Multi-Component, Time-Course screening to develop combination cancer therapies based on synergistic toxicity Authors: Michele Ceribelli, Frances Anne Tosto, Xiaohu Zhang, Christopher J. Melani, Mark Roschewski, Erin Beck, Carleen Klumpp-Thomas, Cody J. Peer, Kelli M. Wilson, Lu Chen, Crystal McKnight, Sam Michael, Zina Itkin, Paul Shinn, William D. Figg, Wyndham H. Wilson, Louis M. Staudt, Craig J. Thomas Published: 2024 ABSTRACT SignificanceCytotoxicity screening of oncology-focused drug libraries versus cancer cell models represents a powerful method to identify promising therapies for translation into human clinical trials. However, these efforts rarely produce the data needed to associate diverse in vitro cellular responses with the multiple parameters that define the concentration ceilings, time limits, and combination interactions which govern in vivo cellular responses and therapeutic indexes. The data we present reveal how a targeted multicomponent therapy in lymphoma benefits from synergy-driven combination responses that align drug concentration and timing with drug pharmacokinetics in a manner consistent with in vivo drug exposures. The phenotypic screening of mechanistically annotated drug libraries is a practice first implemented by the pharmaceutical industry to compare recently developed agents versus best-in-class approved and investigational drugs (1, 2). The technique has gained practitioners in academia owing to the highly informative results and experimental feasibility. The data that arise from these efforts can reveal distinctive mechanism-to-phenotype associations, uncover additional polypharmacology for existing drugs, and complement functional genomics to provide systems-level insights into cellular dependencies (3–5). Moreover, the outcomes of these screens can initiate a translational path toward innovative clinical investigations. Clinical trials, in their best form, are built on a sound mechanistic rationale whereby a drug’s mechanism of action exploits unique features of cellular/tissue dysregulation associated with disease origin or cause. The diverse mutational landscape of human cancers creates multiple possibilities for mechanistically driven drug activity, and cytotoxicity-based screens can uncover these opportunities. Several public interfaces are available that present cytotoxicity-based drug screening data across large, heterogeneous panels of cancer cell lines, including the Welcome Trust “Genomics for Drug Sensitivity,” the Broad Institutes “Drug Repurposing Hub,” and the NCI “CellMiner” online resource (6–9). Our teams’ contributions typically focus on targeted screening outcomes from a cell line panel that models a specific disease/cancer type (10–13). For these studies, we utilize a mechanistically annotated, oncology-focused library of nearly 3,000 approved and investigational small molecules referred to as the Mechanism Interrogation PlatE (MIPE) library (12). The MIPE library embraces mechanistic redundancy for key oncogenic targets (e.g., EGFR, HSP90) while simultaneously achieving exceptional mechanistic diversity (>850 unique mechanisms of action interrogated). In single-agent studies, we explore this library in quantitative high-throughput screening (qHTS) settings, where each agent is tested at 11 concentrations (from 46 µM to 0.8 nM) generating full concentration–response curves for each library member (14). qHTS outcomes routinely evolve into targeted, pairwise combination assessments involving several agents of interest (i.e., an “all-versus-all” screen) or into the combinatorial examination of a drug of choice versus the entire collection (i.e., a “drug-versus-all” screen). The first reported study from our screening platform evaluated the combination landscape of the Bruton’s tyrosine kinase (BTK) inhibitor Ibrutinib in diffuse large B cell lymphoma (DLBCL), a molecularly heterogeneous disease derived from transformed B lymphocytes (10). Gene expression profiling subcategorizes DLBCL into two main subtypes (15–18). The more aggressive subtype is referred to as the activated B cell–like (ABC DLBCL) and is defined by chronic active B cell receptor (BCR) signaling which promotes abnormal growth and survival. The second subtype is known as germinal-center B cell (GCB) DLBCL and is characterized by tonic-level BCR signaling and overall favorable outcomes in response to standard multicomponent chemotherapies. More recently, the co-occurrence of genetic alterations within DLBCL tumors was used to define genetic subtypes with distinct pathogenesis and response to chemotherapy and targeted therapies (17–21). Genetic subtypes of ABC DLBCL include “MCD” (27%), “BN2” (13%), A53 (10%), and “N1” (6%) while the genetic subtypes of GCB DLBCL include EZB MYC– (29%), EZB MYC+ (11%), BN2 (12%), ST2 (9%), and A53 (9%). EZB MYC+ is also known as high-grade B cell lymphoma “double hit” with BCL2 and MYC translocation (dHIT) and has an inferior prognosis among GCB DLBCL cases. BTK is the proximal signaling hub for acute and chronic BCR signaling, and BTK inhibition by Ibrutinib is directly related to its clinical activity in chronic lymphocytic leukemia (CLL), mantle cell lymphoma (MCL), Waldenstrom macroglobulinemia, and ABC DLBCL (22–25). The first reported study from our combination screening platform identified many synergistic drug combinations with Ibrutinib, establishing the primary rationale for the development