Retinoblastoma mutation predicts poor outcomes in advanced non small cell lung cancer

Abstract The retinoblastoma gene (RB1) encodes the retinoblastoma (RB) pocket protein that plays an important role in cell cycle progression. Here we determine the frequency and prognostic significance of RB1 mutation in non small cell lung cancer (NSCLC), restricting inclusion to Stage III and IV patients with linked genomic and clinical data. The primary outcome was median overall survival (OS). We identified RB1 mutation in 8.2% of NSCLC patients. The median OS for wild‐type (wt) RB1 was 28.3 months vs 8.3 months for mutant RB1 (Hazard Ratio = 2.59, P = 0.002). Of special interest, RB1 mutation also correlated with lack of response to immunotherapy. Our study focused on RB1 mutation in locally advanced and advanced non small cell lung cancer to better facilitate comparisons with small cell lung cancer (SCLC). In our SCLC cohort, RB1 mutation was identified in 75% of patients and wt RB1 was associated with significantly shorter OS (P = 0.002). The different outcomes of RB1 mutation observed among lung cancer subtypes suggest a more complicated mechanism than simple regulation of cell cycle or response to chemotherapy.


| INTRODUCTION
Genomic sequencing of tumor DNA has changed the therapeutic landscape for non squamous, non small cell lung cancer (NSCLC) with the discovery of recurrent oncogenic driver mutations in EGFR, ALK, and ROS-1 that can be specifically and effectively targeted by new drugs. Additional targetable oncogenic driver mutations in BRAF, NTRK1, HER2, RET, and MET are found at lower frequency in lung adenocarcinoma patients. 1 Although KRAS mutation is found in ~25-30% of lung adenocarcinoma patients and remains largely untargetable, 2 there is a suggestion these patients demonstrate favorable responses to immunotherapy, although co-mutation in the tumor suppressor gene STK11/LKB1 identifies a subset of KRAS mutant patients that show poor response to immunotherapy. 3 Alteration in the tumor suppressor NF1 was also suggested as a potential druggable target in NSCLC. 4 Ultimately, about 30%-40% of adenocarcinoma lacks a clearly identifiable oncogenic alteration. 5 Genomic studies in small cell lung cancer (SCLC) have also identified subgroups with MYC amplification, SOX-2 amplification, FGFR-1 amplification, PTEN loss, RICTOR amplification, and NOTCH inactivation. 6,7 Genomic identification in SCLC has clearly lagged behind that of NSCLC in part due to tissue availability. Our group previously published on the genomics of small cell lung cancer and identified retinoblastoma (RB1) gene mutation status through targeted exome sequencing as a predictor of outcomes. 8 Retinoblastoma was the first tumor suppressor gene to be discovered based on an association with a rare childhood tumor, retinoblastoma, that occurs at a frequency of 1 in 20 000 live births. 9 The Knudson hypothesis of a second hit in retinal cells of children with germline mutation led to the understanding of how tumor suppressor genes drive the development of cancer. About 40% of retinoblastomas are hereditary and hereditary retinoblastoma survivors are at risk for second malignancies like osteosarcoma, melanoma, and epithelial malignancies like lung, bladder, and breast cancer. 10,11 Somatic alterations in RB1 are known to occur in various malignancies including lung, breast, bladder, and prostate cancer.
Retinoblastoma encodes the retinoblastoma pocket protein (RB) that regulates the cell cycle by binding to E2F transcription factors in its unphosphorylated form to repress their activity. In response to mitogenic stimuli, the cyclin dependent kinases (CDK) phosphorylate RB, causing release of the binding to E2F and progression through the cell cycle. p16 INK4A and other CDK inhibitors maintain RB in the unphosphorylated, active form. The role of RB1 is most understood in the regulation of G1 to S transition and cell proliferation. There are other roles attributed to RB like regulation of epithelial to mesenchymal transition 12,13 and a possible role in immune response. 14 Here, we explore the association of RB1 mutation status to outcome in advanced NSCLC. Our study focused on locally advanced and advanced NSCLC to better facilitate comparisons with SCLC, a disease with a defined role for RB1.

| PATIENTS AND METHODS
We have an IRB approved institutional database that includes all lung cancer patients diagnosed at our institution and referred for thoracic oncology opinion. Patients were staged according to TNM7 staging (American Joint Committee on Cancer staging manual, 7th edition). All patients are discussed at our multidisciplinary tumor board and a TNM staging is assigned after the multidisciplinary discussion.
