Immunohistochemical Loss of LKB1 Is a Biomarker for More Aggressive Biology in KRAS-Mutant Lung Adenocarcinoma Free

The methodology for classifying non–small cell lung cancer (NSCLC) is evolving from use of histology alone to incorporating genomics. This change is based on the discovery of genomic alterations that define molecular subtypes, which can be targeted therapeutically (1). Lung cancer patients with identified driver genomic alterations treated with a targeted therapy survive longer than those who did not receive a targeted therapy (2). KRAS mutations are the most prevalent oncogenic driver mutation, detected in 30% of smokers and 10% of never-smokers with lung adenocarcinoma in Western populations (3–5). In contrast, for patients with KRAS mutations, targeted therapies have been largely unsuccessful and chemotherapy remains the current standard of care.

Therapeutic efforts at targeting KRAS have focused on inhibition of its intrinsically activated GTPase activity (6, 7), the use of farnesyl transferase inhibitors to impair the association of KRAS with the cell membrane where KRAS mediates its action (8–10), or synthetic lethality by inhibiting targets that impair the function of dependent genes (11–13). More recently, inhibition of MEK, the immediate downstream effectors of KRAS-mediated signaling, either alone or in combination with chemotherapy, has been evaluated both in preclinical models and in clinical trials (14, 15). In a randomized phase II clinical trial, previously treated patients with KRAS mutations who were treated with the MEK inhibitor selumetinib in combination with docetaxel had clinically meaningful improvement in progression-free survival, tumor responses, and patient-reported outcomes, compared with patients who were treated with docetaxel alone (16). Currently, this combination is being evaluated in a phase III clinical trial compared with docetaxel alone (ClinicalTrials.gov identifier: NCT01933932). Another MEK inhibitor, trametinib, has also been compared with docetaxel in previously treated advanced KRAS-mutant NSCLC patients (17). In this study, both trametinib and docetaxel had a similar response rate (12%) and progression-free survival (17).

The clinical efficacy of single-agent trametinib or the selumetinib/docetaxel combination is only observed in a subset of KRAS-mutant patients and suggests that there may be heterogeneity among KRAS-mutant lung cancers. Genetic mutations that coexist with the KRAS mutations, especially inactivation of tumor suppressors, may contribute to both tumor aggressiveness and response to treatment. In a genetically engineered mouse model (GEMM) of Kras G12D NSCLC, concurrent inactivation of the tumor suppressor liver kinase B1 (Lkb1^−/−) confers a more aggressive phenotype, and leads to more tumor metastases compared with cancers in Kras G12D GEMMs (18). Consistent with the clinical findings, the selumetinib/docetaxel combination was more effective than docetaxel alone in Kras G12D GEMMs of lung cancer (19). However, in Kras G12D/Lkb1^−/− GEMMS, selumetinib/docetaxel was significantly less effective (19). Collectively these findings highlight LKB1 status as a relevant marker in the biologic heterogeneity of KRAS-mutant NSCLC.

LKB1, also known as the serine/threonine protein kinase 11 (STK11), regulates cell energy homeostasis, cell polarization, and apoptosis (20, 21). LKB1 comprises 9 exons located on chromosome 19p13.3. Germline mutations in LKB1 cause the autosomal dominant Peutz–Jeghers syndrome (PJS; OMIM 175200; refs. 22, 23). PJS is characterized by the development of hamartomatous polyps in the gastrointestinal (GI) tract, mucocutaneous pigmentation and an increased risk of developing cancer. LKB1 functions as a tumor suppressor gene, and has been associated with several different types of sporadic cancer. Within lung cancer, LKB1 ranks as the third most common mutation found in lung adenocarcinoma (24–27). Both LKB1 alleles are somatically inactivated in 30% of lung adenocarcinomas and more commonly within KRAS-mutant NSCLC compared with EGFR-mutant NSCLC (24–27). Similarly to KRAS-mutant NSCLC, LKB1 mutations are linked to smoking history (24).

