Development of Constitutively Synergistic Nanoformulations to Enhance Chemosensitivity in T-Cell Leukemia

Abstract

Advances in multiagent chemotherapy have led to recent improvements in survival for patients with acute lymphoblastic leukemia (ALL); however, a significant fraction do not respond to frontline chemotherapy or later relapse with recurrent disease, after which long-term survival rates remain low. To develop new, effective treatment options for these patients, we conducted a series of high-throughput combination drug screens to identify chemotherapies that synergize in a lineage-specific manner with MRX-2843, a small molecule dual MERTK and FLT3 kinase inhibitor currently in clinical testing for treatment of relapsed/refractory leukemias and solid tumors. Using experimental and computational approaches, we found that MRX-2843 synergized strongly—and in a ratio-dependent manner—with vincristine to inhibit both B-ALL and T-ALL cell line expansion. Based on these findings, we developed multiagent lipid nanoparticle formulations of these drugs that not only delivered defined drug ratios intracellularly in T-ALL, but also improved anti-leukemia activity following drug encapsulation. Synergistic and additive interactions were recapitulated in primary T-ALL patient samples treated with MRX-2843 and vincristine nanoparticle formulations, suggesting their clinical relevance. Moreover, the nanoparticle formulations reduced disease burden and prolonged survival in an orthotopic murine xenograft model of early thymic precursor T-ALL (ETP-ALL), with both agents contributing to therapeutic activity in a dose-dependent manner. In contrast, nanoparticles containing MRX-2843 alone were ineffective in this model. Thus, MRX-2843 increased the sensitivity of ETP-ALL cells to vincristine in vivo. In this context, the additive particles, containing a higher dose of MRX-2843, provided more effective disease control than the synergistic particles. In contrast, particles containing an even higher, antagonistic ratio of MRX-2843 and vincristine were less effective. Thus, both the drug dose and the ratio-dependent interaction between MRX-2843 and vincristine significantly impacted therapeutic activity in vivo. Together, these findings present a systematic approach to high-throughput combination drug screening and multiagent drug delivery that maximizes the therapeutic potential of combined MRX-2843 and vincristine in T-ALL and describe a novel translational agent that could be used to enhance therapeutic responses to vincristine in patients with T-ALL. This broadly generalizable approach could also be applied to develop other constitutively synergistic combination products for the treatment of cancer and other diseases.

INTRODUCTION

Combination approaches to cancer therapy have greatly improved treatment outcomes since their inception in the mid-1950s [1–3] and today, a majority of cancer patients receive some form of multiagent therapy [4] that can encompass a wide range of drug types including small molecules, proteins, nucleic acids, viral vectors, or engineered cells. In addition to diversity of size and structure, these agents also vary widely in their route or frequency of administration and, as a result, large time-dependent fluctuations in both tissue and plasma drug concentration—and therefore drug ratio—are commonly observed following the administration of combination therapies [5–7].

Drug synergy, which is the observation of combined drug effect that exceeds the expected sum of component drug effects, is a common basis for the use of multiagent therapies in cancer, infectious diseases, and neurological disorders. Yet, despite the strong therapeutic potential of synergistic multiagent approaches, prior large-scale screening studies suggest that the identification of synergistic drug pairs is in fact rare (4–10%) [8], while others indicate that synergistic effects, when they do occur, are highly ratio-dependent, whereby some ratios of particular drugs may be supra-additive (i.e. synergistic) while others are additive or sub-additive (i.e. antagonistic) [9–11].

Recent approaches to maximize the therapeutic potential of multiagent cocktails have aimed to co-deliver therapeutics in a manner that constitutively maintains drug synergy [12]. Such approaches include nanometer-scale drug carriers that deliver agents in a time-staggered [13, 14] or ratiometric [15–19] fashion, as well as those that harness synthetic lethal gene interactions [20] or target adaptive drug resistance mechanisms a priori [21, 22]. Liposomal cytarabine and daunorubicin (CPX-351, Vyxeos^®) is one such example and was originally discovered through in vitro screens that identified synergistic cell growth inhibition at a 5:1 mole ratio of cytarabine:daunorubicin [23, 24]. By constitutively maintaining ratiometric drug synergy between these two agents via lipid nanoparticle formulation, CPX-351 improved treatment outcomes relative to free combination chemotherapy in both mouse models and clinical trials, and is now FDA-approved for the treatment of both adult and pediatric patients with leukemia [25–27].

The receptor tyrosine kinase, MERTK, is ectopically expressed in a majority of acute myeloid leukemias (AML) and approximately 50% of T-cell acute lymphoblastic leukemias (T-ALL) and we have described key roles for this protein in leukemia cell survival and leukemogenesis [28–31]. More recently, we developed small molecule inhibitors that target MERTK and FLT3, a clinically-validated therapeutic target in AML [32]. These agents improve survival in murine ALL and AML models [33–5] and the lead compound, MRX-2843, is currently being tested in Phase 1b clinical trials (NCT03510104, NCT04762199, NCT04872478, NCT04946890). We hypothesized that the therapeutic potential of combined MERTK/FLT3 inhibition and chemotherapy for leukemia could be maximized via ratiometric drug screening and formulation. To this end, here we describe a novel and systematic approach to high-throughput combination drug screening and nanoscale drug delivery that maximizes drug synergy between MRX-2843 and cytotoxic chemotherapy. This novel, broadly generalizable approach to formulation discovery could lead to the development of constitutively synergistic combination products for the treatment of cancer and other diseases.

RESULTS

Guided by prior reports indicating the potential for synergy between MERTK inhibition and methotrexate or vincristine [28, 31, 33], we developed a high-throughput drug combination screen (HTS) in which we sought to identify ratiometric synergy between these compounds following cell exposure to drug interaction matrices comprising concentration gradients of MRX-2843, vincristine and methotrexate (Fig 1a). As an in vitro model, we selected a genotypically and phenotypically diverse set of leukemia cell lines spanning B- and pro-B ALL, as well as T- and early T-cell precursor (ETP-) ALL, exposing each of these cell lines to drug interaction matrices comprising >530 discrete ratiometric drug combinations, and their equivalent single-agents, in 384-well plates using high-throughput liquid handling robotics (Fig 1b). Cell density was measured in quadruplicate via luminescent cell viability assay using Z’≥0.5 as a cutoff for data quality.

Figure 1. Combined MRX-2843 and vincristine synergize to inhibit T-ALL cell growth in high-throughput combination drug screens.

a) Illustration of a ratiometric drug screen combining the dual MERTK/FLT3 inhibitor, MRX-2843, with methotrexate and/or vincristine in a panel of 13 B- and T-ALL cell lines as measured by luminescent viability assay (72 h, Z’≥0.5). b) Single-agent dose response curves and (c) lineage-specific frequencies of combined drug synergy, additivity, or antagonism for 537 unique pairwise or triplet drug combinations as measured in parallel and assessed via the Bliss Independence model. d) Scatter plots of mean drug synergy as a function of mean growth inhibition grouped by cell lineage. Ratiometric drug synergy is conserved across T- and ETP-ALL cell lines with synergy among distinct molar ratios of MRX-2843 and vincristine. e) Drug synergy and antagonism appear at distinct doses and molar ratios in cells from the T- and ETP-ALL lineage. (c-e) Synergy represents percent reduction in cell density greater or lesser than that predicted by the Bliss Independence model with synergy (>1%) and antagonism (<−1%). (e) Sphere size and saturation correspond to the mean magnitude of synergy observed across T- and ETP-ALL cell lines (n=6) for each pairwise and triplet drug combination among 537 tested.