The inclusion criteria were locally advanced or metastatic (stage III and stage IV or recurrent NSCLC) disease, age greater than 18, seen at our institution from 2013 to 2016. We used the Foundation One sequencing platform which interrogates 315 genes exomes as well as introns of 28 genes involved in rearrangements (Supporting Information, Table  S1). 15 Patients who did not have next generation sequencing of tumor DNA were excluded. We collected data on age, sex, race, smoking status, stage, histological subtype for NSCLC, treatment with systemic chemotherapy, immunotherapy, and mutations. Smoking status was defined as yes for current smoker (who is smoking at the time of diagnosis or quit within 12 months of diagnosis) or former smoker (who quit at 12 months prior to diagnosis). Never smoker was defined as someone who smoked <100 cigarettes over their lifetime. The date of diagnosis is the date of biopsy and pathological confirmation of disease. The overall survival (OS) was calculated from the date of diagnosis to date of death and censored at the date of last follow-up for survivors. Survivor distribution was estimated using Kaplan-Meier methods and the difference of OS between the groups was examined by a log-rank test. The effect of continuous measurements, including age, on OS was estimated using the Cox model. The effect of RB1 mutation status in our NSCLC cohort on OS was further evaluated using the multivariable Cox model controlling for the effects of age, sex, stage, smoking, and chemotherapy. All tests are two-sided and P ≤ 0.05 was considered statistically significant. The specific type of RB1 mutation was not considered for outcome analysis, just its presence or absence. The characteristics of our SCLC cohort have been previously described. 8,16 The mutation distribution along the RB protein was plotted using cBioPortal mutation mapper. Immunohistochemistry (IHC) was performed on formalin-fixed paraffin-embedded (FFPE) specimens to evaluate RB expression using Cell Signaling Technology 4H1 mouse antibody (catalog number 9309). p16 INK4A IHC was done using the CINtec histology kit. p16 INK4A expression has been proposed as a surrogate for loss of RB protein expression or dysfunctional protein. [17][18][19] IHC scoring was done by a thoracic pathologist. The intensity of IHC staining was graded as absent (0), weak (1+) or strong (2+) and focused on nuclear staining for RB and cytoplasmic staining for p16. In addition, the percent of tumor cells showing staining was scored separately.

| RESULTS
One hundred and ninety-five patients met the inclusion criteria for NSCLC and had available both clinical and genomic data. The mutation frequency of RB1 in our cohort of NSCLC was 8.2%, which is consistent with prior reports and the TCGA database. 5 The baseline characteristics (Table 1) of RB1 mutant compared to RB1 wt patients were well balanced between the 2 groups, except for a higher number of stage 3 patients in the RB1 mutant NSCLC group.
In NSCLC, RB1 mutant status when compared to RB1 wt was associated with worse OS (8.3 months vs 28.3 months, Hazard Ratio (HR) = 2.59, 95% Confidence Interval (CI) = 1.4-4.79, P = 0.002) and this was statistically significant ( Figure 1). On multivariate analysis (Table 2), after adjusting for age, sex, stage, smoking status, receipt of chemotherapy, and other gene mutations, RB1 mutant status was still predictive of worse outcomes in NSCLC (HR = 3.07, 95% CI = 1.54-6.14, P = 0.002). While the mutation status of seven other genes also predicted worse outcomes, only MLL2 and KEAP1 were more significant than RB1. We focused on RB1 here to pursue comparisons with SCLC.
The RB1 wt NSCLC group had 12 patients with EGFR Exon 18-21 alterations and 4 patients with ALK gene rearrangement. When we excluded these patients with targetable, driver mutations due to a more favorable outcome, RB1 mutant NSCLC patients still had worse outcomes. There was one NSCLC patient with EGFR exon 19 deletion and RB1 alteration in our cohort. The RB1 alteration in this patient was found on repeat biopsy at the time of progression on first generation EGFR tyrosine kinase inhibitor (TKI), along with an acquired EGFR T790M mutation and a histopathology of adenocarcinoma. Transformation of EGFR mutant NSCLC to SCLC has been associated with loss of RB1. 20 The EGFR exon 19 deletion at the time of initial diagnosis was found on a limited gene panel analysis. It is unclear if the RB1 alteration represented a new event at the time of progression.