The role of LKB1 loss-of-function mutations as either a prognostic or predictive marker within human KRAS-mutant NSCLC is currently unknown. One potential reason for this observation is the wide range of LKB1 genomic alterations observed in sporadic cancers, making it a challenge to develop one single assay capable of detecting all such alterations. Several types of somatic mutations in LKB1 in sporadic cancers have been identified, including insertions, deletions (including large intragenic deletions), nonsense, frameshift, missense, and splice site alterations, and more infrequently hypermethylation of the promoter (28). Moreover, homozygous deletions of LKB1 have been frequently reported in NSCLC (29, 30). Detection of genomic loss by DNA sequencing of clinical specimens is challenging because of the admixture with DNA from nontumor cells. However, most of the genomic alterations in LKB1 result in either a truncated protein (and therefore inactive) or the absence of the protein (28). Thus, an IHC assay represents a potentially reliable, simple, and cost-effective method for evaluating for LKB1 loss by querying the expression status of LKB1.

We report on the optimization and validation of an IHC assay for detection of LKB1 expression that can be applied to clinical NSCLC tumor specimens. We used this assay to determine the frequency of LKB1 loss in KRAS-mutant NSCLC, and to understand the clinical characteristics and the prognostic implications of LKB1 in advanced KRAS-mutant NSCLC.

We identified patients with KRAS-mutant NSCLC through an institutional database of patients that had undergone genotyping as previously described (31, 32). Systematic genotyping for KRAS mutations for all advanced nonsquamous NSCLC was initiated at DFCI in 2008 and is performed in a CLIA-certified laboratory (32). We manually reviewed the list of the selected cases and the staging workup at the time the patient was identified to have systemic disease. Brain metastasis either pathologically confirmed or detected by imaging (CT scan or brain MRI) at any point of the evolution of the disease was considered for the analysis. All patients were consented to an Institutional Review Board (IRB)-approved protocol allowing collection of specimens and clinical data. The tumor content of each of the specimens was confirmed to be sufficient for IHC analyses. Pack-years of smoking was defined as (average number of cigarettes per day/20) × years smoking. Former smokers were defined as those who had quit smoking ≥1 year from the time of diagnosis of lung cancer. Never-smokers were defined as patients who had smoked ≤ 100 cigarettes in their lifetime.

Lung cancer cell lines A549, H460, DMS-53, UMC-11, H2122, H23, 11-18, H322, HCC827 were obtained from ATCC and grown in RPMI-1640 supplemented with 10% FBS and 1% 1× antibiotic-antimycotic (Life Technologies, cat# 15240-062). DMS-53 cells were grown in Waymouth Medium (Life Technologies, cat# 31220-023) and supplemented as above. Ectopic LKB1 expression in the LKB1-mutant A549 human NSCLC was performed by lentiviral transfection as previously described (33). All cells were cultured at 37°C in a humidified incubator with 5% CO2 and tested negative for mycoplasma infection. All cell lines were authenticated in September 2014 using the Promega Geneprint 16 cell ID system and were performed at the Research Technology Support Facility at Michigan State University (East Lansing, MI).

Four-micron thick formalin-fixed, paraffin-embedded (FFPE) tissue sections on charged glass slides were prepared per standard protocols for IHC. Epitope retrieval was performed on the Leica Bond III autostainer at pH 9 for 20 minutes. LKB1 antibody (Abcam, cat# ab15095, clone Ley 37D/G6) was added at 1:15,000 for 30 minutes at room temperature and antibody detected using the Bond Polymer Refine Detection kit, cat# DS9800. Slides were counterstained with hematoxylin. LKB1 antibody clone Ley37D/G6 is a murine monoclonal IgG2b antibody anti-human LKB1, raised against recombinant LKB1 of human origin targeting an epitope located at the C-terminal region of the protein (34). For antibody optimization studies, A549 and A549 LKB1 cells were FFPE into a block. Sections from the FFPE block were subject to IHC analogous to tumor sections.

For clinical specimens, LKB1 expression was evaluated in the background nonneoplastic tissue, for example, alveolar macrophages or respiratory epithelium in lung specimens, providing an internal negative staining control. LKB1 staining was scored as intact or lost, with any degree of expression qualifying as intact. In some cases, the staining was heterogeneous, which was also noted. All the cases tested in the optimization and validation set showed cytoplasmic staining, and the cellular localization of protein expression in the clinical set was hence not further classified.

Mass-spectrometric–based cancer gene mutation profiling was performed in 440 lung adenocarcinoma specimens as part of an institutional genotyping platform, termed OncoMap, as previously described (35). LKB1 mutations were found in six cases and we used these specimens for IHC optimization purposes. For IHC validation, clinical next-generation sequencing (NGS) was performed in FFPE clinical tumor samples from newly diagnosed NSCLC patients in our CLIA-certified laboratory as part of an IRB-approved research protocol. We utilized high-throughput exon capture and massively parallel sequencing, termed OncoPanel, to detect single nucleotide variants, insertion-deletion events, and copy number alterations, as previously described (36).