Consistent with other large scale screening studies [9, 36, 37], we observed that synergy among these ratiometric combinations was rare (<6.7% of combinations overall) as assessed using the Bliss Independence model of drug synergy (Fig 1c,d) [38]. Interestingly, drug synergy was largely absent when averaged over B lineage ALL cells (Fig S1); however, we observed strong synergy and antagonism that was ratio-dependent across T and ETP lineage ALL cell lines (Fig 1e). The consistency of these trends in synergy was confirmed using the Response Additivity and Zero Interaction Potency synergy models [38] (Fig S2a). HTS results were reproduced in a validation study using an alternative 96-well assay format and new drug and cell stocks (Fig S2b,c).

To guide the selection of a drug combination with maximum potency in T-ALL, we next determined whether synergy observed in our prior screen was principally attributable to either higher order (3-drug) or lower order (2-drug) effects. By comparing concordance between measured cell viabilities and those predicted from synergy models [39], we found that synergy observed in 3-drug combinations via HTS was predominantly attributable to lower order (2-drug), rather than higher order (3-drug), interactions (Fig 2a, S3a-c). To confirm which among the three pairwise drug combinations was most synergistic, we conducted a microwell assay in which hydrogel-embedded Jurkat (T-ALL) cells were exposed to continuous gradients of each drug pair and assessed for cell viability. Comparing viability data to corresponding additivity isoboles, we found that the combination of MRX-2843 and vincristine was consistently synergistic and most effective among these pairwise combinations (Fig 2b, S2d,e).

Figure 2. In silico and in vitromodels prioritize pairwise drug synergy between MRX-2843 and vincristine among screened combinations.

a) Comparison of higher order (3-drug) and lower order (2-drug) synergy models to high-throughput combination drug screening data indicates that experimentally observed drug synergy is predominantly attributable to pairwise drug interactions. b) Comparison of expected (dashed) and observed (solid) Jurkat (T-ALL) cell viability following exposure to continuous pairwise drug gradients demonstrates that MRX-2843 and vincristine are most consistently synergistic. (a) Data points show 378 triplet drug combination responses measured in T- and ETP-ALL cell lines (n=6) via HTS. Data in (b) represent mean cell viability (72 h) as measured by nuclear dye exclusion.

To better define the biological underpinnings of synergy between MRX-2843 and vincristine in T-cell ALL, we performed transcriptomic profiling on Jurkat (T-ALL) cells treated with MRX-2843:vincristine or their respective single-agent and vehicle controls (Fig S4a). Although the set of differentially expressed genes (DEGs) associated with a synergistic combination was unique, it was also limited in mapping to enriched ontologies after adjusting for multiple comparisons (Fig S4b,c). However, the union of DEGs in cells treated with MRX-2843 and vincristine monotherapies mapped with high significance to a network of similar and interconnected shared biological processes (GO, Reactome, KEGG and WikiPathways) (Fig S4d-g) associated with respiratory electron transport, RNA metabolism, ribosomal complex biogenesis, and anaphase initiation via the degradation of securin (Fig S4d,e). Insight into these ontological relationships may lead to development of novel drug combinations targeting shared mechanisms of action.

Having determined that MRX-2843 and vincristine act in a ratio-dependent manner to decrease T-ALL cell density, we next sought to develop drug formulations that both deliver and constitutively maintain ratiometric synergy between these compounds. Using a cheminformatics-guided approach, we determined pH-dependent physicochemical properties associated with MRX-2843 and vincristine via ACD/Labs software (see Cheminformatics Computation, Methods), and then devised a pH-gradient based method of drug loading (Fig 3a,b) in which the compounds were simultaneously co-encapsulated within lipid nanoparticles composed of DSPC (1,2-distearoyl-sn-glycero-3-phosphocoline), DSPG (1,2-distearoyl-sn-glycero-3-phosphoglycerol, sodium salt) and cholesterol lipids. Using this approach, we synthesized drug formulations of MRX-2843 and vincristine that recapitulated drug ratios which we had identified via HTS as synergistic (8.9 – 90 mol:mol and 4.7 – 48 mg:mg MRX-2843:vincristine), additive (90 – 160 mol:mol; 48 – 85 mg:mg), or antagonistic (≥160 mol:mol; ≥85 mg:mg) in their inhibition of T-ALL and ETP-ALL cell expansion (Fig 3c), referred to hereafter as Syn, Add and Antag nanoparticles, respectively. Characterization of dual-agent lipid nanoparticle formulations by dynamic light scattering, transmission electron microscopy, and liquid chromatography-mass spectrometry (LC-MS) indicated that pH-gradient co-loading achieved size uniformity with low polydispersity indices (PDI 0.11 – 0.14), efficient drug loading (6 wt% for dual-agent nanoparticles) and precise ratiometric encapsulation using these methods (Fig 3d,e, S5a,b, see Methods for details). In addition, we evaluated nanoparticle stability following lyophilization. Dual- and single-agent nanoformulations demonstrated zeta-potentials <−30 mV (Fig S5c) and lost <3% of total drug following storage (Fig S5d). A dual-agent nanoformulation released <7.5% of MRX-2843 over 72 hours of dialysis in PBS at 37°C without any apparent loss in vincristine, demonstrating stable ratiometric maintenance, whereas both drugs were rapidly released and at varying rates in acidic dialysis buffer (pH=5.5) (Fig S5e). Dual-agent nanoformulations maintained potent activity in cell-based assays for up to 1 year (Syn range: 1.36 – 5.30 nM (vincristine); Add range: 0.96 – 1.70 nM (vincristine)) (Fig S5f). Suggesting further formulation stability, nanoparticles showed minimal changes in size and PDI following rehydration or in solution with serum (Fig S6a,b).