of multitargeted combination therapies centered around BTK inhibition (10). Synergistic partner agents included drugs that block the BCR-signaling cascade at alternate targets (e.g., SYK inhibitors) as well as parallel and convergent signaling pathways (e.g., PI3K inhibitors). Inhibitors of the antiapoptotic BCL-2 family proteins (e.g., Navitoclax, Venetoclax) also possessed strong synergy with Ibrutinib. The outcomes of that study, alongside deep mechanistic insights and clinical erudition, were used to design the VIPOR multicomponent targeted therapy trial in lymphoma (NCT#03223610) (Fig. 1A) (26). Each of the VIPOR drugs [i.e., Venetoclax, Ibrutinib, Prednisone, Obinutuzumab, and Revlimid (Lenalidomide)] has strong mechanistic rationale for targeting specific oncogenic survival dependencies in B cell lymphomas. Venetoclax, Ibrutinib, Lenalidomide, and Prednisolone (Prednisone’s active metabolite) each possess cell-intrinsic, single-agent activity in selected lymphoma models (in vitro and in vivo), as well as against primary human disease. Pairwise combination activity of VIPOR drugs in DLBCL. (A) Visual schematic of the mechanistic targets of the VIPOR multicomponent therapy. (B) Synergy heatmap representation of all the possible pairwise combinations of Venetoclax, Ibrutinib, Prednisolone, and Lenalidomide in four GCB DLBCL (Left) and four ABC DLBCL (Right) cell lines. For each pair, viability was assessed 48 h posttreatment in 10 × 10 matrix blocks, and the corresponding ΔBliss values were then used to generate this synergy heatmap. (C) Heatmap view of a selected Lenalidomide+Venetoclax 10 × 10 block in the OCI-Ly10 cell line. Both % response and ΔBliss heatmaps are shown. (D) As in C, but the HBL1 cell line is shown. Herein, we expand on our previous work by exploring all possible VIPOR small-molecule drug+drug combinations in multiple DLBCL cell models utilizing multiple concentration ranges and time points. Moreover, we established a broad experimental framework to granularly map the multidrug interactions of the VIPOR agents referred to as Multi-Component, Time-Course (MCTC) screening. The resulting data allow for a systematic comparison of in vitro HTS outcomes with clinical drug responses of the VIPOR regimen and highlight the value of expanded formats for in vitro drug screening to uncover promising multicomponent therapies for the treatment of cancer. Results Single-Agent and Pairwise Combination Responses to VIPOR Drugs. We first conducted a single-agent survey of the small-molecule VIPOR drugs [i.e., Venetoclax, Ibrutinib, Prednisone (and its active metabolite Prednisolone) and Revlimid (Lenalidomide)] in a panel of 14 B cell lymphoma cell lines including models of DLBCL, follicular lymphoma, and marginal zone lymphoma using a cytotoxicity informing assay (CellTiterGlo™ @ 48 h post drug treatment) (SI Appendix, Fig. S1A and Dataset S1). Single-agent drug responses were not uniformly driven by disease subtype. For instance, two of the cell lines most sensitive to Venetoclax were the ABC DLBCL cell line Riva and the GCB DLBCL cell line OCI-Ly8 (SI Appendix, Fig. S1A and Dataset S1). Response to Ibrutinib was also heterogeneous across subtypes at the time point and concentrations tested. The prodrug Prednisone was inactive in these in vitro screenings (as expected), whereas its active metabolite Prednisolone was cytotoxic in selected cell lines, again spanning across disease subtypes. Lenalidomide did not show substantial single-agent activity at the concentrations and time point tested. Next, we performed combination assessments for all possible pairs of VIPOR drugs (six pairs), focusing on a panel of GCB DLBCL (OCI-Ly8, RC-K8, SUDHL6, and SUDHL9) and ABC DLBCL (HBL1, OCI-Ly10, Riva, and TMD8) cell lines. Briefly, we utilized the outcomes of the single-agent studies to define the appropriate concentration ranges for each agent and used a 10 × 10 matrix (checkerboard) layout to comprehensively survey the “concentration space” for each unique combination. Multiple techniques, including the delta Bliss method, were used to define synergism/antagonism within each matrix. To enable comparative assessments of drug+drug interactions across cell lines with divergent single-agent responses, we utilized multiple concentration ranges for each drug (except Lenalidomide), for a total of 18 pairs that we assessed with both a viability assay (CellTiterGlo™, 48 h after treatment) and an apoptosis induction assay (CaspaseGlo™, measured both at 9 h and 18 h after treatment) (Fig. 1B, SI Appendix, Fig. S1 B and C, and Dataset S2 A–I). Drug+drug synergism was dominant in the ABC DLBCL subtype, with significant enrichment in the three “MCD” cell line models (HBL1, OCI-Ly10, and TMD8) (Fig. 1 B, Right section of the heatmap). This included several instances of exceptional synergy, where strong combinatorial cytotoxicity was achieved at concentrations with limited/partial single-agent activity (Fig. 1C). Conversely, synergism was more limited in GCB DLBCL cell lines, with only the dHIT/GCB cell lines, OCI-Ly8 and SUDHL6, displaying any appreciable synergy when exposed to selected drug+drug combinations (Fig. 1 B, Left section of the heatmap). Among the most unexpected and intriguing findings was the degree of response heterogeneity, even within well-defined genetic subtypes. For example, the combination of Venetoclax and Lenalidomide possessed remarkable