Background
Renal cell carcinoma is eighth in incidence and mortality among all cancers as measured in the Surveillance, Epidemiology, and End Results database [
1]. Clear cell renal cell carcinoma (ccRCC) is the most common and well-studied histologic subtype of RCC and carries the highest risk of metastatic spread [
2]. Eighty percent of ccRCC have inactivating mutations in the
Von Hippel Lindau gene which stabilizes hypoxia inducible factors and leads to overexpression of vascular endothelial growth factor receptor and platelet-derived growth factor receptor which promote angiogenesis, tumor growth and metastasis [
3].
Treatment of ccRCC has changed dramatically over a short period of time. Historically ccRCC has been one of a select number of tumors where cytokine therapies such as interleukin-2 or interferon-α have been used to treat metastatic disease [
4,
5]. mTOR inhibitors and tyrosine kinase inhibitors (TKI) against VEGF receptor 2 (VEGFR2) developed as disease specific targeted therapies and remain therapeutic standards [
6]. More recently immune-checkpoint inhibition (ICI), centered on programmed death-1 (PD-1) or programmed death ligand-1 (PD-L1) as well as cytotoxic T lymphocyte antigen 4 (CTLA4), has become a backbone of therapy. In previously untreated metastatic ccRCC, anti-PD1/L1 antibodies combined with VEGFR2 TKI is emerging [
7,
8].
Despite promising activity, many patients have tumors that are refractory to ICI or suffer early progression on treatment. Identifying predictive molecular and clinical markers of resistance is a priority to guide optimal treatment selection. Established biomarkers, such as tumor PD-L1 expression and tumor mutational burden (TMB), have not been demonstrated to have a highly predictive utility in ccRCC [
9,
10]. Expression of PD-L1 in ccRCC is somewhat correlated with improved outcomes to ICI however patients with tumors without PD-L1 expression also achieve responses [
11‐
13]. Composite gene expression profiling (GEP) across tumor types has identified gene signatures that associate with treatment response. The T-cell inflamed GEP comprised of IFNγ signaling and T-cell related genes has correlated with treatment response to immunotherapy in multiple tumor types [
14‐
16]. A gene signature of six VEGF-dependent genes validated as a predictive biomarker for anti-VEGF therapy has been used to assess angiogenic activity in ccRCC [
17,
18].
Clinical variables may more easily be identified in association with treatment resistance and more favorable outcomes to immunotherapy. ECOG and Karnofsky performance status define functional status of cancer patients and are predictive of outcomes to systemic chemotherapy [
19,
20]. The RCC International Metastatic Database Consortium (IDMC) Risk Score defines adverse clinical prognostic risk factors in patients with ccRCC treated with VEGF-targeted therapy [
21,
22]. Smoking status and serum albumin correlate with immunotherapy outcomes in some tumors [
23]. Increased body-mass index (BMI) was found to correlate with improved outcomes in colorectal and lung cancers as well as immunotherapy response in melanoma and other cancers [
24‐
26]. A prognostic signature incorporating both clinical and genomic variables has been developed in breast cancer [
27]. In ccRCC, duration of prior anti-VEGFR2 TKI therapy and neutrophil:lymphocyte ratio (NLR) have been found to be independent predictors of survival [
28].
As immunotherapy has become a backbone therapy in the treatment of ccRCC, predictors of lack of response, or primary resistance, are needed. Here we assessed for clinical characteristics that correlate with outcomes in a cohort of patients with ccRCC treated with anti-PD1/PD-L1 and The Cancer Genome Atlas Kidney Renal Clear Cell Carcinoma cohort (TCGA-ccRCC). Recent biomarker analysis found correlation between angiogenesis, T-effector and myeloid inflammatory gene expression and treatment response in ccRCC [
18,
29]. VEGF signaling has known immunosuppressive activity and preclinical work has suggested anti-VEGF treatment might enhance efficacy of ICI [
30‐
34]. As such, we interrogated the TCGA cohort with angiogenesis and T-cell inflammation gene signatures and identified clinical, neoplasm histologic grade and stage specific associations that may inform treatment selection and support adjuvant or neo-adjuvant use of ICI and/or VEGFR2-TKI.
