Genetic and epigenetic profiling indicates the proximal tubule origin of renal cancers in end-stage renal disease.

End-stage renal disease (ESRD) patients on dialysis therapy have a higher incidence of renal cell carcinomas (RCCs), which consist of 2 major histopathological types: clear-cell RCCs (ESRD-ccRCCs) and acquired cystic disease (ACD)-associated RCCs. However, their genetic and epigenetic alterations are still poorly understood. Here, we investigated somatic mutations, copy number alterations (CNAs), and DNA methylation profiles in 9 ESRD-ccRCCs and 7 ACD-associated RCCs to identify their molecular alterations and cellular origins. Targeted sequencing of 409 cancer-related genes, including VHL, PBRM1, SETD2, BAP1, KDM5C, MET, KMT2C (MLL3), and TP53, showed ESRD-ccRCCs harbored frequent VHL mutations, while ACD-associated RCCs did not. CNA analysis showed that ESRD-ccRCCs had a frequent loss of chromosome 3p while ACD-associated RCCs had a gain of chromosome 16. Beadarray methylation analysis showed that ESRD-ccRCCs had methylation profiles similar to those of sporadic ccRCCs, while ACD-associated RCCs had profiles similar to those of papillary RCCs. Expression analysis of genes whose expression levels are characteristic to individual segments of a nephron showed that ESRD-ccRCCs and ACD-associated RCCs had high expression of proximal tubule cell marker genes, while chromophobe RCCs had high expression of distal tubule cell/collecting duct cell marker genes. In conclusion, ESRD-ccRCCs and ACD-associated RCCs had mutation and methylation profiles similar to those of sporadic ccRCCs and papillary RCCs, respectively, and these 2 histopathological types of RCCs were indicated to have originated from proximal tubule cells of the nephron.

INTRODUCTION

End‐stage renal disease (ESRD) patients on dialysis therapy have a 4‐ to 40‐fold higher incidence of renal cell carcinomas (RCCs) than the general population. , , , RCCs arising in ESRD generally consist of 2 major histopathological types: (a) clear‐cell renal cell carcinomas (ESRD‐ccRCCs), and (b) acquired cystic disease (ACD)‐associated RCCs. The involvement of chronic inflammation, induced by uremic toxins and oxidative stress, , and impaired immune surveillance, , has been implicated in the increased incidence. , As advanced RCCs arising in ESRD respond poorly to systemic therapy such as molecular‐targeted therapy, , a greater understanding of their pathophysiology is essential.

In sporadic ccRCCs (ie, developing in the general population), the most common histopathological type of RCCs, genetic and epigenetic alterations have been extensively investigated. , , Dysregulation of VHL is an almost universal finding, and subsequent genomic alterations of PBRM1, SETD2, KDM5C, or BAP1 are involved in the pathogenesis of sporadic ccRCCs. , Furthermore, their cellular origin is thought to be proximal tubule cells of the nephron, based on findings from immunohistochemical staining patterns and mouse ccRCC models. Conversely, alterations in RCCs arising in ESRD, namely ESRD‐ccRCCs and ACD‐associated RCCs, remain largely unexplored. With regards to ACD‐associated RCCs in particular, given that their histopathological diagnosis was only recently added to the World Health Organization (WHO) 2016 classification, , the molecular alterations and cellular origin remain unknown.

We aimed to clarify genetic and epigenetic alterations in the 2 major histopathological types of RCCs arising in ESRD, namely ESRD‐ccRCCs and ACD‐associated RCCs, and to identify their cellular origin.

MATERIALS AND METHODS

Renal cell carcinoma and kidney tissue samples

Sixteen RCCs arising in ESRD, including 9 ESRD‐ccRCCs and 7 ACD‐associated RCCs) (Table 1), 33 sporadic RCCs (14 ccRCCs, 8 papillary RCCs [PRCCs], 11 chromophobe RCCs [ChRCCs]), and 7 noncancerous kidney surgical specimens (Table S1) were obtained via radical or partial nephrectomy at Tokyo Women's Medical University and Tokyo Women's Medical University Medical Center East. Each specimen was split into 2 pieces to produce fresh frozen and formalin‐fixed, paraffin‐embedded (FFPE) samples. The fresh frozen samples were used for molecular analyses, and the FFPE samples were used for H&E‐staining to confirm histology and estimate cancer cell fraction by microscopic cell counting. All RCC specimens were confirmed to have cancer cell fractions ≥40%. Histopathological diagnoses were conducted by a certified and experienced pathologist (Y. N.) based on the 2016 WHO classification. In this study, ESRD was defined as a disease which required maintenance hemodialysis or peritoneal dialysis therapy, regardless of duration. This study was approved by the Institutional Review Board of the Tokyo Women's Medical University (approved no. 382) and the National Cancer Center Research Institute (approved no. 2018‐024), and all the specimens were obtained with written informed consents.