synergy in OCI-LY10 cells whereas the same combination was not synergistic in HBL1 cells (compare Fig. 1C with Fig. 1D) or TMD8 cells. Collectively, these data align with previous reports suggesting that the ABC DLBCL subtype will be the most likely to respond to the VIPOR regimen. At the same time, these data suggest that unknown determinants of heterogeneity can influence the nature of drug combination responses. Pairwise Combinations Promote Drug Activity at Concentrations Achievable in Humans. To fully appreciate the translational value of these data, it is essential to estimate whether the drug concentrations that elicit a response in each cell line model can be achieved in vivo. At safe human doses, each VIPOR drug has a defined human exposure, i.e., a systemic concentration over time based upon the drug’s pharmacokinetic (PK) profile (Fig. 2A). Ideally, the active in vitro concentration ranges for each drug would be below safely achievable in vivo concentrations. To assess this, we reviewed published PK data for Venetoclax, Ibrutinib, Prednisone (and its active metabolite Prednisolone), and Lenalidomide (Fig. 2B and Dataset S3). The amount of drug that systemically dispersed cancer cells may experience is a function of drug PK parameters, alongside the amount of drug not bound by serum proteins [fraction unbound (fu)] and the degree to which the drug distributes beyond the circulatory system [volume of distribution (VD)]. Consequently, the true active and sustained bioavailable concentration range for most drugs at the site of disease is likely to be considerably below reported PK parameters. VIPOR pairwise combinations are active at clinically achievable concentrations. (A) A generic PK drug concentration versus time (PK) plot profile highlighting key PK parameters (Cmax: maximum concentration measured; Tmax: time of Cmax) establishing the therapeutically relevant concentration range. (B) The reported maximum achieved concentration (Cmax) and area under the curve (AUC) values for Venetoclax, Ibrutinib, Prednisolone, and Lenalidomide (see Dataset S3 for appropriate references). (C) Heatmap view of % response (cytotoxicity) for the Lenalidomide+Venetoclax combination in OCI-Ly10 cells. Drug exposure (Cmax) levels are highlighted in the plot. (D) As in C, but the Venetoclax+Ibrutinib combination in SUDHL6 cells is shown. (E) As in C, but the Prednisolone+Venetoclax combination in HBL1 cells is shown. (F) As in C, but the Lenalidomide+Ibrutinib combination in TMD8 cells is shown. Utilizing each agent’s Cmax and the calculated 5% Cmax as a stringent proxy of a more attainable in vivo exposure, we reassessed the in vitro activity of each VIPOR agent to evaluate its translational potential. Based on these parameters, only a limited set of single-agent responses to VIPOR agents would be predictive of clinical effectiveness (SI Appendix, Fig. S2 A–D). In pairwise-combination settings, however, synergy-driven cytotoxicity was observed at attainable in vivo concentrations. For example, the 800 mg dose of Venetoclax used in the VIPOR trial is reported to yield a Cmax range of 3.44 to 3.95 µM (Fig. 2B) (26–28). Venetoclax, however, is highly bound to serum proteins (fu < 0.1%) making it unlikely that these concentrations are experienced system-wide. At the more realistic concentration range of 5% of the Cmax (~180 nM), Venetoclax as a single agent only achieves ~10% cytotoxicity in OCI-Ly10 cells. However, in combination with Lenalidomide the cytotoxicity of 180 nM of Venetoclax exceeds 80% (Figs. 1C and 2C). Importantly, this happened across multiple Lenalidomide doses that did not show any single-agent activity. Similar dose-reduction shifts were noted for selected combination responses to Ibrutinib, Prednisolone, and Lenalidomide (Fig. 2 D–F). Altogether, these data highlight how combinatorial drug activities are instrumental in eliciting cell responses at concentrations that are achievable in human subjects. Pairwise Combinations Promote Rapid Drug Activity. The systemic exposure to small-molecule drugs following dosing is time limited. To achieve clinical effectiveness, drug-induced cytotoxicity should ideally happen within the exposure window of each drug (Fig. 3A). Exceptions include covalent inhibitors, such as Ibrutinib, that possess a beneficial decoupling of pharmacodynamic effect from the actual PK exposure. To systemically map the temporal onset of drug activity within VIPOR, we performed a granular kinetic assessment of pairwise drug interactions measuring apoptosis induction every 3 h and viability every 6 h up to 36 h of combination treatment. We performed these experiments in two ABC DLBCL (HBL1 and OCI-Ly10) and two GCB DLBCL (OCI-Ly8 and SUDHL9) cell lines. Overall, we observed excellent correlation between synergistic cytotoxicity and synergistic induction of apoptosis, identifying programmed cell death as a central determinant of VIPOR-driven cytotoxicity (Fig. 3B and Dataset S4 A–D). Importantly, apoptosis responses in synergistic drug+drug pairings were rapid, initiating as early as 6 h posttreatment in selected cell models. As previously observed for end point viability assays (Fig. 1B, cytotoxicity at 48 h posttreatment), VIPOR drug+drug apoptotic synergy was dominant in the ABC DLBCL subtype, with several instances of cell line–specific drug interactions (e.g., Lenalidomide-containing pairs in OCI-LY10). VIPOR pairwise