Methods
Data collection
We performed an international multi-center data collection from patients with stage IV ccRCC who received at least one dose of anti-PD-1/PD-L1 (pembrolizumab, nivolumab, atezolizumab) between 01/01/2011 and 06/01/2018. The participating centers included the University of Chicago Comprehensive Cancer Center (n = 22), Laura and Isaac Perlmutter Comprehensive Cancer Center at NYU Langone (n = 21), Massey Cancer Center at Virginia Commonwealth University (n = 14), London Health Sciences Centre at Western University, Ontario, Canada (n = 17) and Marqués de Valdecilla University Hospital in Santander, Spain (n = 16). Local institutional review board approval, including waiver of consent where appropriate, was obtained at all participating sites using a master data collection protocol.
Study design
De-identified demographic and clinical variables including but not limited to age, gender, BMI, smoking status, performance status as defined by the Eastern Cooperative Oncology Group (ECOG) were collected. Clinical and laboratory data required for calculation of the RCC IMDC risk score were also collected [
21,
22]. Information regarding additional treatments including prior radiation, prior therapies, additional therapies after anti-PD1/L1 treatment and the occurrence of an immune-related adverse event (irAE) were recorded. Clinical outcomes from each center were obtained from their respective electronic medical records with identifiers and dates removed prior to data aggregation.
Radiologic tumor assessment for each patient included computed tomography (CT) scans performed every 8–12 weeks unless otherwise clinically indicated. Patients with either clinical progression or disease progression on first CT evaluation by investigator assessment or death due to cancer prior to first CT evaluation were defined as having primary resistance. Characteristics of patients alive with clinical benefit by investigator assessment on first CT, i.e. those found to not have primary resistance, were then evaluated to identify factors associated with subsequent progression or secondary resistance and survival outcomes.
Statistical analysis
Baseline demographic data was used to generate descriptive statistics. Tabular summaries were presented for overall patient population and those with primary resistance vs. those with clinical benefit. Continuous variables were summarized using descriptive statistics (n, mean, standard deviation, standard error median, minimum, and maximum). Categorical variables were summarized showing the number and percentage (n, %) of patients within each category. Baseline characteristics between patients with primary resistance and those with clinical benefit were compared using Two sample t-test and Chi-square test (or Fisher’s Exact test).
The Kaplan–Meier method was used to estimate the distribution of progression-free survival (PFS) and overall survival (OS) in all patients and the subsets of patients with clinical benefit or primary resistance. PFS was defined as time from start of treatment to progression by investigator assessment or death from any cause. OS was defined as time from the start of treatment until death from any cause. If no event had occurred, patients were censored at their last follow up visit or 6/30/18, whichever came sooner. Cox’s regression model was used to evaluate the correlation between survival endpoints and the variables of interest. Univariable Cox’s model was first implemented to examine the relationship between survival endpoint and each covariate. Covariates with p-value less than 0.10 were then included in the multivariable models, and a backward selection was performed to derive the final multivariable model for PFS and OS. Statistical analyses were done with SAS 9.4. All statistical tests were two-sided and considered significant at p < 0.05.
Analysis of TCGA-ccRCC data set
RNAseq gene expression data (release date January 28, 2016) were downloaded from Broad Institute’s GDAC Firehose website [
35] for primary tumor of 517 patients from TCGA Kidney RCC (TCGA-ccRCC) database. Harmonized survival data were extracted from the previously published TCGA Pan-Cancer study [
36] for progression-free interval event (PFI, defined as “for patient having new tumor event whether it was a progression of disease, local recurrence, distant metastasis, new primary tumors all sites, or died with the cancer without new tumor event, including cases with a new tumor event whose type is N/A” [
36]) and OS. Additional demographic and clinical information were downloaded from Genomic Data Commons data portal (GDC) [
37]. All tumor samples were used for analysis.
For the TCGA-ccRCC patients, the RSEM [
38]—summarized gene level read counts were upper quartile normalized across all tumor samples and log
2 transformed. For each tumor, the scores for the T cell-inflamed (Tinfl) or Angiogenesis (Angio) gene signatures were calculated by averaging the expression level of all genes from each signature after centering and scaling across samples for each gene. The probability of survival including PFI and OS were compared between designated groups by log-rank test. The Cox proportional hazard (PH) model was used to evaluate significance of factor of interest in multivariate model with p-values computed by Wald test in function
coxph from R library survival (v2.43-3).