TABLE 1

Patient and tumor characteristics of the 16 RCC surgical specimens arising in ESRD

Sample ID	Histopathological type	Other tumors in the removed kidney	Sex	Age	pStage	Duration of dialysis therapy	Cause of ESRD
ESRD‐ccRCC1	Clear‐cell RCC	Papillary adenoma	Male	71	I	2 y 10 mo	Unknown
ESRD‐ccRCC2	Clear‐cell RCC	None	Male	39	I	3 mo	Unknown
ESRD‐ccRCC3 a	Clear‐cell RCC	None	Male	39	I	2 y 1 mo	IgA nephropathy
ESRD‐ccRCC4	Clear‐cell RCC	Clear‐cell RCC, clear‐cell papillary RCC	Male	74	I	17 y	Nephrosclerosis
ESRD‐ccRCC5 a	Clear‐cell RCC	None	Male	66	I	3 y 1 mo	Unknown
ESRD‐ccRCC6	Clear‐cell RCC	None	Male	62	I	11 y	Nephrosclerosis
ESRD‐ccRCC7	Clear‐cell RCC	ACD‐associated RCC	Male	64	I	10 y 5 mo	Diabetic nephropathy
ESRD‐ccRCC8	Clear‐cell RCC	Clear‐cell RCC	Male	58	I	6 mo	Diabetic nephropathy
ESRD‐ccRCC9	Clear‐cell RCC	None	Male	46	I	1 mo	Unknown
ACD‐RCC1	ACD‐associated RCC	None	Male	50	I	18 y	Focal segmental glomerular sclerosis
ACD‐RCC2	ACD‐associated RCC	None	Male	43	I	18 y	IgA nephropathy
ACD‐RCC3	ACD‐associated RCC	None	Male	42	I	18 y	IgA nephropathy
ACD‐RCC4	ACD‐associated RCC	None	Male	43	I	19 y	IgA nephropathy
ACD‐RCC5	ACD‐associated RCC	None	Male	63	III	14 y 9 mo	Diabetic nephropathy
ACD‐RCC6	ACD‐associated RCC	None	Male	47	I	15 y	Diabetic nephropathy
ACD‐RCC7	ACD‐associated RCC	Clear‐cell RCC, oncocytoma	Male	66	I	16 y	Unknown

Developed from the original non‐functioning kidney in ESRD patients treated with kidney transplantation.

In addition to our own samples, we randomly downloaded the data of 100 sporadic RCC samples (40 ccRCCs, 30 PRCCs, and 30 ChRCCs) and 25 noncancerous kidney tissue samples from The Cancer Genome Atlas (TCGA) database; and those of 2 normal kidney tissue samples (GEO accession: GSE79100) from the Gene Expression Omnibus (GEO) database (Table S2).

Analysis of somatic mutations and loss of heterozygosity

A multiplex PCR was conducted on genomic DNA for 15 991 amplicons in 409 cancer‐related genes using the Ion AmpliSeq Library Kit 2.0 with the Comprehensive Cancer Panel (Life Technologies) in accordance with the manufacturer's protocol. The 409 cancer‐related genes included VHL, PBRM1, SETD2, BAP1, KDM5C, MET, KMT2C (MLL3), and TP53, which are known to be frequently mutated in sporadic ccRCCs, , PRCCs, and ChRCCs. The synthesized library was loaded onto an Ion PI Chip v3 (Life Technologies) using Ion Chef (Thermo Fischer Scientific), and was sequenced using an Ion Proton sequencer (Life Technologies). The sequences were aligned to the human reference genome hg19 with Torrent Suite 5.0 (Life Technologies). Using CLC Genomics Workbench 8.5 (CLC bio), a variant was considered a somatic mutation in accordance with the following criteria: (a) its frequency in cancer samples was >10%, (b) its frequency in normal samples was <1%, (c) its homopolymer length was <3, (d) it was present in both at least 1 forward and 1 reverse reads, and (e) its coverage was >50. Pathogenicity of mutations was assessed using the COSMIC and ClinVar databases. When information regarding the pathogenicity of a mutation was registered in at least one of them, we considered the mutation as “pathogenic.” When not registered, we described it as “no information.”

Tumor mutation burden (TMB) was calculated based upon a previous report showing that the calculation of TMB was accurate when genomic regions ≥1.1 Mb were targeted. The TMB was calculated as the number of synonymous and nonsynonymous somatic mutations, including single nucleotide variant (SNV), multiple nucleotide variant (MNV), and small InDels per megabase.

