Purpose:

Genomic alterations in key components of PI3K/mTOR pathway have been proposed as candidate predictive markers for rapalog therapy in renal cell carcinoma (RCC). We tested this hypothesis in patients from a randomized phase II trial of everolimus versus sunitinib.

Patients and Methods:

Archival specimens collected at baseline were analyzed with targeted next-generation sequencing (NGS). Focus of interest were alterations in key PI3K pathway components. PTEN expression was assessed by IHC. Association between molecular findings and treatment outcomes was investigated; same associations were tested for 2 everolimus-treated trial cohorts in gastric and hepatocellular carcinoma (HCC).

Results:

Among 184 everolimus-treated patients with RCC with NGS data, mutation rates in genes of interest were 6% (TSC1), 4.4% (TSC2), and 8.2% (mTOR); 44% harbored alterations in ≥1 PI3K pathway component. For subjects with presence versus absence of mutations in TSC1, TSC2, or mTOR progression-free survival (PFS) neither differed on univariate analysis (HR, 1.0; P = 0.895) nor on multivariate testing stratified by MSKCC risk group and other established prognostic factors (HR, 1.1; P = 0.806). Everolimus-treated patients with retained (n = 50) versus lost (n = 50) PTEN IHC expression had median PFS of 5.3 months versus 10.5 months (HR, 2.5; P < 0.001). Such differences were not seen with sunitinib (10.9 months vs. 10.3 months; HR, 0.8; P = 0.475). Molecular findings did not correlate with outcomes in gastric and HCC cohorts.

Conclusions:

Association between mutation status for TSC1/TSC2/mTOR and therapeutic outcome on everolimus was not confirmed. Clinically meaningful differences in PFS were seen based on PTEN expression by IHC, lost in >50% of patients.

Translational Relevance

Previously, retrospective reports have highlighted the presence of acquired mutations in PI3K pathway components (TSC1, TSC2, and MTOR) in tumors of patients with RCC with unusual benefit from mTOR inhibitors. It is often implied that these findings constitute a predictive biomarker signal that can guide choice of agents. However, analyses of prospectively collected cohorts have been lacking. This study constitutes the first report to apply NGS testing in an unselected cohort of rapalog treated patients with RCC. Specimens and data were collected prospectively on a randomized clinical trial of first line everolimus versus sunitinib. We could not confirm association between TSC1/TSC2/mTOR mutation status and benefit from everolimus therapy suggesting such information alone should not determine choice of rapalog therapy. Contrarily, IHC analysis showed correlation between loss of PTEN expression and favorable PFS for everolimus, not for sunitinib. This predictive signal deserves further study in independent RCC cohorts.

The inhibitors of the mTOR complex 1 (mTORC1), also termed as rapalogs, have proven antitumor activity in patients with advanced renal cell carcinoma (RCC). Everolimus was originally approved for patients with advanced RCC after the failure of treatment with sunitinib or sorafenib, based on the randomized, placebo-controlled, RECORD-1 trial (1), conducted in a molecularly unselect population.

Subsequent case series demonstrated somatic alterations in key components of the PI3K/mTOR signaling pathway, particularly mTOR, TSC1, and TSC2 in outlier cases with exceptional therapeutic benefit (2, 3), suggesting predictive value for rapalog therapy. In the current analysis, we performed next-generation sequencing (NGS) analyses of archival specimens collected prospectively on RECORD-3 (NCT00903175), a randomized phase II trial investigating noninferiority of the first-line everolimus versus sunitinib in patients with advanced RCC (1). In the primary analysis, the preset noninferiority margin was not achieved. Our interest was to investigate the correlation of baseline tissue biomarkers with therapeutic outcomes.

Patients

The study design including eligibility has been reported previously (1). Treatment with everolimus versus sunitinib was continued until disease progression, at which, patients were offered to cross-over to the other agent. The study was conducted in accordance with the International Conference on Harmonization Good Clinical Practice guidelines and was approved by Institutional Review Boards or independent ethics committees of each center. All patients gave written informed consent.

Given the common suggestion that oncogenomics may supersede histologic classification in novel clinical trial designs (4, 5), we sought to extend our investigation beyond RCC and analyzed data from 2 additional, placebo-controlled, phase III trials of everolimus. One was GRANITE-1 conducted in chemotherapy-pretreated gastric cancer (6); the other EVOLVE-1 conducted in sorafenib-pretreated HCC.

Next-generation sequencing

For RECORD-3 patients, paraffin-embedded specimens were microdissected for DNA extraction from tumor plus adjacent normal tissue and subjected to an analysis using IMPACT, a NGS pull-down assay across >340 genes. (7, 8) Our analysis focused on genes within the PI3K/mTOR pathway. In samples with loss of function (LOF) mutations in TSC1 or TSC2, we investigated loss of heterozygosity (LOH) through allele-specific copy-number analysis using open-source Fraction and Allele-Specific Copy Number Estimates from Tumor Sequencing (FACETS) tool (9). The same approach was used to determine PTEN copy-number status.

For the 2 non-RCC cohorts, NGS analysis was performed at an average coverage depth of 100× on a targeted cancer gene panel containing 296 genes (10). Single-nucleotide variants, small insertions or deletions, and copy-number alterations (CNA) were interrogated. Our analysis focused on the alterations in genes encoding key components of the PI3K/mTOR pathway, particularly those deemed likely to have functional significance resulting in hyperactivation of mTORC1 signaling. In considering the individual genes of interest, patients were classified as “mutant” (MT) if an exonic mutation was identified; otherwise, they were labeled as “wild-type” (WT) for the gene in questions. The total and allele-specific copy number was estimated from the IMPACT data using FACETS where diploid level was either calculated using the default parameters or if fewer than 15% of genome had 2 copies under default level, set to median. The LOH at the specific genes was determined based on the copy-number level of the overlapping segment.

PTEN IHC

PTEN expression status has previously been investigated for its prognostic and predictive value in RCC with the conflicting findings (11–13). For the purpose of this report, in addition to NGS/FACETS analyses, PTEN status was assessed by IHC to determine expression levels semiquantitatively. PTEN (138G6) IgG rabbit monoclonal antibody (Cell Signal) was used for PTEN expression testing by IHC. The staining intensity of positive cells was scored from 0 to 3+. The proportion of positive staining cells (% cells) at each intensity level was recorded and the total H-score was tabulated. PTEN positive was defined as the total H-score > 0, that is with any percentage of positively stained cells.