Figure 3. Lipid nanoparticles co-encapsulating MRX-2843 and vincristine constitutively maintain defined drug ratios following co-formulation and intracellular delivery.

a) Illustration of a pH gradient-based method of drug co-loading that relies upon (b) pH-dependent MRX-2843 and vincristine lipophilicity and charge. c) A plot of mean drug synergy versus MRX-2843:vincristine mole ratio defines synergistic (8.9 – 90), additive (90 – 160), and antagonistic (≥160) drug ratios. d) T-ALL tailored combination drug formulations encapsulating synergistic (Syn), additive (Add), and antagonistic (Antag) drug ratios shown at nanometer-scale size and uniformity as measured by dynamic light scattering and (inset) transmission electron microscopy e) Drug co-loading as measured by LC-MS. f) Intracellular drug delivery kinetics following treatment of LOUCY (ETP-ALL) cells with Antag nanoparticles over 24 h as measured by LC-MS. g) Comparison of growth inhibition following treatment with Syn nanoparticles, as well as combined and free MRX-2843/vincristine demonstrates that nanoparticle encapsulation further enhances in vitro drug synergy as measured by luminescent cell viability assay in Jurkat (T-ALL) cells (72 h) and (h) LOUCY (ETP-ALL) (72 h). (c) The line traces the trend in mean response between tested ratios, and hashes delineate the range of +/− 1% synergy, which we define as additivity. (d) Scale bars measure 80 nm, and PDI stands for polydispersity index. (e) Data points represent drug ratios from distinct nanoparticles batches as measured by LC-MS. Best fit by linear regression is shown with coefficient of linearity. (f) Vertices are mean values ± SD of intracellular drug concentrations of 3 replicates per time point as measured by LC-MS. (g,h) Individual values (dots) and mean values ± SD of 3 replicates from a single experiment are shown, where the dashed line represents the expected effect for an additive interaction (GI_exp). Mean differences were evaluated by one-way ANOVA with Tukey’s correction for multiple comparisons, where * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001.

To measure the ratiometric delivery of MRX-2843 and vincristine, we exposed LOUCY (ETP-ALL) cells to drug-loaded lipid nanoparticles for 24 hours and measured intracellular drug delivery via LC-MS (Fig 3f). We found that drug-loaded nanoparticles not only achieved highly efficient delivery of MRX-2843 to T-lineage ALL cells, but that they also co-delivered MRX-2843 and vincristine at concentrations that monotonically increased in parallel over time. To further characterize the therapeutic performance of dual-agent ratiometric nanoparticles, we examined their impact on LOUCY and Jurkat cell density via luminescent viability assay. Strikingly, drug synergy was even further enhanced in both cell lines following nanoparticle encapsulation, whereby growth inhibition far exceeded that of the already synergistic combination of free MRX-2843 and vincristine (Jurkat, P=0.0090; LOUCY, P<0.0001) (Fig 3g,h, S5g). Syn, Add and Antag dual-agent nanoformulations demonstrated higher potency than both the expected effect for an additive interaction and the observed effect in cultures treated with the free drug combination (Fig S5g).

We next investigated the clinical relevance of these combination drug formulations using primary pediatric T-ALL patient samples collected at time of diagnosis. Mononuclear cells from bone marrow or blood were treated with liposomal MRX-2843, free vincristine, vehicle lipid nanoparticles, or Syn or Add nanoparticles that reflect optimal synergy or additivity in LOUCY (ETP-ALL) cells (Fig 4a). GI₅₀ values of cell density, measured via luminescent viability assay, demonstrated growth inhibition mediated by both Syn and Add nanoparticles (Fig 4b), and both formulations recapitulated their expected synergism or additivity in patient samples compared to expectation based on the effects of liposomal MRX-2843 and free drug vincristine measured in parallel (Fig 4c,d).

Figure 4. Comparative potency of ratiometric drug formulations in primary T-ALL samples ex vivo.

a) Approach to testing ratiometric nanoformulations in T-ALL or near ETP-ALL samples obtained from patient blood or bone marrow (BM) at initial diagnosis. b) Expected (Exp.; dashed) and observed (Obs.; solid) dose-dependent growth inhibition from synergistic and additive ratiometric drug formulations as measured by luminescent viability assay (96 h). c) Heatmap of GI₅₀s and P values from (c) Syn and (d) Add drug formulations shown in (b). (b) Mean values ± SD of growth inhibition scaled against the concomitant MRX-2843 doses in the nanoparticles. Exp. dose response curves in (b) were calculated with the Bliss Independence model of drug synergy between single-agent liposomal MRX and vincristine free drug dose responses. (c,d) Heatmaps show GI50 ± standard error (nM) of Syn and Add potencies scaled by MRX-2843 dose. Bar charts show associated log transformed P values of unpaired parametric t tests. Asterisks (*) denote extrapolated GI50 values. All Obs. dose responses were run in triplicate, while Exp. dose responses were run either in duplicate (1172a) or triplicate (1147a, 1369b) in a single experiment.

Next ratiometric lipid nanoparticles were compared in a murine ETP-ALL xenograft model. For these studies, a luciferase-expressing derivative of the LOUCY cell line (LOUCY-luc) [40] was injected intravenously into NRG mice to establish orthotopic xenografts, and mice were treated with nanoparticles containing Syn, Add and Antag formulations of MRX-2843 and vincristine, vincristine (Vinc) or MRX-2843 (MRX) monotherapies, or empty liposomes (vehicle, Veh) once weekly for four weeks to mimic the schedule of administration for vincristine in patients with leukemia (Fig S7a,b). In an initial study various formulations were administered to span a range of MRX-2843 and vincristine doses and Vinc only particles contained the same dose as the Syn formulation. Mice treated with Syn and Vinc only nanoparticles delivering a dose of 1.152 mg/kg vincristine were removed from the study by 80 days after initiation of treatment due to poor health, despite a lack of significant disease burden (Fig S7c-e). Add particles delivering a dose of 0.57 mg/kg vincristine were well-tolerated and bioluminescence intensity increased coincident with removal of mice from study, suggesting leukemia as a cause of death. Thus, the maximum tolerated dose of vincristine nanoparticles in this model is between 0.57–1.152 mg/kg. When treatment was initiated with minimal disease burden, bioluminescence intensity was significantly decreased in mice treated with the Add formulation compared to mice treated with Antag nanoparticles (p=0.0159) (Fig S7c) and median survival was significantly prolonged (74 versus 121 days after tumor inoculation, p=0.0059) (Fig S7d). When initiation of treatment was delayed until the mice had developed higher disease burden, the Add nanoparticles were no longer sufficient to provide therapeutic activity, while the Syn particles still prevented disease progression and significantly reduced disease burden compared to treatment with Add particles (p=0.0317) (Fig S7f). Thus, both toxicity and therapeutic efficacy varied with vincristine dose. In contrast, MRX nanoparticles were well tolerated at doses of up to 60 mg/kg twice weekly, but did not have therapeutic activity in mice with LOUCY-luc xenografts (Fig S8a-g).

To determine whether MRX-2843 contributes to the therapeutic activity mediated by the combination nanoparticles in this model, Syn and Add nanoparticle formulations containing vincristine at a fixed, tolerated and effective dose (0.713 mg/kg) were directly compared in mice with LOUCY-luc xenografts (Fig 5a,b). Both formulations significantly decreased disease burden compared to treatment with empty liposomes (p=0.0106 and p=0.0042, respectively) (Fig 5c). Leukemia was also significantly decreased in mice treated with Add compared to Syn particles (p=0.0204). Similarly, median survival was significantly prolonged in mice treated with Add nanoparticles (85 days post-treatment) compared to Veh-treated mice (41 days; p = 0.0287) and survival was prolonged but the difference was not significant in mice treated with Syn particles (71 days) compared to Veh (p=0.1269) (Fig 5d). Thus, both Add and Syn nanoparticle formulations of MRX-2843 and vincristine were therapeutically effective in mice with LOUCY-luc xenografts, but the Add particles, which deliver a higher dose of MRX-2843, provided greater therapeutic benefit.

