Functional drivers of resistance to anti-CD19 CAR-T cell therapy in diffuse large B cell lymphoma

Abstract

Chimeric antigen receptor T-cell therapy targeting CD19 (CAR-19) promotes impressive durable remissions for relapsed or refractory (rel/ref) large B-cell lymphoma (LBCL) patients with historically poor prognoses. Despite this, over half of patients still fail to respond or eventually progress. Studies to reveal mechanisms of resistance have examined host clinical parameters, CAR-19 product composition, and tumor microenvironment (TME) alterations, while a relative paucity of studies has analyzed contributions by genomic alterations in tumor cells. Factors associated with outcome include increased tumor volume, specific characteristics of infused CAR-T products, infiltration by myeloid cells in tumor microenvironments, and markers of complexity in LBCL genomes. Functional laboratory studies of resistance are largely absent in the current literature, illustrating a need for experiments in genetically accurate immunocompetent systems to confirm candidate alterations’ roles in resistance and inform future improvements. In this review, we highlight key studies that have elucidated biomarkers of resistance in hosts, CAR products, TMEs, and comparatively understudied tumor-intrinsic mediators encoded by tumor genomes. We conclude with an experimental framework suitable for CAR-19 resistance biomarker identification and laboratory functional validation.

Keywords:

• CAR-T
• resistance
• mechanisms
• lymphoma

Introduction

CD19-directed chimeric antigen receptor T-cells (CAR-19) were first tested in heavily pretreated patients with diffuse large B-cell lymphoma (DLBCL) and related aggressive B-cell lymphomas in 2012 [1,2]. In 2017, based on the results of the ZUMA-1 phase 2 trial, the United States Food and Drug Administration (FDA) approved the CAR-19 product axicabtagene ciloleucel (axi-cel) for patients who failed by two or more lines of prior therapy [3]. Since then, two additional CD19-directed CAR products have been FDA-approved for these patients, tisgenlecleucel (tisa-cel) and lisocabtagene maraleucel (liso-cel) [4,5]. Subsequently, the randomized phase 3 trials ZUMA-7 and TRANSFORM demonstrated superior outcomes for axi-cel and liso-cel, respectively, when compared to standard salvage in the second-line for LBCL patients either refractory to frontline therapy or relapsed within 12 months [6,7]. Both products are now approved as second-line options in this setting. Recently, longer follow-ups of ZUMA-7 revealed improved overall survival for axi-cel-treated patients compared to traditional second-line salvage chemotherapy followed by autologous stem cell transplant (ASCT) [8]. Advantage was not seen, however, for tisa-cel in BELINDA, a similar phase 3 comparison to standard salvage. This was possibly due to longer production times, the optional inclusion of bridging therapy that resulted in the inclusion of patients with more aggressive disease, and a control arm where salvage therapy at week 12 was allowed and not counted as an event. Overall, however, CAR-19 therapy has emerged as a key new treatment option for DLBCL-NOS and other rel/ref LBCL types. Expansion now to the highest-risk second-line patients emphasizes the need for a comprehensive understanding of resistance to make progress toward extending benefits to larger numbers of patients.

Importantly, there is increasing evidence that complete responses (CRs) to CAR-19 may be long-term. Five-year follow-up of ZUMA-1 patients, in which 58% achieved CR, shows that 30% were still in CR at data cutoff (median follow-up 63.1 months). This plateau in PFS is evidence of cure for these high-risk LBCL patients. Real-world data continue to mature but appear to corroborate. For example, the US Lymphoma CAR-T Consortium’s retrospective analysis of patients with LBCL (68.1% with DLBCL) demonstrated that overall response rate (ORR) and CR rate among the 275 patients who received axi-cel were 82% (95% CI, 77% to 86%) and 64% (95% CI, 58% to 69%), respectively [9]. The largest published set of real-world data comes from the French DESCAR-T registry, which reported similar findings in 418 propensity score-matched patients, with ORR/CR of 80.4%/60.3% for axi-cel and 66.0%/42.1% for tisa-cel [10]. For all 729 LBCL patients who underwent lymphodepletion in the DESCAR-T data, the median OS and PFS from infusion were 19.0 and 5.6 months, respectively, but with a clear plateaus on the PFS and OS curves, though median follow-up was only 11.7 months. Such data must continue to mature to provide further clarity, but the current consensus is that approximately 30–40% of CAR-19-treated LBCL patients will achieve long-term CRs, likely amounting to cures for most of these individuals. CAR-19 therefore represents a major advance for rel/ref LBCL patients considering that only 1 in 10 overall ultimately can be cured through transplant-dependent approaches [11]. CAR-19 outcome data, however, also demonstrate how much more progress still needs to be made, with 60-70% of patients ultimately failing the approach [3,4,6]. Outcomes after failure unfortunately appear abysmal [12]. In this review, we will explore known factors associated with CAR-T failure including product-related and clinical factors. We will also untangle the effect of the tumor microenvironment (TME) on CAR-19 and highlight the under-studied impact of alterations in tumor genomes.