DNA methylation analysis

Genome‐wide DNA methylation analysis was conducted using an Infinium MethylationEPIC BeadChip array (Illumina) covering 865 860 probes (862 927 CpG sites and 2932 non‐CpG sites). The array was scanned with an iScan System (Illumina) and the data were processed using the GenomeStudio Methylation Module Software (Illumina). To adjust for probe design biases, intra‐array normalization by a peak‐based correction method (BMIQ) was performed using a web tool, MACON.

To combine the DNA methylation data obtained from 2 different platforms, MethylationEPIC and Methylation450K (TCGA and GEO), we normalized DNA methylation data from the 2 platforms separately using MACON, and selected 439 031 probes common between the 2 platforms and at CpG sites on autosomes. From the 439 031 probes, we first eliminated 36 117 probes that had differences of β values (≥.1) between the 2 platforms in any of the human leukocytes from 2 healthy volunteers and in 293FT cells (data not shown). From the remaining 402 914 probes, we further excluded 84 884 probes whose β values were presented as “NA” in any of the specimens and samples. Finally, 318 030 probes were considered as eligible for genome‐wide DNA methylation analysis. VHL methylation was evaluated at a CpG site located in a CpG island (CGI) in its exon 1 (chr3: 10 183 328 based on hg19) (Table S3). CDKN2A methylation was evaluated at a CpG site located in a CGI in its exon 1 (chr9: 21 974 704) (Table S3). , A β value ≥ .3 was considered to be methylated.

Analysis of copy number alterations

Copy number alterations (CNAs) were assessed using the Infinium MethylationEPIC BeadChip array with a simplified data‐processing method used in a previous report. In brief, after the exclusion of probes located on sex chromosomes, the remaining probes were combined into a window by joining 1000 adjacent probes, resulting in 844 windows throughout the genome. A ratio of signal intensity in a window to the whole genome was first calculated in a normal sample (leukocytes from a healthy volunteer) (Sn). Next, that in a cancer sample (Sc) was calculated, and the Sc was divided by the Sn (Sc/Sn), reflecting the CNA in the window. This normalization with the intensity in the whole genome was to minimize biases of intensity between normal and cancer samples. Finally, the log2 ratio of the divided signal was calculated, and >0.15 and <−0.15 were defined as chromosomal gain and loss, respectively.

mRNA expression analysis

Quantitative RT‐PCR (qRT‐PCR) was performed in the CFX connect Real‐Time PCR Detection System (Bio‐Rad Laboratories) using SYBR Green I (BioWhittaker Molecular Applications). A copy number of a gene transcript was normalized to that of GAPDH. The primer sequences for target genes and GAPDH are shown in Table S4.

Immunohistochemistry

Immunohistochemistry was performed on FFPE specimens. For antigen retrieval, paraffin sections (3 µm thick) were autoclaved using a microwave for 15 min by adding Tris/EDTA pH 9.0. Anti‐HNF1a (1:3000 dilution; ab242140, Abcam), anti‐HNF4a (1:3000 dilution; ab201460, Abcam), and anti‐CLDN8 (1:2000 dilution, ab183738, Abcam) antibodies were used as the primary antibodies. The staining findings were evaluated by 2 pathologists (ST and YN).

Gene ontology analysis

Gene ontology (GO) analysis was conducted using a web tool, Gene Ontology enRIchment anaLysis and visuaLizAtion (GOrilla). , Enrichment of GOs in the 3 classes, namely, “Process,” “Function,” and “Component,” was separately analyzed.

Cluster analysis

Unsupervised hierarchical clustering analysis was performed using R 3.61 with the Heatplus package from Bioconductor. Dimension reduction and visualization was performed by Uniform Manifold Approximation and Projection (UMAP) [Leland Mclnnes (2018), URL https://arxiv.org/abs/1802.03426] using the R with the umap package.

Statistical analysis

Continuous and categorical variables were analyzed using the Mann‐Whitney U test and the Fisher exact test, respectively. The analyses were performed using PASW statistics version 18.0 (SPSS Japan Inc), and a difference with a P‐value of <.05 was considered statistically significant.

