Immunoglobulin light-chain toxicity in a mouse model of monoclonal immunoglobulin light-chain deposition disease

Monoclonal gammopathies of renal significance (MGRSs) are characterized by renal lesions resulting from monoclonal immunoglobulins produced by a nonmalignant B- or plasma-cell (PC) clone.¹  They comprise tubulopathies caused by light chain (LC) accumulating in the proximal or distal tubules, or glomerulopathies, most frequently amyloid LC (AL) amyloidosis and Randall-type monoclonal immunoglobulin deposition disease (MIDD).²  MIDDs are characterized by linear amorphous deposits of a monoclonal LC (LC deposition disease [LCDD]) or a truncated heavy chain (HC; HC deposition disease [HCDD]) along the tubular and glomerular basement membranes (BMs), around arteriolar myocytes, and in the mesangium, eventually leading to diabetic-like nodular glomerulosclerosis. Clinical manifestations include glomerular proteinuria with progressive kidney failure.^3-5  In LCDD, the most frequent form of MIDD, involved LCs are mostly of the κ isotype, with a striking overrepresentation of the Vκ4 variable (V) domain. Cationic V domains and/or complementarity-determining regions could account for their propensity to deposit along negatively charged BMs.^5-7  However, little is known about the pathogenic mechanisms involved in glomerular damage leading to proteinuria and glomerulosclerosis. Russell et al,⁸  Keeling et al,^9,10  and Herrera et al¹¹  showed the acquisition of a myofibroblast phenotype by mesangial cells exposed to purified LCDD LC. As in diabetic nephropathy (DN),¹²  these changes occurred along with the increased production of platelet-derived growth factor β (PDGFβ) and transforing growth factor β (TGFβ), resulting in excessive generation of extracellular matrix (ECM) proteins. However, because most of these experiments relied on acute exposure to pathogenic LC, they should be confirmed in a model reproducing the progressive development of glomerular lesions observed in MIDD. We recently developed a transgenic mouse model of MIDD in which the continuous production of a truncated HC from an HCDD patient by PCs led the main pathologic features of HCDD, but neither glomerulosclerosis nor kidney dysfunction was observed.¹³  In the present study, we developed a transgenic mouse model of LCDD (double homozygous DH-LMP2A [DH] and κF strain [κF-DH]) using targeted insertion of a human LC gene into the mouse κ locus, coupled with a deletion of the HC genes, to obtain high levels of circulating pathogenic free LC (FLC).¹⁴  This strategy proved efficient, with the induction of full-blown LCDD, and provided new insights into the pathophysiology of the disease and into the toxicity of monoclonal LC for PCs.

Gene targeting into the murine immunoglobulin κ locus was performed as previously described.¹⁵  The procedure is detailed in the supplemental Methods (available on the Blood Web site) and Figure 1A. As described elsewhere¹⁴  and in the supplemental Methods, LC transgenic mice (κF) were backcrossed with the DH strain to produce only free LC (Figure 1A). LCDD mice, corresponding to κF-DH were analyzed throughout this study. Primers used are listed in supplemental Table 1. DH mice¹⁶  were kindly provided by S. Casola (Istituto FIRC di Oncologia Molecolare, Milan, Italy). All animals had a mixed genetic background (BALB/C × 129SV × C57/BL6) and were maintained in pathogen-free conditions and analyzed at 2, 6, and 8 months except when otherwise stated. All experimental procedures were approved by our Institutional Review Board for Animal Experimentation, and by the French Ministry of Research (#7655-2016112211028184).

For long-term treatment, mice were treated with 0.75mg/kg of bortezomib^13,17  (Velcade; Janssen Cilag) and 2 mg/kg of cyclophosphamide (Endoxan; Baxter) once per week over 2 months. Bortezomib was injected subcutaneously, and cyclophosphamide was injected intraperitoneally. To test the sensitivity of PCs to proteasome inhibitor (PI), we injected for 2 consecutive days a suboptimal dose of bortezomib as previously described.¹³  Biochemical parameters were measured on a Konelab 30 analyzer with a creatinine enzymatic test (Thermo Fisher Scientific). Urine albumin concentrations were measured using a mouse albumin enzyme linked immunosorbent assay (ELISA) kit (Abcam).

Kidney samples were processed for light microscopic examination, immunofluorescence, and electron microscopic studies as previously described.¹³  Some specific techniques and the lesion score protocol are described in the supplemental Methods. Serum was analyzed for the presence of κLC by ELISA and western blots as previously described.^13,15  Antibodies are listed in supplemental Table 2.

