Comprehensive analysis of somatic copy number alterations in clear cell renal cell carcinoma

Abstract Somatic copy number alterations (SCNAs) are important biological characteristics that can identify genome‐wide alterations in renal cell carcinoma (RCC). Recent studies have shown that SCNAs have potential value for determining the prognosis of RCC. We examined SCNAs using the Affymetrix platform to analyze samples from 59 patients with clear cell RCCs (ccRCCs) including first cohort (30 cases) and second cohort (validation cohort, 29 cases). We stratified SCNAs in the ccRCCs using a hierarchical cluster analysis based on SCNA types, including gain, loss of heterozygosity (LOH), copy neutral LOH, mosaic, and mixed types. In this way, the examined two cohorts were categorized into two subgroups (1 and 2). Although the frequency of mixed type was higher in subgroup 1 than in subgroup 2 in the two cohorts, the association did not reach statistical significance. There was a significant difference in the frequency of metachronous metastasis between subgroups 1 and 2 (subgroup 2 > 1). In addition, subgroup 2 was retained in multivariate analysis of both cohorts. We examined whether there were specific alleles differing between subgroups 1 and 2 in both cohorts. We found that there was indeed a statistically significant difference in the 3p mixed types. Among the 3p mixed type, we found that 3p24.3 mixed type was inversely correlated with the presence of metachronous metastasis in ccRCC. The association was also retained in multivariate analysis in second cohort. We suggest that the 3p24.3 mixed type may be a novel marker to predict a favorable prognosis in ccRCC.

classification is limited in its ability to predict patient outcome, even in clear cell RCC (ccRCC). 4,5 In that regard, biological markers that predict patient outcome in ccRCC are available. [6][7][8] Recently, somatic copy number alterations (SCNAs) have been used to predict tumor aggressiveness in ccRCC. 9,10 Several studies have demonstrated that the presence of multiple SCNAs is associated with overall patient survival, tumor stage, and development of metastasis. 11,12 Genome-wide assessment using somatic SCNAs provides useful information for identifying overall genomic profiles of cancer cells. 11,12 The accumulation of SCNAs contributes to tumor heterogeneity and consequently striking differences in the presence of SCNAs between primary and metastatic sites. 13 The accumulation of SCNAs plays important roles in tumor progression in RCC as well as other types of cancer. 14 Here, we examined the clinicopathological impact of SCNAs in a Japanese cohort using an array-based comprehensive genomic methodology. The differences in molecular profiles and clinical outcome were carefully analyzed in the present study. Our aim was to identify the association of SCNAs present in ccRCC with clinicopathological findings.
In addition, we attempted to show the genomic heterogeneity that occurred in ccRCC and the clinical impact of SCNA patterns.

| Patients
A total of 59 patients undergoing renal mass excision for ccRCC between January 2011 and June 2017 were enrolled in this study at Iwate Medical University. The 59 patients we examined were divided into two categories, including a first cohort (30 cases) and then a second cohort (29 cases) to validate the results of the first cohort. The fresh tissues were frozen in liquid nitrogen immediately after dissection. All tissue samples were confirmed to be ccRCC type based on their pathology and they were diagnosed according to the "WHO guidelines for tumors of the urinary system and male genital organs" with a slight modification. 15,16 The clinicopathological findings included sex, age, tumor size, tumor location, Fuhrman grade, necrosis, venous invasion, TNM stage, and the presence of metachronous metastasis as indicated by the Japanese Classification for Renal Cell Carcinoma (Table 1). 17 The median duration of follow-up of metachronous metastasis was 49 months (range, 3-82 months). During this follow-up period, 11 patients with metachronous metastasis died. In the present study, no patients with additional treatment, such as chemotherapy and radiotherapy, were included. Protocols were approved by the ethics committees and institutional review boards of participating centers (HG2018-519).

| DNA extraction
DNA was extracted from isolated normal and tumor tissues by sodium dodecyl sulfate lysis and proteinase K digestion, followed by a phenol-chloroform procedure as reported previously. 18

| Estimation of tumor DNA content
Pathologic examination of hematoxylin and eosin stained tissue directly adjacent to the area used for single nucleotide polymorphism (SNP) array was performed to ensure that the region of tumor examined was as phenotypically homogenous as possible, to maximize tumor percentage in the tissue sampled and to minimize the presence of stroma and normal tissue. In addition, necrotic tissue was avoided.
The area of the selected tumor tissue that was adjacent to the sample site was quantitatively analyzed using digital pathology with Aperio Software (Leica Biosystems). Tissue sections were scanned on an Aperio AT2 scanner with an average scan time of 120 seconds (compression quality, 70). Images were analyzed using color deconvolution and colocalization. Aperio Image Analysis software (for measurement of tumor tissue area) was used. The ratio was obtained by dividing the area of the whole tumor tissue that included interstitial tissue by the area of tumor tissue without such interstitial tissue. The ratio was generally greater than 0.8 (80%-90%). A representative figure is shown in Figure S1.