Gene expression comparison between groups were performed using two-sided Student’s t-test. For multiple comparisons, p-value was adjusted using Benjamini–Hochberg false discovery rate (FDR) correction [
39]. Spearman’s correlation ρ was used for measuring statistical dependence between normalized and log
2-transformed expression level of different gene signatures and was applied in 2 biologically relevant sets: one within non-T cell-inflamed plus intermediate and the other within the T cell-inflamed group.
p < 0.05 was considered statistically significant. Statistical analysis was performed using R (v3.5.2) and Bioconductor.
Discussion
Treatment with immune-checkpoint inhibition (ICI) has changed the treatment paradigm in ccRCC however many do not respond to these treatments and no reliable molecular biomarker exists to predict response to ICI in individual patients. PD-L1 immunohistochemistry and TMB have emerged as relevant biomarkers for ICI across many tumors however have not been relevant in ccRCC. Data across tumor types however suggests that some clinical features associate strongly with clinical outcomes to ICI. This study reviewed the course of 90 patients with ccRCC treated with anti-PD1/L1 and identified factors associated with primary resistance including a worse ECOG performance status at the start of ICI, an earlier stage at diagnosis, no prior nephrectomy and no occurrence of an irAE. Overall survival in these patients was correlated with IMDC Risk Score and pre-treatment NLR. In patients with initial benefit, increased BMI and overweight BMI status correlated with improved progression free and overall survival, respectively. Occurrence of an irAE correlated with longer time to progression while brain metastasis correlated decreased OS.
Higher BMI has correlated with a survival advantage to ICI in melanoma and other cancers [
24‐
26,
43]. A recent study demonstrated a survival benefit in patients with higher BMI in ccRCC treated with ICI [
44]. The mechanisms by which BMI impact clinical outcomes remain poorly understood. Lalani et al. did not find differences in genomic alteration frequency or tumor mutational burden by BMI status. Hyperadiposity may drive a tumorigenic immune-dysfunction that is more effectively reversed by ICI [
45,
46]. However, BMI may not adequately reflect the complexities of body composition. Computerized tomography-based body composition (CTBC) and bioelectrical impedance analysis have defined phenotypes which correlate with outcomes such as high visceral adipose tissue, skeletal muscle density, and sarcopenia [
47‐
49]. Further studies are needed to characterize mechanisms by which these phenotypes overlay with known biomarkers.
Our study demonstrated increased likelihood of response to ICI and improved PFS in patients who experienced and irAE, consistent observations in multiple solid tumors [
50,
51]. Studies have demonstrated association between outcomes and incidence of vitiligo and dermatitis in patients with melanoma as well as thyroiditis in NSCLC [
52‐
55]. No specific irAE were associated with improved outcomes in this study. Mechanisms by which irAEs correlate to tumor regression need to be further clarified. One proposed mechanism is cross reactivity of activated T-cells against antigens specific to both tumors and normal tissue, known as antigen sharing [
56].
Tissue based biomarkers for ICI across tumor types and especially in ccRCC are evolving. Exploratory biomarkers, such as gene expression profiling suggest that it may be possible to identify sub-populations of patients most likely to benefit to particular treatments. Angiogenesis, T-effector gene (similar to T cell-inflamed) expression signatures associated with outcomes in a recent clinical trial in ccRCC [
18,
29]. The Angio
low and T-eff
high GEP subgroups had improved outcomes to ICI whereas the Angio
high subgroup had worse outcomes to ICI but improved outcomes to VEGFR2 TKI. Our analysis of the TCGA revealed an inverse correlation between angiogenesis and T-cell inflammation signatures in tumors of high T cell-inflamed gene expression, a pattern not observed in non-T cell-inflamed tumors. An inverse association between the angiogenesis signature and histologic grade was demonstrated and a positive association between the T-cell inflammation signature and pathologic stage. This data suggests a suppressive role of angiogenesis on T cell-inflammation and may support further development of VEGFR2-TKI in combination or sequential therapy with ICI in earlier stage ccRCC. In clinical trials in non-metastatic ccRCC, perioperative systemic treatment with VEGFR2-TKI was not shown to increase overall survival versus surgery alone [
57‐
59]. Benefit from ICI in the adjuvant and neoadjuvant setting has been observed in multiple cancers including NSCLC, breast cancer and melanoma and multiple phase III clinical trials evaluating ICI in ccRCC in both adjuvant and neoadjuvant settings are ongoing [
60‐
64] (NCT03024996, NCT03142334, NCT03055013).
