Research ArticleCancer

Acute kidney injury promotes development of papillary renal cell adenoma and carcinoma from renal progenitor cells

See allHide authors and affiliations

Science Translational Medicine  25 Mar 2020:
Vol. 12, Issue 536, eaaw6003
DOI: 10.1126/scitranslmed.aaw6003

Kidneys’ immortal wounds

Tissue injury and repair can contribute to subsequent development of cancer, especially in the setting of chronic damage and repeated cycles of tissue repair. It appears that acute injury may have a similar effect, at least in the kidney. Peired et al. first analyzed data from multiple independent cohorts of human patients and demonstrated a correlation between acute kidney injury and subsequent development of papillary renal cancer and recurrence of cancer in patients who had undergone surgery to remove their initial tumors. The authors then studied human renal progenitor cells and mouse models to determine the underlying mechanism for kidney injury–induced carcinogenesis.

Abstract

Acute tissue injury causes DNA damage and repair processes involving increased cell mitosis and polyploidization, leading to cell function alterations that may potentially drive cancer development. Here, we show that acute kidney injury (AKI) increased the risk for papillary renal cell carcinoma (pRCC) development and tumor relapse in humans as confirmed by data collected from several single-center and multicentric studies. Lineage tracing of tubular epithelial cells (TECs) after AKI induction and long-term follow-up in mice showed time-dependent onset of clonal papillary tumors in an adenoma-carcinoma sequence. Among AKI-related pathways, NOTCH1 overexpression in human pRCC associated with worse outcome and was specific for type 2 pRCC. Mice overexpressing NOTCH1 in TECs developed papillary adenomas and type 2 pRCCs, and AKI accelerated this process. Lineage tracing in mice identified single renal progenitors as the cell of origin of papillary tumors. Single-cell RNA sequencing showed that human renal progenitor transcriptome showed similarities to PT1, the putative cell of origin of human pRCC. Furthermore, NOTCH1 overexpression in cultured human renal progenitor cells induced tumor-like 3D growth. Thus, AKI can drive tumorigenesis from local tissue progenitor cells. In particular, we find that AKI promotes the development of pRCC from single progenitors through a classical adenoma-carcinoma sequence.

INTRODUCTION

Chronic tissue injury related to toxin exposure or viral infections are proven factors in carcinogenesis (1). Whether acute tissue injury and remodeling may promote cancer is purely speculative. Acute tissue injury causes DNA damage and repair involving extensive proliferation of resident stem cell populations (2) and/or polyploid genome alterations of differentiated cells (3). Both conditions may facilitate cancer formation. In addition, repair processes trigger signaling pathways that control cell cycle activation, cell survival, and proliferation, stimulating pathways that drive cancer development (3, 4). Here, we examine a putative link between acute kidney injury (AKI) and subsequent cancer development. The kidney provides a unique opportunity to study this possibility because patients with AKI survive for a long time even when developing end stage renal disease with supportive or renal replacement therapies, such as dialysis and kidney transplantation. In both situations, the native kidneys remain in situ and allow assessing tumor rates (5).

AKI is a global public health concern resulting in 1.7 million deaths per year (6). If not lethal in the acute phase, AKI is largely considered reversible as indicated by resumed urine production or biomarkers assessing injury resolution (7). However, even mild AKI episodes give a substantial risk for subsequent chronic kidney disease (CKD) due to irreversible nephron loss (8). In this study, we hypothesized that AKI, particularly ischemia-reperfusion injury (IRI), may be a risk factor for one or more types of renal cell carcinoma (RCC) by inducing tissue repair signaling pathways that may drive neoplastic growth.

RESULTS

AKI is associated with RCC in two different human cohorts

To assess the potential role of AKI insults as a risk factor for RCC development, we reviewed the hospital database of the Careggi University Hospital, Florence, Italy, for all the patients with incident RCC (n = 1618) who underwent surgery between 2012 and 2018 (fig. S1). We retrieved laboratory information for 468 patients with a minimum follow-up of 1 year before tumor diagnosis. AKI episodes were identified in the patients’ clinical history in 76 of 468 (16.2%), corresponding to an incidence rate estimate of 420 AKI episodes per 10,000 person-years, whereas the expected rate in the general population is 50 to 150 per 10,000 person-years (9, 10). In detail, comparing our population with the one described by Sawhney et al. (9) yielded an AKI incidence ratio of 3.2 [95% confidence interval (CI), 2.6 to 4.1].

To confirm the observed association between AKI episodes and RCC, we analyzed a nationwide cohort study in Denmark (population of 5.4 million) using population-based medical databases (11, 12). We identified all patients with a hospital diagnosis of AKI between 1977 and 2006 (fig. S2). We obtained data on subsequent incident RCCs and quantified the risk of RCC after AKI. Among 12,480 patients diagnosed with AKI, 7 (0.06%) were diagnosed with RCC within the first year, and 14 (0.1%) were diagnosed with RCC more than 1 year after the first AKI episode. Comparing the observed number of RCC with the expected number computed on the basis of age- and sex-specific cancer incidence rates in the general Danish population (1.6 in the first year and 6.7 cancer cases in the subsequent years) led to calculation of standardized incidence ratios for developing RCC of 4.4 (95% CI, 1.7 to 9.0) 1 year after the AKI episode and of 2.1 afterward (95% CI, 1.1 to 3.5). These results suggest that AKI episodes associate with RCC.

AKI episodes are a risk factor for pRCC development and recurrence in a single-center study

We evaluated whether AKI episodes associated with clear cell RCC (ccRCC) and/or papillary RCC (pRCC), the two most frequent tumor histotypes accounting for approximatively 75 and 15% of RCCs, respectively (13). We retrieved clinical and laboratory information for 56 patients with pRCC and 75 patients with ccRCC stage T1 or T2 with a minimum follow-up of 1 year before tumor diagnosis (Table 1 and fig. S1). The frequency of the exposure (that is, previous AKI episodes) was significantly higher in patients with pRCC compared with a cohort of 101 in-house nontumor controls (21.4 versus 6.9%, P = 0.008; Table 1). For ccRCC, there was no significant difference (12.0 versus 6.9%, P = 0.247; Table 1). To further investigate the association with AKI episodes, a binary logistic regression was performed including known risk factors for RCC such as male gender, age, presence of diabetes mellitus (DM), and CKD as covariates (14). With multivariate analysis, a previous AKI episode was significantly associated with pRCC development [odds ratio (OR), 3.48; 95% CI, 1.1 to 10.67] compared with controls (Table 1), but there was no correlation in patients with ccRCC (Table 1). pRCC can be classified into types 1 and 2 based on histology (56 patients with pRCC: 21 type 1, 28 type 2, and 7 undefined). Among the 28 patients affected by type 2 pRCC, 7 had a previous AKI episode (25 versus 6.9% of controls, P = 0.012), whereas of the 21 type 1 patients, 3 had AKI (14.3 versus 6.9%, P = 0.37). The remaining two patients with AKI had unclassified pRCC histotype. These findings suggest that AKI could be a risk factor for pRCC development and particularly for the type 2 subtype.

Table 1 AKI episodes are a risk factor for the development of pRCC but not for ccRCC in humans.

Clinical characteristics of pRCC, ccRCC, and controls are described. Between-group comparisons (a) were performed using the Mann-Whitney U test for continuous variables or the chi-squared test for categorical variables. Binary logistic regression analysis (b) was used for the association between previous acute kidney injury (AKI) episodes (yes/no) and pRCC development adjusted for age, gender, preoperative chronic kidney disease (CKD), and diabetes mellitus (DM). eGFR, estimated glomerular filtration rate; OR, odds ratio; RCC, renal cell carcinoma; pRCC, papillary RCC; ccRCC, clear cell RCC; CI, confidence interval; y, yes; n, no; F, female; M, male. Continuous variables are presented as mean ± SD. P values ≤0.05 are indicated in bold.

View this table:

Resection of localized pRCC is frequently associated with postoperative AKI because of renal artery clamping, parenchymal mass reduction, intraoperative hypotension, and decreased renal perfusion pressure (15). We investigated whether AKI upon pRCC surgery could be a risk factor for tumor recurrence. In 83 patients, a follow-up of at least 1 year after nephrectomy for pRCC (mean follow-up, 3.31 ± 1.84 years) was available (fig. S1). In this cohort, postoperative AKI occurred in 18 of 83 (21.7%) patients, similar to previous reports (16). Although the cumulative recurrence frequency was 14 of 83 (16.9%), it was significantly more frequent among patients who experienced postoperative AKI compared with non-AKI patients [7 of 18 (38.9%) versus 7 of 65 (10.8%), P = 0.005)] (table S1). Five-year recurrence-free survival was significantly (P = 0.008) lower in postoperative patients with AKI (Fig. 1A). A binary logistic regression was performed to evaluate the association between postoperative AKI and recurrence, including age at surgery, clinical T stage, and R.E.N.A.L. (radius, exophytic/endophytic properties, nearness of deepest tumor portion to collecting system or sinus, anterior/posterior, and location relative to polar line) nephrometry score, a surgical indicator of tumor complexity previously associated with recurrence (17), as covariates. By multivariate analysis, postoperative AKI was significantly associated with recurrence (OR, 7.24; 95% CI, 1.65 to 31.86) (table S1). Together, these results suggest that an AKI episode could be a previously unrecognized major risk factor for pRCC recurrence.