Associations between categorical data were tested using the Fisher exact test; Wilcoxon rank-sum test was used to test for association with continuous measures. Using the Fisher exact test, this study had 80% power for comparing a rate of LKB1 loss of 35% among smokers with a rate of 10% among never smokers while testing at the one-sided 0.025 level. Overall survival is defined as the time in months from date of diagnosis of metastatic disease to death from any cause, and patients not experiencing an event are censored at their last date of follow-up. Event time distributions were compared using the log-rank test, and Cox proportional hazards models were fitted to estimate HRs. All tests were conducted at the two-sided 0.05 level, and no adjustments have been made for multiple comparisons.

Between August 2008 and January 2014, we identified 514 KRAS-mutant patients among 1818 patients who had undergone genotyping (28%). Forty-two (8.2%) were never-smokers. We retrieved archival FFPE tumor specimens from never-smokers and smokers in a 1:3 ratio. We prioritized tumors from smokers with KRAS-mutant NSCLC with surgical specimens amenable to IHC. From the original 42 never-smokers, tissue could not be collected in 11 patients (tumor block exhausted/not located = 7 patients, lack of consent = 2 patients, confirmed remote story of light-smoking = 2 patients). The clinical and molecular characteristics of the patients and tumors selected for further analyses did not differ significantly from the remainder of the KRAS-mutant patients as two of the categories (sex and histology) approached P value of 0.05 (Supplementary Table S1).

We first set to optimize and validate the LKB1 IHC assay by using both lung cancer clinical samples and cell lines with known LKB1 mutations. LKB1-mutant cell lines included DMS-53 (harboring the most common missense mutation, located in the kinase domain), and the H23 cells (nonsense mutation at exon 8; predicted to encode for a shorter LKB1 protein that lacks the C-terminal region; Supplementary Table S2; ref. 37). A549 cells (KRAS G12S, LKB1 p.Q37*) and its variant transfected with active LKB1 were used as positive and negative controls, respectively (Supplementary Table S2). EGFR-mutant cell line HCC827 was also used as negative control because it contains an intact LKB1 (27). We detected a high degree of correlation with LKB1 loss and lack of LKB1 protein expression by IHC (Fig. 1A and 1B and Supplementary Table S2). The one exception was the H322 cells in which LKB1 was detected by IHC despite an LKB1 mutation. These cells are known to harbor an interstitial in-frame deletion of exons 2 and 3, which is expected to encode for a smaller protein (38), preserving the C-terminal portion of the protein, making it possible to be detected by IHC. A549/LKB1 cells demonstrated positive staining by IHC compared with A549 cells and staining was both nuclear and cytoplasmic (Fig. 1A and 1B). This dual cellular localization of LKB1 has been previously described for ectopic expression of LKB1 (39) and has been postulated being related to the lack of STRADα required for cytosolic localization of LKB1 complex (34). Some cell lines only demonstrated staining in the Golgi, characterized by a paranuclear dot-like staining, and were categorized as LKB1 negative (Supplementary Table S2).

Next, we tested the LKB1 IHC assay in a panel of FFPE specimens from patients with NSCLC. We first used tumor and nontumor tissue sections from two PJS patients who developed lung adenocarcinoma (Fig. 1D–I). Both patients had previous resections of hamartomatous polyps from the GI tract. LKB1 expression was detected in the cytoplasm of the epithelium but not in the stroma of the GI polyps (Fig. 1D and 1G). Similarly, positive staining was observed in the normal respiratory epithelium and the alveolar macrophages in their lung cancer surgical specimens (Fig. 1E, F, H, and I, respectively). In contrast, LKB1 expression was lost in the lung adenocarcinomas of both patients (Fig. 1F and 1I). The presence of LKB1 expression in nontumor tissue is likely due to the presence of one remaining intact LKB1 germline allele and is restricted to previously known patterns of expression while in the tumor cells both alleles were lost (40). We next tested for LKB1 expression in 6 LKB1-mutant tumors from patients with NSCLC where the mutation was detected by allelotyping (OncoMap; ref. 35; Supplementary Table S3). Two specimens with p.P281L and p.E57fsX7 mutations, respectively, revealed weak staining and were considered as positive. The lack of information on other LKB1 allele did not allow further interpretation of these results. In addition, two cases had equivocal staining, although one of them did not contain any cells that could be used as an internal negative control. The positive control—a tumor specimen with a biallelic deletion of LKB1 by NGS—had loss of LKB1 expression, whereas the negative control—an EGFR mutant lung adenocarcinoma—showed strong LKB1 staining (Fig. 1J and 1C, respectively).