Figure 5. Nanoparticles containing MRX-2843 and vincristine at an additive ratio mediate superior antileukemia activity in a murine ETP-ALL xenograft model.

a,b) NRG mice were inoculated with a luciferase-expressing ETP-ALL cell line (LOUCY-luc; 2 million cells). Mice were randomized to groups with equal starting disease burden (n=7–8) and treatment with synergistic (Syn) or additive (Add) liposomal formulations of MRX-2843 and vincristine or vehicle nanoparticles without drug (Veh) was initiated 20 days later. Arrows indicate treatment administration. c) Disease burden was determined at intervals by bioluminescence imaging. Data are reported as mean ± SEM. Differences were determined by two-way ANOVA. d) Differences in median survival (MS) were calculated by the log-rank method. e,f) NRG mice were inoculated intravenously with 3 million LOUCY-luc cells, then randomized to groups (n=9–10) and treatment with the indicated liposomal formulations of MRX-2843 and vincristine or Veh was initiated 20 days later. f) Differences in median survival were calculated by the log-rank method.

In contrast, nanoparticles containing an even higher dose of MRX-2843 (Max, MRX-2843:vincrstine = 159 mol:mol) provided decreased therapeutic efficacy compared to the Add nanoparticles in this same model (Fig 5e,f). Both Add and Max nanoformulations significantly prolonged survival compared to vehicle (p=0.0027 and p=0.0129, respectively) and median survival was also significantly decreased from 92 days in mice treated with the Add nanoformulation to 74 days in mice treated with Max nanoparticles (p=0.0384). Of note, any changes in body weight in mice treated with nanoparticles containing 0.713 mg/kg vincirstine in combination with MRX-2843 were coincident with advanced disease (Fig S9). Thus, among dual-agent nanoformulations of MRX-2843 and vincristine, the Add nanoparticles provided optimal therapeutic activity in vivo.

DISCUSSION

Here we address the urgent and unmet need to develop new, more potent therapies for patients with relapsed/refractory T-ALL. While advancements in chemotherapy dose-intensification and patient risk-stratification have led to gradual improvements in patient survival for T-ALL in recent years, a substantial proportion of patients have treatment refractory disease or later relapse. As a result, mortality rates for pediatric and adult T-ALL patients are 20% and >50%, respectively [41]. Patients diagnosed with ETP-ALL experience particularly high rates of induction therapy failure, necessitating intensified treatments associated with increased toxicity and increasing the likelihood for bone marrow transplant [42–46]. To address the challenge of improving outcomes in these patients, we devised a systematic approach to maximize drug synergy between small molecule anticancer drug combinations and to tailor these effects to a specific disease indication, in this case TALL.

Using a novel, combinatorial high-throughput drug screen, we found that combined treatment with the tyrosine kinase inhibitor, MRX-2843, and the cytotoxic chemotherapy, vincristine, synergized in a ratio-dependent manner to inhibit T-ALL cell expansion, a finding that is significant because (i) vincristine is a mainstay of frontline induction and consolidation therapies for T-ALL [47] and (ii) both MRX-2843 targets (MERTK/FLT3) are frequently actionable in T-ALL. For example, in our previous studies, approximately 50% of pediatric T-ALL patient samples ectopically expressed MERTK [28] and 50% of both pediatric and adult samples were sensitive to treatment with a MERTK/FLT3 inhibitor [48]. Similarly, approximately 15% and 35% of pediatric and adult ETP-ALLs, respectively, carry activating FLT3 mutations [49, 50].

Given the dual targets of MRX-2843 and the multifarious effects of MERTK on oncogenic signaling pathways [51], the mechanism(s) by which MRX-2843 synergizes with vincristine in T-ALL are speculative; however, pharmacologic inhibitors of several pathways and proteins that are known to be downstream targets of MERTK signaling also synergize with vincristine, suggesting their potential as mediators of synergy. MERTK promotes expression of the anti-apoptotic protein BCL-XL in ALL and other cancer cells and can inhibit expression of the related protein, BCL-2, in some circumstances [31, 52]. Gossypol, a BCL-2 and BCL-XL inhibitor, synergized with vincristine in non-Hodgkin’s lymphoma cells [53]. Likewise, inhibition of BCR/ABL whose downstream signaling, like MERTK, converges on the MAPK pathway, synergized with vincristine against ALL cells [54]. Lastly, inhibitors of PI3K – positively regulated by MERTK signaling – synergized with vincristine in T-ALL cell culture and murine models [55].

Based on the findings from our screen and subsequent analyses indicating ratiometric drug synergy between MRX-2843 and vincristine, we developed multiagent lipid nanoparticles closely related in lipid composition to FDA-approved products such as AmBisome and Vyxeos, both of which are comprised of phosphatidylcholine, phosphatidylglycerol and cholesterol [56]. To further enhance the feasibility of clinical-scale manufacturing, we also devised a method to co-load defined ratios of MRX-2843 and vincristine using a gradient-based method which is notable in its typically high encapsulation efficiency, scalability and reproducibility, as well as its utilization in the manufacture of FDA-approved products, such as Onivyde and Doxil [57, 58]. In addition, we demonstrated a scalable method of freeze-dried storage that stabilizes drug activity for at least 1 year, retains encapsulated drug upon reconstitution, and maintains nanoparticle size and dispersion. Given the safe and effective clinical use of these lipid excipients, together with a commercially viable method of manufacturing and storage, these findings further support the translational potential of liposomal MRX-2843 and vincristine.

Having developed lipid nanoparticle formulations co-loading these compounds, we further examined the extent to which combination formulations deliver synergistic and antagonistic drug ratios intracellularly. Indeed, the Antag formulation delivered an antagonistic drug ratio of MRX-2843 and vincristine in vitro that varied by only 11.43-fold over 24 hours. These findings support the success and importance of the pH gradient-based loading method described here. Drugs contained within the lumen of lipid vesicles are typically released less rapidly than those that reside in the lipid bilayer [14], where lipophilic compounds often reside following passive drug loading or incomplete gradient-based loading [57, 58]. The comparable intracellular drug release profiles between MRX-2843 (clogD_7.4=3.07) and vincristine (clogD_7.4=1.14) observed here thus suggest that our loading method was not only successful in luminal drug entrapment, but that it was also critical to maintaining ratiometric drug synergy following cell delivery. Further, the intracellular drug concentrations achieved were well in-excess (3x – 100x) of their corresponding GI_{50, 72h} values in LOUCY (ETP-ALL) cells. Together, these findings support both the therapeutic utility and clinical relevance of our described approach to co-formulating MRX-2843 and vincristine.