Fig. 1 Recurrence-free survival in patients with pRCC according to the presence or absence of a postoperative AKI episode.

(A and B) Recurrence-free survival of patients with pRCC based on the presence or absence of a postoperative AKI episode in (A) a single-center study (n = 65 no postoperative AKI and n = 18 postoperative AKI) and (B) a multicentric study (n = 43 no postoperative AKI and n = 46 postoperative AKI). Censored patients and patients at risk are shown. Kaplan-Meier curves are compared by log-rank test (7.143 and 5.36, respectively) and Cox regression [HR(A) = 3.6 (1.3 to 10.2); HR(B) = 3.4 (1.1 to 10.3)]. Statistical significance was set for a P value <0.05. PO AKI, postoperative AKI.

A multicentric study confirms that AKI episodes are a risk factor for pRCC recurrence

To validate the association between AKI and pRCC, we analyzed the Italian Registry of Conservative Renal Surgery (RECORd1 project), a 4-year prospective observational multicentric study promoted by the Leading Urological Non-Profit Foundation for Advanced Research (LUNA) of the Italian Society of Urology (18). Overall, 769 consecutive patients underwent partial nephrectomy for a renal tumor at 19 Italian urological institutions from 2009 to 2012 (fig. S3). For the current study, we extracted data on 594 patients with a histologic diagnosis of RCC and with a minimum of 1-year follow-up after surgery. Renal function was evaluated at baseline and at first postoperative day. Table S2 summarizes the baseline features of the entire cohort.

In 89 patients with pRCC histotype, 18 (20.2%) patients experienced local recurrence, whereas no systemic recurrences were recorded. Forty-six (51.7%) patients experienced postoperative AKI. In these patients, a significantly higher rate of local recurrence was found (30.4 versus 9.3% in non-AKI patients; P = 0.01) (table S3). Survival analysis s showed a significantly (P = 0.021) lower 5-year recurrence-free survival in patients experiencing a postoperative AKI episode compared with those without (Fig. 1B). By multivariate analysis, the association between postoperative AKI and tumor recurrence remained significant (P = 0.023) after adjustment for age at surgery and clinical T stage (table S3). Together, these results confirm that AKI at surgery is associated with pRCC recurrence.

Long-term follow-up reveals that AKI induces papillary tumors in mice

To experimentally test the hypothesis that AKI promotes tumor development, we induced unilateral IRI in wild-type mice and examined their kidneys in the long term at different time points (fig. S4A). Development of types 1 and 2 papillary tumors occurred in a time-dependent manner (Fig. 2, A to H). Over the course of 36 weeks, the lesions were detectable by ultrasound imaging (Fig. 2, A and B). Papillary tumors ranged from small adenomas with low nuclear grade to pRCC appearing as larger papillary tumors with high nuclear grade (Fig. 2, C to F). Type 1 papillary tumors appeared as papillae and tubular structures covered with monolayers of small cells containing basophilic cytoplasm and small, uniform oval nuclei, often forming cysts (Fig. 2C). Type 2 papillary tumors appeared as multilayered papillae covered by large cells with eosinophilic cytoplasm and large spherical nuclei with prominent nucleoli (Fig. 2D) and frequent high nuclear grade (Fig. 2E) as described in the most recent World Health Organization pRCC classification (19). In addition, focal accumulation of foamy macrophages in the stroma of the papillary stalks (a highly characteristic pRCC finding) was also seen in papillary mouse tumors (Fig. 2F). No tumors were found 4 weeks after AKI, but 55.6% of mice had papillary adenomas (pAdenomas) at 12 and 36 weeks (Fig. 2G). At 36 weeks, 22.2% of mice showed pRCC (Fig. 2G). Overall, 68.75% of the tumors were of type 1 and 31.25% were of type 2 (Fig. 2H). In contrast, mice that did not undergo IRI did not develop any tumors (fig. S4B). These results suggest that AKI is a risk factor for types 1 and 2 mouse papillary tumors, similar to findings in humans.

Fig. 2 Long-term follow-up reveals that AKI induces papillary tumors in mice.

(A) Ultrasound image with power Doppler mode (red) of a wild-type mouse kidney (displayed in transverse plane) at 36 weeks after IRI (representative of n = 3 mice). The region of interest (dashed line) indicates tumoral tissue. Scale bar. 1 mm. (B) Length (micrometers) of the lesions in wild-type mouse kidneys at 36 weeks after IRI (n = 5 mice). Bar indicates median value. (C to F) Representative histological images with H&E staining of papillary tumors from wild-type mouse kidneys after IRI. (C) Type 1 papillary tumor. (D) Type 2 papillary tumor. (E) Nuclear atypias. (F) Foamy macrophages within a papillary lesion. Scale bars, 25 μm. (G) Percentages of mice at 4 weeks (n = 8), 12 weeks (n = 9), and 36 weeks (n = 9) after IRI presenting with papillary adenomas (pAs) and pRCCs. Numbers above graph represent P values. (H) Percentage of type 1 versus type 2 papillary tumors in wild-type mice at 36 weeks after IRI (n = 4 mice). Bars indicate mean values. No significant difference between the groups. (I) Representative confocal image of a type 1 papillary tumor from Pax8/Confetti mouse kidneys at 36 weeks after IRI. Scale bar, 25 μm. (J) Split confocal images of CK7 staining (orange) of a type 1 papillary tumor from Pax8/Confetti mouse kidneys at 36 weeks after IRI. Scale bar, 25 μm. (K) Split confocal images of CK7 staining (orange) of a type 2 papillary tumor from Pax8/Confetti mouse kidneys at 36 weeks after IRI. Scale bar, 25 μm. (L) Detail of nuclear atypias in a type 2 papillary tumor from Pax8/Confetti mouse kidneys at 36 weeks after IRI. Scale bar, 25 μm. (M) Clonal analysis in type 1 and 2 papillary tumors from Pax8/Confetti mouse kidneys at 36 weeks after IRI (n = 7 mice). Bars indicate mean value. (N and O) 3D reconstruction of representative pRCC in Pax8/Confetti mouse kidneys at 36 weeks after IRI. (N) Collagen IV (COL4) staining (cyan). Arrows indicate invasion. (O) Von Willebrand Factor (VWF) staining (purple). Arrows indicate tumor vascularization. Scale bars, 25 μm. DAPI (4′,6-diamidino-2-phenylindole) (white) counterstains nuclei. Signals for fluorescent Confetti proteins are as follows: green fluorescent protein (GFP), green; red fluorescent protein (RFP), red; yellow fluorescent protein (YFP), yellow; and cyan fluorescent protein (CFP), blue. Statistical power was assessed using Fisher's exact test (G) or Mann-Whitney test (H). N.S Not significant (P > 0.05).

AKI-driven papillary tumors originate from proliferation of single mouse tubular epithelial cells

To study the origin of papillary tumors, we evaluated their clonality using a lineage tracing approach based on the Confetti reporter stochastic labeling of tubular epithelial cells (TECs) at single-cell level with different fluorochromes, that is, Pax8.rtTA;TetO.Cre;R26.Confetti (Pax8/Confetti) mice (fig. S4, C and D). Both type 1 (Fig. 2, I and J) and type 2 (Fig. 2, K and L) papillary tumors tagged at 36 weeks after IRI were mostly of single colors. Clonal analysis revealed that >95% of types 1 and 2 papillary tumors were monoclonal (Fig. 2M). Of note, type 1 tumors stained positively for cytokeratin 7 (CK7) (Fig. 2J), whereas type 2 did not (Fig. 2K), as reported in patients with pRCC (13). In addition, type 2 pRCCs presented extensive nuclear atypia, as shown in Fig. 2L. Three-dimensional (3D) analysis confirmed that clonal populations from the primary tumors invaded the kidney parenchyma (Fig. 2N) and that papillary tumors showed characteristic aberrant vascularization, confirming invasive growth typical of carcinomas (Fig. 2O). Thus, papillary tumors developing after postischemic AKI originate from clonal amplification of single TECs.