Statistical methods

Patients were classified for objective response based on RECIST v1.0 assessments conducted on the trial (14). Patients were classified into responders and nonresponders on the basis of the best overall response and changes in tumor volume. “Responders” had achieved complete response (CR), partial response (PR), or stable disease (SD) with no tumor growth in sum of target lesions, per RECIST v1.0. Best response for “nonresponders” was progression of disease (PD), SD with any tumor growth, or “unknown” response with PFS < 5.5 months on the first-line everolimus, PFS < 2 months on the second-line everolimus. Others with “unknown” response status were not included in the best overall response assessment. PFS was defined as the time from randomization to the date of the first radiographically documented disease progression or death due to any cause during or after the first-line treatment.

Associations between biomarker status (“MT” vs. “WT” for individual genes; “PTEN positive” vs. “PTEN negative”) and response categories (“responder” vs. “nonresponder”) were tested via Fisher exact test. Kaplan–Meier estimates of the median PFS were generated for everolimus patients dichotomizing the study population by biomarker status for individual genes of interest. Estimates of the HR (95% CI) and 2-sided P values were obtained using a Cox proportional hazards model stratified by Memorial Sloan Kettering Cancer Center (MSKCC) risk criteria at randomization. The multivariate models incorporating gender, number of metastatic sites (1 vs. ≥2), and the underlying histologic subtype (clear vs. non–clear cell) as covariates were also fit to these data.

Out of 471 patients treated with everolimus on RECORD-3, 258 contributed NGS data (128 receiving first-line everolimus and 56 second-line everolimus), 213 PTEN IHC data (100 receiving first-line everolimus, 49 second-line everolimus), and 168 patients both (80 and 37 patients receiving everolimus in first-line and second-line, respectively).

Clinical characteristics and demographics for everolimus-treated patients are summarized in Table 1 and were reflective of the overall RECORD-3 population (1, 15). Of the 238 patients treated with the first-line everolimus, 164 (75.5%) were deemed nonresponders and 53 (24.4%) were responders. The median PFS with the first-line everolimus across patients entering the biomarker analyses was comparable to the overall population [all receiving the first-line everolimus, 7.9 months (95% confidence interval [CI], 5.6–8.3) vs. those with NGS data, 8.1 months (95% CI, 5.4–10.5), or IHC data, 8.1 months (95% CI, 5.5–10.2)].

Table 1.

Patient characteristics for first-line everolimus and first-line sunitinib switch to second-line everolimus, by NGS and PTEN availability

Patients treated with first-line everolimusPatients treated with first sunitinib switch to second-line everolimus
Patients, n (%)FAS total N = 238NGS available N = 128PTEN available N = 100FAS total N = 99NGS available N = 56PTEN available N = 49
Age 
 Mean (SD) 61 (12) 62 (13) 63 (11) 61 (10) 62 (10) 61 (10) 
Gender 
 Male 166 (70) 92 (72) 68 (68) 74 (75) 44 (79) 37 (76) 
 Female 72 (30) 36 (28) 32 (32) 25 (25) 12 (21) 12 (24) 
Metastatic sites 
 ≤1 76 (32) 42 (33) 35 (35) 32 (32) 18 (32) 18 (37) 
 >2 162 (68) 86 (67) 65 (65) 67 (68) 38 (68) 31 (63) 
Histology 
 Clear cell 205 (86) 109 (85) 87 (87) 83 (84) 45 (80) 39 (80) 
 Non–clear cell 31 (13) 18 (14) 13 (13) 16 (16) 11 (20) 10 (20) 
 Missing 2 (1) 1 (1) 0 (0) 0 (0) 0 (0) 0 (0) 
MSKCC risk category at the time of starting first-line therapy 
 Favorable (0 risk) 85 (36) 45 (35) 41 (41) 31 (31) 21 (38) 16 (33) 
 Intermediate (1–2 risk) 124 (52) 69 (54) 48 (48) 55 (56) 28 (50) 27 (55) 
 Poor (≥ 3 risk) 29 (12) 14 (11) 11 (11) 13 (13) 7 (13) 6 (12) 
Best overall response to everolimus therapy 
 CR 1 (0) 0 (0) 1 (1) 0 (0) 0 (0) 0 (0) 
 PR 18 (8) 12 (9) 9 (9) 2 (2) 1 (2) 0 (0) 
 SD 137 (58) 72 (56) 60 (60) 39 (39) 22 (39) 22 (45) 
 PD 49 (21) 30 (23) 22 (22) 37 (37) 18 (32) 16 (33) 
 Unknown 33 (14) 14 (11) 8 (8) 21 (21) 15 (27) 11 (22) 
Patients treated with first-line everolimusPatients treated with first sunitinib switch to second-line everolimus
Patients, n (%)FAS total N = 238NGS available N = 128PTEN available N = 100FAS total N = 99NGS available N = 56PTEN available N = 49
Age 
 Mean (SD) 61 (12) 62 (13) 63 (11) 61 (10) 62 (10) 61 (10) 
Gender 
 Male 166 (70) 92 (72) 68 (68) 74 (75) 44 (79) 37 (76) 
 Female 72 (30) 36 (28) 32 (32) 25 (25) 12 (21) 12 (24) 
Metastatic sites 
 ≤1 76 (32) 42 (33) 35 (35) 32 (32) 18 (32) 18 (37) 
 >2 162 (68) 86 (67) 65 (65) 67 (68) 38 (68) 31 (63) 
Histology 
 Clear cell 205 (86) 109 (85) 87 (87) 83 (84) 45 (80) 39 (80) 
 Non–clear cell 31 (13) 18 (14) 13 (13) 16 (16) 11 (20) 10 (20) 
 Missing 2 (1) 1 (1) 0 (0) 0 (0) 0 (0) 0 (0) 
MSKCC risk category at the time of starting first-line therapy 
 Favorable (0 risk) 85 (36) 45 (35) 41 (41) 31 (31) 21 (38) 16 (33) 
 Intermediate (1–2 risk) 124 (52) 69 (54) 48 (48) 55 (56) 28 (50) 27 (55) 
 Poor (≥ 3 risk) 29 (12) 14 (11) 11 (11) 13 (13) 7 (13) 6 (12) 
Best overall response to everolimus therapy 
 CR 1 (0) 0 (0) 1 (1) 0 (0) 0 (0) 0 (0) 
 PR 18 (8) 12 (9) 9 (9) 2 (2) 1 (2) 0 (0) 
 SD 137 (58) 72 (56) 60 (60) 39 (39) 22 (39) 22 (45) 
 PD 49 (21) 30 (23) 22 (22) 37 (37) 18 (32) 16 (33) 
 Unknown 33 (14) 14 (11) 8 (8) 21 (21) 15 (27) 11 (22) 

Abbreviations: FAS, full analysis set; PD, progressive disease.

