Gm7278 Cmtm6 Akr1c14 Keg1 Odc1 Cndp2 Kap Intercept p B per 100 copies β p adjusted R2 Kap -5,212 0,065 0,004 0,910 <0,001 0,807 Cndp2 -1,760 0,181 0,046 0,962 <0,001 0,917 Odc1 0,062 0,965 0,047 0,936 <0,001 0,861 Cd36 -6,084 0,048 0,114 0,907 <0,001 0,801 Akr1c14 -4,216 0,101 0,215 0,911 <0,001 0,808 mtm6 -7,809 0,024 0,296 0,911 <0,001 0,809 Gm7278 -0,212 0,881 0,168 0,939 <0,001 0,866 Ces2c -7,646 0,021 0,229 0,919 <0,001 0,824 Cyp7b1 -4,094 0,029 0,288 0,956 <0,001 0,903 Myo5a -2,599 0,050 0,329 0,970 <0,001 0,934 Nhp2 -8,466 0,031 2,354 0,894 <0,001 0,775 Slc39a11 -13,998 0,003 3,325 0,927 <0,001 0,842 Tmem237 -4,408 0,056 2,100 0,933 <0,001 0,854 Slco1a42 -4,959 0,068 3,461 0,913 <0,001 0,813 Tc2n -3,106 0,155 3,079 0,920 <0,001 0,828 Tspan13 -3,189 0,034 2,596 0,967 <0,001 0,926

ENSMUSG00000033715

10000

Pear. r= 0.9371 C20 ● Spear. r = 0.9663

C24 ● C22 ● HP14 8000 ●

HP12 ● HP11 ●

6000 raw read counts raw CR4Wo4 ●

4000 CR4Wo1 CR4Wo2● ●

CR4Wo3 ●

4 8 12 16 total.score ENSMUSG00000002944

Pear. r= 0.9081 C24 ● Spear. r = 0.9663 20000

C20 ●

C22 HP14 ● 15000 ● HP12 HP11 ● ● raw read counts raw

CR4Wo4 ● CR4Wo1 ● 10000

CR4Wo2 ●

CR4Wo3 ●

4 8 12 16 total.score ENSMUSG00000061825

Pear. r= 0.945 C24 ● Spear. r = 0.9847

C20 ●

10000

C22 ●

8000

HP14 ●

raw read counts raw HP12 ●

HP11 CR4Wo1 ● ● CR4Wo4 6000 ●

CR4Wo2 CR4Wo3● ●

4000 4 8 12 16 total.score ENSMUSG00000032434

Pear. r= 0.9299 C24 ● Spear. r = 0.9724 C20 8000 ●

7000 C22 ●

HP14 HP11 ● ●

6000

HP12 ● raw read counts raw

5000 CR4Wo1 ● CR4Wo4 ●

4000 CR4Wo2 ● CR4Wo3 ●

4 8 12 16 total.score ENSMUSG00000024644

Pear. r= 0.9543 C24 ● Spear. r = 0.9663

C20 40000 ●

C22 ●

30000

HP14 ●

HP12 ● raw read counts raw HP11 ● 20000

CR4Wo4 ● CR4Wo1 ●

CR4Wo2 ● CR4Wo3 ● 10000

4 8 12 16 total.score ENSMUSG00000039519

Pear. r= 0.9315 C24 ● Spear. r = 0.954

C20 ●

HP14 C22 6000 ● ●

HP11 ● HP12 ●

raw read counts raw CR4Wo4 4000 ●

CR4Wo1 CR4Wo2● ●

2000 CR4Wo3 ●

4 8 12 16 total.score ENSMUSG00000080950

Pear. r= 0.9349 C24 ● Spear. r = 0.9724

C20 ●

9000

C22 ●

6000 HP14 HP11 ● ●

raw read counts raw HP12 ●

CR4Wo1 ● CR4Wo2 ● CR4Wo4 3000 ●

CR4Wo3 ●

4 8 12 16 total.score ENSMUSG00000032758

6e+05

Pear. r= 0.8981 C24 ● Spear. r = 0.9601

C20 5e+05 ●

HP14 ●

C22 ● HP11 HP12 ● ● 4e+05 raw read counts raw

3e+05 CR4Wo4CR4Wo1 ● ●

CR4Wo2 ●

2e+05

CR4Wo3 ●

4 8 12 16 total.score ENSMUSG00000034593

Pear. r= 0.9645 C24 6000 ●C20 Spear. r = 0.9663 ●

5000

C22 ● HP14 ●

4000

HP12 ● HP11 ● raw read counts raw

3000

CR4Wo4 ●

CR4Wo1 ●

2000 CR4Wo2 ●

CR4Wo3 ●

4 8 12 16 total.score ENSMUSG00000001056 1200 Pear. r= 0.8727 C24 ● Spear. r = 0.9724

1000 C20 ●

C22 ● HP14 800 ●

HP11 ● HP12 ● raw read counts raw CR4Wo1 ● CR4Wo4 ●

CR4Wo2 600 ●

400 CR4Wo3 ●

4 8 12 16 total.score ENSMUSG00000011179

40000 Pear. r= 0.9286 C24 ● Spear. r = 0.9724

C20 ●

30000

C22 ●

20000 HP11 HP14 ● ●

raw read counts raw HP12 ●

CR4Wo1 ● CR4Wo2 ● CR4Wo4 10000 ●

CR4Wo3 ●

4 8 12 16 total.score ENSMUSG00000041654

1000 Pear. r= 0.8819 C24 ● Spear. r = 0.954

C22C20 ● 800 HP14 ●

HP11 HP12 ● ● CR4Wo4 ● raw read counts raw CR4Wo1 ●

600

CR4Wo2 CR4Wo3● ●

4 8 12 16 total.score ENSMUSG00000030237

Pear. r= 0.9278 C24 ● Spear. r = 0.9601 600

C20 ●

HP14 ● 500 C22 ●

HP12 ● HP11 ●

400 raw read counts raw

300 CR4Wo1 ● CR4Wo4CR4Wo2 ●

CR4Wo3 ● 200

4 8 12 16 total.score ENSMUSG00000021187

Pear. r= 0.9506 C20 700 ● Spear. r = 0.9847

600 C24 ●

C22 500 ●

HP14 ●

400 HP12 ● raw read counts raw

HP11 ●

CR4Wo1 ● 300 CR4Wo4 ●

CR4Wo2 CR4Wo3● ● 200

4 8 12 16 total.score ENSMUSG00000038079

Pear. r= 0.921 C24 ● Spear. r = 0.9785

1000 C20 ●

800

C22 ● HP14 ●

HP12 ● HP11 ● raw read counts raw 600

CR4Wo4CR4Wo1 ● ●

CR4Wo2 ● 400

CR4Wo3 ●

4 8 12 16 total.score ENSMUSG00000020577

Pear. r= 0.9496 C24 ● 800 Spear. r = 0.954 C20 ●

C22 ● 600 HP14 ●

HP11 ● raw read counts raw HP12 ● 400 CR4Wo4 ●

CR4Wo1 ● CR4Wo2 ●

CR4Wo3 ●

200

4 8 12 16 total.score