Isolation of kidney glomeruli from mice was derived from Takemoto et al.¹⁸  Briefly, mice were perfused in the heart with magnetic Dynabeads 4.5 µm in diameter. Kidneys were minced, digested by Dnase I (Roche) and collagenase IV (Sigma-Aldrich), and filtered, and glomeruli were isolated using a magnetic rack. We could not apply this technique to kidneys with advanced sclerosis, because of the lack of access of the beads to the glomerular capillaries.

Intracellular staining was performed using the Intraprep Kit (Beckman Coulter). Flow cytometric analysis was performed on a BD Pharmingen LSRFortessa cytometer. Data were analyzed with BD FACSDiva software (BD Biosciences). PC enrichment was performed using a mouse CD138⁺ PC Isolation Kit (Miltenyi Biotec).

Total RNA from purified glomeruli or PCs was obtained using the miRNeasy Mini Kit (Qiagen), and purity (RNA integrity number >7) was assayed with the Bioanalyzer RNA 6000 Nano (Agilent). RNA sequencing (RNA-seq) libraries were generated with 500 ng of RNA. Alternatively, reverse transcription was performed using a high-capacity complementary DNA (cDNA) Reverse Transcription Kit (Applied Biosystems), with random hexamers. Relative quantification was performed with Premix Ex Taq or SYBR Premix Ex Taq (Takara) on cDNA samples (10 ng per reaction). Quantification of the gene of interest was analyzed by the ΔCt method, with immunoglobulin J used as the housekeeper gene. Probes are listed in supplemental Table 1.

Libraries were generated from 500 ng of total RNA using the Truseq Stranded mRNA Kit (Illumina), quantified with the Qubit dsDNAHS Assay Kit (Invitrogen), and pooled; 4 nM of this pool were loaded on a Nextseq 500 high-output flow cell and sequenced on a NextSeq 500 platform (Illumina) with 2 × 75–bp paired-end chemistry. Bioinformatic pipelines for mapping and filtering and tools for pathways analysis are described in the supplemental Methods. Gene Expression Omnibus accession numbers for the RNA-seq data sets reported in this paper are as follows: GSE119049 for glomeruli and GSE119048 for PCs (Gene Expression Omnibus superseries GSE119050).

The statistical tests used to evaluate differences between variables were carried out using Prism software (GraphPad Software). Statistical significance was determined using the nonparametric Mann-Whitney test to assess 2 independent groups and the log-rank (Mantel-Cox) test for survival analyses. Values of P < .05 were considered significant.

LCDD mice, as described in "Methods" (Figure 1A-B), closely recapitulate the features of monoclonal gammopathy, with an elevated number of PCs producing the human/mouse chimeric monoclonal FLC. Indeed, LCDD mice presented with ∼10% of PCs in spleen (as compared with ∼1% in WT mice), of which 95% produced the transgenic LC (Figure 1C-D). Serum FLC levels were similar to those observed in patients with LCDD,¹⁹  progressively increased with age, and lasted the lifespan of the animals. Bence-Jones proteinuria was readily detectable in 5-month-old mice (Figure 1E).

κF-DH mice (n = 12 in 2 independent cohorts) were monitored daily from month 5 for external signs of morbidity (see "Methods"). Median survival was 8.5 months, with the first mouse euthanized at 6 months and the last at 14 months (Figure 2A). DH mice (n = 10 in 2 independent cohorts) were used as control (similarly to κF-DH mice, they do not produce complete antibodies). At 14 months, all control mice were healthy (Figure 2A). Increased albuminuria was observed in 8-month-old κF-DH mice (Figure 2B). Mice euthanized at humane end points showed heavy albuminuria, reaching up to 24.80 g/L (mean, 8.30 ± 2.63 g/L; Figure 2B), together with high serum creatinine (135.58 ± 19.23 vs 7.89 ± 0.47 and 9.20 ± 1.33 µM/L in 8-month-old WT and DH control mice, respectively), confirming severe kidney failure in these mice (Figure 2C). Macroscopic observations revealed pale and atrophic kidneys (supplemental Figure 1). Altogether, these results confirm that the short lifespan of κF-DH mice is due to end-stage renal failure.