Human pRCCs associate with altered AKI-activated pathways

To understand how AKI could drive pRCC formation, we speculated that AKI-activated signaling pathways could promote pRCC pathogenesis. To this aim, we analyzed AKI-activated pathways included in the protein array deposited in The Cancer Genome Atlas (TCGA) (https://cancergenome.nih.gov/cancersselected/kidneypapilary) from 161 patients with pRCC (20). Among the pathways involved in post-AKI repair (table S4), a decreased expression of mammalian target of rapamycin (mTOR), β-catenin, Von Hippel-Lindau (VHL), and an increased expression of NOTCH1 were associated with advanced pRCC stage (T3-T4) at diagnosis (Fig. 3A). A survival analysis indicated that among them, only increased NOTCH1 associated with worse prognosis (Fig. 3B and table S4). Low amounts of phosphatase and tensin homolog (PTEN) and SMAD4 also associated with worse prognosis (Fig. 3, C and D, and table S4). Types 1 and 2 pRCC associated with differences in expression of distinct AKI-related signaling pathways (Fig. 3E). In particular, high amounts of mTOR, β-catenin, and VHL associated with type 1, whereas high amounts of NOTCH1 strongly associated with type 2 (Fig. 3E). These results suggest that pRCC is associated with altered AKI-activated pathways.

Fig. 3 NOTCH1 overexpression is specific for human type 2 pRCC and predicts poor prognosis.

(A) Log2 RPPA signal for proteins differently expressed in tumor stage T1-T2 (black boxes, n = 29 patients) versus T3-T4 (pink boxes, n = 66) pRCC. (B to D) Overall survival of patients with pRCC based on high (≥50th percentile, n = 62) or low (˂50th percentile, n = 61) expression of different proteins: NOTCH1 (B), PTEN (C), or SMAD4 (D). (E) Log2 RPPA signal for proteins different in tumor type 1 (black box, n = 58) versus type 2 (red box, n = 49) pRCC. (F) NOTCH1 log2 RPPA signal in patients with pRCC of clusters C1, C2a, C2b, and C2c/CIMP (n = 71, 27, 18, and 9, respectively). (G) NOTCH1 log2 RPPA signal in patients with nonmutated (black boxes) or mutated (green boxes) MET (nonmutated n = 110, mutated n = 15), FH (nonmutated n = 4, mutated n = 5), CDKN2A (nonmutated n = 88, mutated n = 37), SETD2/BAP1/PBRM1 (nonmutated n = 111, mutated n = 11), or NRF2-ARE pathway (nonmutated n = 108, mutated n = 17). (H) Differences in overall survival of patients with type 1 or 2 pRCC based on high (≥50th percentile, type 1 n = 14 and type 2 n = 40) or low (˂50th percentile, type 1 n = 44 and type 2 n = 9) NOTCH1 protein expression. β-cat, β-catenin. RPPA, reverse-phase protein array. Box-and-whisker plots: line, median; box, 25 to 75%; whiskers, minimum or maximum. Statistical significance was calculated by Mann-Whitney test. Kaplan-Meier curves are compared by log-rank test and Cox regression [HR(B) = 4.0 (1.1 to 14.3); HR(C) = 3.3 (1.0 to 10.8); HR(D) = 3.2 (1.0 to 10.1); HR(F) = 4.6 (1.0 to 20.5)]. Numbers above graphs represent P values.

NOTCH1 overexpression is specific for human type 2 pRCC and predicts poor prognosis

NOTCH1 strongly associates with disease severity and patient survival in pRCC, is highly specific for type 2 subtype, and has critical importance in AKI repair (21). These specificities prompted us to examine by TCGA Research Network data, focusing our attention on NOTCH1 (20). The results presented by TCGA Research Network indicate that type 2 pRCC consists of at least three subtypes based on molecular features (clusters C2a, C2b, and C2c-CIMP) (20). NOTCH1 protein was increased in all these subgroups, but not in type 1 pRCCs (cluster C1) (Fig. 3F). Types 1 and 2 pRCCs are both associated with several gene mutations (20). Germline or somatic mutations of the tricarboxylic acid cycle enzyme gene fumarate hydratase (FH), or the nuclear factor erythroid 2 (NRF2)/antioxidant response element (ARE) pathway, including CUL3, KEAP1, NF2, and NFE2L2 (NRF2) (22), can cause sporadic pRCC that is usually type 2. In contrast, activating germline MET mutations cause hereditary type 1 pRCCs (20). Type 1 pRCCs with MET mutations did not show increased NOTCH1 (Fig. 3G). In contrast, type 2 pRCCs characterized by CDKN2A silencing, or by SETD2BAP1/PBRM1 mutations or by mutations in FH or the NRF2-ARE pathway (20), showed increased NOTCH1 protein in comparison to nonmutated tumors (Fig. 3G). Lifetime survival analysis revealed that NOTCH1 protein amounts correlated with survival in patients with type 2 pRCC (Fig. 3H). Last, multivariate analysis showed that NOTCH1 expression in tumors correlated with prognosis even when compared to known risk factors for pRCC (table S5). These results indicate that the AKI-associated pathway NOTCH1 is a critical prognostic factor for all subgroups of type 2 but not for type 1 pRCC.

NICD1 overexpression in mouse tubular cells drives clonal type 2 papillary tumors

To verify whether NOTCH1 overexpression could trigger type 2 pRCC development, we generated Pax8.rtTA;TetO.Cre;R26.NICD1.GFP (Pax8/NICD1) transgenic mice with overexpression of the intracellular portion of the Notch1 gene (NICD1) in Pax8+ TECs after induction by administration of doxycycline in drinking water (fig. S5A). To verify the effective activation of the transgenes and subsequent expression of the NICD1 and green fluorescent protein (GFP) proteins, we stained Pax8/NICD1 kidney sections with an anti-GFP antibody. We observed that TECs expressed high amounts of GFP, whereas interstitial cells remained negative (fig. S5B). We analyzed the animals at 4, 12, and 36 weeks after transgene induction (fig. S5C). Blood urea nitrogen (BUN) measurements revealed a progressive decline of renal excretory function in induced versus uninduced Pax8/NICD1 mice (fig. S5D). Kidney weights and volumes were higher and increased over time in induced Pax8/NICD1 mice (fig. S5, E and F). Histological analysis revealed cystic changes (fig. S5G), with or without epithelial hyperplasia or tumors (fig. S5, H and I). In particular, kidneys of Pax8/NICD1 mice developed numerous papillary tumors in the cortex and outer stripe of the outer medulla at 12 weeks that increased in size, number, and complexity over time (fig. S5, H and I). The uninduced Pax8/NICD1 mice did not develop CKD, renomegaly, or other lesions (fig. S5J).

Animals that were homozygous for the R26.NICD1.GFP effector transgene (Pax8/NICD1+/+) had significantly more pAdenomas/pRCCs than heterozygotes (Pax8/NICD1+/−) at 24 to 36 weeks (P = 0.03 and P = 0.005, respectively), suggesting that high NICD1 amounts are critical for generation of papillary tumors (fig. S5K).

All Pax8/NICD1 mice had CKD (fig. S5D), and 93.75% developed papillary tumors at 36 weeks (Fig. 4A), reaching up to 1.6 mm in size (Fig. 4B) and detectable by power Doppler imaging as vascularized masses (Fig. 4C). Papillary tumors consisted of both adenomas (Fig. 4D) and pRCC (Fig. 4, E and F) but were only type 2 (100% of the tumors). Papillary tumors were negative for carbonic anhydrase IX (Fig. 4G), as reported in human pRCC (13). Strong BCL-2 expression was detected, suggesting resistance to cell death (Fig. 4H).

Fig. 4 NICD1 overexpression in mouse tubular cells induces clonal type 2 papillary tumors.

