Concurrent MEK and CDK4/6 inhibition shows promise in clinical trials for patients with advanced-stage mutant BRAF/NRAS solid tumors. The effects of CDK4/6 inhibitor (CDK4/6i) in combination with BRAF/MEK-targeting agents on the tumor immune microenvironment are unclear, especially in melanoma, for which immune checkpoint inhibitors are effective in approximately 50% of patients. Here, we show that patients progressing on CDK4/6i/MEK pathway inhibitor combinations exhibit T-cell exclusion. We found that MEK and CDK4/6 targeting was more effective at delaying regrowth of mutant BRAF melanoma in immunocompetent versus immune-deficient mice. Although MEK inhibitor (MEKi) treatment increased tumor immunogenicity and intratumoral recruitment of CD8+ T cells, the main effect of CDK4/6i alone and in combination with MEKi was increased expression of CD137L, a T-cell costimulatory molecule on immune cells. Depletion of CD8+ T cells or blockade of the CD137 ligand–receptor interaction reduced time to regrowth of melanomas in the context of treatment with CDK4/6i plus MEKi treatment in vivo. Together, our data outline an antitumor immune-based mechanism and show the efficacy of targeting both the MEK pathway and CDK4/6.
Although targeted therapies and immune checkpoint inhibitors (ICI) are FDA approved for treatment of late-stage cutaneous melanoma, challenges remain. The combination of BRAF inhibitors (BRAFi) and MEK inhibitors (MEKi) produces response in 63%–75% of patients and 14.9 months median progression-free survival (1). ICIs such as anti-CTLA4, anti–PD-1, and anti–PD-L1 elicit durable tumor regression but many patients with melanoma do not respond and/or suffer immune-related adverse effects (2, 3). Combination strategies that modulate the tumor immune microenvironment or leverage the durability of ICI could overcome the short-term efficacy of targeted therapies (4).
Aberrant G1–S cell-cycle progression in melanoma is driven by multiple mechanisms. Highly selective CDK4/6 inhibitors (CDK4/6i) are currently FDA approved for estrogen receptor–positive (ER+) breast cancer in combination with hormone signaling blockade (5). In melanoma, the combination BRAFi and/or MEKi plus CDK4/6i shows promise in preclinical mouse models (6–9). Furthermore, the addition of CDK4/6i (ribociclib) to MEKi (binimetinib) improved response rates in patients with mutant NRAS melanoma by more than 2-fold compared with MEKi alone (10).
Tumor-infiltrating lymphocytes (TIL) regulate tumor progression and response to therapies (11). In addition to direct effects on the tumor, CDK4/6i may also alter the tumor immune microenvironment through upregulation of coinhibitory/stimulatory molecules such as PD-L1 and MHC class I (MHC-I), release of tumor-secreted cytokines, and suppression of immune-suppressive populations such as regulatory T cells (12–15). MEKi alone promotes infiltration of CD8+ T cells into the tumor but may impair T-cell priming and effector functions (16). Effects of combined MEK pathway and CDK4/6 targeting on the tumor immune microenvironment are not well understood. We therefore sought to elucidate the effects of combined treatment on different immune cell populations.
Materials and Methods
Mouse D4M3.A cells (donated by Dr. Constance E. Brinckerhoff, Dartmouth University, Hanover, NH, 2016) were cultured in DMEM/F12 advanced with 5% FBS, 1% penicillin/streptomycin, and 1% l-glutamine (17). Human 1205Lu cells (donated by Meenhard Herlyn, The Wistar Institute, Philadelphia, PA, 2005) were grown in WM media (MCDB153 with 2% FBS, 10% Leibowtiz L-15 medium, 5 μg/mL insulin). Human SKMEL207 cells (donated by David Solit, Memorial Sloan Kettering Cancer Center, New York, NY, 2010) were cultured in RPMI with 10% FBS. Cells were validated by short tandem repeat analysis using the Genomics Facility at The Wistar Institute. The BRAF mutational status of human melanoma cells lines used was confirmed by Sanger sequencing. Cell lines were tested for Mycoplasma contamination every 2 months using the MycoScope Kit (Genlantis).