combinations are active in clinically relevant timeframes. (A) A generic PK drug concentration versus time profile plot highlighting key PK parameters establishing the therapeutically relevant concentration range, the time to peak concentration (Tmax) and the reported half-life time (T1/2) values for Venetoclax, Ibrutinib, Prednisolone, and Lenalidomide from the literature (see Dataset S3 for appropriate references). (B) Heatmap representation of the temporal onset of synergy for all the possible pairwise combinations of Venetoclax, Ibrutinib, Prednisolone, and Lenalidomide in two GCB DLBCL and two ABC DLBCL cell lines. For each 10 × 10 drug pair, viability was assessed every 6 h (6, 12, 18, 24, 30, and 36 h posttreatment) and apoptosis (casp3/7 activation) every 3 h (3, 6, 9, 12, 15, 18, 21, 24, 27, 30, 33, and 36 h posttreatment). The corresponding ΔBliss values were then used to generate these synergy heatmaps. (C) Heatmap view of apoptosis induction over time (3, 6, 9, 12, 15, 18, 21, and 24 h posttreatment), for the 10 × 10 combination of Prednisolone+Venetoclax in the HBL1 cells. % response blocks are shown. The combination of Venetoclax+Prednisolone (V+P) in HBL1 cells offers a useful example for reviewing these data. Whereas Venetoclax has a long human exposure (T1/2 ~ 40 h), the human exposure to Prednisolone is limited (T1/2 ~ 2 to 4 h) (Fig. 3A and Dataset S3). Therefore, synergy-driven apoptosis and cytotoxicity associated with concurrent exposure to these agents must occur rapidly to influence therapy response in the clinical setting. Our kinetic assessment revealed rapid synergy onset in response to V+P in ABC DLBCL, including HBL1 cells (Fig. 3B). Even at the earliest viability time point assessed (6 h), the combination of V+P begins to demonstrate synergistic cytotoxicity. By the 24 h time point, this combination induces strong cytotoxicity, at concentrations below those that are systemically achievable for both agents (SI Appendix, Fig. S3 A and B). The apoptosis induction data perfectly align with these outcomes, again showing synergy onset at 6 h and strong apoptotic synergy across most concentrations tested by 24 h (Fig. 3C and Dataset S5 A–D). Altogether, these experiments revealed that apoptosis is the major contributor to VIPOR responses and that combination-based synergies are a driving element of rapid onset of both apoptosis and cytotoxicity. VIPOR Responses Are Driven by Multidrug Synergism. Methods for in vitro evaluation of three- and four-drug combinations are still emerging and defining “synergy metrics” for combinations involving more than two agents remains difficult. We reasoned that a time-course assessment of apoptosis induction following exposure to multiple agents could provide a more straightforward way to systematically assess multidrug interactions. Therefore, we designed an experiment whereby selected ABC DLBCL and GCB DLBCL cell lines were exposed to all the possible VIPOR drug-treatment sets (i.e., four single agents, six pairs, four triplets, and the four-drug combination of Venetoclax, Ibrutinib, Prednisolone, and Lenalidomide) (Fig. 4A). To cover a wide dose range, we established eight unique, fixed dose levels for each drug-treatment set; each level was based on equivalent concentrations of Venetoclax and Ibrutinib and slightly higher, equal concentrations of Prednisolone and Lenalidomide. Drug concentrations spanned from single-digit nM dosing (dose level #1: 2 nM Venetoclax and Ibrutinib; 5 nM Prednisolone and Lenalidomide) to µM dosing (dose level #8: 5 µM Venetoclax and Ibrutinib; 20 µM Prednisolone and Lenalidomide). Each drug-treatment set (N = 4) was assessed every 2 h over a 24 h period, using a luciferase-based Caspase 3/7 activation assay (CaspaseGlo™). The proteasome inhibitor Bortezomib (20 µM) was used as a positive control and relative luminescence unit (RLU) values were scaled to the maximum bortezomib signal achieved for each cell line’s time course (Fig. 4 B and C). We refer to this screening format as MCTC screening. Multicomponent assessment of VIPOR drug interactions in DLBCL. (A) Experimental layout of an exemplar 1,536-well plate examining eight dose/concentration levels for all the possible VIPOR drug-treatment sets (i.e., four single agents, six pairs, four triplets, and the four-drug combination of Venetoclax, Ibrutinib, Prednisolone, and Lenalidomide). One plate was used for each time point for each cell line. (B) Experimental design and heatmap view of normalized apoptosis induction by the V+P across all dose levels and time points in the HBL1 cell line. (C) Heatmap view of normalized apoptosis induction by all VIPOR drug-treatment sets in ABC DLBCL (HBL1 and OCI-Ly10) and GCB DLBCL (OCI-Ly8 and SUDHL9) cell lines. C1 = vehicle; C2 = 20 μM Bortezomib; 1 to 8 correspond to increasing dose levels. (D) Dose level–specific plot (dose #4) of normalized apoptosis induction in response to selected drug-treatment sets in HBL1 cells. (E) As in D, but responses for OCI-Ly10 cells are shown. (F) Heatmap view of differential apoptosis induction, obtained by subtracting single agent Venetoclax activity from the normalized apoptosis signal for V+P and V+I doublets. (G) Similar to F, heatmap view of differential apoptosis induction, this time displaying normalized apoptosis induction values of the V+I+P triplet minus those of the highest corresponding doublet (V+P). The single-agent and two-drug