(A) Percentages of Pax8/NICD1 mice at 4 weeks (n = 8), 12 weeks (n = 10), and 36 weeks (n = 16) presenting with papillary adenomas (pAs) and pRCCs. Statistical power was assessed using Fisher's exact test: a versus c, P = 0.004; a versus e, P = 0.0001; c versus e, P = 0.2642; and b versus d, P = 0.0128; b versus f, P = 0.002; d versus f, P = 0.6924. (B) Length (micrometers) of the lesions in Pax8/NICD1 mouse kidneys at 36 weeks (n = 5 mice). Bar (red) indicates median value. (C) Ultrasound image with power Doppler mode (red) of a wild-type mouse kidney (displayed in transverse plane) at 36 weeks (representative of n = 4 mice). Inset: High magnification of the ultrasound appearance of a tumor area (dashed line). Scale bars, 1 mm. (D) Representative image of H&E staining of papillary adenoma from Pax8/NICD1 mouse kidney at 36 weeks. Scale bar, 25 μm. (E) Type 2 papillary carcinoma from Pax8/NICD1 mouse kidney at 36 weeks. Scale bar, 100 μm. (F) H&E staining of foamy macrophages within a papillary lesion from Pax8/NICD1 mouse kidney at 36 weeks. Scale bar, 25 μm. (G) Representative image of a carbonic anhydrase IX (CAIX) negative papillary tumor (brown) from Pax8/NICD1 mouse kidney at 36 weeks. Scale bar, 25 μm. (H) BCL-2–positive papillary tumor (brown) from Pax8/NICD1 mouse kidney at 36 weeks. Scale bar, 25 μm. (I) Representative type 2 papillary tumor from Pax8/NICD1/Confetti mouse kidney at 36 weeks. Scale bar, 25 μm. (J) Representative type 2 papillary tumor from Pax8/NICD1/Confetti mouse kidney at 36 weeks. Scale bar, 25 μm. (K) Clonal analysis of papillary tumors in Pax8/NICD1/Confetti mice at 36 weeks (n = 6 mice). Bars indicate mean values. (L and M) 3D reconstruction of representative lesions in Pax8/NICD1/Confetti kidney. (L) COL4 staining (cyan). Arrows indicate invasion. (M) VWF staining (purple). Arrows indicate vascularization of the tumor. Scale bars, 100 μm. (N and O) Low-power images of a papillary tumor with rhabdoid features. (N) H&E staining. (O) PAX2 staining (brown). Scale bars, 100 μm. (P to T) Histologic images of a subcutaneous metastasis. (P) H&E staining. In the upper part of the image, subcutaneous adipose tissue; in the lower part, metastatic tissue with rhabdoid features. (Q) PAX2 staining (brown). (R) Vimentin staining (brown). (S) MUC1 staining (brown). (T) PanCK staining (brown). Scale bars, 25 μm. DAPI (white) counterstains nuclei. Signals for fluorescent Confetti proteins are as follows: GFP, green; RFP, red; YFP, yellow; and CFP, blue.

We then evaluated the clonal composition of NOTCH1 overexpression–induced type 2 papillary tumors by generating Pax8.rtTA;TetO.Cre;R26.NICD1.GFP;R26.Confetti (Pax8/NICD1/Confetti) mice (fig. S5L). After induction, clonal analysis revealed that over 80% of the papillary tumors in Pax8/NICD1/Confetti mice at 36 weeks were either mono- or biclonal (Fig. 4, I to K, and fig. S5, M to O). 3D analysis confirmed that multiple papillary tumors were interconnected and generated by a clonal population spreading from the primary pRCC and invading the surrounding kidney tissue (Fig. 4L and movie S1). Primary tumor masses also showed vascularization (Fig. 4M).

Tumors invading the surrounding kidney tissue or the kidney capsule and/or metastasizing at distant sites were observed in all (100%) mice. Metastasis was associated with areas that showed rhabdoid-like morphology, characteristic of high-grade pRCC (23), within the dominant pRCC mass (Fig. 4, N and O), as well as in the metastasis (Fig. 4P). Metastases were also positive for PAX2, confirming their renal origin (Fig. 4Q). Cytokeratins (PanCK), vimentin, and epithelial membrane antigen [Mucin 1, cell surface associated (MUC1)] were also expressed, as reported for rhabdoid variants of human RCC metastases (Fig. 4, R to T) (23). These results suggest that NOTCH1 overexpression promotes development of pAdenomas and highly aggressive type 2 pRCCs derived from clonal amplification of single TECs.

AKI accelerates clonal type 2 papillary tumor development in NOTCH1-overexpressing mice

To investigate whether AKI could also act as a tumor-accelerating factor in Pax8/NICD1 mice, we performed unilateral IRI in induced and uninduced Pax8/NICD1 mice (fig. S6A). Although activated NOTCH1 expression was virtually absent in uninduced healthy mice (Fig. 5A), IRI alone induced strong NOTCH1 activation in uninduced mice 1 day after injury (Fig. 5B); it remained higher compared with controls thereafter (Fig. 5C). Strong nuclear NOTCH1 immunostaining was seen, indicating diffuse NOTCH1 activation in TECs of Pax8/NICD1 mice (Fig. 5D). AKI increased NOTCH1 in Pax8/NICD1 mice (Fig. 5, E and F). We then checked whether AKI could accelerate pRCC formation when acting together with NICD1 overexpression. Expectedly, the combination of IRI and NICD1 overexpression in induced Pax8/NICD1 and Pax8/NICD1/Confetti mice accelerated the formation of cysts, pAdenomas, CKD (fig. S6B), and pRCCs in just 4 weeks (Fig. 5, G to I, and fig S6, C and D). pRCC included foamy macrophages (Fig. 5G), stained positive for PAX2 (Fig. 5H), similar to human pRCCs, and were mostly of type 2 (Fig. 5J). By contrast, no tumors were observed in uninduced Pax8/NICD1 or Pax8/Confetti mice with AKI (fig. S6, E and F) or in Pax8/NICD1 and Pax8/NICD1/Confetti mice without AKI at 4 weeks (Fig. 5I and fig. S6, G and H). Papillary tumors induced by AKI were clonal (Fig. 5, K and L). 3D analysis confirmed that papillary tumors showed vascularization and clonal population spread from the primary tumors into the surrounding parenchyma (Fig. 5, M and N). These results establish AKI as an accelerating factor for papillary tumors.

Fig. 5 AKI accelerates clonal type 2 papillary tumor development in mice.

(A) Representative image of activated NOTCH1 staining (brown) and lotus tetraglonolobus lectin (LTA; blue) in healthy uninduced Pax8/NICD1 mouse kidney at 1 day. Scale bar, 25 μm. (B and C) Representative images of activated NOTCH1 staining (brown) and LTA (blue) in uninduced Pax8/NICD1 mouse kidney 1 day (B) and 7 days (C) after IRI. Scale bars, 25 μm. (D) Representative image of activated NOTCH1 staining (brown) and LTA (blue) in healthy induced Pax8/NICD1 mouse kidney at 1 day. Scale bar, 25 μm. (E) Representative image of activated NOTCH1 staining (brown) and LTA (blue) in induced Pax8/NICD1 mouse kidney 1 day after IRI. Scale bar, 25 μm. (F) Graph of the reciprocal intensity of activated NOTCH1 per nucleus in uninduced and induced Pax8/NICD1 mice 1 day after IRI (n = 4 mice for each group). a. u., arbitrary units. Bars indicate mean values. (G) Papillary tumor with foamy macrophages in the lumen, present in Pax8/NICD1 mice 4 weeks after IRI. Scale bar, 25 μm. (H) PAX2-positive papillary tumor (brown) from Pax8/NICD1 mouse kidneys 4 weeks after IRI. Scale bar, 25 μm. (I) Percentages of induced Pax8/NICD1 mice with or without IRI presenting with pAdenomas and pRCCs (pAs/pRCCs, n = 5 uninduced and n = 11 induced mice) and CKD (n = 7 uninduced and n = 5 induced mice) after 4 weeks. (J) Percentages of type 1 versus type 2 papillary tumors in Pax8/NICD1 mice 4 weeks after IRI (n = 5 mice). Bars indicate mean values. (K) Representative papillary tumor from Pax8/NICD1/Confetti mouse kidney 4 weeks after IRI. Scale bar, 25 μm. DAPI (white) counterstains nuclei. (L) Clonal analysis of papillary tumors from Pax8/NICD1/Confetti mice 4 weeks after IRI (n = 5 mice). Bars indicate mean values. (M and N) 3D reconstruction of representative lesions in Pax8/NICD1/Confetti kidneys after IRI. (M) COL4 staining (cyan). Arrow indicates invasion. (N) VWF staining (purple). Arrows indicate vascularization. DAPI (white) counterstains nuclei. Signals for fluorescent Confetti proteins are as follows: GFP, green; RFP, red; YFP, yellow; and CFP, blue. Scale bars, 25 μm. Statistical power was assessed using Fisher's exact test (I) or Mann-Whitney test (F and J); numbers above graphs represent P values.