Exon coverage was >100× in >97% of samples. Among the 184 everolimus-treated patients with the NGS data, mutation rates in TSC1, TSC2, and mTOR were 6% (11 of 184), 4.4% (8 of 184), and 8.2% (15 of 184), respectively. Forty-four percent of patients (81 of 184) harbored alterations in at least 1 PI3K pathway component (Supplementary Table S1). Radiographic response status could be determined in 166 patients with the known NGS status (first-line, n = 128; second-line, n = 38). Based on prior reports (2, 3), we were most interested in 3 genes: TSC1, TSC2, and mTOR. Although the proportion of patients labeled as responders was numerically higher in patients harboring mutations in TSC1, TSC2, or mTOR (36%) compared to those patients with WT for all 3 genes (21.3%), this difference was not significant on Fisher exact test (P = 0.1268; see Fig. 1A). Similarly, no correlation with the response status was seen when broadening NGS criteria to include alterations in TSC1, TSC2, mTOR, or PTEN (Fisher exact P = 0.6823), or alterations in any 1 of 17 PI3K pathway component genes included in IMPACT (Fisher exact P = 1.0000; see Table 2; Supplementary Fig. S1).

Figure 1.

A, Response to everolimus (first-line or second-line) and genomic status. Responder: CR; PR/SD with no tumor growth. Nonresponder: PD/SD with any tumor growth. “Unknown” with PFS < 5.5 months for first-line and PFS < 2 months for second-line. *Patients with best overall response as “unknown” that did not meet the criteria stated above were not included in the assessment. The pathways were defined as mutated if 1 or more alterations are detected in any of the genes in that pathway. B, Kaplan–Meier curves for PFS by mTOR/TSC1/TSC2 status in 128 patients receiving first-line everolimus. HR (95% CI) and P value from a multivariate Cox model, stratified by MSKCC risk groups with terms for pathway, gender, cell histology, and number of metastatic sites. Median PFS (95% CI) estimated by Kaplan–Meier estimator. Mutation status = MT if 1 or more alterations are detected in mTOR, TSC1, or TSC2. WT = wild type for all 3 genes.

Figure 1.

A, Response to everolimus (first-line or second-line) and genomic status. Responder: CR; PR/SD with no tumor growth. Nonresponder: PD/SD with any tumor growth. “Unknown” with PFS < 5.5 months for first-line and PFS < 2 months for second-line. *Patients with best overall response as “unknown” that did not meet the criteria stated above were not included in the assessment. The pathways were defined as mutated if 1 or more alterations are detected in any of the genes in that pathway. B, Kaplan–Meier curves for PFS by mTOR/TSC1/TSC2 status in 128 patients receiving first-line everolimus. HR (95% CI) and P value from a multivariate Cox model, stratified by MSKCC risk groups with terms for pathway, gender, cell histology, and number of metastatic sites. Median PFS (95% CI) estimated by Kaplan–Meier estimator. Mutation status = MT if 1 or more alterations are detected in mTOR, TSC1, or TSC2. WT = wild type for all 3 genes.

Close modal
Table 2.

Response to everolimus (first-line or second-line) and genomic status

PathwayMutation statusNonresponder (n = 127)Responder (n = 39)TotalFisher exact P value
n (%)n (%)
mTOR/TSC1/TSC2 Alt 16 (64) 9 (36) 25 0.1268 
 WT 111 (78.7) 30 (21.3) 141  
mTOR/TSC1/TSC2/PTEN Alt 32 (74.4) 11 (25.6) 43 0.6823 
 WT 95 (77.2) 28 (22.8) 123  
PI3K pathway Alt 54 (76.1) 17 (23.9) 71 1.0000 
 WT 73 (76.8) 22 (23.2) 95  
PathwayMutation statusNonresponder (n = 127)Responder (n = 39)TotalFisher exact P value
n (%)n (%)
mTOR/TSC1/TSC2 Alt 16 (64) 9 (36) 25 0.1268 
 WT 111 (78.7) 30 (21.3) 141  
mTOR/TSC1/TSC2/PTEN Alt 32 (74.4) 11 (25.6) 43 0.6823 
 WT 95 (77.2) 28 (22.8) 123  
PI3K pathway Alt 54 (76.1) 17 (23.9) 71 1.0000 
 WT 73 (76.8) 22 (23.2) 95  

The data from both first-line everolimus and second-line everolimus patients are combined.

Responder*: CR; PR; SD with no tumor growth.

Nonresponder*: PD; SD with any tumor growth. “Unknown” with PFS < 5.5 months for first-line and PFS < 2 months for second-line.

*Patients with best overall response as “unknown” that did not meet the criteria stated above were not included in the assessment.

The pathways were defined as mutated (MT) if 1 or more alterations are detected in any of the genes in that pathway.

Percentage is computed with row totals in the denominator.

PI3K pathway was defined as the following 17 genes: AKT1, AKT2, mTOR, TSC1, TSC2, PTEN, PIK3CA, PIK3C2G, PIK3C3, PIK3CB, PIK3CD, PIK3CG, PIK3R, PIK3R2, PIK3R3, RICTOR, and RPTOR.