Western blot analysis
Protein lysates were prepared in Laemmli sample buffer, separated by SDS-PAGE, and proteins were transferred to polyvinylidene difluoride membranes. Immunoreactivity was detected using horseradish peroxidase–conjugated secondary antibodies (CalBioTech) and chemiluminescence substrate (Thermo Fisher Scientific) on a Versadoc Imaging System (Bio-Rad). Primary antibodies were as follows: anti–PD-L1 (sc-19095, Santa Cruz Biotechnology) and ERK2 (sc-1647, Santa Cruz Biotechnology); anti–PD-L1 (human #13684, Cell Signaling Technology), anti–phospho-RB1 (S780, #9307, Cell Signaling Technology) and (S807/811, #9308, Cell Signaling Technology), anti-HSP90 (#4877, Cell Signaling Technology), phospho-ERK1/2 (T202/Y204, #9101, Cell Signaling Technology), and anti-actin (A2066, Sigma).
In vivo experiments
All studies were reviewed and approved by the Institutional Animal Care and Use Committee at Thomas Jefferson University (Philadelphia, PA). Male C57BL/6 mice (Jackson Labs) and NSG mice were used. Tumors were measured by digital caliper. Mice bearing D4M3.A tumors were fed with control chow (AIN-76A), MEKi chow (AIN-76A containing 7 ppm PD0325901), CDK4/6i chow (429 ppm palbociclib), or combination chow (AIN-76A diet with both PD0325901 and palbociclib). Combination chow was intermittently dosed for three weeks then switched to single-agent MEKi for 1 week.
Flow cytometry of tumor and spleen samples
Samples were analyzed on the BD Fortessa and data quantified with FlowJo software. Tumors were removed and dissociated into single-cell suspensions. Spleens were processed mechanically using a 70-μm nylon filter and a syringe plunger. For immunogenicity studies, cells were stained with a fixable live/dead stain (Zombie UV, BioLegend) per company instructions, fixed and permeabilized using BD Cytofix/Cytoperm Kit (BD Biosciences), followed by antibody staining. For TIL studies, cells were first stained with a fixable live/dead stain followed by surface antibody staining using the following antibodies: CD45.2 (clone 104), PD-L1 (clone 10F.9G2), IDO-1 (clone mIDO-48, eBioscience), CD137L (clone TKS-1), H-2 (MHC-I, clone M1/42), I-A/I-E (MHC-II, clone M5/114.15.1), LGalS9 (clone RG9-35), OX40L (clone RM134L), CD8α (clone 53.6.7), CD8β (clone YTS156.7.7), CD4 (clone RM4.4), NK1.1 (clone PK136), PD-1 (clone 29F.1A12), CD44 (clone IM7), CD11c (clone N418), GrzB (clone GB11, Invitrogen). For regulatory T cells (Treg) staining and Ki67 staining, cells were fixed and nuclear permeabilized using the eBioscience Foxp3/Transcription factor buffer staining set and FoxP3 antibody (clone FJK-16s) or Ki67 (clone 16A8) following manufacturer's instructions. Representative FACS plots showing gating strategy can be found in Supplementary Fig. S1.
Melanoma patient sample analysis
Patient studies were conducted according to the Declaration of Helsinki and written informed consent was obtained from all patients. Tumor samples were collected and analyzed according to University of Zurich Institutional Review Board–approved protocols (EK647 and EK800). Patient samples were collected from two clinical trials. Patients 1 and 3 were treated with LGX818 + MEK162 + LEE001, as part of the NCT01543698/CMEK162 × 2110 trial. Patient 2 was treated with LGX818 + MEK162 + LEE001 after progression on LGX818 + MEK162, as part of the NCT02159066/LOGIC2 trial. For NCT01543698, LGX818 and MEK162 were administered on a continuous schedule and LEE011 was administered in a 3-week-on, 1-week-off schedule. For NCT02159066, patients progressing on LGX818 and MEK162 were treated with LEE011. Samples were analyzed by IHC for CD8+ T cells using a mouse mAb to human CD8 (clone C8/144B from Agilent). CD8+ cells were counted in the entire tissue sections by imaging analysis using QuPath software (qupath.github.io). Only areas with tumor were included in the analysis. The operator was blinded to the clinical outcome.
In vivo CD8 depletion
To deplete CD8+ T cells, mice were treated with 300 μg of anti-CD8α (clone 53-6.72, BioXCell) every 3 days for the duration of the experiment, starting 2 days before tumor implantation. Mouse IgG2a antibody (clone 2A3, BioXCell) was used as a control. An intermittent dosing schedule was used in which mice were fed chow dosed with a combination CDK4/6i + MEKi for 3 weeks and then switched to chow dosed with only MEKi for 1 week.