Our analysis of the TCGA revealed a positive correlation between T-cell inflammation signature and pathologic stage and in our ICI cohort, patients diagnosed at earlier stage were more likely to experience primary resistance to ICI. It should be noted that patients diagnosed at an earlier stage likely received ICI at time of metastatic recurrence in which there was indeed a longer time from initial diagnosis to treatment than those diagnosed at Stage IV disease (60 mo vs. 3 mo,
p < 0.0001). This latency may account for the increased likelihood of resistance. Perhaps metastatic recurrences progress predominately from tumors diagnosed de-novo and recur in non-inflamed, immunosuppressed or immune-exhausted environments. Investigation of primary ccRCC and ccRCC lung metastases demonstrated differential expressions of immunosuppressive molecules between primary and metastatic tumors [
65]. Further work to define the immune microenvironment of metastatic recurrences is warranted.
Patients who did not undergo nephrectomy were also more likely to suffer primary resistance. Pre-clinical work has suggested the primary tumor may produce T-cell inhibitory cytokines that divert antitumor immune response away from metastasis [
66]. Correlation between the morphologic immune character of the resected primary tumor, such as Teff/Treg ratio, with outcomes to ICI in ccRCC has previously been demonstrated [
67]. A recent study found increased response rate in patients who underwent cytoreductive nephrectomy or metastasectomy while receiving ICI [
68]. These conclusions highlight the need to further investigate the correlation between the immune character of the resected tumor and treatment outcomes.
While our study sheds light on factors associated with treatment response to immunotherapy and survival in ccRCC, we acknowledge limitations. This is a retrospective study and while we included all identified patients at each institution, a selection bias cannot be excluded. While we aggregated data across 5 centers internationally, we acknowledge that the sample size could be larger which could expand our potential findings. While our report predominately focused on response, we note that our follow-up time potentially did not fully assess long-term survival outcomes in patients receiving immunotherapy. Finally, the TCGA dataset did not record BMI and these patients were not treated with immunotherapy. We alternatively employed gene expression signatures which have been strongly associated with treatment outcomes however this narrows the conclusions that could be reached.
Conclusions
This international, multi-institutional effort supports conclusions that clinical factors, notably BMI and occurrence of an irAE, strongly associate with treatment outcomes in patients with ccRCC treated with ICI. We have identified novel predictors as well as variables that support previous studies, all of which may help guide clinical selection criteria for immunotherapy treatment. On the gene expression level we identified biologically relevant gene signatures including the T-cell inflammation and angiogenesis signatures that associate with histologic grade, pathologic stage and survival. Given the suppressive role of angiogenesis on T cell-inflammation, these data may support further development of VEGFR2-TKI in combination or sequential therapy with ICI in earlier stage ccRCC emphasizing the importance of adjuvant and neo-adjuvant strategies.
Acknowledgements
Not applicable.
Disclosures
BWL: none; JJL: Data and Safety Monitoring Board: TTC Oncology, Scientific Advisory Board: 7 Hills, Actym, Akrevia, Alphamab Oncology, Mavu, Pyxis, Springbank, Tempest, Consultancy: Abbvie, Array, Astellas, AstraZeneca, Bayer, Bristol-Myers Squibb, Compugen, EMD Serono, IDEAYA, Immunocore, Incyte, Janssen, Jounce, Leap, Merck, Mersana, Novartis, RefleXion, Spring Bank, Tempest, Vividion, Research Support: (all to institution for clinical trials unless noted) AbbVie, Array (Scientific Research Agreement; SRA), Boston Biomedical, Bristol-Myers Squibb, Celldex, CheckMate (SRA), Compugen, Corvus, EMD Serono, Evelo (SRA), Delcath, Five Prime, FLX Bio, Genentech, Immunocore, Incyte, Leap, MedImmune, Macrogenics, Novartis, Pharmacyclics, Palleon (SRA), Merck, Tesaro, Xencor, Travel: Array, AstraZeneca, Bayer, BeneVir, Bristol-Myers Squibb, Castle, CheckMate, EMD Serono, IDEAYA, Immunocore, Incyte, Janssen, Jounce, Merck, Mersana, NewLink, Novartis, RefleXion, Patents: (both provisional) Serial #15/612,657 (Cancer Immunotherapy), PCT/US18/36052 (Microbiome Biomarkers for Anti-PD-1/PD-L1 Responsiveness: Diagnostic, Prognostic and Therapeutic Uses Thereof).
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.