Human renal progenitors are transformed by NOTCH1 overexpression

The monoclonality of papillary tumors as well as their association with AKI suggested a possible origin from renal progenitors. We previously reported that renal progenitors consist of a multipotent progenitor population (24) characterized by coexpression of CD133 and vascular cell adhesion molecule–1 (VCAM-1), and a committed tubular progenitor that is CD133+VCAM1 (24). CD133+VCAM-1+ cells are localized at the urinary pole of the Bowman’s capsule and represent a rare subset of the scattered CD133+ cells localized mostly in proximal tubule (Fig. 6, A and B). To evaluate the contribution of tubular cell subsets to pRCC in humans, we performed immunomagnetic sorting of CD133+ cells and stained them with the fluorescent label carboxyfluorescein diacetate succinimidyl ester (CFDA-SE). We then seeded labeled CD133+ cells in coculture with unlabeled CD133 cells in a ratio of 1:1. After 5 days of culture, the labeled cell population had undergone many cell divisions, becoming more than 98% of cultured cells at day 10 (Fig. 6C). This enrichment of CD133+ cells related to their ~50-fold higher proliferative capacity compared with CD133 cells (Fig. 6D, P = 0.00003), as also demonstrated by the high number of cell divisions observed with CFDA-SE labeling (Fig. 6E). Accordingly, overgrown traced cells were virtually all CD133+VCAM-1+ (Fig. 6F), in agreement with previous studies (24). We then overexpressed NICD1 in primary cultures of human renal progenitors by ZsGreen1-NICD1 lentivirus infection (25). In comparison to mock-infected cells (Fig. 6, G and H), NICD1-infected renal progenitors acquired an enlarged, eosinophilic cytoplasm with prominent nucleoli and nuclear grooves (Fig. 6I), typical of pRCC histology (26), and a more developed cytoskeleton (Fig. 6J). When mock-infected renal progenitors were cultured in a microfluidic device under fluid shear stress, they reconstituted a 3D tubule with a lumen (Fig. 6K), whereas NICD1 overexpression changed the orientation of renal progenitor cell divisions, leading to generation of papilla-like structures within the lumen (Fig. 6L) that elongated, filling the lumen and generating tumor-like masses (Fig. 6M).

Fig. 6 Human renal progenitors are transformed by NOTCH1 overexpression.

(A and B) Representative confocal images of healthy human kidney labeled with CD133 (green) and VCAM-1 (red) (n = 5 patients). (A) Glomerulus. (B) Tubules (arrows indicate CD133+VCAM-1+ cells). Scale bars, 25 μm. (C) Graph representing the evolution over time of the percentages of CD133+ and CD133 populations (n = 3 biological replicates). (D) Graph representing the number of CD133 and CD133+ cells over time (n = 3 biological replicates). Bars indicate median value. Statistical power was assessed using t test; number above graph represents P value. (E) Representative cell division analysis of CFDA-SE–labeled CD133+ cells at day 10, out of three biological replicates. In the graph, the far right peak indicates undivided cells, and each peak toward the left-hand side represents one cell division or generation. The percentages of cells in each division are shown in the graphs. (F) Representative fluorescence-activated cell sorting analysis for the contemporaneous expression of CD133 and VCAM-1 in CD133+ fraction (right) and corresponding isotype controls (left) (n = 3 biological replicates). IgG2a, immunoglobulin G2a. (G and H) Representative H&E (G) and confocal (H) images of human renal progenitors infected with ZsGreen1-mock (n = 3). In red, phalloidin staining. Scale bars, 25 μm. (I and J) Representative H&E (I) and confocal (J) images of human renal progenitors infected with ZsGreen1-NICD1 (n = 3) (black arrow indicates nuclear groove). In red, phalloidin staining. Scale bars, 25 μm. (K to M) 3D reconstruction of a confocal z-stack of human renal progenitors infected with either ZsGreen1-mock (K) or ZsGreen1-NICD1 (L and M) and showing tubular morphology with a lumen (K) or growth toward inside the tubule lumen (L), generating a tumor-like mass (M) (one representative image of n = 3 biological replicates each). In red, phalloidin staining. Scale bars, 50 μm. DAPI counterstains nuclei. (N) Representative confocal images of biopsies from patients with AKI (n = 4). In red, CD133; in green, activated NOTCH1. Scale bar, 25 μm. (O) Higher magnification of a representative confocal image of a biopsy from a patient with AKI (n = 4). In red, CD133; in green, activated NOTCH1. Scale bar, 25 μm. (P) UMAP representation of 3327 cells, with 1479 pRCC (blue) and 1848 human renal progenitor (orange) cells. (Q) Expression of canonical PT1 signature and renal progenitor (PROM1) genes. (R) Violin plots of scRNAseq data from pRCC (blue) (32) and from human renal progenitor cell cultures (orange), showing the expression of selected marker genes. In red, genes of the PT1 signature. VCAM1, vascular cell adhesion molecule 1; EPCAM, epithelial cell adhesion molecule; SLC17A3, solute carrier family 17 member 3; and SLC7A13, solute carrier family 7 member 13; CFDA-SE, carboxyfluorescein diacetate succinimidyl ester.

NOTCH1 overexpression induces renal progenitor proliferation and aberrant mitosis

We next analyzed how NOTCH1 overexpression could promote renal progenitor transformation. We previously reported that the Notch pathway is active in renal progenitors in culture and controls their proliferation (25). NICD1 infection further enhanced renal progenitor proliferation (fig. S7A). Mitoses in mock-infected cultures were normal (fig. S7, B and C), whereas NICD1-infected progenitors frequently showed abnormal mitotic spindle distribution (fig. S7, D and E). These results suggest that NOTCH1 activation induced normal renal progenitor mitosis, whereas NOTCH1 hyperactivation disrupted cell polarity signaling, leading to a notable number of aberrant mitoses (fig. S7, B to F). Transcriptome analysis identified numerous genes differentially expressed between NICD1-infected renal and mock-infected progenitor cells (fig. S7G). Gene ontology analysis demonstrated down-regulation in NICD1-infected progenitors of several pathways involved in checkpoints and/or mitotic spindle control, including E2F targets (27), G2M checkpoint, and MYC targets V1 and V2 (fig. S7H) (28). In particular, several genes that control mitotic spindle polarity, such as INSC, DYNC2H1, PRKAA2, DLG1, PARD3, PARD6G, STK33, HNF1B, and CCNE1, were deregulated (fig. S7I) (29, 30). Of note, survival analysis of 285 patients with pRCC from the Human Protein Atlas (31) (www.proteinatlas.org) showed that the modulation of transcripts of some of these mitotic spindle genes in tumor tissue of patients with pRCCs correlated with reduced survival (fig. S7, J to M). Together, NOTCH1 overactivation deregulates the expression of genes controlling mitotic spindle polarity and promotes renal progenitor proliferation and aberrant mitosis.

Renal progenitors express NOTCH1 after AKI and represent the cell of origin of pRCC in humans

To evaluate NOTCH1 expression in humans after AKI, we analyzed biopsies of four patients. Patients with AKI showed nuclear expression of NOTCH1 in 77 ± 4.7% of CD133+ versus 51.3 ± 6.9% of CD133 tubular cells (Fig. 6, N and O; P = 0.03).

Recently, Young et al. (32) identified the putative cell of origin of pRCCs using single-cell RNA sequencing (scRNAseq) as a rare VCAM-1+ SLC17A3+ (solute carrier family 17 member 3+) SLC7A13 (solute carrier family 7 member 13) proximal tubular cell. This cell cluster signature was defined as PT1. We then analyzed their pRCC scRNAseq dataset together with scRNAseq data obtained from wild-type human renal progenitor cell cultures from our laboratory to evaluate whether they represented the same population. After correcting for batch effects by matching mutual nearest neighbors (33), we reduced dimensionality by running a principal components analysis on the most highly variable genes, performed graph-based clustering on the significant (P < 0.05) principal components, and, lastly, visualized distinct cell subgroups using uniform manifold approximation and projection (UMAP) algorithm. Unsupervised clustering of the entire pooled dataset identified seven transcriptionally distinct populations (Fig. 6P), four deriving from Young’s (1, 3, 4, and 5) and three from our data (0, 2, and 6). The pRCC cell cluster 5 characterized by PT1 signature markers clustered together with the renal progenitor cell clusters 0 and 6, which showed a similar pattern of mRNA expression (Fig. 6, Q and R). All three clusters expressed CD133 (PROM1) and PAX2, as well as the epithelial cell adhesion molecule epithelial cell adhesion molecule (EPCAM), further suggesting that the cell of origin of pRCC corresponded to renal progenitor cells. Together, these results suggest that NOTCH1 overexpression may induce pRCC and papillary tumor-like growth from human renal progenitors.