combination outcomes were consistent with previous experiments reported here (Fig. 4C and Dataset S6 A–D). Responses to higher-order drug combination sets were broader in ABC DLBCL, and importantly, synergy-driven apoptosis was detected at dosages where all single agents failed to elicit meaningful responses (dose level#4 or below). In HBL1 cells, for example, the combinations of Venetoclax+Ibrutinib (V+I) and V+P drove a strong dose and time-dependent apoptotic response that was superior to Venetoclax alone (single agent Venetoclax promoted apoptosis only at dose level #5 (200 nM) or above). This potentiation phenotype can more easily be appreciated by zooming into specific dose levels (Fig. 4 D and E) or by calculating the RLU difference between a given drug pair and the corresponding most active single agent (Fig. 4F). The same approach can be used to systematically highlight the contribution of drug triplets with respect to drug pairs, and more broadly to compare/contrast any (n) drug set to the corresponding, most active (n − 1) set (SI Appendix, Fig. S4A). For example, the V+I+P triplet achieved stronger and faster apoptotic induction at lower doses than the matched V+P pair (Fig. 4G), whereas the addition of Lenalidomide on top of the V+I+P did not provide any additional benefit (compare green and red lines in Fig. 4D) in this cell model. Consistent with pairwise combination outcomes, heterogeneity was observed within the well-defined ABC DLBCL MCD subtype. Whereas Venetoclax, Ibrutinib, and Prednisolone behaved similarly in HBL1 and OCI-Ly10 cells, only OCI-Ly10 cells benefited from Lenalidomide-based combinations including V+L, V+I+L, and V+P+L (Fig. 4 C and E). Notably, the V+P+L combination was the strongest triplet for OCI-Ly10 cells, with the addition of Ibrutinib on top of V+P+L further potentiating apoptotic outcomes (Fig. 4E). Consistent with our previous observations, the two GCB DLBCL cell lines we explored in this format (OCI-Ly8 and SUDHL9) did not show significant combination benefit (Fig. 4 C, Right section of the heatmap). Whereas Venetoclax was active in both cell lines as a single agent (including strong activity in the dHIT OCI-Ly8 model), the apoptotic signal observed in Venetoclax-containing treatment sets paralleled single-agent Venetoclax activity, with no significant potentiation by partner drugs (SI Appendix, Fig. S4 B and C). These outcomes illustrate the utility of the MCTC screening format and shed light on the complex nature of drug interactions within the VIPOR multicomponent therapy. The distinct benefit of selected two-drug and three-drug combinations in the two ABC DLBCL cell lines reveals fundamental information on therapy response while simultaneously uncovering a level of heterogeneity which is only revealed through the substantial concentration and temporal granularity of this experiment. In more static experimental formats, much of these data would be obscured or overwhelmed by the activity of a highly active single drug or drug combination. Multicomponent Chemotherapy Responses Are Driven by Individual Drugs. The current standard of care for non-Hodgkin’s lymphoma depends on multicomponent chemotherapy regimens, such as R-CHOP and EPOCH-R regimens [C = Cyclophosphomide, H = Hydroxydaunorubicin (also known as Doxorubicin), O = Oncovin (Vincristine), P = Prednisone/Prednisolone, E = Etoposide, R = Rituximab] (Fig. 5A) (29, 30). To gain comparative insight between VIPOR agents and these historical standards of care, we conducted a MCTC evaluation of drug interactions within CHOP/EPOCH regimens in one ABC DLBCL and one GCB DLBCL cell line (HBL1 and OCI-Ly8, respectively) (Fig. 5 A–E, SI Appendix, Fig. S5 A–D, and Dataset S7 A and B). Noticeably, even though most CHOP/EPOCH drugs possessed detectable, although variable, single-agent activity, synergy was extremely limited throughout the dataset, and antagonism was observed for selected combinations of these agents (Fig. 5C and SI Appendix, Fig. S5 A–C). In HBL1 cells, for instance, Vincristine was active at relevant in vivo concentrations (<50 nM) and time points (<20 h). However, most Vincristine-based combinations were either equivalent or less active than Vincristine alone, apart from a modest benefit from the addition of Prednisolone (Fig. 5 D and E and SI Appendix, Fig. S5A). The addition of Doxorubicin was particularly detrimental to apoptosis induction in Vincristine containing drug-treatment sets. In OCI-Ly8 cells, apoptotic responses were more limited and, intriguingly, driven solely by Doxorubicin (SI Appendix, Fig. S5 B–D). These results add further validation to the concept that R-CHOP and EPOCH-R elicit clinical responses primarily through patient-specific, single-agent drug responses. Time-course multicomponent chemotherapy outcomes in DLBCL are driven by individual drugs. (A) Definition of the small-molecule drug components of the R-CHOP and EPOCH-R regimens. Please note that Mitomycin was used in vitro as an “alkylating proxy” for Cyclophosphamide. (B) Experimental layout of a 1,536-well plate examining eight dose/concentration levels for all possible EPOCH/CHOP drug-treatment sets. (C) Heatmap view of normalized apoptosis induction by all possible EPOCH/CHOP drug-treatment sets across all dose levels and time points in the HBL1 cells. (D) Heatmap view of normalized apoptosis induction by Vincristine, Vin.+Pred., and Vin.