Purpose: Tumor heterogeneity may represent a barrier to preoperative genomic characterization by needle biopsy in clear cell renal cell carcinoma (ccRCC). The extent of heterogeneity in small renal tumors remains unknown. Therefore, we set out to evaluate heterogeneity in resected large and small renal tumors.

Experimental Design: We conducted a study from 2013 to 2016 that evaluated 47 consecutive ccRCC tumors resected during radical or partial nephrectomy. Cases were designated as small (<4 cm) and large (>7 cm) tumors. Each tumor had three regions sampled. Copy-number variation (CNV) was assessed and gene expression analysis was performed to characterize the clear-cell A and B (ccA/ccB) profile and the cell-cycle progression (CCP) score. Genomic heterogeneity between three regions was evaluated using CNV subclonal events, regional expression profiles, and correlation between gene expression.

Results: Twenty-three small and 24 large tumors were analyzed. Total CNVs and subclonal CNVs events were less frequent in small tumors (P < 0.001). Significant gene expression heterogeneity was observed for both CCP scores and ccA/ccB classifications. Larger tumors had more variance in CCP scores (P = 0.026). The distribution of ccA/ccB differed between small and large tumors with mixed ccA/ccB tumors occurring more frequently in the larger tumors (P = 0.024). Analysis of five mixed tumors (with both ccA/ccB regions) demonstrated the more aggressive ccB phenotype had greater CNV events (P = 0.014).

Conclusions: Small renal tumors have much less genomic complexity and fewer subclonal events. Pretreatment genomic characterization with single-needle biopsy in small tumors may be useful to assess biologic potential and may influence therapy. Clin Cancer Res; 24(17); 4137–44. ©2018 AACR.

Translational Relevance

Tumor heterogeneity is believed to represent a barrier to preoperative genomic characterization in kidney cancer. Previous studies of heterogeneity in clear cell renal cell carcinoma (ccRCC) evaluated only large and metastatic tumors. In small renal tumors in which multiple biopsies are not feasible, the extent of heterogeneity remains unknown. In this study, we evaluated how the extent of genomic heterogeneity in small and large ccRCC. A total of 23 small (<4 cm) and 24 large (>7 cm) ccRCC had three regions sampled for evaluation of copy-number, clear-cell A/clear-cell B (ccA/ccB) classification, and cell-cycle progression (CCP) score. Small tumors have less genomic complexity and significantly fewer subclonal CNV events. Pretreatment genomic characterization based on a single biopsy in small ccRCC can provide insight into biologic potential to make clinical decisions.

The past 3 decades have seen a dramatic increase in the incidence of renal cell carcinoma (RCC) largely attributed to the increased rate of cross-sectional imaging. The rising incidence of RCC has largely been due to small renal masses (SRMs) ≤4 cm with most of them asymptomatic at diagnosis (1). With the changing pattern of detection, now approximately 60% of new diagnoses are stage I RCC (2). The clinical outcome in patients with early-stage RCC is generally excellent when treated surgically, although approximately 25% of high-grade stage I RCCs still metastasize (3). However, the relatively stable mortality rate, despite the increased incidence, has suggested overdiagnosis and overtreatment (4). Overtreatment may pose significant health risks, as the majority of patients with kidney cancer are elderly and/or have preexisting comorbidities. Besides the immediate morbidity of treatment, overtreatment may pose long-term adverse health outcomes, as there are increased renal and cardiovascular events associated with all forms of therapy (5).

For elderly individuals who are not candidates for treatment, nonoperative management led to the recognition that many SRM exhibit a low growth rate (6, 7). In recent years, the concept of active surveillance has become more widely accepted (8). Despite the safety, the adoption of this approach has been low and is currently implemented in approximately 10% of new diagnoses (9). Despite the acceptance of this treatment modality by the American Urologic Association (AUA; ref. 10), a significant barrier exists in the widespread adoption of active surveillance as there are no reliable methods to determine biologic potential.

The introduction of cancer genomics to clinical practice has revolutionized the diagnostic, prognostic, and therapeutic approach in various malignancies. In kidney cancer, the molecular events associated with clear cell RCC (ccRCC) have been characterized due to large-scale genomic sequencing efforts such as The Cancer Genome Atlas (TCGA). Similarly, transcriptomic profiling has identified specific clusters of tumors with worse behavior after nephrectomy, which may be useful in the study of SRM prior to treatment. The clear cell A and B (ccA/ccB) designation and the cell-cycle progression (CCP) score are two mRNA classifiers that may be useful in the evaluation of recurrence-free survival and overall survival and are being further developed for clinical care (11, 12).

