Resistance to Ibrutinib in B Cell Malignancies: One Size Does Not Fit All
Ibrutinib resistance, as a result of coordinated rewiring of signaling networks and enforced tumor microenvironment (TME)–lymphoma interactions, drives unrestrained proliferation and disease progression. To combat resistance mechanisms, we must identify the compensatory resistance pathways and the central modulators of reprogramming events. Targeting the transcriptome and kinome reprogramming of lymphoma cells represents a rational approach to mitigate ibrutinib resistance in B cell malignancies. However, with the apparent heterogeneity and plasticity of tumors shown in therapy response, a one size fits all approach may be unattainable. To this end, a reliable and real-time drug screening platform to tailor effective individualized therapies in patients with B cell malignancies is warranted. Here, we describe the complexity of ibrutinib resistance in B cell lymphomas and the current approaches, including a drug screening assay, which has the potential to further explore the mechanisms of ibrutinib resistance and to design effective individualized combination therapies to overcome resistance and disable aggressive lymphomas (see Outstanding Questions).
Complexities of Ibrutinib Resistance (IR) in B Cell Lymphomas
Ibrutinib, a Bruton tyrosine kinase (BTK) inhibitor, has unprecedented activity in a variety of B cell lymphomas; however, relapse in B cell lymphomas is virtually universal and associated with dismal outcomes [1,2]. IR involves mutations and reprogramming and reactivation of key intracellular signal networks, which allow the bypass of targeted kinase signaling [3,4]. Inhibition of BTK leads to complex rewiring of adaptive signaling that evolves from kinome remodeling and changes in gene expression. These events are modulated in part by external signaling from the TME [5–10]. Interactions between malignant cells and the TME dictate the clinical behavior, prognosis, and response to treatment, thus contributing to disease relapse. Chemokine- and integrin-mediated migration and adhesion of malignant cells to nonmalignant stromal cells within the TME confer protection against drug-induced apoptosis and drug resistance [6].
BTK inhibitors have molecular effects that are not explained by the classic role of BTK in lymphoma biology [11]. In particular, ibrutinib can rapidly induce lymphocytosis (an increase in the number of circulating lymphocytes) in patients with chronic lymphocytic leukemia (CLL) and Mantle cell lymphoma (MCL), in parallel with lymph node regression, suggesting that ibrutinib impairs the homing and retention of B cells within the lymph nodes, bone marrow, and other sites. This phenomenon has been also observed with inhibitors of Syk and phosphoinositide 3-kinase (PI3K), which also target B cell receptor (BCR) signaling pathways [12]. Indeed, the initial success of these inhibitors has been attributed to attenuation of BCR-dependent lymphoma–TME interactions [3,13], including their effects on BCR- and chemokine-controlled integrin-mediated adhesion and homing of malignant B cells to their growth- and survival-supporting lymph node and bone marrow TME [6]. Clearly, TME is integral to lymphoma survival, progression, and development of IR. Understanding the breadth of TME-mediated (de novo) and acquired drug resistance, as well as their unifying signaling networks and mecha- nistic links, is crucial to developing appropriate strategies for durable clinical efficacy.
Remodeling of the Kinome Network Drives IR Many tumors respond to kinase inhibitors with rapid adaptive changes in signaling networks that bypass the targeted oncoprotein [14]. This occurs because the linear and the parallel kinase architectures allow targeted therapies to be circumvented by signaling crosstalk through alterna- tive pathways. Multiplexed inhibitor bead-quantitative mass spectrometry technology, which can measure hundreds of kinases, have shown that tumor cells acquire resistance to targeted therapiesthrough adaptivekinome networkreprogramming [15–18]. Specifically, it is thisadaptive reprogramming that contributes to the ultimate failure of therapies. However, whether kinase- mediated resistance mechanisms are involved in IR has been up to now largely unknown [19].