Recognizing that established radiographic response criteria per RECIST have previously been deemed suboptimal to assess the benefit from rapalog therapy in this disease (16), we were interested to investigate the correlation of the biomarker status with PFS, an outcome measure deemed to better reflect rapalog outcomes. NGS and PFS data were correlated in 128 patients treated with the first-line everolimus. For 16 patients (12.5%) harboring mutations in mTOR, TSC1, or TSC2, the median PFS with everolimus was 10.3 months (95% CI, 2.2–13.9) versus 8.1 months (95% CI, 5.4–10.5) in 112 patients whose tumors were WT for all 3 genes (Fig. 1B). This difference was neither significant on univariate analysis (Cox PH model; HR, 1.0; 95% CI, 0.5–1.8; P = 0.8951) nor on multivariate testing stratified by MSKCC risk groups and incorporating gender, RCC histology, and the number of metastatic sites (Cox PH model; HR, 1.1; 95% CI, 0.6–2.1; P = 0.8056). Similarly, when increasing the sample size by including patients receiving everolimus in the first-line and second-line settings (n = 184), the median PFS did not differ significantly between 29 patients (15.8%) with the alterations vs 155 patients (84.2%) without alterations in mTOR, TSC1, or TSC2 on stratified multivariate testing (Cox proportional hazard, 1.0; 95% CI, 0.6–1.6; P = 0.8515; Supplementary Fig. S2).

Several caveats must be considered when considering the tumor mutation profiles with respect to their functional effects (17), and we added further analyses to address these to our best ability.

The majority of mTOR mutations detected here have previously been deemed activating, per publicly available databases and/or prior investigations by our group (18). Two patients in the RECORD-3 dataset harbored mTOR mutations of unclear functional significance and were removed from the dataset for a repeat analysis. TSC1 and TSC2 being tumor suppressors, we hypothesized that LOH was likely critical for complete functional loss of the hamartin–tuberin complex with downstream mTORC1 hyperactivation. Eleven patients had retained the second WT allele on FACETs analysis (6 with TSC1 mutations, 5 with TSC2) and were removed for this added investigation. We repeated our analyses correlating mTOR, TSC1, TSC2 mutation status, and the outcome with everolimus by applying these considerations for the 3 genes, removing 13 patients with a variant of unclear functional significance, per the reasoning above. Despite these more stringent molecular criteria, no association between mutation status and outcome on first-line everolimus was seen, neither for response (Fisher exact test, P = 0.2561) nor for PFS (multivariate Cox model, P = 0.9446; Supplementary Table S2; Supplementary Fig. S3A).

Concurrent mutation in other drivers may supersede the biological effects of mTORC1-activating mutations. In prior analyses from RECORD-3, we found that mutations in BAP1 correlated adversely with PFS in everolimus-treated patients (15). Three patients harbored mutations in BAP1 and were removed for a repeat analysis. Despite exclusion of BAP1-mutant patients, no correlation was seen between TSC1/TSC2/mTOR mutation status and PFS, neither in all patients with NGS data treated in the first-line (multivariate Cox PH, 1.2; 95% CI, 0.6–2.5; P = 0.6922; Supplementary Fig. S3B) nor those with functionally annotated mTOR mutations and/or biallelic loss of TSC1/TSC2 (multivariate Cox PH, 1.2; 95% CI, 0.5–2.7; P = 0.6763; Supplementary Fig. S3C).

A total of 213 patients in the RECORD-3 dataset provided PTEN IHC data. PTEN was considered to be “lost” if PTEN IHC H-score was 0, whereas patients were categorized as “PTEN-positive” if the IHC score was >0. Overall, PTEN loss was seen in 114 patients (53.5%). Across 100 patients treated with first-line everolimus, PTEN expression was retained in 50 (50%); their median PFS was 5.3 months, significantly inferior to 50 patients (50%) with the lost protein expression who achieved a median PFS of 10.5 months (multivariate COX PH, 2.5; 95% CI, 1.5–4.1; P = 0.0004; Fig. 2A). The distribution of RECIST responder/nonresponder differed numerically, but the difference was not significant (PTEN-null patients, 29.6% of responders; PTEN-positive patients, 18.8% of responders; Fisher exact test, P = 0.3279; Fig. 2C). Notably, such differences were not observed in 113 patients treated with the first-line sunitinib with valid IHC data. Here, the median PFS with PTEN retained versus lost was 10.9 months versus 10.3 months (multivariate COX PH, 0.8; 95% CI, 0.5–1.4; P = 0.4752; Fig. 2B); the proportion of sunitinib responder/nonresponder did not correlate with the PTEN expression status (PTEN-null patients, 29.8% of responders; PTEN-positive patients, 38.9%; Fisher exact test, P = 0.3773; Fig. 2D). We compared distribution of PTEN IHC status in patients with underlying clear cell versus non–clear cell histology. Nonparametric testing suggested higher median H-scores in patients with non–clear cell RCC (Wilcoxon rank-sum, P = 0.047), but this observation was limited by small sample size (183 vs. 30 cases) and overlapping CIs (see also Supplementary Fig. S4). We found no association between PTEN expression status and the presence of mutation events for PI3K pathway components: in 168 patients who contributed both NGS and IHC data, rate of PTEN loss by IHC was 40% and 47% for PI3K pathway mutant versus WT tumors (Fisher exact P = 1.0). NGS status for individual genes of interest distributed similarly for patients with lost versus retained PTEN on IHC (see Supplementary Table S3). Further, we tested whether PTEN loss by IHC was related to PTEN copy-number loss, determined via NGS. Thirty-five patients had PTEN status by IHC and copy-number status by NGS (FACET analysis off of IMPACT data) available. Fisher exact test suggested a possible association between PTEN IHC status and PTEN copy number by NGS (P = 0.02; see also Supplementary Table S4).

Figure 2.

A, KM curves for PFS by PTEN expression status in patients treated with first-line everolimus. HR (95% CI) and P value from a multivariate Cox model, stratified by MSKCC risk groups with terms for biomarker, gender, cell histology, and number of metastatic sites. The median PFS (95% CI) was estimated by Kaplan–Meier estimator. PTEN loss: PTEN H-score of 0. PTEN positive: PTEN H-score > 0. B, KM curves for PFS by PTEN expression in patients treated with first-line sunitinib. C, Box plot comparing responder/nonresponder in first-line everolimus PTEN expression status. D, Box plot comparing responder/nonresponder first-line sunitinib and PTEN expression status. Responder: CR; PR/SD with no tumor growth. Nonresponder: PD/SD with any tumor growth. “Unknown” with PFS < 5.5 months for first-line. *Patients with best overall response as unknown that did not meet the criteria stated above were not included in the assessment.