Although genomic alterations can be reliably detected from a single renal tumor biopsy, the impact of precision medicine has unfortunately been limited. One of the perceived obstacles to this approach is the concern of significant tumor heterogeneity. Previous studies of heterogeneity in RCC tumors evaluated only large and metastatic kidney tumors (13, 14). In these tumors, approximately 75% of driver alterations (copy number and mutations) were found to be subclonal, that is, not shared throughout all tumor regions. Thus, a single biopsy may not sufficiently identify particular genetic events that may be relevant to clinical management. Although it may be safe to perform multisite biopsies within a large primary tumor, this is not feasible in an SRM in which obtaining sufficient tissue on a single biopsy can be challenging. Although heterogeneity may limit genomic characterization from a single biopsy in the large renal mass, its clinical relevance is currently unknown in the SRM that may be eligible for active surveillance.

To define the feasibility of preoperative genomic characterization of the SRM, we set out to compare the extent of genomic heterogeneity in a consecutive series of resected small and large ccRCC tumors. We evaluated subclonal events involving driver copy-number variations (CNVs) and determined prognostic gene expression signatures to better define the limitation of a single region biopsy.

Research subjects and specimen processing

All patients undergoing surgery were provided with informed consent and offered enrollment into a biospecimen repository approved by a Yale University Institutional Review Board. Between 2013 and 2016, 100 consecutive research subjects undergoing radical or partial nephrectomy had tumor procurement. Prior to resection, the surgeon denoted clinical stage of the tumor and cases were designated as small (<4 cm, cT1a) and large (>7 cm, cT2+) renal tumors. Intermediate-sized tumors (4–7 cm) were excluded from further analyses. Research material was procured within 30 minutes of removal and processed in accordance with institutional and College of American Pathology guidelines. Upon tumor sectioning, clinical pathology assistants sampled non-necrotic components ≥1 cm apart and designated regions 1 to 3 (R1–R3). Tumors were snap-frozen with liquid nitrogen. Members of the genitourinary pathology team (AA, PH) reviewed all cases and those determined to be ccRCCs were selected for our analysis. The highest Fuhrman grade was assigned and individual regions did not take specific regional grade assessment. DNA and RNA were extracted from individual regions using a Maxwell-16 SimplyRNA Tissue Kit, and Maxwell-16 DNA Purification Kit (Promega). Nucleic acids were quantified using a Nanodrop 2000 (Thermofisher).

Analysis of copy number

DNA samples were applied to an Illumina CytoSNP 12 Bead Array (Illumina), using a total of 500 ng input. Fifteen recurrent CNVs identified in TCGA project were selected for review. Chromosomal locations included 11 regions of loss (3p25, 14q24, 9p21.3, 6q26, 8p11, 10q23, 1p36, 4q35, 13q21, 15q21, and 2q37) and four regions of gain (5q35, 8q24, 3q26, and 1q32). CNVs were assessed using Blue Fuse Genomic Studio (Illumina) and manually confirmed by inspection of the LogR ratio and B allele frequency. We defined CNVs as being present in an individual tumor region if ≥15% of the sample had evidence of a gain or loss (15). The presence of subclonal/branch events was assessed only when three regions had successful profiling and was given this term when an alteration was not shared among all regions. For TCGA samples, CNV data were downloaded from cbioportal.org and analyzed using the Integrated Genomic Viewer (IGV, version 2.3.72; ref. 16).

NanoString and data analysis of classification ccA/ccB and CCP score

Gene expression profiling was performed on 200 ng of RNA using a custom designed hybridization probe set to evaluate 34 genes involved in designation of ccA/ccB and 31 genes involved in calculation of the CCP score (11, 17). We included the same house keeping genes (five and 15 for ccA/ccB and CCP scores, respectively) used in the development of both classification systems. Gene expression analysis was performed using the NanoString nCounter system and analyzed with nSolverAnalysis Software (NanoString Technologies). A previously described prediction analysis of microarrays (PAM) classifier was applied to the normalized microarray data for designation of ccA/ccB using R version 3.3.0 (12). Although the CCP score calculation was initially designed using qPCR, it has successfully been calculated using microarray data (18); and in a similar way, we calculated the CCP score using NanoString data. Briefly, after normalization with R package DESeq2 and log2 transformation, the gene expression was z-normalized across samples, and a CCP score was created for each sample as the mean expression value of the normalized CCP genes.

Data access

The data have been deposited at the European Genome-phenome Archive (EGA), which is hosted by the EBI and the CRG, under accession number EGAS00001002919.