+Dox. (yellow heatmaps, Left) in HBL1 cells. The corresponding differential apoptosis induction heatmaps, obtained by subtracting single agent Vincristine from the normalized apoptosis signal of Vin.+Pred. and Vin+Dox. doublets, are shown on the Right. (E) Dose level–specific plot (dose #4) of normalized apoptosis induction over time in response to all possible EPOCH/CHOP drug-treatment sets in HBL1 cells. Discussion Multicomponent cancer therapies first emerged over 50 y ago utilizing those drugs that were available and reasonably effective as single agents. These included DNA-damaging agents, cell cycle inhibitors, and steroids. Through iterative and pain-staking adjustments of dose and schedule, effective multicomponent therapies emerged that demonstrated durable responses in non-Hodgkin’s lymphoma including the CHOP and EPOCH regimens. Modern versions of these regimens include the B cell–depleting, CD20 targeting antibody Rituximab (R-CHOP and EPOCH-R) (Fig. 5A) (29, 30). These “chemoimmunotherapy” regimens remain the standard first-line treatment for DLBCL. They can be curative in selected subtypes of DLBCL, such as GCB DLBCL, but they are less effective in achieving durable responses in certain aggressive subtypes, such as ABC DLBCL. As an alternative to chemotherapy, many current treatment regimens involve targeted agents, whereby a drug’s activity is dependent on aligning its mechanism (pharmacology) with disease-associated, genomically manufactured vulnerabilities. Multicomponent therapies involving targeted agents can simultaneously attack co-occurring vulnerabilities. In lymphoma, significant effort has delivered a cadre of targeted therapies including Ibrutinib and Venetoclax that promise to replace chemotherapy regimens. The need is especially high in the relapsed/refractory setting where chemoimmunotherapy yields few durable responses. In B cell malignancies, monotherapy and doublet combination approaches involving targeted agents (e.g., Ibrutinib and Venetoclax) have been reported with promising activity, but few durable responses or cures (23–25, 31–33). The VIPOR clinical trial was designed to expand the promise of multicomponent targeted therapy in lymphoma. Each of the VIPOR drugs possesses a strong mechanistic basis for activity in selected lymphoma subtypes (Fig. 1A). To shed light on the higher-order drug interaction profile of the VIPOR agents we developed a multilayered, high-throughput drug screening pipeline that allowed us to evaluate in vitro drug response through the lens of key PK parameters that govern systemic drug exposure (concentration and time) and strongly influence clinical success. Based on these assessments, the single-agent activity of individual VIPOR drugs was surprisingly limited across cell models. Conversely, pairwise synergy among VIPOR drugs was prevalent in ABC DLBCL (Fig. 1B and SI Appendix, Fig. S1C), particularly in the three cell lines corresponding to the “MCD” subtype (i.e., HBL1, OCI-Ly10, and TMD8). Synergistic cytotoxicity in these models was largely driven by synergistic apoptosis induction. We observed several instances of exceptional synergy (akin to genetic synthetic lethality), where strong combinatorial activity was achieved at multiple concentrations where there was no detectable single-agent activity (Fig. 1C). These potentiation phenotypes are highly relevant from a translational perspective, as synergy-driven cytotoxicity and apoptosis were observed at concentrations and time points that are systemically achievable in vivo (Figs. 2 A–F and 3 A–C and SI Appendix, Figs. S2 A–D and S3 A and B). Using a unique experimental format called MCTC screening we investigated the proapoptotic potential of all possible VIPOR exposure scenarios (single agent → two-drug combinations → three-drug combinations → four-drug combination) examined every 2 h for 24 h to further elaborate on these data (Fig. 4 A–E). Collectively, these data highlight important lessons for associating in vitro screening outcomes with real-world clinical experience. Consistent with our knowledge of the oncogenic drivers of the ABC subtype of DLBCL, drug+drug synergism among VIPOR agents was enriched in HBL1, OCI-Ly10, and TMD8 cell lines. It was also noteworthy that synergistic induction of apoptosis was the major driver of strongly cytotoxic drug+drug interactions, identifying active killing of DLBCL cells as a likely contributor to VIPOR clinical success. MCTC screening revealed that multidrug exposure promoted rapid apoptosis onset at concentrations and time points that are safely achievable in humans. The kinetic benefit of multidrug synergy in ABC DLBCL cell models was particularly evident in Venetoclax-containing combinations, identifying the presence of a direct apoptotic sensitizer as a likely determinant of responses to VIPOR. Overall, these data align with the VIPOR clinical outcomes, in which non-GCB DLBCL had the highest complete response rate (62%) (26). These data also revealed a surprising level of heterogeneity for VIPOR drug+drug interactions, even within well-defined genetic subtypes (Figs. 1 B–D and 6 A–D and SI Appendix, Fig. S1 A–C). It is plausible that similarly heterogenous drug responses are occurring in the clinical setting. Individual patients, like the cell models used herein, may be responding to a subset of the VIPOR agents rather than receiving a collectively additive benefit from each agent (Fig. 6E). In