Figure 2.

A, KM curves for PFS by PTEN expression status in patients treated with first-line everolimus. HR (95% CI) and P value from a multivariate Cox model, stratified by MSKCC risk groups with terms for biomarker, gender, cell histology, and number of metastatic sites. The median PFS (95% CI) was estimated by Kaplan–Meier estimator. PTEN loss: PTEN H-score of 0. PTEN positive: PTEN H-score > 0. B, KM curves for PFS by PTEN expression in patients treated with first-line sunitinib. C, Box plot comparing responder/nonresponder in first-line everolimus PTEN expression status. D, Box plot comparing responder/nonresponder first-line sunitinib and PTEN expression status. Responder: CR; PR/SD with no tumor growth. Nonresponder: PD/SD with any tumor growth. “Unknown” with PFS < 5.5 months for first-line. *Patients with best overall response as unknown that did not meet the criteria stated above were not included in the assessment.

Close modal

Everolimus has been studied in other diseases, and the additional analyses were undertaken to test the same candidate biomarkers in gastric cancer and HCC. For 47 everolimus-treated gastric cancer patients with NGS data on GRANITE-1, TSC1/TSC2/mTOR mutation status did not correlate with PFS (log-rank, P = 0.7998; Table 3; Supplementary Fig. S5A; ref. 6). We found no association between PFS and PTEN expression across 121 everolimus-treated patients from the same trial (stratified Cox model, P = 0.5408; see also Table 3; Supplementary Fig. S5B). Notably, expression was lost in only 15 tested tumors (12.4%). Similarly, in 80 HCC patients treated with everolimus on EVOLVE-1 (19), we found no association between TSC1/TSC2/mTOR status and time to progression (TTP) with numerically longer TTP in patients with WT tumors (stratified Cox PH model, P = 0.6476; see Table 3; Supplementary Fig. S6).

Table 3.

Association between PFS and mTOR/TSC1/TSC2 and PTEN expression status and association between TTP and mTOR/TSC1/TSC2 status

Randomized treatment groupNMedian PFS (95% CI)HR (95% CI) P value Alt/WT
GRANITE-1: Everolimus + best supportive care in patients with advanced gastric cancer 47   
mTOR/TSC1/TSC2 mutation status 
 Alt 1.6 (1.4–2.7) 0.9 (0.6–1.5); P = 0.7998 
 WT 38 1.5 (1.4–2.1)  
PTEN group 121   
 Loss 15 2.3 (1.4–4.2) 1.2 (0.7–2.2); P = 0.5408 
 Positive 106 1.5 (1.4–1.8)  
EVOLVE-1: Everolimus in patients with advanced HCC  Median TTP (95% CI)  
mTOR/TSC1/TSC2 mutation status 80   
 Alt 16 1.7 (1.3–3.9) 1.2 (0.6–2.2); P = 0.6476 
 WT 64 2.8 (1.5–4.5)  
Randomized treatment groupNMedian PFS (95% CI)HR (95% CI) P value Alt/WT
GRANITE-1: Everolimus + best supportive care in patients with advanced gastric cancer 47   
mTOR/TSC1/TSC2 mutation status 
 Alt 1.6 (1.4–2.7) 0.9 (0.6–1.5); P = 0.7998 
 WT 38 1.5 (1.4–2.1)  
PTEN group 121   
 Loss 15 2.3 (1.4–4.2) 1.2 (0.7–2.2); P = 0.5408 
 Positive 106 1.5 (1.4–1.8)  
EVOLVE-1: Everolimus in patients with advanced HCC  Median TTP (95% CI)  
mTOR/TSC1/TSC2 mutation status 80   
 Alt 16 1.7 (1.3–3.9) 1.2 (0.6–2.2); P = 0.6476 
 WT 64 2.8 (1.5–4.5)  

HR (95% CI) and P value from a univariate Cox model, stratified by stratum STU2A.

The median PFS (95% CI) and median TTP (95% CI) were estimated by Kaplan–Meier estimator.

Mutation status = MT if 1 or more alterations are detected in any of the genes in that pathway.

PTEN loss: PTEN H-score of 0. PTEN positive: PTEN H-score > 0.

The advent of molecularly targeted therapies for RCC predated innovations in diagnostic molecular pathology, which could have allowed for broad-scale genomic testing during drug development. Consequently, none of the many targeted agents approved for patients with advanced RCC received registration in a defined genomic setting (20). Hyperactivation of mTOR signaling had long been recognized as a central element of RCC biology (21, 22), ultimately paving the way for the development and broad application of mTOR inhibitors everolimus and temsirolimus in RCC (23).