Pathologic studies were carried out in 2- and 6-month-old κF-DH mice and in mice euthanized at humane end points. As early as 2 months, immunofluorescence analysis revealed linear LC deposits along glomerular (Figure 3A) and medullary tubule BMs (supplemental Figure 2). At 6 months of age, intense LC staining was observed along all BMs in the renal medulla and cortex and into the mesangial space of glomeruli (Figure 3A; supplemental Figure 2), but kidneys showed normal tubular and glomerular structures (Figure 3A; supplemental Figure 2). In contrast, mice euthanized at humane end points displayed an absence of detectable LC reabsorption in proximal tubules, with massive staining of tubular BMs, mesangium, and Bowman’s capsule and diffuse nodular appearance of glomerular deposits (Figure 3A; supplemental Figure 2). Corticomedullary dedifferentiation suggestive of severe renal failure was observed (supplemental Figure 2). Periodic acid Schiff staining confirmed thickening of tubular BMs, mesangial extracellular matrix expansion with mesangial cell proliferation and mesangiolysis, and nodular glomerulosclerosis with glomeruli enlargement (Figure 3B). Focal areas of interstitial fibrosis with inflammatory infiltrates mainly composed of granulocytes/macrophages (CD11b⁺) and T cells (CD3⁺; supplemental Figure 3) were observed around sclerotic glomeruli and dilated tubules (Figure 3B). Electron microscopy revealed characteristic linear powdery punctuate deposits on the inner aspect of glomerular BMs and the outer aspect of tubular BMs, diffuse thickening of tubular and glomerular BMs, and expanded mesangial areas with massive accumulation of electron-dense material (Figure 3C). Finally, immunoelectron microscopy using gold-labeled anti-mouse κLC confirmed the κ composition of the deposits (Figure 3D). Lesions of thrombotic microangiopathy with widening of the subendothelial zone by electron lucent material were seen in some glomeruli from 6-month-old mice and in most glomeruli from mice euthanized at humane end points (Figure 3C-D). Similarly to the patient from whom the LC gene was extracted,²⁰  κF-DH mice also yielded LC deposits in other organs, including lung, liver, and heart (supplemental Figure 4). Collectively, kidney lesions observed in κF-DH mice fully recapitulate the human pathologic features of MIDD and are consistent with premature death resulting from end-stage kidney failure.

Having demonstrated that most 6-month-old κF-DH mice presented with early renal lesions of LCDD (Figure 3), we sought to compare their glomeruli transcriptome with that of control mice. RNA-seq performed on purified glomeruli revealed a total of 360 and 475 genes expressed differentially in κF-DH mice compared with WT and DH controls, respectively (abs[log₂FC] ≥0.7, false discovery rate (FDR) ≤0.05; Figure 4A; supplemental Figure 5A). Among these genes, 192 were deregulated compared with both controls, with 155 upregulated and 37 downregulated genes (supplemental Figure 5A-B). We performed a C2 canonical pathway analysis from the Molecular Signatures Database (MSignDB v.6.2),^21,22  which revealed that the cell cycle (mitotic) was the main modified pathway (Figure 4B; supplemental Table 3). All deregulated genes involved in the cell cycle were upregulated (Figure 4A; supplemental Figure 5A-B; supplemental Table 3). In addition to the cell cycle, we found 24 deregulated genes involved in the matrisome pathway (upregulated, n = 18; downregulated, n = 6), including the previously described upregulation of tenascin C (Tnc) and CTGF, known to be involved in glomerulosclerosis and interstitial fibrosis (Figure 4A-B; supplemental Figure 5B; supplemental Table 3).^10,12,23  However, we did not find any activation of pathways related to TGFβ, despite its established role in glomerulosclerosis. Immunostaining for Ki67 (Figure 4C-D) and Tnc (supplemental Figure 6) in 6-month-old mice confirmed extensive cell-cycle and ECM remodeling, respectively, in the glomeruli of κF-DH mice.