Clinical factors associated with CAR-19 response

Many clinical parameters assessed in routine practice can help predict response to CAR-19. For example, increased tumor burden associated with increased risk of failure. In ZUMA-1, elevated pre-infusion metabolic tumor volume (MTV) or lactate dehydrogenase (LDH), a surrogate for both burden and proliferation, correlated negatively with durable responses [13]. High inflammatory markers such as C-reactive protein (CRP), ferritin, and IL-6 also predict inferior outcomes [13]. Beyond routine laboratory makers, Next-Generation Sequencing (NGS) techniques provide a dynamic detection method of lymphoma circulating tumor DNA (ctDNA) allowing tracking of minimal residual disease (MRD) based on lymphoma-specific variable, diversity, and joining gene segments (VDJ) clonotypes [14]. Here, higher pre-CAR-19 treatment ctDNA levels correlated with progressive disease, while MRD-negative status at days 7 and 28 correlated with better outcomes. ctDNA was detected at or before radiographic relapse in 94% of patients, while it remained undetectable in patients with durable responses at 3 months. This finding of higher pretreatment ctDNA predicting CAR-T failure was corroborated by a recent study that added parallel monitoring of cell-free T-cell receptor DNA and axi-cel vector fragments to profile the immune dynamics of patient responses [15]. These studies collectively highlight ctDNA as a useful prognostic marker that is amenable to implementation in clinical practice. Overall, the key prognostic principle for CAR therapies emerges from these findings is that within each patient, the higher the ratio of tumor burden to CAR-product expansion, the worse the outcome is likely to be.

Many other clinical factors, however, show no association with CAR-19 response. These include age and international prognostic index (IPI) scores [6,7]. Indeed, at least two studies surprisingly found improved outcomes with higher age, though it remains unclear the degree to which selection bias vs. biology drove these findings [9,16]. Similarly unpredictable are biologic factors with clear significance for newly diagnosed patients such as the cell of origin (COO) molecular subtypes of DLBCL [6,7,17]. This extends also to aggressive double or triple-hit lymphoma with MYC, and BCL2 and/or BCL6 rearrangements. Current molecular classifications, i.e. the Chapuy [18] and LymphGen [19,20] classification systems, also carry no predictive value for CAR-19 response [15,16].

Taken together, clinical markers reflect lymphoma burden with varying degrees of sensitivity, whether through positron-emission tomography (PET) imaging, LDH, or ctDNA. These clinical factors, however, provide little insight into mechanism. In the following sections, we will discuss CAR-product factors that have been the subject of several high-profile studies followed by a review of tumor-intrinsic factors that are understudied by comparison and likely hold many insights into CAR-19 resistance that remain to be discovered.

CAR-19 product composition shapes responses

A growing body of research focuses on quality, cellular subsets, and other characteristics of CAR-19 infusion products to better understand patient responses (Table 1). These studies rely primarily on transcriptomics and/or multiplex immunofluorescence to evaluate products at a single-cell level both pre- and post-infusion. An early study by Rossi et al. performed single-cell RNA sequencing (scRNA-seq), revealing a T-cell polyfunctionality strength index (PSI) score that predicted clinical responses [21]. A more granular assessment of pre-infusion products was conducted by Deng et al. in 2020 [22]. Here, scRNA-seq on pre-infusion axi-cel products from 24 LBCL patients identified CD8+ central memory T-cells enriched in patients with CRs at 3 months. This finding of enriched memory T-cells was confirmed by Locke et al. [13] and is in line with recent data from tisa-cel-treated patients in which pre-infusion peripheral blood subsets of CD3 + CD27-CD28- effector memory cells predicted favorable responses [23]. Recently, evidence for CD57 + T-bet + CAR-T cells in driving remission arose from the analysis of samples at day 7 post-infusion [24]. The authors propose that these senescent, but highly cytolytic, cells are not present upon infusion but differentiate and enrich throughout the days leading up to peak expansion.