AKI induces papillary tumors by promoting clonal expansion of renal progenitors in mice

To evaluate the role of renal progenitors in pRCCs in vivo, we used Pax2.rtTA;TetO.Cre transgenic mice, which specifically track renal progenitors in the mouse kidney, where CD133 cannot be used because it is species specific (34, 35). For this purpose, we generated conditional Pax2.rtTA;TetO.Cre;R26.NICD1.GFP (Pax2/NICD1) and Pax2.rtTA;TetO.Cre;R26.NICD1.GFP; R26.Confetti (Pax2/NICD1/Confetti) mice (fig. S8A), allowing NICD1 overexpression in the Pax2+ subpopulation of TECs (renal progenitors) and their clonal tracing upon doxycycline induction (fig. S8B), as previously described (34, 35). To verify the effective activation of the transgenes, we stained Pax2/NICD1 kidney sections with an anti-GFP antibody (fig. S8C). We observed positive cells in the Bowman’s capsule and scattered in the tubules, consistent with Pax2+ cell distribution (34, 35). Four weeks after IRI, more than 85% of induced Pax2/NICD1 mice presented with type 2 papillary tumors that stained positive for BCL-2 (Fig. 7, A to E, and fig. S8, D and E), similar to those observed in Pax8/NICD1 mice at 4 weeks after IRI (Fig. 5). Most of the papillary lesions were of type 2 (Fig. 7, A to D and F). Histological analysis of uninduced Pax2/NICD1 mice 4 weeks after IRI and Pax2/NICD1 mice without AKI did not show neoplastic lesions (fig. S8, F and G). 3D analysis confirmed papillary tumor vascularization and clonal invasive tumor growth in the surrounding parenchyma (Fig. 7, C and D). Clonal analysis of papillary tumors in Pax2/NICD1/Confetti mice 4 weeks after IRI (Fig. 7, B to D, and fig. S8H) revealed that 88% of pAdenomas/pRCCs were monoclonal, the rest being bi- and triclonal (Fig. 7G). A direct quantitative comparison of type 2 papillary tumors in Pax8/NICD1 and Pax2/NICD1 4 weeks after IRI showed similar numbers of these lesions in both mice (Fig. 7H). In addition, no CKD was seen in Pax2/NICD1 mice after 4 weeks (Fig. 7E and fig. S8I), whereas a large proportion of Pax8/NICD1 mice showed reduced kidney function (Fig. 7E and fig. S6B). To further verify whether the role of AKI in papillary tumor induction relates to endogenous NOTCH1 activation, we induced IRI in Pax2/NICD1/Confetti mice and treated them for 1 week with N-[N-(3,5-difluorophenacetyl)-l-alanyl]-S-phenylglycine t-butyl ester (DAPT). In this setting, DAPT blocks only the endogenous AKI-induced NOTCH1 activation, but not the transgene-derived NICD1. DAPT-treated Pax2/NICD1/Confetti mice developed significantly (P = 0.008) fewer papillary tumors, compared with vehicle-treated mice 4 weeks after IRI (Fig. 7I). These results suggest that single renal progenitors are the source of papillary tumors induced by NOTCH1 overexpression.

Fig. 7 AKI induces papillary tumors by promoting clonal expansion of renal progenitors in mice.

(A) Representative histological image with H&E staining of a type 2 papillary carcinoma present in Pax2/NICD1 mice 4 weeks after IRI. Scale bar, 25 μm. (B) Representative confocal image of type 2 papillary tumor of Pax2/NICD1/Confetti mice 4 weeks after IRI. Scale bar, 25 μm. (C and D) 3D reconstruction of representative lesions in Pax2/NICD1/Confetti kidneys 4 weeks after IRI. (C) COL4 staining (cyan). Arrows indicate invasion. (D) VWF staining (purple). Arrows indicate vascularization of the tumor. Scale bars, 25 μm. (E) Percentages of Pax2/NICD1 and Pax8/NICD1 mice presenting with pAdenomas and pRCCs (pAs/pRCCs) (n = 11 for Pax2/NICD1 and n = 11 for Pax8/NICD1) and CKD (n = 6 for Pax2/NICD1 and n = 5 for Pax8/NICD1) 4 weeks after IRI. (F) Percentages of type 1 versus type 2 papillary tumors in Pax2/NICD1 mice 4 weeks after IRI (n = 6 mice). Bars indicate mean values. (G) Clonal analysis of papillary tumors in each Pax2/NICD1/Confetti mouse 4 weeks after IRI (n = 7 mice). Bars indicate mean value. (H) Graph of the number of papillary tumors per kidney section in Pax2/NICD1 (n = 6) and Pax8/NICD1 (n = 6) mice 4 weeks after IRI. Bars indicate mean values. No significant difference between the groups. (I) Graph of the number of papillary tumors per kidney section in Pax2/NICD1/Confetti mice treated with vehicle (n = 5 mice) or DAPT (n = 5 mice) and euthanized 4 weeks after IRI. Bars indicate mean values. (J) Representative confocal image of activated NOTCH1 (cyan) in Pax2/Confetti mice 2 days after IRI (n = 4). Arrows indicate Pax2+NOTCH1+ cells. Scale bar, 25 μm. (K) Percentages of Pax2/Confetti mice at 4 weeks (n = 8), 12 weeks (n = 3), and 36 weeks (n = 9) after IRI presenting with papillary adenomas (pAs) and pRCCs. (L) Clone frequency analysis in papillary tumors of Pax2/Confetti mouse kidneys at 36 weeks (n = 7 mice). Bars indicate mean values. (M) Percentages of type 1 versus type 2 papillary tumors in Pax2/Confetti mice at 36 weeks (n = 5 mice). Bars indicate mean values. (N) Graph of the numbers of papillary tumors per kidney section in Pax2/Confetti mice treated with vehicle (n = 7) or DAPT (n = 5) and euthanized 36 weeks after IRI. Bars indicate mean values. (O and P) Representative confocal images of type 1 (O) and type 2 (P) papillary tumors of Pax2/Confetti mice 36 weeks after IRI. Scale bars, 25 μm. (Q and R) 3D reconstruction of representative lesions in Pax2/Confetti kidney 4 weeks after IRI. (Q) COL4 staining (cyan). Arrows indicate invasion. (R) VWF staining (purple). Arrows indicate vascularization of the tumor. Scale bars, 25 μm. DAPI (white) counterstains nuclei. Signals for fluorescent Confetti proteins are as follows: GFP, green; RFP, red; YFP, yellow; and CFP, blue. Data are means ± SEM. Statistical power was assessed using Fisher's exact test (E and K) or Mann-Whitney test (F, H, I, M, and N); numbers above graphs represent P values.

To address whether clonal tumors developing spontaneously in mice after AKI also originated from renal progenitors, we first confirmed that in Pax2/Confetti mice NOTCH1 was expressed by renal progenitors at a percentage significantly higher than that of Pax2 cells after AKI (P = 0.03) and of renal progenitors in healthy mice (P = 0.03) (fig. S9A). A representative staining is shown in Fig. 7J. We then performed lineage tracing for 36 weeks after IRI induction using Pax2/Confetti mice (Fig. 7, K to R). Papillary tumors spontaneously developed in Pax2/Confetti mice after IRI (Fig. 7K). As previously observed in Pax8/Confetti mice (Fig. 2M), Pax2/Confetti mice showed a clonal origin (Fig. 7L). Type 1 papillary tumors were prevalent (Fig. 7M). DAPT-treated Pax2/Confetti mice developed significantly (P = 0.03) fewer papillary tumors, compared with vehicle-treated mice 4 weeks after IRI (Fig. 7N). This reduction was related to a decrease in type 2 papillary tumors (fig. S9, B and C). 3D analysis confirmed that papillary tumors showed vascularization and that clonal populations spread from the primary tumors into the surrounding parenchyma, promoting invasive clonal tumor growth (Fig. 7, Q and R). These results suggest that papillary tumors that develop after AKI derive from clonal proliferation of single renal progenitors in a Notch-dependent manner.

DISCUSSION

We hypothesized that AKI may drive the development of one or more subtypes of RCC. Our results reveal that (i) AKI is a risk factor for pRCC development and for pRCC recurrence in humans in different cohorts. (ii) AKI promotes long-term development of papillary tumors in mice by activating tumor growth–promoting pathways in a type 1– or type 2–specific manner. (iii) Overexpression of the AKI-associated pathway NOTCH1 is a specific feature of type 2 pRCC, which correlates with outcome in humans, and drives type 2 papillary tumors in an inducible mouse model of pRCC. (iv) pRCC originates from clonal proliferation of renal progenitors—the cellular source of tubule regeneration upon AKI—in a classical adenoma-carcinoma sequence, leading to invasive pRCC growth and even remote metastasis.