Statistical analysis

For the purpose of determining sample size, we hypothesized that subclonal CNV events were three times more common in larger tumors. Using an 80% power and P ≤ 0.05, we determined we needed a cohort of 20 small and 20 large tumors to show a three-fold increase in subclonal CNV events for larger tumors. Continuous variables were compared using the independent t test. Chi-squared or Fisher exact tests were used to evaluate differences in categorical variables. Nonparametric testing of continuous variables was performed using the Mann–Whitney U test. Correlation between continuous variables was determined using linear regression. Intratumoral correlation of gene expression was evaluated by individual and combined gene sets for each tumor region using principal component analysis and heatmaps. The intratumoral correlation was determined using the Spearman correlation. The determination whether the distributions of correlation coefficients between large and small samples differ significantly was evaluated using the Kolmogorov–Smirnov test. Statistical significance was taken at P ≤ 0.05 for all comparisons. When multiple hypotheses were tested, the FDR correction was applied. Statistical calculations were performed using R V3.4 (R Core Team), Matlab V2014b (The MathWorks, Inc.), and SPSS V24 (IBM).

A total of 23 small and 24 large renal tumors were included in the study. Subject and tumor characteristics are shown in Table 1. Small tumors were more frequently stage T1 [95.7% (22 of 23)] and a lower Fuhrman grade [95.7% (22 of 23) grade 1 and 2]. Large tumors were higher grade [79.1% (19 of 24) grade 3+], higher stage [79.1% (19 of 24) T3+], and more frequently had necrotic areas [62.5% (15 of 24); P < 0.001, all associations].

Table 1.

List of tumor characteristics for small and large clear cell renal tumors in the cohort

Small tumor (≤4 cm)Large tumor (≥7 cm)
(n = 23)(n = 24)P
Age, median (range), years 60.5 (33–82) 66.5 (49–94) 0.04 
Sex 
 Male, n (%) 14 (60.9) 17 (70.8) 0.47 
 Female, n (%) 9 (39.1) 7 (29.2)  
Tumor sizes, median (range), cm 2.5 (1.2–3.9) 9.25 (7–19) <0.001 
Stage n (%) 
 T1 22 (95.7) 0 (0) <0.001 
 T2 0 (0) 5 (20.8)  
 T3 1 (4.3) 17 (70.8)  
 T4 0 (0) 2 (8.3)  
Tumor grade, n (%) 
 G1 2 (8.7) 0 (0) <0.001 
 G2 20 (87.0) 5 (20.8)  
 G3 1 (4.3) 11 (45.8)  
 G4 0 (0) 8 (33.3)  
Sarcomatoid component, n (%) 
 Negative 23 (100) 21 (87.5) 0.08 
 Positive 3 (12.5)  
Rhabdoid component, n (%) 
 Negative 23 (100) 21 (87.5) 0.08 
 Positive 3 (12.5)  
MVI, n (%) 
 Negative 23 (100) 19 (79.2) 0.021 
 Positive 5 (20.8)  
Necrosis, n (%) 
 Negative 22 (95.7) 9 (37.5) <0.001 
 Positive 1 (4.3) 15 (62.5)  
Regional lymph node metastasis, n (%) 
 Negative 23 (100) 22 (91.7) 0.16 
 Positive 2 (8.3)  
Distant metastasis, n (%) 
 Negative 23 (100) 19 (79.2) 0.021 
 Positive 5 (20.8)  
Small tumor (≤4 cm)Large tumor (≥7 cm)
(n = 23)(n = 24)P
Age, median (range), years 60.5 (33–82) 66.5 (49–94) 0.04 
Sex 
 Male, n (%) 14 (60.9) 17 (70.8) 0.47 
 Female, n (%) 9 (39.1) 7 (29.2)  
Tumor sizes, median (range), cm 2.5 (1.2–3.9) 9.25 (7–19) <0.001 
Stage n (%) 
 T1 22 (95.7) 0 (0) <0.001 
 T2 0 (0) 5 (20.8)  
 T3 1 (4.3) 17 (70.8)  
 T4 0 (0) 2 (8.3)  
Tumor grade, n (%) 
 G1 2 (8.7) 0 (0) <0.001 
 G2 20 (87.0) 5 (20.8)  
 G3 1 (4.3) 11 (45.8)  
 G4 0 (0) 8 (33.3)  
Sarcomatoid component, n (%) 
 Negative 23 (100) 21 (87.5) 0.08 
 Positive 3 (12.5)  
Rhabdoid component, n (%) 
 Negative 23 (100) 21 (87.5) 0.08 
 Positive 3 (12.5)  
MVI, n (%) 
 Negative 23 (100) 19 (79.2) 0.021 
 Positive 5 (20.8)  
Necrosis, n (%) 
 Negative 22 (95.7) 9 (37.5) <0.001 
 Positive 1 (4.3) 15 (62.5)  
Regional lymph node metastasis, n (%) 
 Negative 23 (100) 22 (91.7) 0.16 
 Positive 2 (8.3)  
Distant metastasis, n (%) 
 Negative 23 (100) 19 (79.2) 0.021 
 Positive 5 (20.8)  

Abbreviation: MVI, microvessel invasion.