Table 1. Infusion product characteristics associated with response to CAR-19 therapy.

Study	Sample collection	Cell features associated with CR	Cell features associated with PD
Rossi et al. [21]	Infusion product	CD4+ and CD8+ T-cells with high PSI	Decreased T effector cells
Locke et al. [13]	Infusion product	CCR7 + CD45RA + T-cells with a negative doubling time; increased CD8+ cells	NR
Deng et al. [22]	Infusion product	CD8 + CCR7+ central memory T-cells	LAG3 + TIM3+ CD8+ T-cells
Chen et al. [42]	Pre/post CAR-T biopsies	CD8 + PD-1+ T-cells	NR
Scholler et al. [26]	Pre/post CAR-T biopsies	CD8 + PD-1+ T-cells	NR
Good et al. [24]	Post CAR-T peripheral blood	CD57 + T-bet + CAR-T cells	CD4 + CD57-Helios + Tregs
Worel et al. [23]	Peripheral blood and infusion product	Low CD3 + CD27-CD28- T-cells	CD3 + CD4+ T-helper cells

Abbreviations: CR: complete response; PD: progressive disease; Tregs: T-regulatory cells; NR: not reported.

Increased regulatory T cells (Tregs) by contrast associate strongly with progressive disease in CAR-19 recipients. The notion that CD25 + FOXP3+ Tregs may be responsible for PD was first suspected after the discovery that patient products with more FOXP3 expression showed less neurotoxicity and correspondingly less response [25]. The role of Tregs in blunting efficacy and promoting PD was later defined as CD4 + CD57-Helios + CAR-Ts (Treg signature) at peak expansion [24]. These thymic-derived Tregs have a classic Treg signature with low cytotoxic potential. Collectively, these studies implicate Tregs as key players in reduced CAR-19 efficacy and suggest future manufacturing efforts should aim at reducing these subsets while enriching memory CCR7+ cells. They also support a hypothesis that CAR-19 should be moved up to earlier lines of therapy when patients’ milieu of available T cells may be more robust due to less impact from previous cytotoxic treatments. This approach, however, must be carefully weighed with increased toxicities of more potent CAR-19 products, even if the possibility of cytokine release syndrome is lower with earlier intervention at lower MTVs.

Tumor microenvironment factors impacting CAR-19 response

Approaches to examine TME effects on CAR-19 efficacy include scRNA-seq, gene-expression profiling, and multiplex immunohistochemistry (IHC) analyses on samples collected pre-infusion, or paired at pre and post-infusion (Table 2). Pre-infusion studies have found that infiltration of macrophage, monocyte, and monocytic myeloid-derived suppressor cells (M-MDSC) associated with CAR-19 resistance and may explain the elevated serum inflammatory markers CRP, ferritin, and IL-6 [13,16]. M-MDSCs are known suppressors of cytotoxic T cells and provide a logical resistance mechanism. However, a recent study by Scholler et al. analyzing pre and post infusion samples reported that some level of inhibition, along with a T-cell-rich TME, predicts durable CAR-19 responses [26]. Specifically, patients with a high density of CD8 + PD-1 + LAG-3+/−TIM-3− T-cells interspersed with a modest amount of Tregs achieved the most favorable outcomes. Therefore, while TMEs associated with good outcomes are not surprisingly infiltrated by CD8+ T effectors, results implicate also a positive effect for at least some resident Tregs, likely related to their well-defined role in protecting normal tissues [26]. Analyzing TMEs post-infusion at peak expansion, Chen et al. made the fascinating observation that <5% of T-cells within tumors were CAR +  [27]. T-cell activation was found most strongly in cases destined to have the best outcomes, but most of that activation came from the non-CAR cells. Combined with the findings in Scholler and other observations, these findings strongly support the authors’ interpretation that CAR-19 infiltration initiates a local antitumor response by activating non-CAR cells within LBCL TMEs. Therefore, the therapeutic success of CAR-T cells depends on existing TMEs at infusion, and T-cell-rich, myeloid-depleted TMEs are the most favorable.

Table 2. TME features associated with response to CAR-19 therapy.