this scenario, the multitarget nature of the VIPOR regimen is likely required to overcome heterogenous responses to specific VIPOR components which, at present, cannot be predicted a priori. The heterogeneity of VIPOR drug+drug synergy and the benefit of multidrug contributions to clinical responses (A–C) View of the ΔBliss heatmaps for selected Ibrutinib+Venetoclax (A), Ibrutinib+Lenalidomide (B), and Venetoclax+Lenalidomide (C) combinations in OCI-Ly10, TMD8, HBL1, and Riva cell models. Data for all Prednisone-based combinations are shown in figure S15 of ref. 26. (D) Synergy heatmap representation of all the possible pairwise combination of Venetoclax, Ibrutinib, Prednisolone, and Lenalidomide in the four ABC DLBCL cell lines. For each pair, viability was assessed 48 h posttreatment in 10 × 10 matrix blocks, and the corresponding ΔBliss values were then used to generate this synergy heatmap. (E) A conceptual schematic illustrating the contribution of multidrug synergy to therapy response in hypothetical DLBCL patients modeled by the ABC DLBCL cell lines utilized in this study. Parallel evaluation of the small-molecule components of R-CHOP and EPOCH-R confirmed prior findings demonstrating that the clinical efficacy of these regimens is likely driven by the actions of individual drugs with little or no contribution from drug+drug synergy (Fig. 5 A–E) (34, 35). These outcomes stimulate questions concerning best practices for crafting effective multicomponent therapies for the treatment of DLBCL and other cancers. Decades of clinical success reinforce the concept that chemotherapy-based multicomponent therapies composed of independently effective drugs represent a sound strategy. Multicomponent combinations of targeted agents represent a relatively unique concept. Head-to-head comparison of these strategies will require thoughtful preclinical and clinical consideration. There are conceptual advantages associated with multicomponent therapies that primarily combine targeted agents. Combinations of targeted therapies can benefit from improvements in tolerability by primarily acting at the site of disease through simultaneous attack on nonredundant cancer cell dependencies. Any resulting synergy-derived improvements in efficacy can contribute to a wider therapeutic index. Conversely, toxicity issues constrain the R-CHOP and EPOCH-R dose levels, and the schedules are limited to a single administration every 3 wk over the course of an average of six cycles (18 wk in total). Recovery of immune system function can take over 2 y following R-CHOP and EPOCH-R (36). By comparison, the tolerability of the VIPOR multicomponent therapy allows for standard/high dose ranges for each agent and is given daily over 2 wk with a third-week holiday. This schedule is maintained for six cycles (18 wk in total) with an acceptable adverse event profile. By tolerability metrics, therefore, the VIPOR multicomponent therapy offers a considerable advantage. Recently, the promotion of therapy resistance by pairwise synergistic drug+drug combinations has been explored in the cancer setting (37). Whereas the breadth of this phenomenon will require careful study, the high response rate and durability of response for VIPOR trial participants suggest that higher-order, multicomponent regimens like VIPOR may overcome, rather than promote, therapy resistance. Importantly, no late relapses were observed during the VIPOR clinical trial, highlighting that widespread therapy resistance following continuous exposure to synergistic multidrug combinations is not a universal phenomenon. Collectively, the clinical outcomes and the data reported herein suggest that the VIPOR therapy represents a unique and promising strategy for treating DLBCL and that multicomponent regimens composed of targeted drugs (multitargeted regimens) have curative potential in molecularly defined cancer subtypes. Nonetheless, room for improvement exists. Even in the era of genomic-based medicine and targeted therapies, multiple unknowns obstruct therapy design and success. The continued expansion of well-characterized cell models alongside comparative analysis of in situ disease is needed. Deeper mechanistic exploration of these cell models in conjunction with HTS-enabled drug response screens may reveal additional VIPOR-like multicomponent therapies which can be precisely directed toward patients likely to benefit. The data described herein offer insight into how such precision medicine-based approaches can be structured and the data formats needed to predict a clinical response. These studies make it clear, however, that the level of granularity must be expanded. While it is not possible to fully rationalize the VIPOR clinical outcomes based on the results presented in this study, there is compelling evidence in the data trends to suggest that in vitro screening methods like MCTC can play a role in the design and clinical implementation of multicomponent therapies. Materials and Methods Cell Lines and Reagents. All the DLBCL cell lines used for this study were cultured in advanced RPMI 1640 medium (Thermo Fisher) supplemented with penicillin/streptomycin and 5% fetal bovine serum (Tet tested, Atlanta Biologicals). All the cell lines were maintained/expanded in a standard tissue culture incubator at 37 °C in a 5% CO2 environment. qHTS (Single Agent). For single-agent qHTS, cells were seeded in 5 μL of media using a Multidrop Combi dispenser (ThermoFisher) into 1,536-well white bottom, polystyrene tissue culture–treated, Corning plates at a density of 500 cells/well. After cell addition, 23 nL of MIPE 5.0 compounds were added to individual wells via a 1,536 pin-tool. A fixed, 11-point concentration range, with constant 1:3 dilution was used for all the MIPE 5.0 drugs assessed as single agents. Plates were incubated for 48 h at standard incubator conditions, covered by a stainless steel gasketed lid to prevent evaporation. 