CNVs were observed with similar frequency to previous reports, with all known recurrent alterations present in ≥5 tumors (19). Twenty-two large and 22 small tumors were analyzed for CNVs, 43 (97.7%) had 3p25 loss. The other common findings (Fig. 1A) were 5q35 gain [27 (61.4%)], 14q24 [22 (50%)], 8p11 [18 (40.9%)], 1p36 [17 (38.6%)], and 10q23 [16 (36.4%)] losses. When comparing CNVs in large versus small tumors, there was a significant increase in 14q24 loss in large tumors [17 (77.3%) vs. 5 (22.7%), P < 0.001; Fig. 1B]. An increase in 9p21.3 [10 (45.5%) vs. 2 (9.1%), P = 0.007], 8p11 [13 (59.1%) vs. 5 (22.7%), P = 0.014), 2q37 [5 (22.7%) vs. 0 (0%), P = 0.048] losses and 3q26 [6 (27.3%) vs. 0 (0%), P = 0.021] gain were observed in large tumors, however there were not statistically significant after multiple hypothesis testing.

Figure 1.

Assessment of the frequency of common CNV driver alterations important in clear cell kidney cancer. A, CNV events analyzed per tumor (n = 44) with gains portrayed in red and losses in blue. B, CNV events analyzed between large (n = 22) and small tumors (n = 22). *, P < 0.05; #, P < 0.05 (FDR correction applied). C, Shared CNV events by region (n = 41).

Figure 1.

Assessment of the frequency of common CNV driver alterations important in clear cell kidney cancer. A, CNV events analyzed per tumor (n = 44) with gains portrayed in red and losses in blue. B, CNV events analyzed between large (n = 22) and small tumors (n = 22). *, P < 0.05; #, P < 0.05 (FDR correction applied). C, Shared CNV events by region (n = 41).

Close modal

Analysis of multiregion CNVs demonstrated a greater burden of CNV events for large tumors (median [range], 7.0 [2–11] vs. 2.5 [1–10], P < 0.001; Fig. 2A). A similar analysis within TCGA (using one region) demonstrated similar results with large tumors having more CNV events (median [range], 4 [0–13] vs. 2 [0–11], P < 0.001; Fig. 2B). We noted subclonal CNVs were more frequent in large tumors (median [range], 4 [0–10] vs. 0 [0–8], P < 0.001; Fig. 2C). For each particular CNV, the prevalence of that CNVs among subclones was measured, CNV events occurring at 3p25, 5q35, 6q26, 15q21, and 2q37 appeared to be shared/truncal events as when present, they occurred in ≥50% of cases (Fig. 1C). 13q21, 1q32, 8q24, 3q26 were likely subclonal/branch events as they were shared among all regions in less than 25% of cases.

Figure 2.

Analysis of the frequency and subclonal nature of driver CNV events in small and large clear cell renal tumors. A, Total CNV events for large and small tumors (all regions considered; n = 44). B, Total CNV events for large and small tumors in the TCGA cohort from single region (n = 263). C, Subclonal CNV events for large and small tumors (n = 41).

Figure 2.

Analysis of the frequency and subclonal nature of driver CNV events in small and large clear cell renal tumors. A, Total CNV events for large and small tumors (all regions considered; n = 44). B, Total CNV events for large and small tumors in the TCGA cohort from single region (n = 263). C, Subclonal CNV events for large and small tumors (n = 41).

Close modal

In 21 small tumors and 21 large tumors, ccA/ccB classification was determined for individual regions and by tumor. Overall, 100 regions were ccA, and 26 regions were ccB. The distribution of ccA/ccB differed between small and large tumors with mixed ccA/ccB tumors occurring more frequently in the larger tumors (Fig. 3; P = 0.024). Of 21 small tumors, 19 were classified as ccA, one as ccB, and one as mixed (a tumor consisting both ccA and ccB regions) designation. Of 21 large tumors, 11 were classified as ccA, five as ccB, and five as mixed designation. For the six mixed tumors, five tumors (15 regions) had available CNV data for comparison between ccA and ccB regions. CNVs were more frequent among ccB [median (range), 7 (2–8) vs. 2 (0–6), P = 0.014; Supplementary Fig. S1)].