To capture the complexity of IR, we modeled acquired resistance by generating resistant MCL cell lines and used chemical proteomics to determine the key kinase signaling pathways associated with its development [19]. Activity-based protein profiling was used to identify global kinome adaptations between parental and IR cells. When relative kinase activities were compared between cells at a systems level, we found that the kinome of IR cells was markedly altered, with many kinases having increased activity. IR cells were enriched for kinases within survival/growth pathways, including cyclin-dependent kinases, integrin-linked kinases, DNA damage response proteins, and JAK–STAT (Janus kinase–signal transducer and activator of transcription). When kinome alterations following ibrutinib treatment were compared, we identified kinases/pathways inhibited in sensitive cells that remained unaltered or elevated in IR cells despite ibrutinib treatment. As expected, IR cells displayed a more blunted response to ibrutinib than parental cells, with many kinases showing unchanged or even increased activity in response to ibrutinib, supporting the concept that BTK-independent kinases or BTK bypass pathways drive IR. Intriguingly, this adaptive response was accompanied by an increase in cell–cell adhesion, proliferation, and resistance to chemotherapy after ibrutinib treatment in IR cells, highlighting how these global changes affected the kinome. Therefore, rather than there being a single mechanism of acquired IR for a given therapy, numerous signaling networks are rewired in MCL cells as they become resistant to targeted therapies. Indeed, resistance to targeted therapy, with PKIs or antibodies, involves kinome network remodeling in both solid tumors and hematologic malignancies [19–22].
Influence of TME in Intrinsic IR
Acquired drug resistance is traditionally studied by chronically exposing cancer cells to increasing drug doses [23–25]. These in vitro models have been instrumental to identifying mechanisms that modulate drug response and, in some cases, have aided in the identification of drug targets. However, these models fail to account for the TME in the emergence of the drug-resistant phenotype. Using naïve-IR cell line pairs, we addressed the question of how cellular components of the TME influence acquired drug resistance. We determined the kinome network of cells with (i) de novo (TME-mediated) drug resistance; (ii) acquired drug resistance (chronic exposure); and (iii) IR while adherent to stroma. We delineated functional resistance mechanisms and corresponding kinase activation profiles associated with de novo TME-mediated drug resistance and acquired resistance mechanisms. When MCL cell lines selected in the absence of stroma for acquired IR were compared with parental MCL cell lines after adhesion (coculture) to the stroma (de novo), we found that acquired IR mechanisms were more complex than those of de novo TME-mediated drug resistance. Despite these observed differences, we observed a significant overlap in kinase changes, supporting the notion that TME is essential for initial survival, priming MCL cells for the acquisition of stable drug resistance. Further mapping of the IR-associated kinome identified the PI3K–AKT–mTOR (mammalian target of rapamycin) pathway as a central signaling hub in IR MCL cells. Functional studies demonstrated that activation of PI3K-AKT–mTOR contributes to increased expression of b1 integrin, leading to sustained cell adhesion and stroma-mediated survival in IR MCL cells [6]. Recent studies have also demonstrated the contribution of a sustained AKT–mTOR pathway to IR development [4,26,27].
IR also overlaps with TME-mediated drug resistance by priming IR MCL to be selected and enriched by TME through activation of adhesion/integrin signaling pathways. This process leads to enhanced TME–MCL cell interactions and amplifies de novo TME-mediated drug resistance. As stated above, after chronic exposure to ibrutinib, MCL cells are characterized by enforced activation of the AKT–mTOR pathway and other signaling pathways, with the PI3K–AKT–mTOR signaling pathway identified as a central hub between the extrinsic B-cell microenvironment and the intrinsic signaling pathways [6]. In the presence of ibrutinib, PI3K–AKT–mTOR pathway activation in IR MCL cells resulted in increased b1 integrin expres- sion and a greater potential for IR cells to adhere and interact with TME compared with naïve cells [19]. Drug-naïve cells were released from stroma in the presence of ibrutinib, whereas IR cells were selected, retained, and enriched in the TME, suggesting a cooperative function of TME-mediated and acquired–mediated mechanisms of drug resistance and the aggres- siveness of relapse (Figure 1). Importantly, kinome reprogramming in MCL with selected resistance to ibrutinib facilitates the reception and interpretation of cues from the TME in a different manner, facilitating a more aggressive phenotype. Thus, the evolution of IR and its clinically catastrophic sequelae occurs via a dynamic and bidirectional interplay between the tumor and its milieu.