In addition to ratiometric drug delivery, the synergistic anti-leukemia activity observed in cell cultures treated with the free drug combination was further enhanced following nanoparticle co-encapsulation, a phenomenon that we and others have previously observed with other drug pairs and carriers [15]. While studies to elucidate the mechanisms by which drug co-encapsulation further improves synergy between MRX-2843 and vincristine are ongoing, such effects may be attributable in part to nanoparticle-accelerated intracellular drug transport, augmented target engagement whereby the intracellular kinase domains of both MERTK and FLT3 (enzymatic IC₅₀~ 1.3 and 0.64 nM [34]) are exposed to drug concentrations in excess of those achieved via passive drug diffusion, or differential plasma/serum protein binding. Further, in addition to prominent cell-surface expression, MERTK has also been reported to co-localize with early endosomes [59] where lipid nanoparticles frequently accumulate following cellular uptake [60]. Thus, nanoparticle-mediated drug delivery may augment the magnitude, duration, and/or location of protein inhibition in order to further enhance drug synergy.

Nanoparticles containing MRX-2843 and vincristine were also therapeutically effective against T-ALL cells in blood and bone marrow cultures derived from diagnostic T-ALL patient samples. Both drug synergy and additivity were maintained by Syn and Add drug formulations, respectively, following ex vivo treatment. These findings are significant in that the delivery of drug carriers and transfection reagents to primary T lymphocytes is notoriously difficult [61].

Dual-agent MRX-2843 and vincristine nanoformulations were also therapeutically effective in a murine ETPALL model. While vincristine was the primary driver of efficacy in the studies described here, when the vincristine dose was fixed to allow a direct comparison of particles containing synergistic or additive ratios of MRX-2843 and vincristine, mice treated with Add particles had decreased disease burden and prolonged survival compared to mice treated with Syn particles. Thus, therapeutic efficacy increased with increasing MRX-2843. However, when MRX-2843 was increased even further, therapeutic efficacy decreased relative to the Add particles, consistent with the antagonistic drug interaction observed in cell-based assays at higher concentrations of MRX-2843. These observations have several implications.

First, both vincristine and MRX-2843 contribute to the therapeutic activity mediated by the nanoparticles and the activity of both drugs is dose dependent. In Syn and Add nanoparticles, ratiometric dosing limits exposure to MRX-2843. Thus, the Syn formulation, with its lower dose of MRX-2843, does not attain maximum efficacy, despite a more favorable interaction between drugs. These studies demonstrate that in at least some cases, increasing drug exposure should be prioritized over maintenance of ratiometric dosing to optimize therapeutic benefit for translational applications, even if the interaction between drugs may be less synergistic under these conditions. This is likely to be particularly relevant in cases where toxicity is driven primarily by one drug in a combination. In contrast, interactions between drugs may be more important when the drugs are synergistic at ratios where both can be administered at or near their maximum tolerated or maximally effective doses. In addition, because nanoparticles containing only MRX-2843 do not have therapeutic activity in vivo, but MRX-2843 still provided enhanced efficacy when combined with vincristine, the interaction between MRX-2843 and vincristine is synergistic in vivo, even when the ratiometric dose of drugs provides an additive response in cell-based assays.

In addition, while Add nanoparticles were more effective than Syn nanoparticles, increasing the concentration of MRX-2843 to a molar ratio that is predicted to be more antagonistic reduced therapeutic efficacy, consistent with previous reports comparing ratiometric nanoformulations of other drugs [6, 24, 62, 63]. Thus, the interaction between drugs can drive therapeutic efficacy in vivo in some circumstances and ratiometric dosing is critical to maintain optimal anti-leukemia activity. Ultimately, in vitro modeling alone is insufficient to predict the optimal dose and ratio for application in animal models a priori. Development of computer models that incorporate and interface data describing drug interactions and relative therapeutic activity in cell-based assays as well as toxicity and dosing information for the single agents in mice and humans will be necessary to accurately predict optimal therapeutic doses and ratios for combination therapies.

Taken together, these data demonstrate both the safety and efficacy of nanoparticle formulations of MRX-2843 and vincristine and demonstrate activity mediated by both drugs in this context, with higher doses of either MRX-2843 or vincristine providing greater therapeutic benefit in Syn or Add particles and Add particles providing optimal therapeutic efficacy. Moreover, co-administration of MRX-2843 enhanced sensitivity to vincristine without evidence of increased toxicity. These data suggest that MRX-2843 could be added to current therapeutic regimens utilizing vincristine to improve outcomes for patients with newly diagnosed or relapsed T-ALL. Given the potent anti-leukemia activity and synergistic interaction mediated by MRX-2843 and vincristine and the efficient formulation and delivery of the combination described here, this work may serve to guide future clinical studies of these or related drug formulations incorporating MRX-2843 and/or other anticancer agents for the treatment of ALL.

MATERIALS AND METHODS

Cell Lines and Culture

CCRF-CEM, Jurkat, MOLT-4 and LOUCY were obtained from the American Type Culture Collection (Manassa, VA). 697, PEER and DND-41 were obtained from the Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures GmbH (Braunschweig, Germany). Nalm-6, KOPN-8 and UoC-B1 were provided by Dr. Christopher Porter (Emory University). REH, RCH-ACV, and RS4;11 were from Dr. Lia Gore (University of Colorado Denver). Cells were cultured in RPMI 1640 with L-Glutamine supplemented with 10–20% heat-inactivated fetal bovine serum or as recommended by the original supplier (ATCC, DSMZ). All cells were cultured in a 5% CO2 humidified atmosphere and were tested for mycoplasma every four months.

High-Throughput Combination Drug Screening

MRX-2843 (Meryx, Inc.; Chapel Hill, NC) and methotrexate (Sigma-Aldrich; Burlington, MA) were reconstituted in dry DMSO and stored in checkerboard array within 384-well polypropylene plates (VWR; Radnor, PA) covered in sterile adhesive plate film prior to use. Vincristine sulfate (Cayman Chemical; Ann Arbor, MI) was solvated in molecular biology grade water (VWR) on the day of experimentation. 15,000 viable cells per well were dispensed into white polystyrene 384-well microplates (Greiner Bio-One) using a Multidrop Combi Reagent Dispenser (Thermo Fisher Scientific; Waltham, MA). Vincristine was subsequently dispensed onto these cell-containing microplates. MRX-2843 and methotrexate were then transferred to these cell + vincristine microplates using a Beckman NX Liquid Handler (Beckman Coulter; Brea, CA) at the Emory Chemical Biology Discovery Center. Cell- and drug-containing plates were then centrifuged for 5 minutes at 135 x g (Eppendorf 5810R; Eppendorf, Hamburg, Germany), then incubated for 72 hours at 37°C in 5% CO₂ humidified atmosphere with gas permeable plate sealing film (VWR). All wells had a DMSO concentration of 0.5% v/v with drugs MRX-2843 (25 – 800 nM), methotrexate (2.9375 – 752 nM), and vincristine (0.0875 – 5.6 nM). After 72 hours, viable cell numbers were assessed using CellTiter-Glo 2.0 Assay (Promega; Madison, WI) and the Integra Viaflo Assist Pipetting Platform (Integra Biosciences; Hudson, NH) according to manufacturer’s instructions. Luminescence was measured using a SpectraMax iD3 plate reader (Molecular Devices; San Jose, CA) and viability data were background corrected to empty wells containing DMSO/water/media and normalized to positive control cells treated with DMSO. All experiments were performed in quadruplicate using the following metrics as criteria for acceptability: CV<20% (positive and negative controls), SNR≥5 (positive controls), Z’≥0.5 (per plate). Drug synergy was calculated using the Bliss Independence model, reported as GI% beyond the expected additive value, and nominally classified using >1% and <−1% synergy as cutoffs for synergistic and antagonistic effects, respectively. Alternative calculations of drug synergy were performed using the Response Additivity [64] and Zero Interaction Potency (ZIP) [65] models.