Study	Sample collection	TME features associated with CR	TME features associated with PD
Locke et al. [13]	Infusion product	NR	High TB, myeloid activation markers, excess type 1 cytokines
Chen et al. [42]	Pre/post CAR-T biopsies	Non-CAR activated T-cells	Non-CAR lymphocytes/monocytes producing IL-6
Jain et al. [16]	Pre CAR-T biopsies	NR	Macrophage infiltration, peripheral blood M-MDSCs, high ferritin, CRP, and IL-6
Scholler et al. [26]	Pre/post CAR-T biopsies	PD1 + LAG3+/-TIM3-CD8+ cells; modest Tregs; high IL-15, IL-17, IL-18, IL-21	NR

Abbreviations: CR: complete response; PD: progressive disease; M-MDSC: monocytic myeloid derived suppressor cell; NR: not reported.

The impact of LBCL genomes on CAR-19 responses

While clinical, CAR-19 product, and TME factors have been widely studied and are clearly correlated with CAR-19 outcomes, there is a paucity of studies exploring the impact of tumor genomic landscapes. In contrast, the impact of genomics on outcomes for untreated DLBCL is well-established. The Chapuy and LymphGen classifications, based on whole-exome sequencing and genomic copy-number alterations, classify the disease into biologic subtypes that are more granular than cell-of-origin and demonstrate differences in overall survival [18–20]. Despite the contribution of these studies to understanding disease biology and pathogenesis, however, they have had little impact on altering treatment options and are used for now primarily as research tools. In addition, as discussed, these classifications have no association with CAR-19 responses. Instead, results from initial genomics studies implicate genomic complexity as a negative factor for CAR-19 and suggest a dynamic interplay among factors (genomes, TMEs, and CAR products) must be elucidated for a comprehensive understanding of resistance.

Early assessments of tumor cell-intrinsic mechanisms focused logically on CD19 antigen escape. In B-cell acute lymphoblastic leukemia (B-ALL), in which CAR-19 therapy also is approved for use, alterations in CD19 including genomic deletions and somatic mutations correlate with lack of response and acquired resistance [28–30]. In LBCL, however, the implications of CD19 alterations and expression are less clear. A study by Zhang et al. identified two patients with the same point mutation in CD19 who did not respond and had extremely poor outcomes [31], but examinations in larger case series suggest CD19 antigen escape explains resistance in a minority of LBCL cases. For example, CD19 assessment in samples from axi-cel-treated patients found pretreatment IHC H-score did not correlate with durable response, and although a greater proportion of relapsed samples contained absent or decreased CD19 expression, this was still a minority of cases [32,33]. The Spiegel et al. study used flow cytometry to quantify CD19 molecules on cells and found <3000 was semi-predictive of CAR-T failure. However, even with this method, the majority of progressors were still CD19+. Diminished CAR-19 efficacy in tumors that lack the CD19 target is expected, but the fact remains that most relapsed cases do not substantially lose the antigen. In sum, alterations in CD19 in LBCL appear to make a relatively minor contribution to the high rates of CAR-19 failure in these malignancies, showing the need to investigate whether additional genomic alterations drive resistance.

Shouval et al. analyzed targeted sequencing data generated on archival tumor specimens from 82 patients treated with CAR-19 and found TP53 alterations associated with progression [34]. This retrospective analysis employed the MSK Impact panel of 685 genes on samples collected a median of 4.7 months before CAR-T infusion, and analysis showed cases with copy-number loss or mutation of TP53 significantly correlated with inferior ORR and CR rates [34]. A similar study albeit with fewer patients did not replicate this finding [35]. The heterogeneity of time points for sample collection in the MSK study may cloud the conclusion of p53 loss as a driver of resistance since other treatments often had occurred between sample collection and CAR-19 infusion. Our study employing whole-genome sequencing (described in more detail below) found 86% of CAR-19 patients had alteration of either TP53 or the closely related tumor suppressor CDKN2A, suggesting loss of the pathway is enriched in cases that end up requiring CAR-19, but in our data loss of the pathway was not prognostic for CAR-19 itself. In addition, in an abstract presented at the 2021 American Society of Hematology annual meeting, Hill et al. presented data on a cohort of 96 patients treated with commercial CAR-19 using WES and transcriptomic analysis of FFPE pre-treatment pretreatment samples. The authors did not report TP53 alterations, highlighting instead mutations in MYC, BCL2, CDKN2A, and KLHL6 as associated with poor PFS after CAR-19 treatment [36]. Recently, Sworder et al. utilized an approach combining ctDNA with RNA-seq from 138 CAR-19-treated DLBCL patients [15]. Besides confirming that high tumor burden indicated by quantitative ctDNA levels associated with disease progression, the authors found specific mutations (TMEM30A, P2RY8, IRF8, TP53) and CNAs (ex. PD-L1 and PD-L2) enriched in relapsed patients. With any focal genomics findings, conclusions are limited until mechanistic experiments assessing specific alterations in vitro and in vivo confirm their function as drivers of resistance. Few such studies are available to date. An exception was a report by Upadhyay et al. in which loss of the extrinsic apoptotic cell-surface molecule FAS emerged as a negative CAR-19 factor in CRISPR screening [37]. Intriguingly, the functional experiments by these authors demonstrated that FAS-mediated apoptotic engagement was important for the clearance of both CD19-antigen-positive tumor cells and antigen-negative bystanders. Functional laboratory experiments therefore show promise in both the characterization and the discovery of CAR-19 resistance mechanisms.