To further study our LKB1 IHC assay using clinical samples, we tested a clinical set of KRAS-mutant NSCLC specimens that had undergone targeted NGS (36). Between June and December 2013, we identified 27 cases harboring KRAS mutations and correlated their LKB1 mutation status with the LKB1 IHC staining results (Table 1). No LKB1 mutation was detected by NGS in 16 of 27 (59%) samples, whereas LKB1 genetic alterations were identified in 11 of 27 (41%) of the samples. LKB1 expression was detected in 14 of 16 LKB1 wild-type samples. In 2 of 16 specimens, no LKB1 expression was detected despite an intact LKB1 locus. Both samples lacked internal negative controls (e.g., macrophages or respiratory epithelium) and it is possible that the assay failed in these two specimens. Alternatively, these cancers may contain other genomic methods of LKB1 inactivation, such as methylation, not captured by NGS. In cases with intact LKB1 staining, there were different degrees of protein expression, varying from weak to strong cytoplasmic staining (data not shown). Among the 11 LKB1-mutant specimens, no LKB1 expression was detected in eight cases (73%). In two cases, LKB1 loss was considered heterogeneous, consisting of some weak staining with areas of multifocal loss of LKB1 expression. In a third case, which harbored a single copy deletion of the LKB1 gene, LKB1 expression was intact by IHC (Table 1). As complete loss of LKB1 expression by IHC requires biallelic inactivation, these mismatches in the LKB1 staining with the LKB1 genotype may be explained by the presence of a wild type LKB1 allele responsible for some degree of protein expression, similar to the PJS patients showed previously. Of note, in three cases, we detected genomic evidence of biallelic inactivation of LKB1, consisting of deletion of one allele and a mutation in the other allele (Table 1). In all of these 3 patients, LKB1 expression was lost by IHC.

We used the LKB1 IHC assay to evaluate the clinical and molecular features of smokers and never-smokers with KRAS-mutant NSCLC. We obtained FFPE tumor specimens on 154 patients (31 from never-smokers, 123 smokers), including the 27 patients previously sequenced used for validation. The demographic features of the patients are summarized in Table 2. Smokers and never-smokers patients were similar in terms of age, gender, race, histology, and stage. A 2:1 predominant female patient distribution is consistent with previous studies in KRAS-mutant population (3, 41). The G12C (40%) and G12V (18%) mutations were the most common KRAS mutation subtype in smokers, whereas G12D (58%) was most common in never-smokers (P < 0.001). Never-smokers were more likely to harbor KRAS transition mutations rather than transversion mutations, while the converse was true for smokers (82% transversion mutations vs. 18% transition mutations, P < 0.001; Supplementary Fig. S1). The most common nucleotide change was G>T in smokers and G>A in never-smokers (P < 0.001). This pattern of KRAS mutations is in accordance with previously reported studies on specific subtypes of KRAS mutations in smokers and never-smokers (4, 42).

Of the 154 samples for LKB1 IHC assay, six cases failed IHC (4%); in three, the tumor tissue quality was not sufficient for staining and in the remaining three cases, we could not interpret the staining due to a lack of internal control. Using our IHC assay, the overall incidence of LKB1 loss was 30% (95% confidence intervals; CI, 24%-38%) and was higher in smokers than in never-smokers (35% vs. 13%, P = 0.017; Fig. 2). Among the cases in which we observed LKB1 staining, the intensity of staining varied from weak to strong (data not shown), and the staining was universally cytoplasmic (Fig. 1). Any degree of LKB1 staining by IHC was considered positive. Heterogeneous staining, composed of areas of intact and lost staining, was observed in eight cases, all of which were classified as LKB1 intact. Overall, we obtained successful staining in a wide variety of clinical biopsies (surgical specimens 73%, core biopsies 16%, and cytology specimens 11%).