Screening Validation

High-throughput screening (HTS) results were validated using an alternate assay plate format and new drug stocks and cell lots. Briefly, MRX-2843 and vincristine sulfate were dissolved in molecular biology grade water (VWR) on the day of experimentation, then serially diluted in cell media to obtain 0.01% v/v of water in all media dilutions. 45,000 viable LOUCY or Jurkat cells were then dispensed into drug/media-containing white polystyrene 96-well μClear microplates and subsequently handled as described for HTS experiments.

2- and 3-drug Effect Models

To determine whether synergy emerged from 2-drug or 3-drug interactions, we calculated pairwise and triplet combination synergy and antagonism using both lower order and higher order predictive models. For the lower order model, an Isserlis-like formula developed by Wood et al. [39] was used,

$$g_{ijk}=g_i{g}_{jk}+g_j{g}_{ik}+g_k{g}_{ij}-2g_i{g}_j{g}_k$$

where g_i, g_ij, and g_ijk refer to cell growth rates when exposed to single, pairwise, or triplet drug combinations, respectively, of drugs i, j, and k. When comparing predictions with measured effects, the Isserlis-like formula may indicate whether drug interactions can be explained by single or pairwise drug responses. For the higher order model, Bliss Independence was similarly used to assess correlations between measured and predicted values.

Microwell Assay

Jurkat cells were embedded in 3 mg/mL collagen hydrogels (Corning; Corning, NY) and exposed concurrently to overlapping concentration gradients of free drug or solvent controls for 72 hours. Following drug exposure, cell viability was assessed using viability dyes Calcein AM (Thermo Fisher) and Propidium Iodide (Thermo Fisher), and nuclei were stained using Hoechst 33342. Images were acquired using a Keyence BZ-X800 Fluorescence Microscope. Cell segmentation was performed using CellProfiler (v3.1, Broad Institute; Cambridge, MA), and cells were classified as live or dead based on mean per cell intensity measurements done in MATLAB (MATLAB 2021a; Mathworks, Natick, MA). A rolling ball averaging method was then used to determine average viability at each drug combination dose ratio. Expected viabilities were computed based on Loewe Additivity isobolograms of single drugs at 10% of the respective maximum doses whereby the line of additivity is assumed to be a linear combination between the maximum effect at the highest concentrations of drug 1 and drug 2. Deviation from the line of additivity towards lower viability indicates synergy.

RNA sequencing (RNAseq)

Three separate Jurkat (T-ALL) cultures were divided and exposed for 72 hours to either a free drug combination comprising MRX-2843 (100 nM) and vincristine (1.4 nM, water), its component monotherapies, or vehicle (DMSO, 0.5% v/v) in separate cell culture flasks (0.6e6 live cells/mL) at 37°C in 5% CO₂. At harvest, cells were centrifuged (200 × g, 5 min) and aspirated three times to remove dead cells. Isolated cell pellets were flash frozen on dry ice and stored at −80°C. RNA was extracted from frozen cell pellets using the miRNEasy mini kit (Qiagen, Cat. No. 217004; Maryland, USA). RNA quantification was determined using the Infinite M200Pro (Tecan; Zurich, Switzerland). RNA quality was assessed using the Bioanalyzer RNA 6000 Nano or Pico reagents (Agilent). RNA extracts were sequenced and analyzed by Novogene. RNA libraries were prepared using the NEBNext^® Ultra^™ DNA Library Prep Kit for Illumina^®. Libraries were then read [66] with the NovaSeq 6000 using PE150 sequencing.

Downstream analysis, including read alignments, reference genome mapping, gene expression quantification and differential expression, was performed using a combination of programs [67–73] including STAR, HTseq, Cufflinks [74] and Novogene wrapped scripts. Alignments were parsed using TopHat program. [72, 73] Reference genome and gene model annotation files were downloaded from genome website browsers (NCBI/UCSC/Ensembl) directly. Indexes of the reference genome were built using STAR and paired-end clean reads were aligned to the reference genome using STAR (v2.5). STAR used the method of Maximal Mappable Prefix (MMP), which can generate a precise mapping result for junction reads. HTSeq v0.6.1 was used to count the read numbers mapped for each gene. FPKM (Fragments Per Kilobase of transcript sequence per Million base pairs sequenced) [75] of each gene was calculated based on the length of the gene and read counts mapped to the same gene, where the threshold for expression was set to FPKM>1. [74, 76] Differential expression analysis was performed using the DESeq2 R package (2_1.6.3). DESeq2 provides statistical routines for determining differential expression in digital gene expression data using a model based on the negative binomial distribution. The resulting P values were adjusted using Benjamini and Hochberg’s approach which controls for the False Discovery Rate (FDR). Genes with an adjusted P value <0.05 found by DESeq2 were assigned as differentially expressed genes (DEGs). TFCat [74] and the COSMIC database were used to annotate differentially expressed genes. TFCat is a curated catalog of mouse and human transcription factors based on a reliable core collection of annotations obtained by expert review of the scientific literature.

Differential Gene Expression Analyses

DEGs identified by RNAseq were organized by hierarchical clustering (Morpheus, https://software.broadinstitute.org/morpheus). DEG expression levels (FPKM) were used as input values, which were then organized using the Euclidean distance of average linkages between clusters. MRX-2843 and vincristine-induced DEGs were analyzed for functional enrichment using the Metascape portal (https://metascape.org). As described by Zhou et al., [77] DEGs were converted into Gene IDs and then queried for membership across multiple databases, including GO, KEGG, Reactome and WikiPathways. DEGs were analyzed for functional enrichment in biological processes using the hypergeometric distribution. Functional terms with P values <0.01 were then hierarchically clustered into trees based on Kappa similarity scores. [78] Functional terms were considered to be similar with gene membership similarity score ≥0.3. [77] Functional terms with high similarity were grouped as clusters, and representative terms within each cluster were selected to eliminate intra-cluster term redundancy while also establishing differences between clusters. Ontology clustering was visualized using Cytoscape (v 3.7.2).

Cheminformatics Computation

All calculations of MRX-2843 and vincristine physicochemical properties (molecular charge, lipophilicity) were performed using ACD/Labs software and SMILES representations of chemical structures. In brief, ionizable microspecies were identified within the parent molecular structures and partial charge distributions were predicted based on orbital electronegativities of constituent atoms. Total molecular charge was then computed as the weighted sum of net charge for each microspecies based on their relative distributions per unit pH (step size: 0.1) and pKa. Lipophilicity was calculated as the logarithm of the Distribution coefficient (cLogD), which is the concentration ratio of all microspecies in octanol compared to water. The partition coefficient (octanol:water) of each microspecies was computed using pre-defined values of structural fragments [79–81].