Whole genome sequencing (WGS) of tumors is an approach that stands unparalleled to targeted and whole-exome sequencing approaches, providing comprehensive pictures of genome complexity. In our study, Jain et al. WGS on 49 CAR-19 treated DLBCL patients revealed that pretreatment genomic events including complex structural variants (chromothripsis and double minutes), specific mutational signatures (APOBEC and SBS18 associated with reactive oxygen species DNA damage) and the del 3p21.31 recurrent copy number alteration predicted CAR-19 failure [16]. When pooled together, these genomic features were far more predictive of response than clinical features (MTV, age, bridging therapy, etc.) and had no correlation with the widely utilized Chapuy and LymphGen genomic subtypes of untreated DLBCL tumors. In a simultaneously published study, Cherng et al. utilized low-pass WGS of cell-free DNA in 122 relapsed/refractory (r/r) DLBCL patients prior to CAR-19 treatment. They found that a high focal copy number alteration (CNA) score indicative of genomic instability correlated with inferior response to CAR-19 [38]. The authors then combined their genomic findings with clinical factors of tumor burden to risk stratify patients, demonstrating a noninvasive method of predicting CAR-19 response. While candidate driver lesions still need to be discovered and validated, both studies highlight genomic complexity as an indicator of CAR-19 response and warrant further evaluation in larger cohorts and functional laboratory models. Table 3 summarizes the findings of studies of genomic factors involved in CAR-19 response.

Table 3. Genomic factors associated with CAR-19 response in r/r DLBCL.

Study	Sample #, type, technique	CAR-19 Product	Genomic Markers	Associated CAR-19 Responses
Zhang et al. [30]	2, tumor, WES	BinD19 (NCT03156101)	CD19: exon 3, p.174 L > V	No response
Spiegel et al. [31]	44, tumor, IHC	Axi-cel	CD19 expression (low/absent)	Relapse
Shouval et al. [33]	82, tumor, WES	Axi-cel, liso-cel, tisa-cel	TP53: mutations or CNAs	Inferior CR & ORR
Hill et al. [35]	96, tumor, WES	Axi-cel, tisa-cel	Mutations in: MYC, BCL2, CDKN2A, KLHL6	Inferior PFS
			Mutations in: BTG2, MYD88, CD79B	Favorable PFS
Sworder et al. [15]	138, ctDNA, WES	Axi-cel	Mutations in TMEM30A, P2RY8, IRF8, TP53, CD19 CNAs in PD-L1 & PD-L2	Inferior EFS
Jain et al. [16]	49, tumors, WGS	Axi-cel	Chromothripsis, double minutes, APOBEC mutational signature, ROS damage, 3p21.31 deletion	Inferior PFS
Cherng et al. [37]	122, cfDNA, WGS	Axi-cel, liso-cel, tisa-cel	high focal CNA, 10q23.3 deletion	Inferior PFS, OS

Abbreviations: WES: whole exome sequencing; IHC: immunohistochemistry; CAN: copy number alteration; CR: complete response, ORR: overall response rate; PFS: progression-free survival; EFS: event-free survival; ROS: reactive oxygen species.