48 h post compound addition, 3 μL of Cell Titer Glo (Promega) were added to each well, and plates were incubated at room temperature for 15 min with the stainless-steel lid in place. Luminescence readings were taken using a ViewLux imager (PerkinElmer) with a 2-s exposure time per plate. Dose–response curves were automatically fitted and classified into four distinct classes based on the quality of curve fit to the data (r2), the magnitude of the response (efficacy), and the number of asymptotes to the calculated curve. Inhibitory concentration 50 (IC50) values were calculated as previously described (14) and used to compare/contrast single-agent drug activities across the different cell lines tested (SI Appendix, Fig. S1A). Quantitative High-Throughput Combination Screening (Pairwise Matrix Format). For pairwise drug-combination assessment in matrix format, compounds (10 nL) were acoustically dispensed into 1,536-well white tissue culture-treated plates with an Echo 550 acoustic liquid handler (Labcyte). A nine-point custom concentration range with 1:2 dilution between points was used for pairwise drug-combination assessments in 10 × 10 matrix format. Dry-spotted plates were sealed and stored at −80 °C until needed. After thawing the plates, cells were added to compound-containing plates at a density of 500 cells/well in a final volume of 5 μL media, using a Multidrop Combi dispenser (ThermoFisher). Plates were incubated for the desired time at standard incubator conditions, covered by a stainless steel gasketed lid to prevent evaporation. For cell viability measurements, 3 μL of Cell Titer Glo (Promega) were added to each well and plates were incubated at room temperature for 15 min with the stainless-steel lid in place. Luminescence readings were taken using a ViewLux imager (PerkinElmer) with a 2” exposure time per plate. For apoptosis activation measurements, 3 μL of Casp3/7 Glo (Promega) were added to each well and plates were incubated at room temperature for 15 min with the stainless-steel lid in place. Luminescence readings were taken using a ViewLux imager (PerkinElmer) with a 10” exposure time per plate. Because both Cell Titer Glo and Casp3/7 Glo are lysis-based reagents, multiple sets of identical plates were used to collect each individual cell line/readout/time point. Measurement of drug-to-drug interactions was performed as previously described (10) and the Delta Bliss metric was used to quantitatively compare/contrast synergy/antagonism/additivity across cell lines. MCTC Screening Assessments. For multidrug combination assessments, we kinetically measured apoptosis induction following exposure to multiple agents as a more straightforward way to systematically measure multidrug interactions. Briefly, we established eight unique, fixed dose levels for each drug of interest and used an Echo 550 acoustic liquid handler (Labcyte) to dispense all the possible drug-treatment sets (singlets, doublets, triplets, four-drugs, and five-drugs combination) into individual wells of a single 1,536-well white bottom, polystyrene tissue culture-treated, Corning plate. Each individual drug-treatment set was plated four times (N = 4) across the eight-preestablished dose levels. A full column (32 wells) contains 30 nL of dimethyl sulfoxide (DMSO) was used for negative control and background subtraction. A full column (32 wells) of 20 μM Bortezomib was used as a positive control for apoptosis induction and for signal scaling. Dry-spotted plates were sealed and stored at −80 °C until needed. After thawing the plates, cells were added to compound-containing plates at a density of 500 cells/well in a final volume of 5 μL media, using a Multidrop Combi dispenser (ThermoFisher). Plates were incubated for the desired time at standard incubator conditions, covered by a stainless steel gasketed lid to prevent evaporation. For apoptosis activation measurements, 3 μL of Casp3/7 Glo (Promega) were added to each well and plates were incubated at room temperature for 15 min with the stainless-steel lid in place. Luminescence readings were taken using a ViewLux imager (PerkinElmer) with a 10” exposure time per plate. Because Casp3/7 Glo is a lysis-based reagent, multiple identical plates were used to collect each individual cell line/time point. For data analysis and heatmap generation (Figs. 4 and 5), the background Casp3/7 luciferase signal of 30 nL DMSO treated cells was subtracted from the luciferase signal of individual drug-treatment sets for each time point. For each cell line, the highest 20 μM Bortezomib luciferase signal measured across all time points was used for “signal-scaling” purposes and set to 100%. All the background-subtracted values were scaled and normalized accordingly. Supplementary Material Appendix 01 (PDF) Dataset S01 (XLSX) Dataset S02 (XLSX) Dataset S03 (XLSX) Dataset S04 (XLSX) Dataset S05 (XLSX) Dataset S06 (XLSX) Dataset S07 (XLSX)