Several studies have demonstrated that the production of a pathogenic monoclonal immunoglobulin or immunoglobulin fragment could influence PC fitness and sensitivity to treatments through a general cellular stress involving endoplasmic reticulum (ER) stress, decreased autophagy, and oxidative stress.^13,24  To determine if such cellular stress occurred in PCs producing LCDD LC, we performed RNA-seq on magnetically sorted spleen PCs (>80%) of κF-DH mice compared with WT mice producing polyclonal complete immunoglobulins and with DH mice producing polyclonal FLC. Our data revealed a total of 1946 and 809 expressed genes deregulated in LCDD LC–producing PCs compared with WT and DH controls, respectively (abs[log₂FC] ≥0.7, FDR ≤0.05); among these deregulated genes, 722 and 376 were upregulated and 1224 and 433 were downregulated, respectively (Figure 5A). Only genes similarly deregulated in both comparisons (κF-DH vs WT and κF-DH vs DH) were considered for further analysis (supplemental Figure 7A). This restricted the analysis to 180 and 213 genes, respectively, upregulated and downregulated in κF-DH vs both controls (supplemental Figure 7A-B). C2 canonical pathway analysis (MSignDB v.6.2) found that among upregulated genes, the only significant pathways were related to UPR (Figure 5B; supplemental Table 4). Genes of the UPR pathway upregulated in κF-DH PCs vs each control (WT and DH) are indicated on the MA plots (Figure 5A), showing that they were not only upregulated but, as expected, also highly expressed in PCs.²⁵  Activated pathways in each comparison are depicted in supplemental Figure 7C. Gene ontology (GO) biologic process analysis showed that 9 of the 20 most significant overlaps and the 3 most significant pathways were related to topologically incorrect proteins and ER stress, further suggesting that the activation of UPR was likely due to the pathogenic LCDD LC (Figure 5C; supplemental Table 5). Downregulated genes were associated with ECM modification, hematopoietic lineage, and G protein–coupled protein signaling (supplemental Table 6). We performed real-time polymerase chain reaction on selected genes (Herpud1, Hspa5, Ddit3, Xbp1s) and confirmed a similar tendency for upregulation of genes involved in UPR, even if, consistent with RNA-seq analysis, no increase in Xbp1s was observed (Figure 5D). We next verified if such ER stress could sensitize PCs to PI.¹³  We treated mice with suboptimal doses of bortezomib for 2 consecutive days to evaluate the level of PC depletion. This protocol led to a partial (approximately twofold) reduction of the absolute number of PCs in WT and DH mice (45.16% and 57.54% of depletion, respectively) as previously demonstrated.¹³  In contrast, PC depletion in κF-DH mice was far more efficient than in control mice, with a mean depletion of 84.46%, corresponding to a 6.44-fold reduction in absolute PC number (Figure 5E).

The efficient response to PI prompted us to evaluate the effect of LC removal on renal lesions evolution. Weekly injections of bortezomib plus cyclophosphamide were administered to 6-month-old mice over 2 months, which were compared with age-matched untreated mice. All mice were subsequently euthanized at the end of the histologic study treatment. The treatment led to an almost complete (>90% of reduction) and sustained depletion of circulating κLC (Figure 6A). Treatment completely protected mice from kidney dysfunction (Figure 6B) and early death (Figure 6C), and treated mice presented significantly fewer abundant renal LC deposits as compared with nontreated mice (Figure 6D-E). Interestingly, number of kidney lesions after treatment (at 8 months) was also significantly lower as compared with younger mice at 6 months (Figure 6D-E). Periodic acid Schiff staining showed glomeruli of smaller size in treated κF-DH mice as compared with 6-month-old mice, and EM analysis revealed less marked glomerulosclerosis and the absence of thrombotic microangiopathy–like lesions, despite persistence of diffuse thickening of glomerular BMs (Figure 6D). Although serum creatinine was significantly increased in nontreated mice associated with an increase of serum FLC, it was maintained at normal levels in treated mice (Figure 6A-B), which, compared with control 6-month-old mice, showed a slight but nonsignificant increase in albuminuria (Figure 6F) and no rise in serum FLC (Figure 6A).

We herein describe the first transgenic mouse model fully reproducing specific renal lesions and kidney dysfunction observed in human LCDD. We previously characterized another model of MIDD using a similar transgenic strategy (ie, targeted insertion of a human truncated γ1 HC obtained from a patient with biopsy-proven HCDD into the mouse κ locus).¹³  In this model, the pathogenic immunoglobulin induced kidney lesions closely resembling those observed in human MIDD, except glomerulosclerosis, a hallmark of HCDD,⁵  and glomerular dysfunction. We first suspected the role of the genetic background (C57/BL6 and 129SV),^26-28  but we could not exclude the possibility that the low level of circulating HC (∼30-40 µg/mL) precluded pathologic progression of the disease. In the present LCDD mouse model, we used a new strategy to obtain high levels of circulating human FLC.^14,29  We inserted the pathogenic V domain of human LC into the κ locus and crossed the mice with DH mice, which display increased PC development in the absence of immunoglobulin HC.^14,16,30  Production of chimeric FLC occurred in virtually all PCs, and κLC serum level reached 1 g/L in κF-DH mice, which is similar to that observed in LCDD patients.¹⁹  Similar productions were obtained in other unpublished models of MGRSs,¹⁴  making this strategy a powerful tool to study the pathogenicity of monoclonal LC in vivo.