In recent years, NGS has become more broadly available, and current efforts in drug development support the notion of studying targeted therapies using genome-driven approaches (24, 25). Alterations within the PI3K pathway are seen recurrently in RCC (26), raising the question whether integration of such information could better inform our use of approved mTOR inhibitors in RCC. Although clinical benefit is only modest for most patients (27, 28), isolated patients can achieve extended benefit from monotherapy (29). We previously presented a retrospective case series with detailed genomic analysis of tumor tissue for 5 patients with outlier responses to everolimus or temsirolimus (30), and were able to detect plausible somatic alterations in 11 of 14 specimens analyzed. This included loss of function mutations in TSC1 with concurrent LOH and/or activating mutations in mTOR itself. Critically, these outlier patients had diffuse activation of the mTOR pathway as determined by multiregional sequencing. Kwiatkowski and colleagues subsequently published a pooled retrospective analysis of 79 patients, including 43 who had achieved benefit, defined as either a PR or SD ≥ 6 months on rapalog therapy (3). Mutations in TSC1, TSC2, or mTOR were numerically more common in patients with treatment benefit (28% vs. 11% in those without benefit), although the difference did not reach statistical significance. Although both reports suggested relevance to these somatic mutations in some patients, particularly those with unusual responses, they were insufficient to prove the predictive value of alterations in TSC1/TSC2/mTOR in an unselected population of patients. Here, we present such data correlating mutation status and therapeutic outcomes across 184 patients treated with everolimus on RECORD-3. We were neither able to show significantly better radiographic effect nor improved PFS in patients harboring mutations of interest. The same was the case when broadening our analysis to include mutation status for other PI3K pathway components. Ultimately, our interpretation is that mutation status can provide the explanation for outlier benefit in some patients; however, it cannot be considered a sufficiently reliable biomarker to guide our choice of agent, recognizing that 64% of patients with somatic mutation(s) in at least 1 of these 3 genes were unable to achieve meaningful radiographic benefit with the first-line everolimus. Rate of radiographic improvement was numerically but not statistically different when comparing everolimus-treated patients by mutation status, and no difference could be detected in PFS. More so, the median first-line PFS of 10.3 months (95% CI, 2.2–13.9) achieved by TSC1/TSC2/mTOR–mutant patients with everolimus was comparable to what sunitinib-treated patients achieved regardless of mutation status on the comparator arm (median PFS, 10.7 months; range, 8.2–11.5 months). We did specifically investigate outlier cases in this dataset and compared mutation frequency for TSC1, TSC2, and mTOR in those everolimus-treated patients achieving >2× the median PFS recorded for first line everolimus on this study (15.8 months). Limited by sample size, we noted no differences in mutation frequency for mTOR, TSC1, or TSC2 (see Supplementary Table S5). None of the above provides compelling evidence that an untreated patient should receive an mTOR inhibitor in lieu of standard tyrosine-kinase inhibitor (TKI) in the first-line setting, based on NGS mutation status for TSC1/TSC2/mTOR. Although sample size may limit our ability to detect a more subtle signal, it seems unlikely that there is the strong predictive implication that earlier reports had hypothesized.

Technical reasons for this lack of association were investigated. We have previously argued that scrutiny is indicated when interpreting genomic data for biomarker development in this setting (30). With this in mind, we pursued additional analyses incorporating individual review of NGS data for patients found to carry mutations of interest in their tumors. For mTOR mutations, we excluded those with mutations previously not annotated and of unclear functional significance; for mutations in the tumor suppressors TSC1/TSC2, we excluded cases without clear evidence of LOH, that is, those that may have retained a functional allele of the gene. Despite the removal of such cases from our dataset, we were unable to detect significant differences in the therapeutic outcome per mutation status. We were interested to assess if concurrent mutations in other genes may have influenced outcome and removed cases with concurrent BAP1 mutations, seen at relevant frequencies in this disease and previously deemed prognostic in this setting; yet again, no association with TSC1/TSC2/mTOR mutation status could be detected. Other pertinent limitations could not be investigated here, including effects of tumor heterogeneity and sampling error, clonal evolution, the age of archival specimens used for this analysis, and the potential relevance of testing tissue from the primary tumor versus metastatic sites. All of these, however, would equally apply in the real-world setting, as clinicians apply NGS data from commercial assays.

Recent efforts in drug development support the notion of studying targeted therapies using genome-driven approaches for genome-driven oncology care (5, 24). However, some such efforts are demonstrating that the same alteration may have different predictive value across disease entities (25). We tested association of mutation status for mTOR/TSC1/TSC2 and the outcome with everolimus in 2 non-RCC cohorts using prospectively collected archival specimens and found no association with PFS or TTP in patients with advanced gastric and HCC, respectively. Conversely, a prior report suggested that correlations with mutation status may exist in patients with HER2—overexpressing advanced breast cancer, in which, everolimus was added to trastuzumab and chemotherapy (31). Altogether, these differences argue that oncogenomics, at least in the case of rapalog therapy and PI3K pathway alterations, should be interpreted in a disease-specific context. This may of course differ for other molecular targets and classes of agents (32).

Surprisingly, PTEN loss by IHC correlated with outcome for patients receiving first-line everolimus, whose median PFS was 10.5 months, almost twice that of the patients with retained PTEN (5.3 months; HR 2.5; P = 0.0004). Such associations were not seen in patients receiving sunitinib on the same trial (median PFS, 10.3 and 10.9 months, respectively; HR 0.8; see Fig. 2B). PTEN loss (IHC) was numerically more common in first-line everolimus patients with PFS > 2× median than those <2× median (71% vs. 43%, respectively; see Supplementary Table S5). These findings suggest that the PTEN expression status may have a clinically meaningful predictive value, and could serve as a putative biomarker, which warrants further investigation in independent RCC cohorts. Notably, loss of expression was seen in >50% of patients, a rate that is high in comparison to other diseases. Although functional loss of PBRM1 was recently implicated as a mechanism to increase MTORC1 signaling in this disease (33), supported by the fact that similar degree of benefit was achieved with sunitinib and everolimus in PBRM1-mutant patients on RECORD-3 (15), we saw no correlation between PTEN expression and NGS mutation status for PI3K pathway components. FACET analysis of IMPACT data to infer PTEN copy-number status was available for only 35 patients with IHC data; with the limitation of this small sample size, an association between IHC and NGS copy-number status was seen, suggesting that genomic loss is likely the cause for low expression status in many if not most cases. PTEN-deficient everolimus patients had comparable outcomes to those receiving sunitinib. Hence, IHC testing, based on these data, may not provide compelling data to deviate from the current standard of care. That being said, a recent randomized trial demonstrated a notable efficacy for the combination of everolimus plus lenvatinib in TKI pretreated patients, ultimately leading to the approval of this combination (34). Certainly, the PTEN IHC data presented here would be of interest for rapalog-containing combination strategies, for example, could be applied as the combination of lenvatinib/everolimus, is being compared with other strategies on a currently ongoing phase III trial (NCT02811861).

In summary, this is the largest and only prospective analysis of patients with advanced RCC receiving rapalog therapy to date that allowed correlation of treatment outcomes with the NGS mutation status for PI3K pathway components. No significant association with the therapeutic benefit and somatic alterations in TSC1/TSC2/mTOR was seen. We did detect clinically relevant differences in PFS based on PTEN IHC expression status, a finding that suggests predictive potential for this tissue biomarker noted in half of the patients studied here. This warrants further study in independent cohorts.