RESULTS

Morphologic features of RCCs arising in ESRD

Representative morphologic features of the 2 major histopathological types of RCCs arising in ESRD, namely ESRD‐ccRCCs and ACD‐associated RCCs, are shown in Figure 1. ESRD‐ccRCCs were composed of alveolar architectures of tumor cells with clear cytoplasm (Figure 1A). Also, there were interposing fine sinusoid‐like vascular networks. These histological appearances were similar to those of sporadic ccRCCs (Figure 1B). ACD‐associated RCCs were composed of papillotubular architectures (Figure 1C‐E) and such appearances resembled those in PRCCs (Figure 1F). At the same time, ACD‐associated RCCs had unique features such as frequent microcystic spaces (Figure 1C), intratumoral hemorrhage (Figure 1D), and deposition of calcium oxalate crystals (Figure 1E).

FIGURE 1

Representative histological findings of RCCs subjected to this study. A, ESRD‐ccRCC (ESRD‐ccRCC9). B, sporadic ccRCC (sporadic ccRCC4). C‐E, ACD‐associated RCCs (C: ACD‐RCC6; D: ACD‐RCC7; E: ACD‐RCC5). F, papillary RCC type II (PRCC6). ESRD‐ccRCCs (A) and sporadic ccRCCs (B) showed similar histological appearance. Both were composed of alveolar architectures of tumor cells with clear cytoplasm. Also, there were interposing fine sinusoid‐like vascular networks. ACD‐associated RCCs (C, D) and PRCCs (F) showed a morphological resemblance, namely papillotubular architectures. At the same time, ACD‐associated RCCs had unique features such as (C) frequent microcystic spaces (arrows), (D) intratumoral hemorrhage, and (E) deposition of calcium oxalate crystals (arrowheads)

Somatic mutations in RCCs arising in ESRD

In total, 16 RCCs arising in ESRD (9 ESRD‐ccRCCs and 7 ACD‐associated RCCs) were analyzed by targeted sequencing to detect somatic mutations in 409 cancer‐related genes. The 9 ESRD‐ccRCCs harbored 29 somatic mutations with amino acid changes (Table 2). Among these, VHL was most frequently mutated (n = 5), which was in line with previous studies , (Figure S1 and Table S5). To evaluate the association between duration of dialysis therapy and VHL mutation frequency, we divided the ESRD‐ccRCCs based on the duration of dialysis (long: ≥10 y; short: <10 y). However, the VHL mutation frequency was not associated with the duration of dialysis therapy (n = 2/5 [40%] vs n = 3/4 [75%], P = .524).

TABLE 2

Somatic mutations detected in the 9 ESRD‐ccRCCs

Sample ID	Gene symbol	Type	Coverage	Variant frequencies (%)	Nucleotide change	Amino acid change	Pathogenicity
ESRD‐ccRCC1	TCF3	Insertion	139	79.1	c.1167+29_1167+30insCA
	PKHD1	SNV	212	25.9	c.5743A>C	p.Asn1915His	No information
	PER1	Deletion	589	18.3	c.2681_2682delGC	p.Gly894fs	Pathogenic
	FGFR1	SNV	222	16.7	c.1285‐45C>A
	HIF1A	SNV	495	16.2	c.1362C>A
	DST	Deletion	558	14.7	c.10312_10314delAAA	p.Lys3438del	Pathogenic
	ZNF521	SNV	554	14.4	c.1628G>A	p.Arg543Gln	Pathogenic
	MLL2	Deletion	226	13.3	c.2892_2893delAG	p.Asp966fs	Pathogenic
	WRN	Deletion	371	13.2	c.725‐10_725‐8delTTC
	PKHD1	Deletion	464	12.7	c.346_347delGT	p.Val116fs	Pathogenic
	GPR124	Deletion	341	11.1	c.1139_1140delGC	p.Cys380fs	Pathogenic
	CRBN	Deletion	267	10.1	c.1119_1120delGC	p.Pro374fs	Pathogenic
ESRD‐ccRCC2	PPARG	MNV	106	27.4	c.568_569delTGinsAA	p.Cys190Asn	No information
	PPARG	SNV	102	24.5	c.571C>A	p.Gln191Lys	No information
	VHL	Deletion	705	18.0	c.399_404delTGAATT	p.Glu134_Leu135del	Pathogenic
ESRD‐ccRCC3	VHL	SNV	512	31.3	c.486C>G	p.Cys162Trp	Pathogenic
	SOX2	SNV	609	21.2	c.810C>A
ESRD‐ccRCC4	VHL	Deletion	540	41.9	c.618_619delTG	p.Ala207fs	Pathogenic
	NFE2L2	SNV	563	33.0	c.223C>A	p.Gln75Lys	No information
ESRD‐ccRCC5	RNF213	SNV	366	25.7	c.13545+9T>G
	NOTCH1	SNV	561	23.9	c.7564A>G	p.Ser2522Gly	No information
	ERBB3	SNV	743	22.5	c.2465T>C	p.Met822Thr	No information
	PDGFRA	SNV	452	21.7	c.2562+53C>A
	JAK3	SNV	551	21.1	c.2422C>G	p.Leu808Val	No information
	IKZF1	SNV	793	19.5	c.1544G>A	p.Arg515His	No information
ESRD‐ccRCC6	WRN	SNV	413	19.1	c.2184T>A
	SYNE1	SNV	1163	17.1	c.3027+43G>A
	TCF12	SNV	620	16.9	c.571C>T	p.Pro191Ser	No information
	SYK	SNV	928	13.9	c.1861G>T	p.Ala621Ser	No information
ESRD‐ccRCC7	VHL	Deletion	501	43.7	c.494delT	p.Val166fs	Pathogenic
	ERCC2	SNV	324	36.1	c.1666‐16C>T
	MYH11	SNV	381	35.4	c.654+7A>G
	JAK1	SNV	528	30.7	c.2455G>A	p.Glu819Lys	No information
	RAD50	SNV	1067	26.4	c.2322C>G
	PKHD1	SNV	816	25.0	c.11785+21T>C
	PDE4DIP	SNV	1878	14.2	c.637‐7226T>A	p.Leu221*	No information
ESRD‐ccRCC8	KDM5C	SNV	74	66.2	c.228+1G>A
	TSC2	SNV	532	37.4	c.1272G>C
ESRD‐ccRCC9	ITGB3	Deletion	147	23.8	c.913delA	p.Asn305fs	Pathogenic
	NOTCH1	MNV	1739	19.0	c.6181‐3_6181‐2delCAinsAC
	VHL	SNV	395	16.5	c.320G>C	p.Arg107Pro	Pathogenic
	MLL2	Deletion	215	14.4	c.2892_2893delAG	p.Asp966fs	Pathogenic
	ADAMTS20	SNV	150	14.0	c.670G>T	p.Asp224Tyr	Pathogenic
	MLL2	SNV	74	12.2	c.14760C>T
	PARP1	Deletion	107	12.1	c.2146delG	p.Val716fs	Pathogenic
	UBR5	Deletion	182	10.4	c.6151delG	p.Ala2051fs	Pathogenic
	NF1	SNV	97	10.3	c.4044T>C