Figure 3.

Transcriptomic classification by the ccA/ccB profile by tumor size in clear cell renal tumors. One hundred regions were ccA, and 26 regions were ccB. The distribution of ccA/ccB differed between small and large tumors (P = 0.024).

Figure 3.

Transcriptomic classification by the ccA/ccB profile by tumor size in clear cell renal tumors. One hundred regions were ccA, and 26 regions were ccB. The distribution of ccA/ccB differed between small and large tumors (P = 0.024).

Close modal

Median CCP scores did not differ between small and large tumors. However, there was increased variance (P = 0.026) for large tumors. Gene expression of both prognostic gene sets showed significant within-sample variation suggested by principal component analysis (PCA) plots (Supplementary Fig. S2) and heat maps (entire gene set; Supplementary Fig. S3). Spearman correlation coefficients for the entire gene set between regions ranged from 0.6 to 0.97. Evaluation of the minimal correlation between regions by size category showed that there was significantly more intratumoral heterogeneity in large tumors (Fig. 4, P = 0.004 Kolomogorov–Smirov test). Within the entire cohort, we found decreasing gene expression correlation with increasing subclonal CNV events (R2 = 0.431; P < 0.001; Supplementary Fig. S4).

Figure 4.

Minimal intratumoral spearman correlation of ccA/B and CCP genes by small and large tumor size (tumor rank from lowest to highest correlation). Evaluation of the minimal correlation between regions by size category showed that there was significantly more intratumoral heterogeneity in large tumors (P = 0.004).

Figure 4.

Minimal intratumoral spearman correlation of ccA/B and CCP genes by small and large tumor size (tumor rank from lowest to highest correlation). Evaluation of the minimal correlation between regions by size category showed that there was significantly more intratumoral heterogeneity in large tumors (P = 0.004).

Close modal

A pretreatment renal mass biopsy may provide pathologic characteristics that could be useful to select patients with increased surgical risk for active surveillance. With current techniques, a renal mass biopsy provides a high diagnostic rate with minimal morbidity (20). However, the accuracy of grading on biopsy is not perfect with 50% to 85% accuracy likely limited by significant grade heterogeneity (15, 21). Therefore, many providers do not routinely biopsy because the results rarely alter anticipated treatment. Despite limitations on pathologic diagnosis, a biopsy provides access to tissue for genomic characterization. However, until recently there were limited testing options due to costs, limited material, difficulty with genomic studies on fixed tissue, and long processing times.

In recent years, the widespread availability and reduced costs of genomic testing have led to increased clinical utilization. In a wide variety of malignancies, testing is accepted as a useful tool in determining the risk of recurrence, prognosis, the utility of adjuvant therapy/radiation, and the selection of molecular targeted therapy. For localized disease, various genomic tests have been introduced. Some of these assays define the likelihood of higher risk features or adverse outcomes at the time of surgery in order to better counsel those interested in active surveillance—something that could also be useful in patients presenting with an SRM. There has been extensive characterization of the recurrent driver alterations associated with the various histologic subtypes based on work from TCGA. Despite the recognition of various CNVs and somatic mutations associated with worse outcome in ccRCC, genomic characterization has not made its way into established clinical guidelines (22).

One of the major obstacles to adoption is the perception of massive intratumor heterogeneity, a perception that is based on the study of advanced kidney cancer (13, 14). Whether this is due to larger tumors being present for longer resulting in an increased number of cell divisions or whether these larger tumors are predisposed to being more aggressive has yet to be determined. How genomic heterogeneity relates to the SRM is critical to understanding how genomic characterization from a single biopsy may guide important clinical decisions for such as selection for surveillance, utility of node dissection, and selection for nephron-sparing surgery.

We set out to characterize genomic heterogeneity in the SRM (cT1a, ≤4 cm) and large renal masses (cT2+, >7 cm) as a comparison to the types of tumors evaluated in similar studies (13, 14). This cohort (n = 47) currently is the largest multisite assessment of tumor heterogeneity in ccRCC and utilized both CNV and mRNA gene expression. Our major finding from CNV analysis is that small renal tumors have significantly less overall genomic events and that subclonal gains/losses are much less frequent in these tumors. Thus, the homogeneous nature of the SRM, may allow sufficient characterization of known adverse CNVs from one or two sites. Although the sensitivity and specificity for an assay based on a single core biopsy may not be 100%, a clinician may be able to better counsel a patient on having a tumor with more aggressive potential and strongly consider treatment versus a more stringent active surveillance protocol.