M.H. Voss reports receiving commercial research grants from Bristol-Myers Squibb and Genentech, and is a consultant/advisory board member for Alexion, Calithera, Eisai, Exelixis, Natera, Novartis, and Pfizer. D. Chen holds ownership interest (including patents) in NVS stock. X. Han holds ownership interest (including patents) in Novartis. J.J. Hsieh is a consultant/advisory board member for Eisai. R.J. Motzer is a consultant/advisory board member for Exelixis, Genentech/Roche, Mercke, Novartis, and Pfizer. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.H. Voss, D. Chen, A. Reising, Y.-B. Chen, J.J. Hsieh, A.A. Hakimi, R.J. Motzer

Development of methodology: D. Chen, J. Shi, J.J. Hsieh, R.J. Motzer

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.H. Voss, D. Chen, A. Redzematovic, Y.-B. Chen, P. Patel, R.J. Motzer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.H. Voss, D. Chen, M. Marker, J. Shi, J. Xu, I. Ostrovnaya, V.E. Seshan, X. Han, J.J. Hsieh, A.A. Hakimi, R.J. Motzer

Writing, review, and/or revision of the manuscript: M.H. Voss, D. Chen, A. Reising, J. Shi, J. Xu, I. Ostrovnaya, V.E. Seshan, Y.-B. Chen, X. Han, J.J. Hsieh, A.A. Hakimi, R.J. Motzer

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Chen, J. Shi, A. Redzematovic, P. Patel, J.J. Hsieh, R.J. Motzer

Study supervision: P. Patel, R.J. Motzer

Other (project management activities and provided support on sample collection, data transfers, data review and contracts with analysis lab): P. Patel

We thank Anuradha Bandaru of Novartis Healthcare Pvt. Ltd., for medical writing and editorial support. The study was designed by academic investigators and representatives of the funder (Novartis Pharmaceuticals Corporation, East Hanover, New Jersey). Patients treated at MSKCC were supported in part by MSKCC Support Grant/Core Grant (P30 CA008748) and funded by J. Randall & Kathleen L. MacDonald Research Fund. All authors contributed to the interpretation of data and the subsequent writing, reviewing, and amending of the report; the first draft of the report was prepared by the first author (Martin H. Voss) and a medical writer employed by the funder. All authors vouch for the accuracy and completeness of the data and attest that the study conformed to the protocol and statistical analysis plan.

The study was designed by academic investigators and representatives of the funder (Novartis Pharmaceuticals Corporation, East Hanover, New Jersey). Patients treated at MSKCC were supported in part by MSKCC Support Grant/Core Grant (P30 CA008748) and funded by J. Randall & Kathleen L. MacDonald Research Fund.