Accordingly, κF-DH mice recapitulate progressive glomerular lesions closely resembling those observed in MIDD patients, with linear monoclonal κLC deposits along glomerular and tubular BMs and in the mesangium, with glomerular enlargement and mesangial hypercellularity, resulting in albuminuria. The lesions eventually lead to nodular glomerulosclerosis and end-stage kidney failure within the first year. We first confirmed that the V domain alone bore the pathogenic properties of LCCD LC, similarly to what we showed in a model of LC-induced Fanconi syndrome.¹⁵  This does not mean that the C domain could not play a positive or negative role in the deposition process, and it would have been interesting to compare with full-length human LC. Unfortunately, our attempts to obtain such a model with insertion of the complete cDNA in the κ locus were unfruitful, with very low expression of the transgene (unpublished data). Tumor graft models could be advantageously used to compare full-length with chimeric LC or to test V domain mutations that could be involved in the deposition of the LC.^20,31  If the pathologic lesions slowly progress during the first months, the occurrence of kidney dysfunction with increased serum creatinine seems to be more sudden and unpredictable, requiring careful follow-up of mice after month 6. All these features make the κF-DH mice an accurate model to study the pathophysiology of MIDD, as well as other kidney diseases featuring nodular glomerulosclerosis, particularly DN. Indeed, κF-DH mice recapitulate all the criteria used to validate the glomerular lesions of DN as established by the Diabetic Complications Consortium³²  and could be advantageously used to decipher common pathways involved in mesangial cell proliferation and ECM expansion leading to glomerulosclerosis. Nodular glomerulosclerosis has been widely studied, especially in the context of DN^32-34  and LCDD,^8,11,35  highlighting a role for TGFβ signaling and PDGF. However, because of the lack of relevant animal models reproducing the chronic progression of kidney lesions, it is still unclear whether TGFβ and PDGF are early mediators of glomerulosclerosis or if they are involved later, enhancing and accelerating terminal fibrosis of the glomeruli. When we performed RNA-seq on glomeruli from 6-month-old LCDD mice, which display abundant LC deposits but no glomerulosclerosis, we did not find any activation of the TGFβ or PDGF pathway despite already-present ECM deregulation, as indicated by the overexpression of Tnc.^10,12,36  Tnc is a ubiquitous component of the expanded glomerular ECM in pathologic conditions, including DN.³⁷  More specifically, it has been characterized as a main component of abnormal ECM produced by mesangial cells after LCDD LC but not AL LC exposure, likely because of the concomitant overexpression of matrix metalloproteinases, which degrade ECM components in the latter case.^9,10  We observed a slight but significant increase of CTGF, known for its profibrotic properties in damaged tissues, including the kidney.^23,38  The most prominent early effect of LC deposition in our model was the induction of proliferation, as observed by RNA-seq and Ki67 staining. This result is consistent with the presence of hypercellularity and the enlargement of glomeruli that precede glomerulosclerosis and kidney dysfunction. Similar studies in mice with more advanced disease stages or in treated vs nontreated mice should be carried out to decipher the sequential molecular events leading to glomerulosclerosis, but our results suggest that TGFβ overproduction and its consequences are secondary events after LC-induced proliferation of mesangial cells and production of abnormal ECM. Understanding the pathways linking LC deposition and cellular proliferation should likely provide new insights into the pathophysiology of the disease.