Our assessment of subclonal CNVs identified significant differences in the frequency of shared events, likely supporting the temporal order of genomic events that occur during tumor evolution. The high incidence of shared CNVs at 3p25, 5q35, 6q26, 15q21, and 2q37 suggest these events occur early in tumorigenesis and are more likely truncal events. Gerlinger and colleagues similarly found differences in the frequency of shared CNVs, however, important differences were observed (13, 14). Although 3p and 5q CNVs were also found to be shared events, losses of 4q, 8p, and 14q were also believed to be early events. We also found that several CNVs were infrequently shared, suggesting these events occur as branch events that occur late in tumorigenesis. These included 13q21 loss and 1q32, 8q24, and 3q26 gains. Similar issues have been raised with mutational profiling in a cohort of 14 ccRCC tumors reported by Sankin (23). In that study, clear differences in the ubiquitous nature of the major driver mutations were observed similar to our findings.

In recent years, there have been efforts to apply transcriptomic profiling in RCC, mostly those with locally advanced or metastatic disease. The ccA/ccB classification has been applied to large cohorts of patients with metastatic disease demonstrating the utility at determining prognosis and prediction of response to sunitinib (24, 25). We performed heterogeneity assessment of genes used in these promising transcriptomic classifiers, including the ccA/ccB system and the CCP score. Transcriptomic RNA classifiers have been considered an attractive option for prognostic models in light of the heterogeneous nature of driver alterations. Global expression may be less sensitive to individual genomic events. However, we demonstrate that heterogeneity of the transcriptomic profile can occur within various tumor regions, especially for large renal masses, which have a mixed pattern in nearly a quarter of tumors. This finding differs from a report by Serie and colleagues, who observed no classification heterogeneity within the resected primary tumor, implying that a single sample per tumor should be sufficient in classification (26). However, similar to the more homogenous nature of CNVs in SRMs, we found that small renal tumors had a more consistent ccA/ccB classification with only one of 21 tumors having a mixed profile. Larger tumors also had increase variance in the CCP score which may need to be taken in context when further evaluating this emerging prognostic marker that has been recently validated in two institutional cohorts (27). Our comparison of transcriptomic profiling similarly shows significantly less heterogeneity for small tumors with nearly all tumors having ≥90% correlation between regions.

The heterogeneity of gene expression has limitations for advanced disease characterization likely related to divergence of clonal populations. For our mixed samples, we demonstrate that CNV events are more common in ccB regions. This could suggest that as tumors grow and evolve, the acquisition of late or subclonal chromosomal changes leads to divergence of gene expression. Analysis of multiregion sampling of metastatic samples in the Gerlinger cohort found that 80% of the tumors contained mixed gene expression profiles of ccA/ccB (14, 28). The more aggressive ccB regions had increased numbers of poor prognostic somatic mutations and CNVs. Both of these findings suggest that increased genomic instability may be driving a change in tumor classification. This is also supported by the Serie cohort, which found that among mixed metastatic samples that underwent longitudinally sampled, 80% of cases progressed to the ccB phenotype (26). Thus, RNA profiling, while perhaps useful for a single biopsy in an SRM, may require multisite biopsies and assessment in larger tumors to improve the sensitivity of detection of the ccB phenotype. In the future, characterization of overall, intratumoral genomic heterogeneity from a single biopsy may be incredibly useful prior to inferring the genomic profile from a single site.

This study has several limitations based on the methodology. Our work is based on a moderate-size cohort (n = 47) and an expanded cohort could improve conclusions and improve the statistical comparisons between groups. Nearly 95% of our SRMs were low grade, and this limits conclusions about the rare high-grade tumor. Ex vivo sampling at the time of radical or partial nephrectomy is different from percutaneous biopsy because it is not done at “random,” and we did not microdissect the most representative area of the tumor which is often done with tumor profiling.

We demonstrate that SRM may be eligible for active surveillance have limited CNV and transcriptomic heterogeneity. Our findings temper some of the concerns that a pretreatment biopsy and characterization cannot overcome massive genomic heterogeneity in ccRCC. Although this may be true in advanced disease and large tumors, it appears less relevant to the majority of small tumors, which are often those detected incidentally. This type of approach could decrease the morbidity associated with surgical management. Further analyses will be useful to confirm our findings and evaluate other RCC subtypes and tumors between 4 to 7 cm, in addition to the evaluation of heterogeneity driver mutations and other mRNA prognostic genes.

No potential conflicts of interest were disclosed.