| SNP array analysis
The Cytoscan HD (Affymetrix, UK) platform was used in all experiments. This array contains more than 1.9 million nonpolymorphic markers and over 740 000 SNP markers with an average intragenic marker spacing of 880 bps and intergenic marker spacing of 1737 bps. These platforms consist of microarrays containing nonpolymorphic probes for copy number variations (CNVs) from coding and noncoding regions of the human genome as well as polymorphic SNP probes. All procedures were carried out as instructed by the manufacturer.
The hybridized slides bearing DNA marked with biotin, were analyzed with a GeneChip Scanner 3000 7G (Affymetrix) and the Chromosome Analysis Suite Software (Affymetrix). Definition of abnormalities required a minimum of (a) 50 consecutively duplicated probes, (b) 50 consecutively deleted probes, or (c) segments of loss of heterozygosity (LOH) larger than 3 Mb.
Smaller alterations involving cancer-associated genes were also investigated.  The regions of gain detected in more than 50% of cases were located at 14q32.33 in the ccRCCs. Additionally, regions of LOH (more than 50% of ccRCC cases) and that of mixed type were at 14q24.3 and 4q13.2, and 3p 24.2, respectively. No copyneutral LOHs or mosaic types showing more than 50% of cases were found.

| Hierarchical clustering based on SCNA patterns in ccRCCs in the first cohort
We assessed the SCNA pattern using hierarchical clustering.
We identified two distinct subgroups (subgroup 1, 13 cases; 414 | subgroup 2, 17 cases) as shown in Figure 1A, in which the copy number alteration (CNA) marker in tumor tissue is indicated by the vertical line, and the horizontal lines denote "relatedness" between samples. Figure S2a was added to facilitate understanding.

| Differences in the clinicopathological findings between subgroups 1 and 2
No statistical differences in clinicopathological findings were found between the subgroups, including sex, age, tumor size, tumor location, Fuhrman grade, necrosis, venous invasion, TNM stage, and overall survival. However, we found that the frequency of metachronous metastasis was greater in subgroup 2 than in subgroup 1 (P = .0067; Table 2a).

| Differences in the SCNA patterns between subgroups 1 and 2
Differences in the SCNAs between subgroups 1 and 2 are shown in Table 3. There were no significant differences between subgroups 1 and 2 in the total number of CNAs, median number of CN gains, LOH, CN-LOH, and CN mosaic types. Although a difference was observed in the median number of CN mixed types between subgroups 1 and 2 (P = .0543), the difference fell short of statistical significance. The results are shown in Figure 2A.
Regions of gains detected in more than 50% of the cases were located at 5q21.2, 5q3, and 14q32. 33 Figure 3A). However, no correlation of each subgroup with overall survival could be found. Cox proportional hazards analysis was performed to determine and compare the diseasefree survival rates. We examined whether the clinicopathological findings and stratified subgroups were independent predictors of patient disease-free survival. We used a univariate analysis for preliminary screening of the variables (Table 5a). This analysis was, in turn, followed by the application of a Cox proportional hazards model. The univariate analysis of patients with ccRCC (   (Table 6a). Of those, only 3p24.3 mixed type remained in multivariate analysis (Table 6a).

| Hierarchical clustering based on SCNA patterns in ccRCCs in the second cohort
Using hierarchical clustering, the SCNA pattern was also examined in the second cohort. Two distinct subgroups (subgroup 1, 13 cases; subgroup 2, 16 cases) were identified as shown in Figure 1B, in which the CNA marker in tumor tissue is indicated by the vertical line, and the horizontal lines denote "relatedness" between samples. Figure S2b facilitates understanding of the data.

| Differences in the clinicopathological findings between subgroups 1 and 2 in the second cohort
In the second cohort, some statistical differences in clinicopathological findings were found between the subgroups, including sex, age, tumor size, tumor location, Fuhrman grade, necrosis, venous invasion, and TNM stage. We found that tumor size, venous invasion, TNM stage, and the frequency of metachronous metastasis were greater in subgroup 2 than in subgroup 1 (tumor size, P = .0283; venous invasion, P = .0200; TNM stage, P = .0032; metachronous metastasis, P = .0025; Table 2b).