Potential Therapeutic Strategies to Combat IR
When designing rational combinations to overcome IR, it is necessary to understand and integrate the multiple mechanisms of tumor cell adaptation and to understand that resistance mechanisms that drive clonal evolution can occur before, during, and after treatment. Lymphoma cells comprise a heterogeneous collection of cell states with distinct signal trans- duction networks for IR MCL tumor growth and survival. Therefore, these diverse IR mecha- nisms pose a significant therapeutic challenge. Targeting bypass mechanisms with combinatorial approaches or inhibiting kinome reprogramming or its common upstream drivers could be a more efficient approach to overcoming drug resistance and improving current therapy in B cell malignancies.
Targeting Bypass Mechanisms with Combinatorial Approaches
A major challenge to the use of targeting bypass mechanisms is identifying the major kinase pathways facilitating IR. In MCL cells, networks are rewired in a plethora of ways during treatment as resistance emerges (Figure 2). In our previous study in which the PI3K–AKT–mTOR pathway was identified as the central signaling hub in IR MCL cells, we found that the downstream targets contributed to increased b1 integrin expression, enforced cell adhesion to TME stromal cells, and enhanced clonogenic growth. This provided a strong rational for combining PI3K–AKT–mTOR inhibitors with ibrutinib to overcome resistance [19]. In preclinical studies, combining PI3K p110 g/d inhibitors with ibrutinib demonstrated activity in IR in MCL, CLL, and other B cell lymphomas [4,28,29]. Preclinical efficacy in CLL and MCL was also shown in studies that targeted the BCRs downstream of BTK, such as protein kinase (PKC) and phospholipase Cg [30,31].
Figure 1. Evolution of IR. (A) After ibrutinib exposure, naïve tumor cells undergo selective pressures in specific niches of the TME, providing a sanctuary for tumor cells and allowing them to survive the insult of therapy via de novo and/or EMDR, resulting in minimal residual disease. Over time, these quiescent residual tumor cells (persisters) expand with acquired-resistance phenotypes including BTK mutations. (B) IR evolves through the expansion of subpopulations with multiple epigenetic/ genetic alterations (acquired or via the selection of pre-existing clones), leading to disease relapse and ultimately aggressive progression. Abbreviations: CAM-DR, cell adhesion-mediated drug resistance; DR, drug resistance; EMDR, environment-mediated drug resistance; IR, ibrutinib resistance; MCL, mantle cell lymphoma; TME, tumor microenvironment.
BCL-2 family proteins have recently emerged as pivotal therapeutic targets in many B cell lymphomas. Venetoclax (ABT-199), a first-in-class, selective, oral inhibitor [32] with subnano- molar affinity for BCL-2, showed antitumor activity in non-Hodgkin lymphoma, CLL, and acute leukemias in vitro [32–35]. In an in vivo mouse xenograft study, it also showed activity against aggressive lymphomas [36]. Thus, venetoclax has emerged as an exciting drug for many aggressive B cell lymphoma subtypes and may play an important role after development of IR. Indeed, our data and those of others have revealed that IR MCL cells are reprogrammed with a greater reliance on BCL-2 and that BTK inhibition enhances mitochondrial BCL-2 dependence [37,38]. This may be attributed to sustained activation of the PI3K–mTOR pathway, which leads to myeloid leukemia cell differentiation protein (MCL)-1 phosphorylation and degradation, culminating in a high BCL-2-to-MCL-1 ratio in IR MCL cells [19]. Indeed, venetoclax and ibrutinib synergistically inhibited the growth of both ibrutinib-sensitive and -resistant CLL cells in vitro and in vivo [39]. This synergism was shown in vitro in MCL [40]. Collectively, these data indicate that targeting compensatory PI3K–AKT–mTOR, PKCb, and BCL-2 pathways may serve as novel combination treatment strategies to overcome IR. However, it is important to note that it is likely that additional potential compensatory signaling beyond these pathways may be selected in a given patient. This underscores the need for continued study to develop tools to identify pathways for appropriate therapy in individual patients (or subgroups of patients).