There were no differences in LKB1 loss by age, sex, race, histology, or pack-years of smoking (Supplementary Table S4). There was a trend toward LKB1 loss in stage IV patients. Although LKB1 was not associated with the type of KRAS mutations, LKB1 loss was preferentially found in tumors with KRAS transversion mutations as opposed to transition mutations (ratio 5:1, P = 0.029). This is in contrast with the never-smokers where the absence of LKB1 expression was almost mutually exclusive with the presence of a KRAS transition mutation. Only one case of 20 with KRAS transition mutations showed loss of LKB1, whereas 5 out of 13 cases with transversion mutations showed loss of LKB1 expression (P = 0.025).

Preclinical mouse models demonstrate that Kras G12D/Lkb1^−/− tumors have a more aggressive phenotype, characterized by increased metastases, compared with Kras G12D–only mice (18). To evaluate whether this same observation could also occur in humans, we evaluated the clinical features of stage IV KRAS-mutant patients based on their LKB1 status. We included patients with both initially diagnosed stage IV disease and those who developed recurrent disease during the follow-up after being diagnosed for early-stage NSCLC. In total, 126 patients were evaluable (74 initially diagnosed with stage IV disease, 52 with recurrent disease). Patients with both a KRAS mutation and LKB1 loss by IHC had higher number of metastatic sites at the time of diagnosis of stage IV disease (median 2.5 vs. 2.0, P = 0.01) and developed brain metastasis more frequently (48% vs. 25%, P = 0.02; Table 3). Patients with intact LKB1 tumors were more likely to have intrathoracic metastases (P = 0.02), whereas those with LKB1 loss tumors had significantly higher incidence of extrathoracic metastases (P = 0.01, Table 3). The higher number of metastatic sites together with the higher frequency of extrathoracic metastasis suggests that LKB1 loss may promote metastasis beyond the lung. Altogether, our data support the preclinical evidence that LKB1 loss confers features of worse prognosis in patients with stage IV KRAS-mutant lung adenocarcinoma.

To confirm whether LKB1 loss has an impact on survival, we analyzed the survival outcomes of the 126 stage IV KRAS-mutant NSCLC patients in which we evaluated LKB1 status by IHC (LKB1 intact 84 patients, LKB1 loss 42 patients; Table 3). The groups were similar in age, gender, race, smoking history, histology, and KRAS genotype (Table 3). The Kaplan–Meier survival curves showed a tendency toward a shorter overall survival in patients with LKB1 loss compared with those with intact LKB1, although this difference was not statistically significant (median OS in LKB1 loss 14.4 vs. LKB1 intact 30.0 months, log-rank P = 0.11) with a HR = 1.44 (95% CI, 0.92–2.28; Fig. 3A). There was no significant difference in outcomes based on the class of KRAS mutation (transversion vs. transition) or type of KRAS genotype (Fig. 3B and Supplementary Fig. S2). The survival effect of LKB1 loss seems to be of the same magnitude regardless of the class (transversion vs. transition) or type of KRAS mutation (G12C/G12V; Fig. 3C and Supplementary Fig. S2, respectively). However, these data did not reach statistical significance in our cohort of patients.

LKB1 is a tumor suppressor gene that is commonly lost in cancers from patients with NSCLC. Previous studies have demonstrated that LKB1 loss is more common in KRAS-mutant compared with in EGFR-mutant lung cancers (24–27). Preclinical murine studies demonstrate that Kras-mutant tumors with Lkb1 loss are more aggressive and less likely to respond to docetaxel/selumetinib treatment (18, 19). The impact of LKB1 loss in human lung cancer is currently not known. Given the increased number of clinical studies focusing on KRAS-mutant NSCLC, including a phase III clinical trial of docetaxel/selumetinib versus docetaxel (ClinicalTrials.gov identifier: NCT01933932), it is critical to be able to evaluate the presence of LKB1 loss and to correlate the findings with outcome. However, no systematic tools exist to evaluate LKB1 loss. In part, this is due to the complexity of LKB1 loss, which can occur through multiple different genomic and non-genomic mechanisms. To overcome this limitation, we evaluated an LKB1 IHC assay as a simple and cost-effective method that can be applied to clinical NSCLC specimens.