* indicates nonsense mutations.

Two specimens (ESRD‐ccRCC1 and ESRD‐ccRCC9) harbored a MLL2 (as also known as KMT2D) mutation, as in sporadic ccRCCs. Conversely, mutations known to be present in sporadic ccRCCs, such as PBRM1, SETD2, or BAP1, were not observed in this study. A mean value of TMB was 3.07/Mb after adding 18 synonymous mutations to the 29 nonsynonymous mutations, and the most frequent base substitutions were C>A and C>T (each, n = 8), followed by C>G and T>C (each, n = 5) (Figure S2A).

The 7 ACD‐associated RCCs harbored 10 somatic mutations with amino acid changes (Table 3), but no mutations were observed in multiple specimens (Table S6). A mean value of TMB was 1.93/Mb after adding 13 synonymous mutations to the 10 nonsynonymous mutations, and the most frequent base substitution was C>T (n = 8), followed by T>C (n = 7) (Figure S2B). One specimen (ACD‐RCC5) harboring a relatively higher TMB (5.88/Mb) was in the advanced disease (Tables 1 and 3).

TABLE 3

Somatic mutations detected in the 7 ACD‐associated RCCs

Sample ID	Gene symbol	Type	Coverage	Variant frequencies (%)	Nucleotide change	Amino acid change	Pathogenicity
ACD‐RCC1	None
ACD‐RCC2	RRM1	SNV	683	30.5	c.1320+48C>T
	LRP1B	SNV	1139	25.6	c.82+55G>A
ACD‐RCC3	SYNE1	SNV	635	25.7	c.20112G>T	p.Gln6704His	No information
	MTOR	Deletion	266	24.4	c.6000delG	p.Glu2000fs	Pathogenic
ACD‐RCC4	BRD3	SNV	365	40.5	c.381A>G
ACD‐RCC5	SF3B1	SNV	559	36.1	c.3013+30T>C
	ERBB2	SNV	629	21.6	c.2844C>T
	RHOH	SNV	472	21.6	c.27G>T	p.Leu9Phe	No information
	PRKDC	SNV	559	19.9	c.4216T>C	p.Leu1405Pro, p.*1406Gln	No information
	MAGI1	SNV	567	19.0	c.*693A>G
	EPHA7	SNV	719	18.8	c.1299T>C
	NTRK3	SNV	1347	17.2	c.2066C>T	p.Ala689Val	No information
	GRM8	SNV	645	15.8	c.2678G>A	p.Ser893Asn	No information
	RALGDS	SNV	484	14.0	c.1585C>T	p.Arg474Trp, p.Arg529Trp	No information
	FANCA	SNV	574	12.0	c.187G>A	p.Glu63Lys	No information
ACD‐RCC6	TCF3	Insertion	451	73.8	c.1167+29_1167+30insCA
	FGFR3	SNV	507	35.9	c.2169‐18T>C
	NLRP1	SNV	53	11.3	c.186G>T
ACD‐RCC7	TCF3	Insertion	169	80.5	c.1167+29_1167+30insCA
	ARID2	SNV	2200	38.9	c.2582T>G	p.Ile861Ser	No information
	COL1A1	SNV	1150	36.6	c.2288G>A	p.Arg763His	Pathogenic
	CSMD3	SNV	1906	36.4	c.2409T>C
	ATR	Deletion	164	15.2	c.3172‐27_3172‐24delATTT