AKI, and particularly IRI, is common and often overlooked. In addition, patients with AKI have a reduced life expectancy (36), which may explain why the link between AKI and pRCC has remained unrecognized until now (6). Autopsy studies suggest that pAdenomas are common, with a prevalence ranging from 5 to 10% before the age of 40 and rising to almost 40% above the age of 70 (37). In contrast, pRCCs are considerably rarer, with the exception of patients on dialysis or with a kidney transplant, two patient categories at high risk of AKI (38, 39), where prevalence increases up to 100 times, reaching 2 to 3% (40). These epidemiologic data are in line with our findings in mice, where pAdenomas were observed in over half of the animals 12 weeks after an AKI episode, whereas pRCCs appeared only 36 weeks after AKI and in only 16% of the animals. These results suggest that the adenoma-carcinoma transformation is a stochastic event that develops only in a minority of patients with pAdenomas and requires long periods of time and likely additional factors to develop. As further evidence that AKI is an important risk factor for pRCC, postoperative AKI episodes also directly increased the risk for pRCC relapse. These results suggest that all patients at high risk for pRCC, that is, patients carrying germinal mutations in pRCC-related genes or patients undergoing pRCC surgery, require rigorous measures to prevent AKI episodes. Once AKI has occurred, careful follow-up for pRCC relapse would be required. These findings may be particularly relevant for dialysis patients or patients with kidney transplant in whom pRCC represents an important cause of death (41). Our results suggest that this risk may relate to AKI episodes that are inherent to advanced CKD and kidney transplantation (5).

How could AKI promote development of pRCC? Our analysis of human tumor samples reported by TCGA study (20) showed that some AKI-associated signaling pathways that critically control healing responses to kidney injury are overexpressed in pRCCs and predict tumor severity and patient survival. Types 1 and 2 pRCCs reveal associations with different pathways, suggesting that these subtypes are promoted by AKI through activation of distinct signaling pathways involved in cell growth control. To verify this hypothesis, we tried to reproduce papillary tumors by overexpressing one of these AKI-associated pathways in mice. We chose to overexpress the NOTCH1 pathway as a proof of concept because our analysis of patients from TCGA study (20) found NOTCH1 overexpression to predict disease severity and outcome in pRCC and to be a specific feature of all three molecular subgroups of type 2 pRCC independently of causative germline or somatic mutations. In agreement with our hypothesis, NOTCH1 overexpression in TECs induced pAdenomas and type 2 pRCC development, and blocking NOTCH hyperactivation after AKI prevented their development in mice. Conversely, a previous study showed that knocking out Notch1 in the kidney epithelium during development can lead to generation of type 1 pAdenomas at birth (42). Consistent with this result, we also found that type 1 papillary tumors had low amounts of NOTCH1 and were not caused by NOTCH overexpression. However, both NOTCH under- and overexpression determine malfunctioning of the mitotic spindle and hyperproliferation (43), as also observed in this study. Together, these results suggest that environmental or genetic factors that promote hypo- or hyperactivation of signaling pathways that control tubular cell proliferation, such as the Notch pathway, can drive papillary tumors. Previous studies reported that NOTCH overexpression in TECs promotes CKD, increased proliferation, and even epithelial dysplasia, but not RCC (4446). However, in these studies, (i) mice were studied in the absence of AKI and only 4 to 8 weeks after transgene activation (44, 45), a time that is not sufficient for tumor development, in agreement with our observations; and (ii) Notch transgene was overexpressed under the control of a promoter (46) that is not expressed in renal progenitors as previously reported (47). Our results demonstrate that NOTCH1 activation induces RCC only after overexpression specifically in renal progenitors.

Until now, the cell of origin for pRCC has been unknown. Because the renal tubule contains 10 to 20 different epithelial cell subtypes (48), not every TEC may be amenable to cancer transformation, making identification of the cell type of pAdenoma/pRCC origin a challenging task. Immunolabeling and mRNA profiling support a possible proximal tubular origin (49, 50). Our lineage tracing studies indicate that pAdenoma/pRCC sequentially develop from an aberrant clonal expansion of renal progenitors occurring after AKI or after genetically induced changes of an AKI-activated pathway specifically in renal progenitors. These results demonstrate a crucial role of the cell of origin in determining the renal papillary tumor type. Our results also provide a possible explanation for recent results obtained from scRNAseq, suggesting the similarity of pRCCs with a scattered, uncharacterized population of tubular cells expressing VCAM-1, defined as PT1 (32). We previously reported that a subset of CD133+ renal progenitors in humans has self-renewal and clonogenic capacity and is VCAM-1+ (3, 24). Our scRNAseq analysis indicated that the PT1 signature closely resembled CD133+ human renal progenitor cells, further confirming that this is the cell of origin of pRCC. In this setting, AKI-induced hypoxia may also be critical. Hypoxia promotes the undifferentiated cell state in various stem and precursor populations by activating Notch-responsive promoters and increasing expression of genes directly downstream of Notch (51). In turn, NOTCH also drives a hypoxia-inducible factor a (HIF1a)–to–HIF2a switch that favors transformation into cancer stem cells and supports tumor growth (52).

Our study also has some limitations. In mouse models, we tested only IRI, the most frequent type of AKI in clinical settings (6). Other studies are needed to determine whether other types of AKI, such as toxic or obstructive, also promote papillary tumor development in animal models. In addition, we did not assess whether the risk of pRCC was related to the severity of AKI. Other studies are needed to address this important point.

In summary, the results of this study provide evidence that acute organ injury can drive tumorigenesis and, in particular, that AKI promotes the development of pRCC from single renal progenitors through a classical adenoma-carcinoma sequence. The process is initiated by overactivation of AKI-associated pathways that promote clonal expansion of renal progenitors as an overshooting healing response. We identify NOTCH1 overexpression as a common pathomechanism of type 2 pRCC formation, suggesting that this pathway could be targeted in this RCC type. Last, the discovery of AKI as a risk factor for pRCC development and relapse suggests that pRCCs unrelated to known risk factors may actually relate to previous AKI episodes, requiring AKI prevention and risk stratification programs.

MATERIALS AND METHODS

Study design

The purpose of this study was to examine the effect of AKI on RCC development and to identify the mechanisms involved in this process. These objectives were addressed by (i) analyzing AKI insults as a risk factor for RCC development in humans, (ii) analyzing the effect of postoperative AKI related to tumor resection in promoting pRCC recurrence in patients, (iii) confirming AKI as a risk factor for pRCC development in mice, and evaluating pRCC clonality, (iv) analyzing the AKI-related pathways that are involved in tumor progression in human pRCC tissues, (v) evaluating the effect of the AKI-related pathway NICD1 on pRCC formation in a transgenic mouse model of pRCC, (vi) evaluating the transforming effect of NICD1 overexpression on human renal progenitors, the critical cell for the response to AKI, and (vii) evaluating renal progenitors as the cell of origin of pRCC in NICD1 overexpressing transgenic mouse and/or in response to AKI. For all clinical studies, all eligible patients from the available observation period were included without previous sample size calculation and were assigned to predefined clinical groups (control/pRCC/ccRCC and AKI/non-AKI) after retrospective clinical adjudication, as per study design.

Single-center study. We analyzed the incidence of previous AKI episodes in the RCC population irrespectively of the histological diagnosis. All patients who underwent surgery for RCC at the Careggi University Hospital, Florence, Italy, from 2012 to 2018 (fig. S1) with available clinical and laboratory information and a follow-up of at least 1 year before tumor diagnosis were included in the analysis. Patients were excluded if no detailed information about the tumor clinical stage was available and in case of TNM (T describes the size of the primary tumor and whether it has invaded nearby tissue; N describes regional lymph nodes that are involved; M describes distant metastasis) stage 3 or 4. Historical clinical records for laboratory, surgical, and histological data of patients with RCC were retrospectively and independently collected by two blinded observers. We evaluated the exposure to previous AKI episodes as an independent risk factor for the development of pRCC or ccRCC. AKI that occurred in the 6 months before surgery were excluded from the analysis. When at least 1 year of follow-up after surgery was available (fig. S1), we evaluated the association between postoperative AKI and tumor recurrence both in a survival analysis and in a multivariate model. This study was approved by the Ethical Committee on Human Experimentation of the Azienda Ospedaliero-Universitaria Careggi, Florence, Italy.

Danish National Registry of Patients. A Danish nationwide cohort study using population-based medical databases (11, 12) between 1977 and 2006 was screened for the association between AKI and RCC (fig. S2). All patients diagnosed with AKI were included in the analysis. Data were extracted by an investigator blinded to the purpose of the analysis. The study was approved by the Danish Data Protection Agency (record no. 2005-41-5281).

Multicentric study. We analyzed the Italian Registry of Conservative Renal Surgery (RECORd1 project) of the Italian Society of Urology (fig. S3) (18) to confirm the association between postoperative AKI and tumor recurrence both in a survival analysis and in a multivariate model. The RECORd1 project includes all patients who underwent conservative surgical treatment for radiologically diagnosed RCC between January 2009 and December 2012 at 19 urological Italian centers. We included all patients with a histologic diagnosis of RCC and with a minimum 1-year oncologic follow-up (fig. S3). Patients were excluded if no detailed information about the tumor clinical stage was available and in case of TNM stage 3 or 4. Data were extracted by an investigator blinded to the purpose of the analysis. The study protocol was approved by the local ethical committee and patients’ acceptance of the written informed consent.