Lipid Nanoparticle Formulation

DSPC (NOF Corporation; Shanghai, China), DSPG (NOF), and cholesterol (Sigma-Aldrich) were combined at a 7:2:1 mole ratio from lipid stocks solvated to 25 mg/mL in chloroform (DSPC, cholesterol) or 5 mg/mL chloroform:methanol (5:1 v/v, DSPG). Solvent from the lipid mixture was removed via rotary evaporation under vacuum at 30°C (Rotavapor R-100; Buchi, New Castle, DE) and the resulting lipid cake was further dried under house vacuum overnight. Lipid films were rehydrated via rapid addition of ammonium sulfate buffer (pH 4.25, 500 – 600 mM) followed by high power cup horn sonication at 60° C (Q700; Qsonica, Newton, CT). Multilamellar vesicles were then extruded ten times at (60°C) through 0.08 μm polycarbonate membranes (Nuclepore; Whatman) supported by polyester drain discs (230600; Whatman) using a high pressure N₂ extruder (Liposofast LF-50; Avestin, Ottawa, ON, Canada). Ammonium sulfate buffer was then exchanged for phosphate buffer (pH 7.7) via centrifugal diafiltration (100kDa Amicon; EMD Millipore, Burlington, MA) or tangential flow filtration (Krosflo KR2i; Repligen, Waltham, MA) and unilamellar vesicles were then sterile filtered through 0.45 μm cellulose acetate membranes (VWR).

To prepare drug-loaded lipid nanoparticles, MRX-2843 (10.0 mg/mL in molecular biology grade water) and vincristine (sulfate) (1.0 mg/mL in molecular biology grade water) were mixed at various molar ratios with freshly prepared, concentrated gradient liposomes. Drug and lipid nanoparticle mixtures were then agitated at 700 rpm (60° C) for 1 hour (Mixer HC; USA Scientific, Ocala, FL). Drug-loaded liposomes were dialyzed twice in phosphate buffered saline using 100 kDa MWCO dialysis bags (Float-A-Lyzer G2; Spectrum Laboratories).

UPLC-MS Analysis of Liposomal Drug Concentrations

Concentrations of MRX-2843, vincristine (sulfate), and DSPC were measured by LC-MS using a Vanquish Horizon UPLC (Thermo Fisher Scientific) fitted with a Waters Corporation ACQUITY UPLC BEH C18 column (2.1×100 mm, 1.7 μm particle size) coupled to a high-resolution accurate mass Orbitrap ID-X Tribrid mass spectrometer (Thermo Fisher Scientific). The chromatographic method for sample analysis involved elution with 20:80 MeCN:water + 10 mM ammonium formate + 0.1% formic acid (mobile phase A) and 10:90 MeCN:IPA + 10 mM ammonium formate + 0.1% formic acid (mobile phase B) with flow rate at 0.4 mL/min and column temperature set to 50°C. In a separate iteration of drug quantification, drug analytes were measured using a Waters Corporation Cortecs UPLC T3 column (2.1×150mm, 1.6 μm particle size). The chromatographic gradient for MRX-2843 and vincristine (sulfate) involved elution with water + 0.1% formic acid (mobile phase A) and MeCN + 0.1% formic acid (mobile phase B) with flow rate at 0.6 mL/min and column temperature set to 40°C. The chromatographic method for DSPC involved elution with water + 0.1% formic acid (mobile phase A) and 90:10 IPA:MeCN + 0.1% formic acid + 10mM ammonium formate (mobile phase B) with flow rate at 0.25 mL/min and column temperature set to 40°C. UPLC-MS² experiments were performed by acquiring mass spectra with targeted MS/MS (tMS²) acquisition. Data processing was performed with Thermo Scientific Xcalibur Version 4.3.73.11. The MSMS transitions were integrated, and the data was exported to excel. The transitions for each adduct were summed and quantified from standard calibration curves.

Alternatively, drug concentrations were measured by LC-MS using an Agilent 6120 mass spectrometer (Agilent Technologies; Santa Clara, CA) with an Agilent 1220 Infinity liquid chromatography inlet. The chromatographic method involved a linear gradient of 5% acetonitrile in water (0.1% formic acid, mobile phase A) and 80% acetonitrile in water (0.1% formic acid, mobile phase B). Elution occurred over 2 minutes an Agilent Zorbax SB-C18 column (1.8 μm particle size, 2.1 mm × 50 mm) with a flow rate of 0.650 mL/min at ambient temperature. Drug concentration was quantified by integrating UV absorbance at 254 nm and normalized against an internal standard (ketoconazole) peak area.

Characterization of Lipid Formulations

For downstream in vitro and ex vivo assays of growth inhibitory synergy, dual-agent MRX-2843:vincristine nanoparticles were mixed at defined ratios that recapitulated observed synergy, additivity, and antagonism in vitro. To determine these molar ratios, combination responses observed via HTS were averaged for all doses tested. Ratiometric responses were analyzed for mean synergistic responses across T-ALL and ETP-ALL cell lines by filtering doses of MRX-2843 (>200 nM) and vincristine (>2.8 nM) near or slightly above the mean lineage GI₅₀s of 208 nM and 2.6 nM, respectively.

Hydrodynamic size and polydispersity were obtained via dynamic light scattering (DLS) (DynaPro, Plate Reader III; Wyatt Technologies, Santa Barbara, CA). Nanoparticles were syringe filtered through 0.45 μm cellulose acetate membranes (VWR) and briefly sonicated prior to size measurements at fixed temperature (25°C) in black polystyrene microwell plates (Greiner Bio-One). Routine observations of all dual-agent, single-agent, and non-drug loaded formulations stored at 4°C over a 120-day period revealed no diameter changes >11% from baseline. Size and polydispersity stability were evaluated for nanoparticles incubating in heat-inactivated fetal bovine serum (0%, 10%, 50% v/v) and at 37°C for up to 72 hours via number abundance.

Nanoparticle morphology was imaged via transmission electron microscopy (TEM) at 80 kV using a Hitachi HT-7700 instrument. Liposomes were diluted in ultrapure water and applied to charge coated formvar/carbon coated copper grids (400 mesh, Electron Microscopy Sciences; Hatfield, PA) for 15 minutes followed by washing for 2 seconds with ultrapure water. Samples were then negatively stained with phosphotungstic acid for 20 seconds and washed again for 2 seconds with ultrapure water.

Lipid nanoparticles were lyophilized to test for the effects of freeze-dried storage on drug activity and drug retention. Formulations were lyophilized by mixing in 99.3 mg/mL Trehalose (Sigma) in PBS and snap freezing in liquid nitrogen. Formulations were then lyophilized under lower pressure (<0.05 mBar) and temperature (−50°C) (Labconco FreeZone; Kansas City, MO), and subsequently stored in −80°C. Long-term storage effects on drug activity were evaluated for Syn and Add formulations by measuring growth inhibitory dose responses in LOUCY cells as described for the Screening Validation. Drug retention was evaluated by comparing drug concentrations of reconstituted dual-agent and single-agent lyophilisates to spin filtered filtrates (100kDa Amicon Ultracel; EMD Millipore) via LC-MS.