In vivo CD137L blocking
To block CD137–CD137L interactions, we utilized anti-CD137L (clone TKS-1). IgG2a isotype (clone 2A3) served as the control. Mice were treated with blocking antibody 2 days before tumor implantation and treatment was continued every 3 days. An intermittent dosing schedule was used in which mice were fed combination CDK4/6i + MEKi chow for 3 weeks and then switched to chow with only MEKi for 1 week.
In vivo OX40 agonist treatment
To stimulate the OX40 pathway, we utilized an agonistic antibody, anti-OX40/CD134 (Bio X Cell InVivoMAb #BE0031), or IgG1 isotype control (clone HRPN, Bio X Cell InVivoMAb #BE0088). Mice were treated twice a week with antibody on a 2-week on/off cycle. An intermittent dosing schedule was used for CDK4/6i + MEKi, as outlined above.
Statistical analysis of tumor growth
The statistical software used was SAS 9.4 (SAS Institute Inc.) and R 4.0.0 (R Foundation for Statistical Computing, Vienna, Austria). The log-transformed tumor volumes were analyzed using the longitudinal linear mixed effects (LME) model with quadratic time trends separate for each treatment group and random effects of an animal. Thus, the fixed effects in the LME model included the treatment group, day and day squared (day2) as well as the interaction between the treatment group and day and day2. All tests were two-tailed. Separate models were fitted for data from C57BL/6 and NSG experiments. The residuals were evaluated to validate the assumptions of the models. On the basis of the fitted LME models, the overall comparisons of treatment groups were performed in terms of the growth rates (testing the global null hypotheses that linear and quadratic coefficients in the quadratic function of day are equal for the compared treatment groups). Standard thresholds of P < 0.05 were used for significance. The P values were adjusted for multiple testing using the method of Hochberg that controls for family-wise type I error.
T cells were excluded in patients who progressed on combined therapy
We have previously shown that acquired resistance to the combination of inhibitors targeting BRAF/MEK plus CDK4/6 occurred via multiple genetic alterations including NRAS mutation and PIK3CA mutation that enhance signaling through the mTOR–pS6 pathway (8, 9). However, effects on tumor-associated immune cells have not been studied in CDK4/6i-resistant melanoma. A limited number of CDK4/6i-treated mutant BRAF melanoma samples are available (9). Analysis of patient samples from two tumors that progressed within one year on BRAFi + MEKi + CDK4/6i therapy showed decreased CD8+ T-cell infiltration compared with the pretreatment patient-matched samples (Fig. 1A). Both patients also demonstrated upregulation of phospho-S6 expression within their progressing tumors (9). In contrast, samples from a patient who showed a prolonged response to BRAFi + MEKi + CDK4/6i displayed increased CD8+ T-cell staining in the posttreatment sample (Fig. 1B). These observations were supported by blinded quantitation of the samples (Fig. 1A and B). These data prompted us to analyze temporal alterations in the tumor immune microenvironment following treatment with inhibitors to the MAPK pathway and CDK4/6.
Combination MEKi and CDK4/6i therapy is more effective in immune-competent models
To study the role of the immune system in the response and resistance to targeted therapies combination, we utilized the Braf V600E/Pten loss D4M3.A mouse melanoma model (17). We focused on CDK4/6i combined with MEKi because this pairing is broadly applicable across many melanoma subtypes with encouraging early clinical data (18, 19) and our previous work has shown the on-target effects of this combination (7, 9). Treatment of D4M3.A cells with MEKi and CDK4/6i in vitro decreased phospho-ERK1/2 and phospho-RB1 amounts, respectively (Fig. 2A).