We optimized our assay using a variety of complementary methods, including evaluation of cell lines and patient specimens that were known to be intact for LKB1 or contain LKB1 mutations. Despite the accuracy of the IHC assay shown in our genotyped samples, this method has also limitations. Although we used NGS of clinical specimens for comparative purposes, genotyping including NGS cannot be considered as gold standard to address for LKB1 loss given the multiple mechanisms leading to its loss. Hence, this limits the ability to determine the sensitivity and specificity of our assay. Second, we cannot extrapolate the accuracy of this assay for all of the different LKB1 mutations beyond the ones included in our panel. For instance, some LKB1 missense mutations that result in functional inactivation of LKB1 could still result on LKB1 protein expression, and hence be misclassified as LKB1 intact using IHC. Third, testing for loss of expression of a protein by IHC may be more challenging than confirming its presence. The interpretation on LKB1 staining depends on the tissue background staining and the presence of internal negative controls: an excessive background or the lack of internal control may lead to either equivocal results or not valid results. In our studies, we observed this in 5% of the cases but we could not assume the same incidence in other conditions and requires further confirmation. In addition, we observed the heterogeneous staining in some cases, likely reflecting the tumor heterogeneity, which cannot be effectively captured with genotyping only. The biologic significance of this heterogeneity is currently unknown and may represent the evolutionary pattern of LKB1 loss as a tumor suppressor gene in the progression of the tumor. Overall, we were able to apply a simple dichotomous and easily reproducible scoring system with cytoplasmic LKB1 either present, or absent, that is preferable other scoring methods, including H-scores, for evaluation of tumor suppressor gene loss (43).

We observed a lower incidence of LKB1 loss in never-smokers with KRAS-mutant NSCLC, consistent with studies suggesting LKB1 is a smoking-induced alteration (28). Moreover, the almost mutually exclusive presence of transition mutations with LKB1 loss could have therapeutic implications. KRAS G12D, the most common mutation in never-smokers, is reported to have better outcomes and responses to selumetinib/docetaxel when LKB1 is intact in murine preclinical models, while the reverse is true when LKB1 is inactivated (19). Thus, it is possible that KRAS-mutant never-smokers may have a differential clinical response to MEK inhibitors or MEK inhibitor combinations (14, 19, 44). The incorporation of tumor LKB1 testing into current and future clinical trials is necessary to help confirm the preclinical observations. Never-smokers with KRAS mutations are usually an underestimated subgroup because of the general belief that KRAS mutations are only found in smokers. Although less common than EGFR mutations, KRAS mutations are detected in 10% of Caucasian never-smokers and are similar in incidence to anaplastic lymphoma kinase rearrangements (3, 4, 45, 46).

Our results using LKB1 IHC are consistent with the murine preclinical observations, which demonstrate an increase in metastases and a worse outcome in the Kras G12D/Lkb1^−/− model (15, 19, 47). The clinical and molecular profiles of the KRAS-mutant patients and their relationship with smoking status are similar to those previously reported (3, 4, 25, 48). We observed a 30% incidence of LKB1 loss in our study, more commonly in smokers, expected for this population suggesting that our large subset of KRAS-mutant patients is representative and consistent with prior studies (18, 26, 29, 30, 47). In patients with LKB1 loss, there was an increase in both the frequency and types of metastases, which translated into a trend toward a worse survival. These observations suggest that LKB1 loss may have a prognostic role, a finding that will require additional prospective validation. Our findings suggest that a simple dichotomous assay for IHC LKB1 loss is both simple and fast and is appropriate for routine use in clinical pathology specimens and/or as part of future prospective trials. If validated, LKB1 status may need to be incorporated into clinical trials of KRAS-mutant patients both for its prognostic and predictive roles.

G.R. Oxnard is a consultant/advisory board member for AstraZeneca, Genentech, and Novartis. P.A. Jänne is a consultant/advisory board member for Abbot, AstraZeneca, Boehringer Ingelheim, Chugai, Clovis, Genentech, Pfizer, and Sanofi. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Calles, S.J. Rodig, P.A. Jänne

Development of methodology: A. Calles, S.J. Rodig, J.L. Hornick

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Calles, S.J. Rodig, A.K. Pelton, J.L. Hornick, M. Butaney, G.R. Oxnard, D.M. Jackman, P.A. Jänne

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Calles, L.M. Sholl, S.J. Rodig, M. Butaney, S.E. Dahlberg, P.A. Jänne

Writing, review, and/or revision of the manuscript: A. Calles, L.M. Sholl, J.L. Hornick, S.E. Dahlberg, G.R. Oxnard, D.M. Jackman, P.A. Jänne

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Calles, M. Butaney, C. Lydon, G.R. Oxnard, D.M. Jackman

Study supervision: P.A. Jänne

A. Calles is a grant recipient from the Spanish Medical Oncology Society. S.E. Dahlberg received funding from the Gallup funds to support her collaboration in this work.

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.