| Differences in the SCNA patterns between subgroups 1 and 2 in the second cohort
Differences in the SCNAs between subgroups 1 and 2 are shown in Table 3. Although a difference was observed in the median number of CN mixed type between subgroups 1 and 2 (P = .0978), the difference did not reach statistical significance. The results are shown in Figure 2B.
Regions of gains detected in more than 50% of the cases were located at 5q23.   (Table 5b). This analysis was in turn followed by application of a Cox proportional hazards model. The univariate analysis of patients with ccRCC (Table 5b) identified four factors (tumor size, venous invasion, TNM stage, and stratified subgroup) that were associated with an increased frequency of metachronous metastasis.
Two factors were identified in the multivariate Cox proportional hazards analysis (Table 5b). Tumor subgroup classifications (subgroup 1 vs 2) remained significant predictors of disease-free survival, even after controlling for the other variables. pT stage was not a factor associated with metachronous metastasis after adjusting for the effects of the other factors.
3.2.5 | Association of disease (metachronous metastasis)-free survival, overall survival and clinicopathological findings with a specific SCNA in the second cohort Kaplan-Meier analysis was performed to compare disease-free survival and overall survival with the 3p24.3 SCNA mixed type ( Figure 3F and 3H). As a result, we found that although the 3p24.3 SCNA mixed type was not associated with overall survival, it was an independent factor to predict favorable prognosis of ccRCC. In univariate analysis, four factors, including tumor size, venous invasion, pT stage, and 3p24.3 mixed type were correlated with metachronous metastasis (Table 6b). Among those, only 3p24.3 mixed type remained after multivariate analysis (Table 6b). However, we found that a 3p CN mixed type (LOH + LOH mosaic) was a significant factor for differentiating the subgroups in our patient cohort.
Next, we examined whether there was an independent prognostic factor among the 3p CN mixed type in ccRCC that we examined. Interestingly, we showed that 3p24.3 mixed type was an independent factor that predicted an excellent disease-free prognosis in ccRCC. Although it is well known that a specific SCNA can be a prognostic factor to predict poor patient prognosis in various cancers including gastric, 26  suggesting inversely low expression of RPL31 may be correlated with good prognosis of ccRCC. Finally, PDCL3 is an interesting molecule in that PDCL3 expression is regulated by hypoxia and plays an important role in the stability of VEGFR-2. This finding may suggest that angiogenesis is a part of the hypoxia-sensing mechanism that maintains physiological angiogenesis. 40 The association of PDCL3 with cancer progression remains unknown, and further study will be needed. To summarize, we hypothesize that the loss or low expression of the above oncogenic genes may be caused by SCNAmixed type with LOH and LOH mosaic.
There are several limitations to this study. First, the number of patients enrolled in the study was small, particularly when compared to comprehensive "big data" analyses, such as those in TCGA. 12 In addition, SCNA types that were used in the present study are different from those of previous comprehensive analyses. 12,41,42 However, a correlation of a specific SCNA with patient prognosis, including disease-free survival, was not observed in a "big data" In conclusion, using cluster analysis of ccRCC, we examined the SCNA patterns including SCNA gain, LOH, CN-LOH, mosaic, and mixed types. We found that they could be stratified into 2 distinct subgroups. Although there was no significant difference between subgroups 1 and 2 in the SCNAs we classified, a statistical difference in the mixed types (LOH and LOH mosaic) at 3p24.3 was found.
The frequency of mixed type (LOH and LOH mosaic) at 3p24.3 in the metachronous metastasis was significantly higher in subgroup 1 than in subgroup 2. The integrated analysis of gene SCNAs pointed to several interesting genes as potential biomarkers for ccRCC although further studies need to be performed. Taken together, these results may be helpful in the understanding of renal carcinogenesis.

ACKNOWLEDGMENTS
We gratefully acknowledge the technical assistance of Ms. E. Sugawara and C. Ishikawa. We also thank members of the Department of Molecular Diagnostic Pathology, Iwate Medical University for their support.

CONFLICT OF INTERESTS
The authors declare that there are no conflict of interests.

AUTHOR CONTRIBUTIONS
TT, who is the first author, constructed the figures and tables and performed statistical analyses. KI and MO performed histological diagnosis and statistical analysis. ES, RK, RT, and WO assisted with clinical data. TS, who is the corresponding author, contributed to the preparation of the manuscript and all aspects of data collection and analysis.

DATA AVAILABILITY STATEMENT
All data generated or analyzed during this study are included in this published article (and its supplementary information files).