Copy number alterations in RCCs arising in ESRD

We assessed the genome‐wide CNAs in the 16 RCCs arising in ESRD using an Infinium MethylationEPIC array. In ESRD‐ccRCCs, the most frequent alteration was loss of chromosome 3p between 3p21.1 and 3p.25.3 (7 of 9) (Figure 2A) (windows: 149‐164, Table S7), which was in line with a previous study. The most frequent alteration in ACD‐associated RCCs was the gain of chromosome 16 (5 of 7) (Figure 2B and Table S8), which was in accordance with previous studies. , , One specimen (ACD‐RCC2) exhibited an intensive CNA throughout the genome.

FIGURE 2

Genome‐wide analysis of copy number alterations in RCCs arising in ESRD. A, Genome‐wide copy number alterations in 9 ESRD‐ccRCCs and (B) 7 ACD‐associated RCCs. The 9 ESRD‐ccRCCs and 7 ACD‐associated RCCs consistently had a loss of chromosome 3p and a gain of chromosome 16, respectively

DNA methylation analysis in RCCs arising in ESRD

We obtained genome‐wide DNA methylation profiles of the 16 RCCs arising in ESRD, and compared them to those of the 100 sporadic RCCs (40 ccRCCs, 30 PRCCs, and 30 ChRCCs), 25 noncancerous and 2 normal kidney tissues obtained from the TCGA and GEO databases. A hierarchical clustering analysis using 1000 probes with the highest standard deviation (SD) among the specimens and samples showed that the DNA methylation profile can clearly classify the specimens and samples, and that the classification was in good agreement with the RCC histopathological types and cancer/normal statuses (Figure 3A). We obtained 2 major clusters (Clusters I and II), of which Cluster I was divided into 3 subclusters (Clusters Ia, Ib, and Ic). Cluster Ia consisted of sporadic and ESRD‐ccRCCs, Cluster Ib of PRCCs and ACD‐associated RCCs, Cluster Ic of noncancerous and normal samples, and Cluster II of ChRCCs. ESRD‐ccRCCs with short (or long) duration of dialysis therapy did not cluster together, suggesting no association between DNA methylation profile and duration of dialysis therapy. A UMAP using the 1000 probes also showed that ESRD‐ccRCCs are grouped with sporadic ccRCCs, ACD‐associated RCCs are with PRCCs, and ChRCCs consisted of their own group (Figure 3B).

FIGURE 3

Genome‐wide DNA methylation analysis in RCCs arising in ESRD. A, Hierarchical clustering analysis and (B) UMAP using 1000 probes with the highest SD among all specimens and samples throughout the genome. ESRD‐ccRCCs clustered together with sporadic ccRCCs, and ACD‐RCCs clustered with PRCCs. C, Hierarchical clustering analysis and (D) UMAP using 1000 probes with the highest SD in TSS200 CGIs. These probes failed to classify the histopathological types. E, Hierarchical clustering analysis and (F) UMAP using 1000 probes with the highest SD in gene bodies. ESRD‐ccRCCs clustered together with sporadic ccRCCs, and ACD‐RCCs clustered with PRCCs

To explore which regions in the genome had the strongest influence on this classification of the histopathological types and cancer/normal statuses, we conducted similar analyses using 1000 probes with the highest SD from promoter CGIs (TSS200 CGIs), gene bodies, and enhancers. The probes from TSS200 CGIs and enhancers failed to classify the samples (Figure 3C,D; and Figure S3A,B), but those from gene bodies did so successfully (Figure 3E,F). These findings suggested that the clear distinction among ESRD‐ccRCCs, ACD‐associated RCCs, and ChRCCs reflected mostly their methylome in gene bodies.