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.
Motzer
RJ
,
Escudier
B
,
Oudard
S
,
Hutson
TE
,
Porta
C
,
Bracarda
S
, et al
Efficacy of everolimus in advanced renal cell carcinoma: a double-blind, randomised, placebo-controlled phase III trial
.
Lancet
2008
;
372
:
449
56
.
2.
Voss
MH
,
Hakimi
AA
,
Pham
CG
,
Brannon
AR
,
Chen
YB
,
Cunha
LF
, et al
Tumor genetic analyses of patients with metastatic renal cell carcinoma and extended benefit from mTOR inhibitor therapy
.
Clin Cancer Res
2014
;
20
:
1955
64
.
3.
Kwiatkowski
DJ
,
Choueiri
TK
,
Fay
AP
,
Rini
BI
,
Thorner
AR
,
de Velasco
G
, et al
Mutations in TSC1, TSC2, and MTOR are associated with response to rapalogs in patients with metastatic renal cell carcinoma
.
Clin Cancer Res
2016
;
22
:
2445
52
.
4.
Tao
JJ
,
Schram
AM
,
Hyman
DM
. 
Basket studies: redefining clinical trials in the era of genome-driven oncology
.
Annu Rev Med
2018
;
69
:
319
31
.
5.
McNeil
C
. 
NCI-MATCH launch highlights new trial design in precision-medicine era
.
J Natl Cancer Inst
2015
;
107
:
4
8
.
6.
Ohtsu
A
,
Ajani
JA
,
Bai
YX
,
Bang
YJ
,
Chung
HC
,
Pan
HM
, et al
Everolimus for previously treated advanced gastric cancer: results of the randomized, double-blind, phase III GRANITE-1 study
.
J Clin Oncol
2013
;
31
:
3935
43
.
7.
Gnirke
A
,
Melnikov
A
,
Maguire
J
,
Rogov
P
,
LeProust
EM
,
Brockman
W
, et al
Solution hybrid selection with ultra-long oligonucleotides for massively parallel targeted sequencing
.
Nat Biotechnol
2009
;
27
:
182
9
.
8.
Wagle
N
,
Berger
MF
,
Davis
MJ
,
Blumenstiel
B
,
Defelice
M
,
Pochanard
P
, et al
High-throughput detection of actionable genomic alterations in clinical tumor samples by targeted, massively parallel sequencing
.
Cancer Discov
2012
;
2
:
82
93
.
9.
Shen
R
,
Seshan
VE
. 
FACETS: allele-specific copy number and clonal heterogeneity analysis tool for high-throughput DNA sequencing
.
Nucleic Acids Res
2016
;
44
:
e131
.
10.
Zhu
AX
,
Chen
D
,
He
W
,
Kanai
M
,
Voi
M
,
Chen
LT
, et al
Integrative biomarker analyses indicate etiological variations in hepatocellular carcinoma
.
J Hepatol
2016
;
65
:
296
304
.
11.
Tang
L
,
Li
X
,
Gao
Y
,
Chen
L
,
Gu
L
,
Chen
J
, et al
Phosphatase and tensin homolog (PTEN) expression on oncologic outcome in renal cell carcinoma: a systematic review and meta-analysis
.
PLoS One
2017
;
12
:
e0179437
.
12.
Figlin
RA
,
de Souza
P
,
McDermott
D
,
Dutcher
JP
,
Berkenblit
A
,
Thiele
A
, et al
Analysis of PTEN and HIF-1alpha and correlation with efficacy in patients with advanced renal cell carcinoma treated with temsirolimus versus interferon-alpha
.
Cancer
2009
;
115
:
3651
60
.
13.
Pantuck
AJ
,
Seligson
DB
,
Klatte
T
,
Yu
H
,
Leppert
JT
,
Moore
L
, et al
Prognostic relevance of the mTOR pathway in renal cell carcinoma: implications for molecular patient selection for targeted therapy
.
Cancer
2007
;
109
:
2257
67
.
14.
Therasse
P
,
Arbuck
SG
,
Eisenhauer
EA
,
Wanders
J
,
Kaplan
RS
,
Rubinstein
L
, et al
New guidelines to evaluate the response to treatment in solid tumors. European Organization for Research and Treatment of Cancer, National Cancer Institute of the United States, National Cancer Institute of Canada
.
J Natl Cancer Inst
2000
;
92
:
205
16
.
15.
Hsieh
JJ
,
Chen
D
,
Wang
PI
,
Marker
M
,
Redzematovic
A
,
Chen
YB
, et al
Genomic biomarkers of a randomized trial comparing first-line everolimus and sunitinib in patients with metastatic renal cell carcinoma
.
Eur Urol
2017
;
71
:
405
14
.
16.
Oudard
S
,
Thiam
R
,
Fournier
LS
,
Medioni
J
,
Lamuraglia
M
,
Scotte
F
, et al
Optimisation of the tumour response threshold in patients treated with everolimus for metastatic renal cell carcinoma: analysis of response and progression-free survival in the RECORD-1 study
.
Eur J Cancer
2012
;
48
:
1512
8
.
17.
Voss
MH
,
Hsieh
JJ
. 
Therapeutic Guide for mTOuRing through the braided kidney cancer genomic river
.
Clin Cancer Res
2016
;
22
:
2320
2
.
18.
Xu
J
,
Pham
CG
,
Albanese
SK
,
Dong
Y
,
Oyama
T
,
Lee
CH
, et al
Mechanistically distinct cancer-associated mTOR activation clusters predict sensitivity to rapamycin
.
J Clin Invest
2016
;
126
:
3526
40
.
19.
Zhu
AX
,
Kudo
M
,
Assenat
E
,
Cattan
S
,
Kang
YK
,
Lim
HY
, et al
Effect of everolimus on survival in advanced hepatocellular carcinoma after failure of sorafenib: the EVOLVE-1 randomized clinical trial
.
JAMA
2014
;
312
:
57
67
.
20.
Choueiri
TK
,
Motzer
RJ
. 
Systemic therapy for metastatic renal-cell carcinoma
.
N Engl J Med
2017
;
376
:
354
66
.
21.
Robb
VA
,
Karbowniczek
M
,
Klein-Szanto
AJ
,
Henske
EP
. 
Activation of the mTOR signaling pathway in renal clear cell carcinoma
.
J Urol
2007
;
177
:
346
52
.
22.
Lin
F
,
Zhang
PL
,
Yang
XJ
,
Prichard
JW
,
Lun
M
,
Brown
RE
. 
Morphoproteomic and molecular concomitants of an overexpressed and activated mTOR pathway in renal cell carcinomas
.
Ann Clin Lab Sci
2006
;
36
:
283
93
.
23.
Voss
MH
,
Molina
AM
,
Motzer
RJ
. 
mTOR inhibitors in advanced renal cell carcinoma
.
Hematol Oncol Clin North Am
2011
;
25
:
835
52
.
24.
Hyman
DM
,
Taylor
BS
,
Baselga
J
. 
Implementing genome-driven oncology
.
Cell
2017
;
168
:
584
599
.
25.
Hyman
DM
,
Puzanov
I
,
Subbiah
V
,
Faris
JE
,
Chau
I
,
Blay
J-Y
, et al
Vemurafenib in multiple nonmelanoma cancers with BRAF V600 mutations
.
N Engl J Med
2015
;
373
:
726
36
.
26.
Cancer Genome Atlas Research N
. 
Comprehensive molecular characterization of clear cell renal cell carcinoma
.
Nature
2013
;
499
:
43
9
.
27.
Hudes
G
,
Carducci
M
,
Tomczak
P
,
Dutcher
J
,
Figlin
R
,
Kapoor
A
, et al
Temsirolimus, interferon alfa, or both for advanced renal-cell carcinoma
.
N Engl J Med
2007
;
356
:
2271
81
.
28.
Motzer
RJ
,
Escudier
B
,
Oudard
S
,
Hutson
TE
,
Porta
C
,
Bracarda
S
, et al
Phase 3 trial of everolimus for metastatic renal cell carcinoma: final results and analysis of prognostic factors
.
Cancer
2010
;
116
:
4256
65
.
29.
Molina
AM
,
Ginsberg
MS
,
Motzer
RJ
. 
Long-term response with everolimus for metastatic renal cell carcinoma refractory to sunitinib
.
Med Oncol
2011
;
28
:
1527
9
.
30.
Voss
MH
,
Hsieh
JJ
. 
Therapeutic guide for mTOuRing through the braided kidney cancer genomic river
.
Clin Cancer Res
2016
;
22
:
2320
2
.
31.
Andre
F
,
Hurvitz
S
,
Fasolo
A
,
Tseng
LM
,
Jerusalem
G
,
Wilks
S
, et al
Molecular alterations and everolimus efficacy in human epidermal growth factor receptor 2-overexpressing metastatic breast cancers: combined exploratory biomarker analysis from BOLERO-1 and BOLERO-3
.
J Clin Oncol
2016
;
34
:
2115
24
.
32.
Drilon
A
,
Laetsch
TW
,
Kummar
S
,
DuBois
SG
,
Lassen
UN
,
Demetri
GD
, et al
Efficacy of larotrectinib in TRK fusion-positive cancers in adults and children
.
N Engl J Med
2018
;
378
:
731
9
.
33.
Nargund
AM
,
Pham
CG
,
Dong
Y
,
Wang
PI
,
Osmangeyoglu
HU
,
Xie
Y
, et al
The SWI/SNF protein PBRM1 restrains VHL-loss-driven clear cell renal cell carcinoma
.
Cell Rep
2017
;
18
:
2893
906
.
34.
Motzer
RJ
,
Hutson
TE
,
Ren
M
,
Dutcus
C
,
Larkin
J
. 
Independent assessment of lenvatinib plus everolimus in patients with metastatic renal cell carcinoma
.
Lancet Oncol
2016
;
17
:
e4
5
.