Discussion

As the use of CAR-19 in the clinic expands, increased knowledge about factors contributing to response will remain an urgent research priority. To date, host, CAR-19 product factors and TME characteristics are reasonably well-characterized, while studies focusing on tumor genomes are more limited and, most acutely, lack functional characterization in laboratory models. In addition to relatively small patient numbers overall, these studies also primarily focus on axi-cel-treated patients and therefore do not yet have the ability to assess for differences across products. Of note, CD28-co-stimulated axi-cel has different expansion and persistence properties than tisa-cel or liso-cel, both of which are 41BB-co-stimulated. Given the expanded clinical use of these products, increased sample size and diverse CAR-19 treatments can be evaluated relatively easily by noninvasive methods such as ctDNA and cfDNA approaches. WGS, however, will provide the most granular picture and fortunately is increasingly available as both its cost and turn-around time are continually decreasing. A pilot study in myeloid cancers showed WGS could be deployed and inform clinical decision-making in real time, though widespread implementation of this approach in any tumor type is far from practical reality [39]. For CAR-19, WGS paired with functional experiments in immunocompetent laboratory models to validate candidate alterations can help bring forward a comprehensive picture of tumor intrinsic drivers of response. With time, these approaches can fill the wide gaps in prognostication left by currently readily available clinical risk parameters like LDH, inflammatory serum markers, and tumor burden.

A key unexplored question is the specificity of specific genomic markers to CAR-19 vs their potential effects on other treatments. Even after CAR-19’s recent approval in second-line settings, all patients who undergo it have received at least frontline chemoimmunotherapy and often one or more additional treatments for their disease. Research is needed to understand if genomic markers associated with CAR-19 resistance are specific to it or if they are general drivers of poor prognosis regardless of therapy. In addition, studies of longitudinal samples from individual patients throughout their clinical courses could identify whether the markers of poor CAR-19 response also are predictive of responses to various chemotherapy regimens or to emerging treatments such as bispecific antibodies or next-generation cellular immunotherapies [40–42]. Characterizing the impact of prior therapies on tumor genomics and subsequently CAR-19 response could provide biological insight into tumor progression and the specificity of these biomarkers to CAR-19 response.

As these findings become better characterized, leveraging biomarkers of response to identify likely responders or non-responders prior to treatment would also provide great benefit to patients. If a set of well-characterized genomic predictors of CAR-19 response were fully validated, predicted responders could be administered CAR-19 while predicted non-responders could be prioritized for alternative therapeutics or clinical trials. In addition, various small molecule inhibitors and other therapies have been tested in DLBCL and although largely unsuccessful clinically, there is an abundance of both clinical and preclinical data on these treatments that could inform repurposing for use alongside CAR-19 if mechanistic studies identified a link to specific targets correlating with resistance. As NGS becomes increasingly accessible, this kind of personalized approach can be applied for DLBCL patients as it has in many other malignancies. NGS and biomarker identification therefore can, with further research, drive the identification and repurposing of efficacious strategies in combination with CAR-19.

In summary, much has been learned regarding resistance to CAR-T therapies in LBCL, revealing unique and novel factors impacting these breakthrough therapies. While clinical features, molecular classification systems, and target antigen loss provide little insight into the mechanisms of LBCL CAR-T resistance, NGS can provide the temporal and granular genomic events employed by CAR-T resistant tumors. However, we are far from the practical implementation of most key molecular findings to inform clinical decision-making in real-time. Clinical markers like LDH, tumor burden, and systemic inflammatory markers are readily available but represent crude instruments for prognostication or choosing among therapeutic options. Thus, more extensive discovery and functional characterizations of genomic factors implicated are needed. Further evaluation combining patient-centered studies with functional laboratory studies to confirm mechanisms will provide powerful data to synergize genomic, host, and TME factors and personalize DLBCL therapy for rel/ref patients (Figure 1).

Figure 1. CAR-19 resistance is driven by infusion product, TME, and tumor-intrinsic mechanisms. Whole-genome sequencing (WGS) revealed complex LBCL genomes in association with poor CAR-19 outcomes, and a variety of studies implicate deregulation of specific genes, but functional laboratory studies are lacking. The relationship between tumor genomes and tumor microenvironments (TMEs) also remains undefined.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by a grant to JHS from the U.S. DoD Congressionally Directed Medical Research Program (CA220385) and by an NIH/NCI grant to the University of Miami Sylvester Comprehensive Cancer Center (P30CA240139).