Methylation of gene bodies is mostly associated with increased gene expression. , To identify possible pathways regulated by the methylome that define histopathological types of RCCs, GO analysis was conducted using the 1000 probes with the highest SD from gene bodies. GO terms related to “regulations of cellular process” and “development process” were significantly enriched in the class of Process (Table S9), showing that the differentially methylated regions are derived from genes involved in processes of cellular regulation or development.

DNA methylation of 2 genes, VHL and CDKN2A, were especially analyzed in the 16 RCCs arising in ESRD because they are known to be epigenetically silenced in small subsets of sporadic ccRCCs (VHL, 7%‐10%), , , , PRCCs (CDKN2A, 6.2%) and ChRCCs (CDKN2A, 6.1%). However, we did not find either VHL or CDKN2A methylation in ESRD‐ccRCCs or ACD‐associated RCCs (Table S3).

Cellular origin of RCCs arising in ESRD

Identification of the cellular origin of a cancer is important for understanding pathogenic mechanisms and to establish novel therapeutic approaches. Methylation Cluster I contained sporadic ccRCCs and PRCCs, both of which are considered to develop from proximal tubule cells in a nephron based on immunohistochemical staining patterns, , , and Cluster II contained ChRCCs, which are considered to develop from distal tubule cells/collecting duct cells (Figure 3A). As ESRD‐ccRCCs and ACD‐associated RCCs were included in Clusters Ia and Ib, we hypothesized that these 2 histopathological types of RCCs originated from proximal tubule cells as sporadic ccRCCs and PRCCs.

The cellular origins of ESRD‐ccRCCs and ACD‐associated RCCs were further analyzed by measuring expression levels of 5 genes whose expression levels were conserved in individual segments of a normal nephron. , HNF1a, HNF4a, and SLC17A3 are known to be expressed in proximal tubule cells, and CLDN8 and SLC4A1 in distal tubule cells/collecting duct cells. , Due to RNA availability, 7 of the 9 ESRD‐ccRCCs, 6 of the 7 ACD‐associated RCCs, and 33 sporadic RCC (14 ccRCC, 8 PRCC, and 11 ChRCC), and 7 noncancer surgical specimens were analyzed. ESRD‐ccRCCs and ACD‐associated RCCs, along with sporadic ccRCCs and PRCCs, had significantly higher expression levels of HNF1a, HNF4a, and SLC17A3 than ChRCCs (Figure 4A). In contrast, ChRCCs had significantly higher expression levels of CLDN8 and SLC4A1 than ESRD‐ccRCCs and ACD‐associated RCCs, along with sporadic ccRCCs and PRCCs (Figure 4B). Collectively, the expression levels of the marker genes of cellular origin also supported that ESRD‐ccRCCs and ACD‐associated RCCs originated from proximal tubule cells.

FIGURE 4

Analysis of gene expression of marker genes of cellular origin in RCCs arising in ESRD. A, Expression levels of 3 proximal tubule cell markers (HNF1a, HNF4a, SLC17A3). These were highly expressed in ESRD‐ccRCCs, ACD‐RCCs, sporadic ccRCCs, and PRCCs. B, Expression levels of 2 distal tubule cell/collecting duct cell markers (CLDN8, SLC4A1). These were highly expressed in ChRCCs. *P < .05

Immunohistochemistry of RCCs arising in ESRD

We examined protein expression of the marker genes of cellular origin in RCCs arising in ESRD by immunohistochemistry to further confirm our findings in genetic, epigenetic, and mRNA analyses. ESRD‐ccRCCs and ACD‐associated RCCs, along with sporadic ccRCCs and PRCCs, had high expression of HNF1a and HNF4a and lacked CLDN8 expression (Figure 5 and Figure S4). In contrast, ChRCCs had high CLDN8 expression and lacked expression of HNF1a and HNF4a. Collectively, the protein expression of marker genes of cellular origin was in line with their mRNA expression levels. These data again supported that ESRD‐ccRCCs and ACD‐associated RCCs originated from proximal tubule cells.