Figure 2. Heterogeneity of IR. (A) Ibrutinib-resistant tumor cells are influenced by ibrutinib exposure and the TME, leading to reprogramming of a number of cellular processes. Importantly, this evolutionary process does not result in a monoclonal population of resistant cells. Instead, IR may be represented by several selected mechanisms. (B) IR manifests via genomic,transcriptomic, apoptotic, and kinomic reprogramming events. Critically, this translates into an unknown number of subpopulations with epigenetic/genetic alterations, each with a different compensatory pathway (e.g., potential vulnerability) to target. To this end, identifying a single combination for all patients with ibrutinib-resistant disease may be impossible. (C) To combat this, novel tools for personalized treatment strategies are needed addressing tumor heterogeneity. For example, automated, in silico, predictive, drug response assays can simultaneously screen primary myeloma/lymphoma cells from biopsies against panels of drugs in an ex vivo reconstruction of the bone marrow TME, including extracellular matrix, patient-derived soluble factors, and patient-derived stroma. Abbreviations: AGC, named after the protein kinase A, G, and C families (PKA, PKC, PKG); BRD4, bromodomain-containing protein 4; BTK, Bruton tyrosine kinase; CAMK, calmodulin/calcium regulated kinases (CAMK) in CAMK1 and CAMK2 families; CDK, cyclin-dependent kinase; CK1, cell kinase 1, originally known as casein kinase 1; CMGC, named after another set of families (CDK, MAPK, GSK3, and CLK); IR, ibrutinib resistance; JAK, Janus kinase; MCL-1, myeloid leukemia cell differentiation protein-1; NEK, NIMA related kinase; PLCG2, phospholipase C gamma 2; PLK1, polo like kinase 1; Pol2, RNA polymerase II; STE, homologs of the yeast STE7, STE11, and STE20 genes, which form the MAPK cascade, transducing signals from the surface of the cell to the nucleus; TK, tyrosine kinase; TKL, tyrosine kinase-like; TLK2, tousled like kinase 2; TME, tumor microenvironment.
Targeting the Transcription Machinery to Abrogate IR-Associated Kinome Remodeling A prominent mechanism of IR in MCL is adaptive kinase network remodeling and resilience, which can overcome the initial drug insult (Figure 2B). We argue that the dysregulation of key transcrip- tional regulators notonly definesthecancer phenotype but is essential forthedevelopment of drug resistance. Recent studies have demonstrated that adaptive kinome reprogramming and altered cancer cell states result from gene expression programs coordinated by epigenetic regulators [e. g., bromodomain-containing protein (BRD)4, cyclin-dependent kinase (CDK)7, CDK9, and RNA polymerase II (RNAP2)] at distinct superenhancer sites [41–44] (Figure 2B). Adaptive responses involve transcriptional upregulation of multiple kinases, as a whole, accounting for increased proliferation and drug resistance, enhanced TME–MCL cell interactions, and amplified TME- mediated drug resistance in IR MCL cells. Adaptive responses from the kinome and transcriptome serve to reactivate the targeted pathway orinitiate by pass tracks, ultimately limiting the durability of targeted therapies and creating therapeutic challenges. With the degree of reprogramming, one could envision a situation in which combinations of two, or even three, kinase inhibitors would be insufficient to suppress the resiliencyofthekinome networks. Tothisend, inhibition ofthecommon transcriptional mechanisms contributing to kinase remodeling could be another strategy to more efficiently overcome drug resistance.
We also argue that targeting the transcriptional machinery is less prone to bypass by alternative mutational events or pathway activation and thus less likely to engender rapid drug resistance emergence. Recently, Zawistowski et al. elegantly demonstrated that inhibition of MEK1/2 with trametinib in patients with triple-negative breast cancer induced dramatic transcriptional responses, including upregulation of global receptor tyrosine kinases. More specifically, MEK inhibition induced genome-wide enhancer formation involving the seeding of BRD4, H3K27 acetylation, and p300, to facilitate transcriptional adaptation. Inhibition of the P-TEFb- associated proteins BRD4 and CDK9 arrested kinome upregulation, overcoming trametinib resistance and producing sustained growth inhibition ex vivo and in vivo [41]. These non- genomic adaptive bypass mechanisms, for example, involving transcriptional upregulation of receptor tyrosine kinases, are not limited to breast cancer. Rather, they are increasingly observed as major mechanisms of clinical resistance in many other cancers [45–47].