We looked at the association of RB1 mutation with response to immunotherapy in NSCLC by analyzing our data of 97 NSCLC patients treated with immunotherapy with second line nivolumab or first line pembrolizumab. We had genomic data available on 66 of these patients. There were 6 patients with RB1 alteration and none of these 6 patients responded to immunotherapy. In contrast, in those without RB1 alteration (N = 60), the response rate was 26.2%. The lack of response in all of the six patients with RB1 mutation requires further evaluation in larger cohorts. The distribution of mutations along the RB protein is depicted in Figure 2. The mutations in NSCLC were concentrated in the amino (N) terminal region. We identified a polyalanine deletion of amino acid [16][17][18] in the N terminal region of RB which was seen in 5 of 16 patients with NSCLC (Table 3). This mutation has been described with an allelic frequency of 0.4% in ClinVar and predicted to be benign using the PolyPhen-2 tool to predict the effect of amino acid alterations on protein structure. RB1 mutation status was still predictive of poorer outcomes in NSCLC when we reanalyzed our data based upon mutations predicted to be benign (missense and deletions) or of unknown significance (splice site mutations) vs those predicted to be damaging (exon loss and nonsense mutations).
We next looked at the expression of RB and the cyclin dependent kinase inhibitor p16 INK4A (CDKN2A) by IHC  in relation to RB1 mutation status in the NSCLC cohort ( Figure 3 and Table 3 We recorded the co-mutations in the 16 RB1 mutant NSCLC. CDKN2A damaging alterations, CDK4 alteration, CCND1 amplification and CCNE1 amplification were seen in 3, 1, 1, and 2 patients, respectively. TP53 mutations were present in 12 patients, followed by STK11 mutations in 6 patients and KRAS mutations in four patients.

| DISCUSSION
We found RB1 mutant status to be strongly associated with worse outcomes in NSCLC. There have been prior studies looking at the clinical correlation of RB1 in NSCLC with varying results, summarized in Table 4. The largest of these by Choi et al used DNA sequencing and identified RB1 mutant status to be associated with shorter disease-free survival only in early stage adenocarcinoma. 21 The incidence of RB1 mutation of 5.9% in their cohort is comparable to ours. Their cohort represented 49% never smokers, which might represent a different patient population than the average 10% never-smokers with NSCLC. 22 Geradts et al 23  outcomes. In another cohort of 106 patients, RB expression in resected NSCLC did not correlate with outcomes. 25 RB1 was altered in 7% of the lung adenocarcinoma TCGA 2014 dataset, 5 but had no effect on survival (203 cases, 13 mutant RB1, P = 0.994) (assessed via cBioPortal 10/2018). This TCGA dataset is most comparable to ours since our cohort was largely adenocarcinoma. A potential difference to explain our highly significant results is that our study focused on locally advanced and advanced NSCLC, as opposed to the early stage disease examined by all the other studies, including TCGA. Our study also used alterations at the DNA level as detected by targeted exome sequencing. The baseline characteristics of the RB1 mutant NSCLC compared to RB1 wt patients were well matched except for a significantly higher number of Stage III patients in the RB1 mutant group. This imbalance had no effect on survival outcomes, however, because multivariate analysis showed the effect of stage on OS was not statistically significant ( Table 2). This lack of statistical significance could be due to the low number of stage III patients in our cohort. An intriguing explanation would be that the recent (2014) addition of immunotherapy to stage IV patient care has increased their survival to 2-3 years. The use of immunotherapy for stage III patients only started in 2018 and therefore is not reflected in this dataset. There is also a suggestion that the proportion of black race is greater in the RB1 mutant group, which may bias survival outcomes. While race and RB1 mutation status were indeed associated (P = 0.011), the effect of RB1 mutation status on survival was essentially unchanged after controlling for the effect of race in the Cox model, with race not significant in predicting survival (P = 0.236).
We noted a concentration of RB1 alterations to the N terminal domain in our NSCLC cohort. Regulation of the cell cycle by RB is primarily attributed to the conserved central pocket (amino acids 379-792) and carboxy (C) terminal region (amino acids 792-928). The crystal structure of the entire RB N terminal domain is not well understood but there is a suggestion that the N terminal domain is well conserved and interacts with the pocket domains. 26 Because a significant number of missense mutations and exon deletions map to the N terminal domain in retinoblastoma patients, similar to our NSCLC cohort, it is possible that the mutations we observed affect an unknown critical function of RB.