We have previously shown MEKi plus CDK4/6i reduced E2F-driven transcription using in vivo reporter models in immune-compromised mice (7). Next, we tested effects in mouse melanoma models. D4M3.A cells were injected into either immune-compromised NOD-scid γ (NSG) mice or immunocompetent syngeneic C57BL/6 mice. Once tumors were formed, mice were treated with chow laced with either vehicle, MEKi alone, CDK4/6i alone, or the combination of MEKi plus CDK4/6i. An intermittent, 3 weeks on/1 week off schedule was used for CDK4/6i, consistent with clinical use. In immune-compromised NSG mice, MEKi alone, and combination treatment led to tumor regressions, whereas CDK4/6i alone elicited little effect (Fig. 2B). However, tumors in both MEKi alone and MEKi plus CDK4/6i combination arms rapidly acquired resistance after 17 days of treatment and the mice had to be sacrificed within 24 and 28 days, respectively. Tumor regressions following MEKi alone and MEKi plus CDK4/6i treatment also occurred in C57BL/6 mice but, in contrast to tumors in NSG mice, there was a more durable inhibition of tumor growth (Fig. 2C). MEKi-treated mice survived up to 38–43 days, whereas MEKi plus CDK4/6i-treated mice survived up to 43–52 days (Fig. 2C). CDK4/6i alone had little to no effect on tumor growth inhibition in either NSG or C57BL/6 mice. Control tumors grew with an average growth rate of 8.6% per day (95% CI: 7.6%–9.7%) in C57BL/6 mice compared with an average of 10.0% per day (95% CI: 8.4%–11.7%) in NSG mice (Fig. 2B and C). The overlap between these two 95% confidence intervals (8.4%–9.7%) provides evidence that the growth rates of untreated tumors are similar in C57BL/6 and NSG mice. Together, these data suggest that the immune system plays a role in the efficacy of MEKi and CDK4/6i with MEKi being the main mediator of response in this model.
CD8+ T cells are required for the efficacy of MEKi and CDK4/6i combination therapy
To determine the mechanism by which the immune system contributes to the antitumor effects of MEKi and CDK4/6i, we treated D4M3.A allografts for four days and profiled changes in immune cell populations. We detected significant upregulation of both MHC-I and MHC-II expression with MEKi plus CDK4/6i treatment, suggesting increased antigen presentation (Fig. 3A; Supplementary Fig. S1). CD8+ T cells were associated with MHC-I expression and were significantly increased within the tumor with both MEKi treatment alone and combination MEKi plus CDK4/6i treatment (Fig. 3B). No significant changes were detected in CD4+ T cells, FOXP3+ regulatory T cells (Tregs), or natural killer (NK) cells within the tumor (Fig. 3B; Supplementary Figs. S1, S2A, and S2B). In vitro, we observed a more pronounced modulation of HLA-ABC expression in mutant BRAF and mutant NRAS human melanoma cell lines treated with MEKi plus CDK4/6i compared with single-agent treatments (Fig. 3C; Supplementary Fig. S1).
Because CD8+ T cells infiltrated into the tumor, we performed phenotypic analysis of this population. There were moderate but nonsignificant increases in markers for proliferation (Ki67), effector/memory T-cell phenotype (CD44), and cytotoxicity (granzyme B) following MEKi treatment alone and combination MEKi plus CDK4/6i treatment (Fig. 3D; Supplementary Fig. S1). CDK4/6 inhibition alone did not repress proliferation of CD8+ T cells (Fig. 3D). To determine whether CD8+ T cells play a functional role in MEK and CDK4/6 targeting, we used antibody to deplete CD8+ T cells in mice prior to implantation of D4M3.A melanoma tumors. We then treated the mice with MEKi plus CDK4/6i (Fig. 3E and F). Although we observed initial regression of the tumors in both IgG and CD8 depletion arms, tumors within the CD8-depleted arm progressed more rapidly and mice had to be sacrificed by day 44 (Fig. 3E and F). Although it is possible that depleting CD8+ cells prior to tumor implantation alters tumor progression when compared with tumor growth in fully immunocompetent animals, the impact should be minimal as there is very little immune infiltration into the lesions without intervention. Taken together, our data suggest a role for CD8+ T cells in the antitumor immunity driven by MEKi plus CDK4/6i treatment.
CDK4/6 inhibition does not affect PD-L1 expression on melanoma or immune cells
Next, we explored whether CDK4/6i-based combinations could modulate T-cell coinhibitory molecules. PD-L1 expression is cell-cycle regulated through the RB1–NF-κB axis and ubiquitination/proteasome pathway proteins (13, 20). When analyzing D4M3.A tumors that were treated for 4 days, we did not observe modulation of PD-L1, IDO1, or LGasS9 expression in any of the treatment groups in either CD45– tumor cells or CD45+ immune cells (Supplementary Figs. S1 and S3A–S3C). However, the average number of PD-L1 molecules per CD45+ immune cell was decreased after MEKi alone and MEKi plus CDK4/6i combination treatment (Supplementary Figs. S1 and S3D). In human melanoma cell lines, targeting CDK4/6 did not lead to modulation of IFNγ-induced PD-L1 amounts (Supplementary Fig. S3E). Overall, these data indicate that CDK4/6 inhibition does not uniformly regulate PD-L1 expression across melanoma models.