TCGA Research Network. We analyzed gender, age, smoking exposure, and mutations on genes previously described as specific for pRCC in relation to expression of NOTCH1 and other pathways associated with AKI (20) in all available patients with pRCC.

Patient dataset from the Human Protein Atlas. We analyzed the survival probability of all patients with pRCC from the Human Protein Atlas (31) (www.proteinatlas.org) based on PRKAA2, PARD6G, HNF1B, and CCNE1 expression.

Patient biopsies. A total of five normal-appearing kidneys (three men and two women; mean age, 64.8 ± 12.4 years) and four kidney biopsies from patients with AKI (two men and two women; mean age, 58.8 ± 5.1 years; time after AKI: 2, 2, 14, and 15 days) were analyzed in agreement with the Ethical Committee on Human Experimentation of the Azienda Ospedaliero-Universitaria Careggi, Florence, Italy. Normal-appearing kidney fragments were obtained from the pole opposite to the tumor of patients who underwent nephrectomy for localized renal tumors.

Animal studies. Mice were randomly assigned to experimental and control groups, but investigators were not blinded. Sample size was calculated on the basis of the primary endpoint rate of tumor formation and other assumptions based on preliminary experiments. No data were excluded from studies in this manuscript. Pathology analysis was performed in a blinded fashion. All procedures were performed in accordance with institutional protocols approved by the Institutional Animal Care and Use Committee of the University of Florence, Italy.

Statistical analysis

Human studies. Results are reported as means ± SD unless otherwise stated. Between-group comparisons for continuous variables were performed by Student’s t test for unpaired data and Mann-Whitney U test according to a normal or nonparametric distribution. Frequencies among groups were compared by chi-squared test for categorical variables, applying Fisher’s exact test when appropriate (Table 1 and tables S1 to S3). A two-sided P value of <0.05 was considered statistically significant.

To test the association between previous AKI episodes and postoperative AKI with pRCC development and recurrence, respectively, binary logistic regression models were constructed. pRCC or ccRCC development and pRCC recurrence served as outcome or dependent variables. Covariates or explanatory variables were treated as categorical (AKI, gender, CKD, DM, and clinical T stage) or continuous (age, RENAL score) as appropriate. The results are presented as OR and 95% CI. A Wald test statistics P value of <0.05 was considered statistically significant. For survival analyses, Kaplan-Meier estimates for postoperative AKI and non-AKI patients were compared by log-rank test with statistical significance set for a P value lower than 0.05. Cumulative 4-year recurrence-free survival and standard error are reported.

Kaplan-Meier estimates were used to generate overall survival curves for

1) NOTCH1, PTEN, and SMAD4 reverse-phase protein array (RPPA) amounts in patients affected by pRCC, or NOTCH1 (using 50th percentile as a cutoff value) in patients affected by type 1 or 2 pRCC obtained from RPPA data from TCGA Research Network (20).

2) PRKAA2, PARD6G, HNF1B, and CCNE1 expression (using best expression cutoff) in patients affected by pRCC from the Human Protein Atlas (31).

Differences between groups were assessed by log-rank test. For each Kaplan-Meier plot, a Cox regression model was constructed to provide hazard ratio and 95% CI. To test the independent effects of different variables on the prognosis of pRCC, a multiple logistic regression model was constructed by entering TNM stages >2 as dependent variable, whereas all parameters previously described as risk factors for pRCC served as covariates.

Animal and in vitro studies. The number of mice used, numbers of replicates, and statistical values (where applicable) are provided in the figure legends. Results were expressed as means ± SEM. Comparison between groups was performed by the Mann-Whitney test or through the analysis of variance (ANOVA) for multiple comparisons (ANOVA for repeated measures) with Bonferroni post hoc analysis, with the Fisher's exact test (two-tailed P value) or with t test (two-tailed P value). A P value <0.05 was considered statistically significant. Statistical analysis was performed using SPSS (RRID:SCR_002865), OriginPro (RRID:SCR_015636), and Stata (RRID:SCR_012763) statistical software.

SUPPLEMENTARY MATERIALS

stm.sciencemag.org/cgi/content/full/12/536/eaaw6003/DC1

Supplementary Materials and Methods

Fig. S1. The flowchart illustrates the selection criteria for the single-center study population.

Fig. S2. The flowchart illustrates the selection criteria for the study population from the Danish National Registry of Patients.

Fig. S3. The flowchart illustrates the selection criteria for the multicentric study population.

Fig. S4. AKI-driven papillary tumors originate from clonal proliferation of single TECs in mice.

Fig. S5. NICD1 overexpression in mouse tubular cells induces papillary tumors and CKD.

Fig. S6. AKI accelerates clonal type 2 papillary tumor development in mice.

Fig. S7. NOTCH1 overexpression enhances cell division and promotes aberrant mitosis.

Fig. S8. AKI induces papillary tumors by promoting clonal expansion of renal progenitors in mice.

Fig. S9. AKI induces papillary tumors from renal progenitors in a NOTCH1-dependent manner.

Table S1. Postoperative AKI is a risk factor for the development of pRCC recurrence in humans in a single-center study.

Table S2. The preoperative characteristics of 594 patients treated with partial nephrectomy for kidney tumors and clinical details of 89 patients with pRCC (the RECORd1 project).

Table S3. Postoperative AKI is a risk factor for the development of pRCC recurrence in humans in a multicentric study (RECORd1 project).

Table S4. AKI-activated pathways are overexpressed in human pRCC in a subtype-specific manner.

Table S5. Multivariate analysis indicates NOTCH1 as a risk factor for pRCC prognosis.

Movie S1. 3D reconstruction of a papillary tumor in Pax8/NICD1/Confetti kidney at 36 weeks.

References (5379)

REFERENCES AND NOTES

Acknowledgments: We gratefully acknowledge the contributors from the RECORd1 Italian Project: Walter Artibani, Nicola Longo, Vincenzo Mirone, Bernardo Rocco, Claudio Simeone, Riccardo Tellini, Carlo Terrone, Alessandro Volpe, and Filiberto Zattoni. We thank L. Calosi, S. Catarinicchia, and D. Guasti from the Department of Experimental and Clinical Medicine, University of Florence, Italy, for technical support with sample preparation for histology. We thank B. Schäfer, University Children's Hospital, Zurich, Switzerland, and A. Burger, University of Colorado Denver, USA, for the Pax2.rtTA mouse. Funding: This study was funded by the European Research Council under the Consolidator Grant RENOIR to P.R. (ERC-2014-CoG), grant number 648274, by the Associazione Italiana per la Ricerca sul Cancro and Fondazione CR Firenze under IG 2018—ID.21821 project—to P.R. A.J.P. is a recipient of the Fondazione Umberto Veronesi fellowship. H.-J.A. was supported by the Deutsche Forschungsgemeinschaft AN372/24-1. Author contributions: P.R. designed the study. D.F., M.G., A. Minervini, S.S., M.C., A. Mari, F.D.M., A.A., F.P., R.Schiavina, V.F., C.F.C., and R.M.R. collected patient clinical data, and M.A., F.G., C.F.C., and A. Mari analyzed them. A.J.P., G.A., M.L.A., A.S., and L.D.C. performed immunohistochemistry and confocal microscopy. A.J.P., G.A., A.S., C.C., B.M., S.N., and M.E.M. designed and carried out experiments in the mouse system. G.J.N. and M.R.R. performed the histopathological tissue analysis of the mouse hematoxylin and eosin (H&E) samples. A.J.P., G.A., M.L.A., A.S., and P.R. analyzed the data. T.L. performed ultrasound imaging. S.L. and S.G. performed scRNAseq. R.Semeraro and A. Magi analyzed the datasets. M.R., A. Mari, M.A., and F.G. performed statistical analysis. A.J.P., M.A., M.L.A., F.G., E.L., L.L., C.F.C., M.R., H.L., H.-J.A., and M.R.R. critically reviewed the manuscript. P.R. and A.J.P. wrote the paper with input from all authors. Competing interests: H.L. receives NIH consultant fees for the CUREGN consortium. H.J.A. does consulting for AstraZeneca, Janssen, GSK, Previpharma, Inositec, Secarna, and NOXXON. All other authors declare that they have no competing interests. Data and material availability: Microarray and scRNAseq data are available in NCBI's Gene Expression Omnibus repository as GSE137704 and GSE137620, respectively. All data associated with this study are present in the paper or the Supplementary Materials or are available from the Human Protein Atlas (https://www.proteinatlas.org/).

Stay Connected to Science Translational Medicine

Navigate This Article