Conception and design: D. Ueno, Z. Xie, K.A. Nguyen, A. Adeniran, H. Kluger, Brian Shuch

Development of methodology: D. Ueno, Z. Xie, K.A. Nguyen, A. Adeniran, Z. Liu, H. Kluger, Brian Shuch

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Ueno, Z. Xie, M. Boeke, J. Syed, K.A. Nguyen, A. Adeniran, P. Humphrey, Brian Shuch

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Ueno, Z. Xie, K.A. Nguyen, P. McGillivray, G.M. Dancik, Y. Kluger, Z. Liu, H. Kluger, Brian Shuch

Writing, review, and/or revision of the manuscript: D. Ueno, M. Boeke, J. Syed, K.A. Nguyen, P. McGillivray, A. Adeniran, P. Humphrey, G.M. Dancik, H. Kluger, Brian Shuch

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Syed, K.A. Nguyen, Brian Shuch

Study supervision: H. Kluger, Brian Shuch

This work was supported by the NIH (1K08CA207845-01 and KL2 TR000140 to B. Shuch; R-01 CA158167 and K24CA172123 to H. Kluger).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Hollingsworth
JM
,
Miller
DC
,
Daignault
S
,
Hollenbeck
BK
. 
Rising incidence of small renal masses: a need to reassess treatment effect
.
J Natl Cancer Inst
2006
;
98
:
1331
4
.
2.
Kane
CJ
,
Mallin
K
,
Ritchey
J
,
Cooperberg
MR
,
Carroll
PR
. 
Renal cell cancer stage migration: analysis of the National Cancer Data Base
.
Cancer
2008
;
113
:
78
83
.
3.
Zisman
A
,
Pantuck
AJ
,
Wieder
J
,
Chao
DH
,
Dorey
F
,
Said
JW
, et al
Risk group assessment and clinical outcome algorithm to predict the natural history of patients with surgically resected renal cell carcinoma
.
J Clin Oncol
2002
;
20
:
4559
66
.
4.
Welch
HG
,
Black
WC
. 
Overdiagnosis in cancer
.
J Natl Cancer Inst
2010
;
102
:
605
13
.
5.
Shuch
B
,
Hanley
JM
,
Lai
JC
,
Vourganti
S
,
Setodji
CM
,
Dick
AW
, et al
Adverse health outcomes associated with surgical management of the small renal mass
.
J Urol
2014
;
191
:
301
9
.
6.
Bosniak
MA
,
Birnbaum
BA
,
Krinsky
GA
,
Waisman
J
. 
Small renal parenchymal neoplasms: further observations on growth
.
Radiology
1995
;
197
:
589
97
.
7.
Rendon
RA
,
Stanietzky
N
,
Panzarella
T
,
Robinette
M
,
Klotz
LH
,
Thurston
W
, et al
The natural history of small renal masses
.
J Urol
2000
;
164
:
1143
7
.
8.
Smaldone
MC
,
Kutikov
A
,
Egleston
BL
,
Canter
DJ
,
Viterbo
R
,
Chen
DYT
, et al
Small renal masses progressing to metastases under active surveillance: a systematic review and pooled analysis
.
Cancer
2012
;
118
:
997
1006
.
9.
Huang
WC
,
Pinheiro
LC
,
Russo
P
,
Lowrance
WT
,
Elkin
EB
. 
Surveillance for the management of small renal masses: utilization and outcomes in a population-based cohort
.
J Clin Oncol
2013
;
31
:
343
3
.
10.
Campbell
S
,
Uzzo
RG
,
Allaf
ME
,
Bass
EB
,
Cadeddu
JA
,
Chang
A
, et al
Renal mass and localized renal cancer: AUA Guideline
.
J Urol
2017
;
198
:
520
9
.
11.
Brooks
SA
,
Brannon
AR
,
Parker
JS
,
Fisher
JC
,
Sen
O
,
Kattan
MW
, et al
ClearCode34: a prognostic risk predictor for localized clear cell renal cell carcinoma
.
Eur Urol
2014
;
66
:
77
84
.
12.
Askeland
EJ
,
Chehval
VA
,
Askeland
RW
,
Fosso
PG
,
Sangale
Z
,
Xu
N
, et al
Cell cycle progression score predicts metastatic progression of clear cell renal cell carcinoma after resection
.
Cancer Biomark
2015
;
15
:
861
7
.
13.
Gerlinger
M
,
Rowan
AJ
,
Horswell
S
,
Larkin
J
,
Endesfelder
D
,
Gronroos
E
, et al