FIGURE 5

Immunohistochemistry of marker genes of cellular origin in RCCs arising in ESRD. Representative immunohistochemical findings of 2 proximal tubule cell markers (HNF1a, and HNF4a) and one distal tubule cell/collecting duct cell marker (CLDN8). ESRD‐ccRCC (ESRD‐ccRCC7) and ACD‐associated RCC (ACD‐RCC2), along with sporadic ccRCC (sporadic ccRCC5) and PRCC (PRCC8), had high expression of HNF1a and HNF4a and lacked CLDN8 expression. In contrast, ChRCC (ChRCC3) had high CLDN8 expression and lacked expression of HNF1a and HNF4a. Noncancerous, noncancerous6. Scale bars indicates 50 µm

DISCUSSION

ESRD‐ccRCCs had mutation and methylation profiles similar to sporadic ccRCCs. The high incidence of VHL mutations in ESRD‐ccRCCs was consistent with previous reports, , and the loss of chromosome 3p was also documented in a previous report. The genome‐wide DNA methylation profile, especially in gene bodies, was shown to be similar between ESRD‐ccRCCs and sporadic ccRCCs. These findings suggested that genetic and epigenetic profiles of ESRD‐ccRCCs resembled those of sporadic ccRCCs, in line with their shared morphologic features (Figure 1). Also, in addition to the DNA methylation analysis, expression analysis of the cellular origin marker genes for the segments of a nephron indicated that ESRD‐ccRCCs were derived from proximal tubule cells.

ACD‐associated RCC was proposed as a unique type of RCC exclusively found in patients with acquired cystic disease on long‐term dialysis, and was established as a new histopathological type based on the WHO 2016 classification. , Morphologically, ACD‐associated RCCs had some similar features to PRCCs , (Figure 1). Immunohistochemistry of ACD‐associated RCCs, namely the positive expression of proximal tubule cell markers, such as AMACR, and the absence of staining of distal tubule cell markers, such as Ksp‐cadherin, indicated that they originated from proximal tubule cells. , In this study, the expression of marker genes for cellular origin gave results in line with the previous immunohistochemical findings. ACD‐associated RCCs consistently had a gain of chromosome 16, which is known to be frequently present in PRCCs. , , Importantly, DNA methylation profiles showed that ACD‐associated RCCs are clustered together with PRCCs (Figure 3). Collectively, the pathogenesis of ACD‐associated RCCs may overlap with PRCCs, thereby suggesting that they are likely to have originated from proximal tubule cells.

The effect of DNA methylation or CNAs on pathogenesis in RCCs arising in ESRD remains unknown. DNA methylation would be aberrantly induced by chronic inflammation in ESRD, as reported in other tissues such as colorectal mucosa by ulcerative colitis or gastric mucosa by Helicobacter pylori. Conversely, hyperhomocysteinemia in ESRD is reported to increase S‐adenosylhomocysteine, which leads to demethylation of the genome and subsequently chromosomal instability. , Indeed, one ACD‐associated RCC (ACD‐RCC2) exhibited extensive CNAs (Figure 2B), suggesting a pathogenic role of DNA methylation and CNAs in RCCs arising in ESRD.

This study has several limitations. First, to conclude that proximal tubule cells are the cellular origins, demonstrating that ESRD‐ccRCCs and ACD‐associated RCCs developed from genetically marked proximal tubule cells using genetically engineered mice (eg, Sglt2‐Cre transgenic mice) would be ideal. Nevertheless, the similar profiles of DNA methylation and expression of marker genes of cellular origin strongly implicated that these RCCs originated from proximal tubule cells. Second, we included ESRD‐ccRCCs with only a short duration of dialysis therapy (eg, ESRD‐ccRCC9) because we hypothesized that genetic or epigenetic alteration would be already induced by chronic inflammation in chronic kidney disease. However, it remains to be confirmed that RCCs arising in ESRD with a short duration of dialysis therapy reflected typical RCCs arising in ESRD. Third, we did not find any mutations in PBRM1, SETD2, or BAP1 in ESRD‐ccRCCs, different from sporadic ccRCCs. However, considering the frequency of these mutations (PBRM1: 30%, SETD2: 10%, BAP1: 10%), , the expected numbers of specimens harboring these mutations would be 0‐5 (30% ± 29.9% [95% CI]), 0‐2 (10% ± 29.6%), and 0‐2, respectively. Nevertheless, we could not eliminate the possibility that the prevalence of mutations of these genes is different between sporadic ccRCCs and ESRD‐ccRCCs.

In conclusion, this analysis of genetic and epigenetic alterations of RCCs in ESRD patients showed that ESRD‐ccRCCs have similar molecular profiles to those of sporadic ccRCCs and ACD‐associated RCCs to PRCCs. In addition, DNA methylation and gene expression profiles indicated that these 2 histopathological types of RCCs originated from proximal tubule cells in the nephron.

DISCLOSURE

The authors declare no conflict of interest.

Fig S1

Click here for additional data file.

Fig S2

Click here for additional data file.

Fig S3

Click here for additional data file.

Fig S4

Click here for additional data file.

Tables S1‐S9

Click here for additional data file.

Supplementary Material

Click here for additional data file.