Recent studies have highlighted the efficiency of PI-based treatment in patients with MIDD, as already demonstrated in AL amyloidosis.^5,19,39-42  We confirmed this observation in our mouse model. This regimen led to a rapid hematologic response with almost complete and sustained depletion of circulating FLC, leading to a significant improvement in kidney lesions. Interestingly, kidney lesions were significantly decreased compared with those observed in animals at ages corresponding to the beginning of treatment, highlighting a possible cleaning mechanism leading to the elimination of LC deposits. This improvement in kidney lesions was also characterized by normalization of glomerular size and absence of thrombotic microangiopathy–like lesions occasionally observed in 6-month-old mice. Although complete recovery was not achieved after 2 months as observed by electron microscopy, the treatment was sufficient to maintain normal serum creatinine levels. This result is in accordance with studies of human diseases, including DN,^43,44  showing that glomerulosclerosis at early stages can be reversed with efficient treatment, although this process in humans can take several years. In mouse models, however, the reversibility was demonstrated after short-term treatment in DN^45,46  or hypertension-related glomerulosclerosis.⁴⁷  Accordingly, we demonstrated the almost complete reversibility of kidney lesions in LCDD-induced glomerulosclerosis after LC depletion. Additional studies are needed to confirm if long-term LC removal could lead to complete renal recovery to decipher the involved mechanisms and test therapeutic strategies that could accelerate this process. In this latter view, it seems that therapy based on mesenchymal stem cells could be a valuable approach, as demonstrated in rodent models of glomerulopathies^48,49  or more recently in ex vivo models of monoclonal LC-induced glomerular damage.⁵⁰  Our model also enabled us to analyze the intrinsic toxicity of pathogenic monoclonal LC for PCs. Similarly to what was shown in patients with MIDD^5,7,19,51  and in the HCDD model,¹³  PCs producing LCDD LC are highly sensitive to PI. We assume that mouse PCs are quite different from dysplastic human PCs, in that genomic alterations other than sole LC production could explain such sensitivity. However, in contrast, differences in PI sensitivity in our model uniquely depended on the presence of human LC, because PCs were otherwise strictly identical. We showed that these PCs display a high ER stress and, more generally, an overexpression of genes involved in the response to misfolded or unfolded proteins. Similar results have been obtained with PCs forced to express HC only,⁵²  truncated HC,¹³  or truncated LC.⁵³  Collectively, we hypothesize that the high sensitivity to PI is likely related to the impaired capacity of PCs to cope with the production of an abnormal, aggregation-prone, monoclonal LC, as observed in AL amyloidosis.²⁴  Using our transgenic strategy, we now have developed several mouse models expressing human LC from other MGRSs, including AL amyloidosis and Fanconi syndrome,¹⁴  which will serve to determine whether all pathogenic LCs are toxic for PCs and decipher the common pathways leading to ER stress and PI sensitivity.

We describe here the first mouse model fully recapitulating the features of MIDD, including glomeruloslerosis and end-stage kidney disease. Compared with other models of chronic kidney diseases, the development of kidney dysfunction relies on the natural causing factor (ie, monoclonal LC), without any chemical or physical induction or genetic mutations artificially triggering the disease.⁵⁴  Consequently, such a model will likely be useful to understand the molecular mechanisms leading to glomerulosclerosis in MIDD, as well as other chronic kidney diseases, such as DN. It also confirms the toxicity of abnormal LC for PCs and will likely serve to design and test new therapeutic strategies for MIDD and other related diseases.

The authors thank the staff of the Biologie Intégrative Santé Chimie Environnement (BISCEm) technical platforms at the University of Limoges (animal, cell cytometry, microscopy, transgenesis, and bioinformatics facilities), the Department of Pathology of Poitiers, L. Magnol and K. Vuillier for access to the Konelab 30 analyzer, and A. Garot for helpful discussions.

This work was supported by grants from Fondation Française pour la Recherche contre le Myélome et les Gammapathies Monoclonales, Limousin Committees of Ligue Nationale contre le Cancer, Fondation pour la Recherche Médicale, Agence Régionale de la Santé, and Institut Universitaire de France. S.B. is supported by Centre Hospitalier Universitaire Dupuytren Limoges and Plan National Maladies Rares. M.V.A., A.B., and M.O.A. were funded by fellowships from Région Limousin (now Région Nouvelle Aquitaine), the French Ministry of Research, and Agence de Valorisation de la Recherche de l’Université de Limoges. M.V.A is funded by a fellowship from Fondation ARC pour la Recherche sur le Cancer.

Contribution: S.B. designed, performed, and analyzed experiments and drafted the manuscript; M.V.A. performed and analyzed experiments and drafted the manuscript; A.B. designed, performed, and analyzed some experiments; V.J. drafted parts of the manuscript and provided general advice; A.R., S.K., C.C., C.O., Z.O., and M.O.A. performed and analyzed some experiments; F.B., B.H., A.P., and N.P. performed the biostatistics and bioinformatics on RNA-seq data; G.T. analyzed and provided advice on kidney pathology; A.J., L.D., and M.C. provided general advice and reviewed the manuscript; F.B. analyzed data and critically reviewed the manuscript; and C.S. designed and analyzed experiments, supervised research, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Christophe Sirac, CRIBL Laboratory, UMR CNRS 7276 INSERM 1262, CBRS Room 110, 2 Rue du Dr Marcland, 87000 Limoges, France; e-mail: christophe.sirac@unilim.fr.