T-cell activation markers are altered by MEK and CDK4/6 cotargeting
To further characterize the mechanisms underlying immune cell–based effects of combined MEK plus CDK4/6 targeting, we analyzed modulation of costimulatory T-cell molecules that may potentially enhance CD8+ T-cell function. First, we detected downregulation of OX40 ligand (OX40L) on CD45+ immune cells and CD11c+ antigen-presenting cells (APC) in both the MEKi alone and in combination with CDK4/6i treatment arms (Fig. 4A; Supplementary Fig. S1). Effects were more marked with MEKi alone and were consistent with other reports that suggest that MEK inhibition may adversely affect long-term T-cell effector function (16, 21). On the basis of OX40L downregulation, we tested the effect of OX40 agonists on the efficacy of the MEKi plus CDK4/6i combination. D4M3.A allografts were treated with OX40 agonist or control antibody before and during MEKi plus CDK4/6i treatment. We compared tumor response, time until tumors regrew to the original size, and animal survival following initiation of treatment. Although tumors regressed in the OX40 agonist and MEKi plus CDK4/6i treatment arm and some of the regressions were durable, the time to regrowth and survival data were not significantly different between the OX40 agonist and control arms (Fig. 4B; Supplementary Fig S4). These data suggest that agonistic anti-OX40 did not improve MEKi and CDK4/6i therapy in the D4M3.A melanoma model. We do not rule out the possibility that OX40L agonists may elicit effects with MEKi alone.
In addition to effects on OX40L expression, we also detected a significant induction of CD137L expression on CD11c+ APCs in D4M3.A allografts following CDK4/6i treatment alone and in combination with MEKi (Fig. 4C; Supplementary Fig. S1). CDK4/6i upregulated CD137L expression on the CD45+ cell population, but this was partially antagonized by the addition of MEKi (Fig. 4C, middle). Nonetheless, these data suggest that CDK4/6 inhibition may enhance activation signals to CD8+ T cells. To determine whether upregulation of CD137L was functionally important, we utilized a blocking antibody to CD137L in mice bearing D4M3.A melanomas and also treated mice with MEKi plus CDK4/6i. Similar to CD8 antibody–mediated depletion, D4M3.A tumor–bearing mice initially regressed in response to MEKi plus CDK4/6i treatment in both IgG control and CD137L blocking antibody arms. However, the majority of tumors in the CD137L blocking arm reached approximately 1,000 mm3 between 34 and 45 days, and all mice had to be sacrificed by day 45 (Supplementary Fig. S4). In contrast, the IgG control group had overall smaller tumors, improved survival, and significantly delayed regrowth of residual tumors to their original size after treatment (Fig. 4D). Together, our data suggest that upregulation of CD137L on CD11c+ APCs by CDK4/6i may contribute to the improved therapeutic outcomes of MEKi plus CDK4/6i combination.
Our studies in a BrafV600E/Pten loss syngeneic mouse model highlight two mechanisms by which MEKi plus CDK4/6i effects associate with antitumor immunity. First, MEKi plus CDK4/6i increased capacity of melanoma cells to present antigen and recruitment of cytotoxic CD8+ T cells to the tumor. Second, CDK4/6 inhibition may enhance MEKi effects by sustaining the activation of infiltrating CD8+ T cells, at least in part, via induction of CD137L (4-1BBL) on APCs. Although increased MHC-I and infiltration of CD8+ T cells (14, 21) have previously been associated with CDK4/6 inhibition alone, a report on a KRAS-mutant lung cancer mouse model treated with combination MEKi, trametinib plus palbociclib, demonstrated that these infiltrating CD8+ T cells did not bear activation markers and did not play a functional role in antitumor immunity (22). Together, our and others' findings suggest that targeted therapies have multiple effects on different cell types within the tumor microenvironment that ultimately enhance the effects of immunotherapy.