Accordingly, targeting proteins involved in transcriptional programs, such as BRD4, are increasingly appreciated as an attractive and novel therapeutic intervention to durably inhibit adaptive resistance. Indeed, it was recently reported that targeting the bromodomain and extra-terminal domain (BET) protein bromodomain with JQ1 attenuated MYC, CDK4/6, RelA, and BCL2 protein levels and the expression of nucler factor (NF)-kB target genes in IR MCL, demonstrating that bromodomain inhibitor-based combinations are highly active against IR MCL. Mechanistically, BET proteins assemble a complex of coregulatory proteins at enhancers, thereby regulating gene transcription [48,49]. The C-terminal-positive transcription elongation factor b-interacting domain of BRD4 interacts with and localizes CDK9 to the superenhancer and promoter regions, further promoting the activity of RNAP2 and gene expression of important MCL-relevant oncogenes, such as MYC, BCL-2, and kinase genes [49–52]. The importance of these findings are further exemplified by recent work demonstrating that MYC overexpression can facilitate resistance to pharmacological inhibitors of BCR sig- naling, such as ibrutinib [26]. This is in agreement with the notion of targeting BRD4 to silence transcription activation and Myc transcript for IR.
Heat shock protein (Hsp)90 inhibitors have also been shown to overcome IR in MCL through the modulation of multiple oncogenic pathways, including alternative NF-kB signaling path- ways [53]. This may be due to Hsp90-associated regulation of sequence-specific transcription
factors, chromatin accessibility, and RNAP2 activity [54,55]. Accordingly, dysregulation of Hsp90 may similarly influence IR processes by modulating the transcriptional machinery, thus serving as an innovative target for IR MCL. Thus, targeting the transcriptional machinery can be an effective strategy to block kinome remodeling and IR development, perhaps leading to another treatment option for MCL and many other types of cancers.
Implementing Cell-Based Drug Screening on Primary Samples as a Clinical Tool to Tailor Therapy for Patients with IR
Although molecularly targeted therapeutics continue to improve outcomes for cancer patients, inter- and intratumoral heterogeneity and tumor plasticity limit their efficacy and durability. The use of cell lines, while important in cancer biology, may not be the best choice for translational application given the genotypic/phenotypic changes necessary for growth in artificial culture conditions. Thus, it is necessary to have a comprehensive functional strategy to directly determine the drug dependency of cancer cells based on ex vivo drug sensitivity and resistance tests. Such a strategy can allow the dynamics of drug response to be studied in the most relevant model system (the patient sample). This would facilitate a more personalized approach and a more clinically relevant system to predict and address drug resistance in cancer patients. To this end, we recently leveraged our newly developed cell-based drug screen assay to optimize and tailor therapeutic strategies (Figure 2C). For this, we capitalized on a novel ex vivo microfluidic cell-based platform termed EMMA (Ex vivo Mathematical Malignancy Advisor) to assess drug responses in a 3D reconstructed TME in sensitive and IR MCL lines and primary MCL and myeloma samples [19,56]. Predicting the clinical response of patients based on ex vivo assays is a major challenge irrespective of how close the assay is to in vivo conditions. The most obvious difficulty is the translation of results from an assay that lasts days into estimates of clinical response across months or even years [57,58].
Recent studies in predictive modeling have suggested different approaches to identifying agents with clinical efficacy in liquid and solid cancers. Pemvoska et al. described a combina- tion of ex vivo chemosensitivity and molecular profiling to determine therapeutic windows for drugs in acute myeloid leukemia [59]. Majumder and colleagues combined ex vivo chemo- sensitivity assays of slices of tumor, immunohistochemistry, and clinical data to create a signature to classify clinical response of patients with solid tumors [60]. Although these assays can predict the initial effects of therapy on tumor burden (responders vs nonresponders), they cannot predict the actual depth, duration, or time to relapse. In B cell disorders (e.g., myeloma and MCL), extent of response is commonly utilized as a surrogate of clinical benefit [58]. As such, a system capable of creating actual clinical trajectories (response) will be central to successfully translating in silico predictions to true clinical outcomes. EMMA provides an instrumental platform to address these issues.