Zeta-potential of nanoparticles dispersed in deionized water were measured with a Zetasizer Nano ZS (Malvern Panalytical Ltd; Malvern, UK) using the Smoluchowski model.

Intracellular Drug Delivery

LOUCY cells were exposed to Antag nanoparticles (10x HTS dose) under normal culture conditions. Cells were washed 3x in PBS and counted prior to snap-freezing on dry ice after 1, 3 and 24 hours of nanoparticle exposure. Frozen cell pellets were prepared for LC-MS analysis by mixing in 500 μL IPA and 50 μL of 500 μm glass beads. Samples were then homogenized in a Tissuelyzer II (QIAGEN) at 30 Hz for 5 minutes and the supernatant was transferred to separate microcentrifuge vials. The original samples in glass beads were washed twice with 100 μL of 80% MeOH, with each wash additive to the IPA supernatant. The cellular extractions were then dried by vacuum centrifugation and reconstituted in 250 μL of 80% MeOH. Reconstitution was assisted by sonication for 5 minutes. The samples were centrifuged at 21,100 x g prior to being transferred to LC vials for LC-MS analysis as described above. LOUCY cells not exposed to nanoparticles were processed as the negative control at 0 hours of exposure.

Drug Release Kinetics

Nanoparticles (1 mL) were suspended in 300 kDa MWCO dialysis bags (Float-A-Lyzer G2; Spectrum Laboratories) in 1 liter of either PBS (pH=7.4) or citrate buffer (pH=5.5) at 37°C under continuous stirring. Aliquots from diluent buffer (filtrates) were withdrawn for drug quantitation (LC-MS) at various time intervals up to 72 hours. Samples were diluted in MeOH and centrifuged at 21,100 x g for 5 minutes prior to LC-MS analysis. MRX-2843 lower limit of detection (LLOD) =0.19 ng/mL< or 0.023% of starting sample mass released into diluent buffer; vincristine LLOD=1.20 ng/mL, or 5.52% of starting sample mass released into dialysis buffer.

In vitro Syn Nanoformulation Dose Responses

LOUCY and Jurkat cells were treated with dose response series of Syn, Add, or Antag nanoparticles, their constituent free drug monotherapies, and free drug combinations matched for vincristine doses. MRX-2843 was dosed in multiples of 30 nM to mimic the dose effects of Syn, Add, and Antag molar ratios (MRX-2843:vincristine) at 30.13, 127.59, and 246.06. Free drug and nanoparticles were dissolved in PBS vehicle (0.05% v/v) in media on white 96-well polystyrene μClear Microplates. 45,000 viable cells (Countess IIl Invitrogen) were added per well and cultured for 72 hours. Cells were then treated with CellTiter-Glo 2.0 (Promega) and analyzed for luminescence with a SpectraMax iD3 (Molecular Devices). Data represent triplicate measurements from a single experiment.

Primary Sample Responses to Ratiometric Nanoformulations

Viably frozen, de-identified blood and bone marrow samples collected from patients with T-ALL at initial diagnosis were obtained from the Aflac Cancer and Blood Disorders Center Leukemia and Lymphoma Biorepository at Children’s Healthcare of Atlanta. Cells (12,500 per 384-well) were cultured in cytokine-sera, consisting of 20 ng/mL human recombinant IL-7, IGF-1, and SCF (StemCell Technologies; Vancouver, BC, Canada) in 90% RPMI + 10% heat inactivated FBS + 0.1 mM beta-mercaptoethanol (Sigma) [82–84] and exposed in triplicate to Syn or Add nanoparticles for 96 hours. Expected GI values for an additive interaction were interpolated from dose response curves (GraphPad Prism) of free vincristine and liposomal MRX-2843 using the Bliss Independence model of drug synergy. Each well contained 0.5% v/v PBS/media. Positive controls were cells treated with PBS/media, and negative controls contained PBS/media but no cells. Following incubation, cultures were treated with CellTiter-Glo 2.0 and relative viable cell numbers were measured as described above.

Animal Studies

A luciferase-expressing ALL cell line (LOUCY-luc) was injected into the tail vein of NOD.Cg-Rag1^tm1Mom Il2rg^tm1Wjl/SzJ (NRG) or NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG) mice (2×10⁶ or 3×10⁶ cells/mouse). Leukemia engraftment and disease burden were monitored by bioluminescence imaging as previously described [85]. Briefly, mice were anesthetized with isoflurane and then sterile filtered D-luciferin in PBS (150 mg/kg) was injected intraperitoneally (i.p.) at a dose of 10 mL/kg. Bioluminescence images were captured on an IVIS Lumina Series 3 (PerkinElmer; Waltham, MA) and quantitated with Living Image software. Mice were randomized to treatment groups with statistically equal starting disease burden (data not shown) and nanoparticles were administered by i.p. injection at a dose of 20 mL/kg unless otherwise noted. Health status was evaluated daily using a standardized scoring system and mice with >20% weight loss or meeting other standardized criteria [33] were removed from study and euthanized. All animal protocols have been reviewed and approved by the Emory University Institutional Animal Care and Use Committee (PROTO201700276).

Statistical Analyses and Software

All dose response series were modeled as nonlinear 4-parameter logistic functions using GraphPad Prism. Multidimensional synergy plots were produced in Tecplot 360. Differences between lipid nanoparticle and free drug dose responses in cell lines were measured by one-way ANOVA with Tukey’s multiple comparisons test. Observed and expected dose responses (GI₅₀) for primary samples were compared by unpaired parametric t test. Heat maps for primary sample potencies were rendered in GraphPad Prism. Experimental diagrams were drawn using BioRender software. Survival comparisons in mice were made with the Mantel-Cox log-rank method. Disease burden was compared by two-way ANOVA. Bioluminescence data were collected until fewer than half of mice in that group remained on study. Maximum bioluminescence and body mass values were carried forward for each subject after removal.

Supplementary Material

1

Fig. S1: B-ALL lineage-specific drug synergy

2

Fig. S2: Experimental validations of high-throughput screening

3

Fig. S3: Computational assessment of lower- and higher-order drug synergy per cell line

4

Fig. S4: RNA sequencing analysis of synergistic MRX-2843 and vincristine ontologies

5

Fig. S5: Characterization of lipid nanoparticle formulations

6

Fig. S6: Nanoparticle size stability post-lyophilization and in serum

7

Fig. S7: Therapeutic efficacy and toxicity in vivo

8

Fig. S8: Activity and safety of MRX-2843 single-agent formulations in vivo

9

Fig. S9: In vivo body mass following treatment

Highlights.

• MRX-2843 and vincristine mediate ratio-dependent synergistic anti-leukemia activity

• Liposomes deliver defined mole ratios of MRX-2843 and vincristine to cells

• Ratiometric nanoparticles constitutively synergize against primary T-ALL cells

• MRX-2843 and vincristine nanoformulations are therapeutically effective in vivo

• In vivo efficacy of the nanoformulations is dependent on drug ratio and drug dose