Intratumor heterogeneity and branched evolution revealed by multiregion sequencing
.
N Engl J Med
2012
;
366
:
883
92
14.
Gerlinger
M
,
Horswell
S
,
Larkin
J
,
Rowan
AJ
,
Salm
MP
,
Varela
I
, et al
Genomic architecture and evolution of clear cell renal cell carcinomas defined by multiregion sequencing
.
Nat Genet
2014
;
46
:
225
33
.
15.
Ball
MW
,
Bezerra
SM
,
Gorin
MA
,
Cowan
M
,
Pavlovich
CP
,
Pierorazio
PM
, et al
Grade heterogeneity in small renal masses: potential implications for renal mass biopsy
.
J Urol
2015
;
193
:
36
40
.
16.
Robinson
JT
,
Thorvaldsdóttir
H
,
Winckler
W
,
Guttman
M
,
Lander
ES
,
Getz
G
, et al
Integrative genomics viewer
.
Nat Biotechnol
2011
;
29
:
24
6
.
17.
Cuzick
J
,
Swanson
GP
,
Fisher
G
,
Brothman
AR
,
Berney
DM
,
Reid
JE
, et al
Prognostic value of an RNA expression signature derived from cell cycle proliferation genes in patients with prostate cancer: a retrospective study
.
Lancet Oncol
2011
;
12
:
245
55
.
18.
Dancik
GM
,
Theodorescu
D
. 
Robust prognostic gene expression signatures in bladder cancer and lung adenocarcinoma depend on cell cycle related genes
.
PLoS One
2014
;
9
:
e85249
9
.
19.
Creighton
CJ
,
Gunaratne
PH
,
Gibbs
RA
,
Chu
A
,
Beroukhim
R
,
Signoretti
S
, et al
Comprehensive molecular characterization of clear cell renal cell carcinoma
.
Nature
2013
;
499
:
43
9
.
20.
Richard
PO
,
Jewett
MAS
,
Bhatt
JR
,
Kachura
JR
,
Evans
AJ
,
Zlotta
AR
, et al
Renal tumor biopsy for small renal masses: a single-center 13-year experience
.
Eur Urol
2015
;
68
:
1007
13
.
21.
Wang
R
,
Wolf
JS
,
Wood
DP
,
Higgins
EJ
,
Hafez
KS
. 
Accuracy of percutaneous core biopsy in management of small renal masses
.
Urology
2009
;
73
:
586
90
.
22.
Motzer
RJ
,
Jonasch
E
,
Agarwal
N
,
Bhayani
S
,
Bro
WP
,
Chang
SS
, et al
Kidney cancer, version 2.2017, NCCN clinical practice guidelines in oncology
.
J Natl Compr Canc Netw
2017
;
15
:
804
34
.
23.
Sankin
A
,
Hakimi
AA
,
Mikkilineni
N
,
Ostrovnaya
I
,
Silk
MT
,
Liang
Y
, et al
The impact of genetic heterogeneity on biomarker development in kidney cancer assessed by multiregional sampling
.
Cancer Med
2014
;
3
:
1485
92
.
24.
Beuselinck
B
,
Job
S
,
Becht
E
,
Karadimou
A
,
Verkarre
V
,
Couchy
G
, et al
Molecular subtypes of clear cell renal cell carcinoma are associated with sunitinib response in the metastatic setting
.
Clin Cancer Res
2015
;
21
:
1329
39
.
25.
de Velasco
G
,
Culhane
AC
,
Fay
AP
,
Hakimi
AA
,
Voss
MH
,
Tannir
NM
, et al
Molecular subtypes improve prognostic value of international metastatic renal cell carcinoma database consortium prognostic model
.
Oncologist
2017
;
22
:
286
92
.
26.
Serie
DJ
,
Joseph
RW
,
Cheville
JC
,
Ho
TH
,
Parasramka
M
,
Hilton
T
, et al
Clear cell type A and B molecular subtypes in metastatic clear cell renal cell carcinoma: tumor heterogeneity and aggressiveness
.
Eur Urol
2017
;
71
:
979
85
.
27.
Morgan
TM
,
Mehra
R
,
Tiemeny
P
,
Wolf
JS
,
Wu
S
,
Sangale
Z
, et al
A Multigene signature based on cell cycle proliferation improves prediction of mortality within 5 yr of radical nephrectomy for renal cell carcinoma
.
Eur Urol
2018
;
73
:
763
69
.
28.
Gulati
S
,
Martinez
P
,
Joshi
T
,
Birkbak
NJ
,
Santos
CR
,
Rowan
AJ
, et al
Systematic evaluation of the prognostic impact and intratumour heterogeneity of clear cell renal cell carcinoma biomarkers
.
Eur Urol
2014
;
66
:
936
48
.