Our study shows that CDK4/6i led to increased expression of CD137L on CD11c+ antigen-presenting cells. CD137L is the ligand for CD137 (4-1BB), which belongs to the TNF receptor family (TNFR9) and serves as a bona fide costimulatory molecule for T cells (23). CD137 is typically expressed after T-cell activation and ligation of CD137 leads to enhanced T-cell survival and proliferation. These functional properties of CD137 signaling are being exploited in the clinic with the use of agonistic antibodies, although their activity has been hampered by liver toxicities and potency issues (24). CD137 receptor expression is upregulated on both primary CD4+ and CD8+ T cells after exposure to CDK4/6i, abemaciclib (14). Our studies highlight an alternative way for providing CD137 ligation to overcome dysfunction or exhaustion of T cells. The adverse downregulation of another costimulatory molecule, OX40L (CD134/TNFRSF4), by MEKi may be compensated for by induction of CD137L from CDK4/6 inhibition (Fig. 4E). Together, this might lead to improved outcomes in the combination arm although more studies are required.
Loss of T-cell infiltration and dysfunction of T cells have been linked to acquired resistance to both MAPK-targeted therapies and immunotherapies (25). Our patient-based studies are complementary to such findings; however, these analyses are limited by the small number of patient-matched samples available to us and by the realization that these biopsies could have been taken from different lesions. Our results indicate that salvage therapies should be carefully selected for use in conjunction with anti–PD-1. Future studies might identify optimal sequencing of CDK4/6i-based combinations and ICI. Studies in melanoma and colorectal cancer models have suggested that phased combinations of CDK4/6i and ICI may be most effective to obtain synergism (14, 26). Overall, our data show that MEKi plus CDK4/6i combinations improve CD8+ T-cell recruitment and function, and they support clinical testing in combination with immunotherapy in melanoma.
Disclosure of Potential Conflicts of Interest
J.L.F. Teh reports a patent for US 9880150B2 issued. D.A. Erkes reports that he is now an employee of the journal publisher, the American Association for Cancer Research, and had no influence over any editorial decisions related to this article. I. Chervoneva reports grants from NIH/NCI R01 CA196278 during the conduct of the study. X. Xu reports personal fees and other from CureBiotech Inc. (cofounder) and other from Exio Bioscience (cofounder) outside the submitted work. R. Dummer has intermittent, project-focused consulting and/or advisory relationships with Novartis, Merck Sharp & Dohme, Bristol-Myers Squibb, Roche, Amgen, Takeda, Pierre Fabre, Sun Pharma, Sanofi, Catalym, Second Genome, Regeneron, and Alligator outside the submitted work. A.E. Aplin reports grants from the NIH (CA196278, CA182635, CA114046, and CA236736), Dr. Miriam and Sheldon G. Adelson Medical Research Foundation, U.S. Department of Defense, Outrun the Sun, and Pfizer during the conduct of the study, as well as a patent for US 9880150 issued. No potential conflicts of interest were disclosed by the other authors.
J.L.F. Teh: Conceptualization, data curation, funding acquisition, validation, investigation, methodology, writing–original draft, writing–review and editing. D.A. Erkes: Investigation, methodology, writing–original draft. P.F. Cheng: Resources, investigation, visualization. M. Tiago: Investigation. N.A. Wilski: Data curation, investigation, writing–review and editing. C.O. Field: Investigation. I. Chervoneva: Data curation. M.P. Levesque: Resources, funding acquisition, project administration. X. Xu: Formal analysis, investigation. R. Dummer: Funding acquisition, project administration. A.E. Aplin: Conceptualization, resources, funding acquisition, writing–original draft, project administration, writing–review and editing.
This work is supported by grants from the NIH (R01 CA196278 and R01 CA182635), and the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation (to A.E. Aplin). A.E. Aplin and X. Xu are supported by an NCI P01 in Targeted Therapies in Melanoma (CA114046). X. Xu was supported by grants from the U.S. Department of Defense (CA170340) and NIH (CA210944 and CA224070). J.L.F. Teh was supported by Outrun the Sun Research Scholar Award. N.A. Wilski is supported by the T32 Training Program in Cancer Biology (CA236736). The Sidney Kimmel Cancer Center Flow Cytometry, Translational Pathology, and Meta-Omics core facilities are supported by an NCI Support Grant (P30 CA056036).