Another reason we focused on advanced stage disease was to better compare the NSCLC results with our recently published genomic studies of small cell lung cancer (SCLC), where we found that mutant RB1 status was associated with favorable survival outcomes. 8 This remained true even in a more recent and expanded analysis examining 64 SCLC patients with largely extensive stage disease, in which RB1 mutation was seen in 75% of SCLC cases and was associated with significantly better OS when compared with RB1 wt status. 16 The contrasting association of RB1 mutation status with outcomes in NSCLC compared with SCLC is intriguing. The role of RB1 in different malignancies and different contexts may be more complicated than simple regulation of cell cycle or response to chemotherapy. 27 RB is inactivated by various mechanisms and there is a complex interplay between cyclin inhibitors and cyclin dependent kinases. It is possible that differences in co-mutations may in part explain the differential outcomes that we saw in RB1 mutated NSCLC compared to SCLC. Mutations in SCLC were distributed throughout the protein (Supporting Information, Figure S1A) and most of these were predicted to be damaging (Supporting Information, Figure S1B).
RB1 mutation status and its relationship to outcomes has also been reported for breast cancer with contrasting results. 28,29 SCLC is a chemosensitive disease with initial response rates to chemotherapy for extensive stage SCLC of about 70%. RB1 wt in SCLC was associated with a chemorefractory response (P = 0.0334) and identifies a subset with poor outcomes. 16 This indicates that loss of RB1, which has so far been considered a hallmark of SCLC, plays a role in making this disease initially chemosensitive. This is likely due to the absence of G1/S regulation in RB1 mutated SCLC and accumulation of DNA adducts caused by platinum agents. Similarly, RB pathway disruption correlated with complete pathological response to neoadjuvant chemotherapy in a study on breast cancer. 29 We also noted a lack of response to immunotherapy in RB1 mutated NSCLC. Tumor genomics likely impacts the immune milleu of the tumors and potentially plays a role in response to immunotherapy. Patients with EGFR mutations and ALK gene rearrangements are known to have poor responses to immunotherapy and have been excluded from recent immunotherapy trials. 30 PDL-1 expression and tumor mutation burden 31 are known biomarkers for predicting response to immunotherapy. High PDL-1 expression predicts a response rate of about 50% to immunotherapy 30 but depending on the cut off, patients with low PDL-1 expression respond to immunotherapy as well, highlighting the imperfections of the currently available markers and the need to identify additional markers.
There is emerging data on the role of RB in mediating immune response in addition to its regulation of the cell cycle. In hepatocellular carcinoma, the retinoblastoma pathway has been proposed to regulate innate immune response. RB depletion in hepatoma cells resulted in a compromised immunological response to multiple stimuli. 32 In bladder cancer, RB under expression was predictive of poor response to bacille Calmette-Guerin (BCG) therapy in concert with interferon-alpha (IFNα) therapy, providing further evidence that RB plays a role in mediating immune response. 33,34 In cervical cancer, human papilloma virus (HPV) oncoprotein E7 is known to bind to RB and cause its inactivation. DNA vaccine where E7 was altered to evade RB binding was more immunogenic compared to unaltered E7, further establishing the role of RB in immune response. 35 Our single institution dataset indicates RB1 mutation to be associated with a lack of response to immunotherapy in NSCLC. Coupled with these observations in other malignancies and biological explanations for the role of RB in mediating immune response, the association of RB1 mutation with lack of response to immunotherapy should be evaluated in larger cohorts of NSCLC. This is a single institution review. Patients who underwent next generation sequencing may represent a specific cohort with access to different treatments. We did not have paired germ line sequencing 36 for these patients to ascertain whether all of these alterations were somatic in nature. Germ line testing at this time is not considered standard of care. We recognize that Stage III lung cancer in itself represents a disease with varying outcomes. Our cohort represents a mixed population of stage III, stage IV, and recurrent NSCLC, but there is general consensus about treatment within a single institution. The majority of our NSCLC patients displayed adenocarcinoma histology, reflecting the practice pattern of genomic sequencing in these patients.