We have begun to bridge this timescale gap through the use of mathematical models accounting for tumor heterogeneity, pharmacodynamics, and pharmacokinetics imputed with a tested phenotype (drug sensitivity). It has been long known that nature selects for phenotype, not genotype, and that multiple genotypes can produce the same phenotype [61]. This nonexclusive relationship makes it challenging to predict clinical outcomes from genotype alone or even gene expression profiles [62]. EMMA directly identifies the phenotypic (or functional) representation of subpopulations regardless of genotypic background, thus remov- ing the ‘middle man’ and producing in silico clinical response outputs. Through nonlinear regression of the ex vivo chemosensitivity results, the model generates trajectories of clinical response demonstrating a high degree of accuracy in predicting outcomes.
With this first-in-class automated, in silico, predictive, drug response assay, primary myeloma/ lymphoma cells were screened from biopsies against a panel of drugs in an ex vivo recon- struction of the bone marrow TME, including extracellular matrix, patient-derived soluble factors, and patient-derived stroma [19,56,63]. Parental (sensitive) or IR MCL cells were then seeded in 384-well plates previously coated with human-derived stroma cells and collagen-1, and a subsequent drug screen was performed using a set of agents and our IR model. AKT and mTOR inhibitors (MK2206 and INK128) demonstrated more dramatic attenuation of cell viability in IR cells versus naïve MCL cells [19,38]. Moreover, the screening results were independently validated by MCL-derived xenograft and MCL patient-derived xenograft models [19]. There- fore, cell-based drug screening could be used to guide the selection of targeted therapy for overcoming IR in MCL. When this drug screen assay was also applied to multiple myeloma patients [56], the assay correctly classified 96% as responders/nonresponders and correctly classified 79% according to International Myeloma Working Group stratification of level of response with a significant correlation between predicted and actual tumor burden measure- ments. This cell-based drug screen assay demonstrated its feasibility for estimating clinical efficacy of drugs.
The accuracy, reproducibility, short turnaround time (5 days), and high-throughput potential of this platform demonstrates that this cell-based drug screen in the setting of TME is a promising tool for effective individualized therapy against MCL and other B cell lymphomas. However, we are aware that continued work is needed to add or develop further modeling tools to predict future occurrence of drug resistance in individuals. With this personalized approach, we would be able to provide the right drug or drugs to the right patient at the right time. Importantly, our findings of drug screening are closely correlated with the results of our functional and chemical proteomic experiments, implicating this drug screen assay as an innovative platform to help to further explore the IR mechanistic insight and design effective individualized combination therapy to overcome IR and disable aggressive lymphomas.
Concluding Remarks
When therapeutic stress is applied, a series of cell responses occur in both lymphoma and stromal cells, creating a positive feedback loop that amplifies the prosurvival and growth signals and eventually acquired IR phenotype. Drug resistance, including IR, rarely relies on a single aberrant pathway; networks are rewired in an untold number of ways in lymphoma cells as they become resistant to targeted therapies. Identifying common signaling nodes or drivers for kinome and transcriptome associated with resistance evolution is warranted. Because of the dynamics and complexity of tumor evolution and the often lack of correlations between mutations with therapy response, the individualized dynamic model in conjunction with a high-throughput, cell-based drug screening platform can be a solution to these complex challenges, potentially allowing precision combinations to be tailored to individual patients. Moving forward, the continued application of this approach in combination with genomes, functional transcriptomes, and proteomes to primary and recurrent tumors will identify novel mechanisms driving tumor progression and therapeutic resistance, thereby facilitating the identification of optimal combinatorial therapeutic strategies in the right patient at the right time. Inhibition of these pathways and resistance networks could be more efficient to overcoming drug resistance and improving current therapy in MCL and other B-cell malignancies.