Ex vivo drug testing of primary tumor cells from individual patients is being evaluated in clinical trials to guide therapeutic decisions for patients with resistant disease1. However, primary cells are generally difficult to maintain viable ex vivo in monoculture. Coculture of primary tumor cells with representative components of the in vivo microenvironment can maintain metabolic homeostasis and viability of primary tumor cells to extended periods. The complexity of the multicellular systems involved is nicely recapitulated using organoids derived from human induced pluripotent stem cells2.
Two-dimensional models have been used to assess the impact of stromal cells in co-culture with malignant cell lines but it remains challenging to accurately measure the viability of primary tumor cells using traditional cell viability assays3. Bone marrow mesenchymal stroma cells (MSCs) maintain the viability of primary acute lymphoblastic leukemia (ALL) cells in ex vivo culture and influence the response to anti-leukemic agents4-6. Even in these simplified models it remains difficult to follow the fate of leukemia cell subpopulations depending on their contact with the microenvironment. Here, we engineer a stroma-based biosensor for specific detection of tumor cells in a coculture model of bone marrow. Using this system, we performed drug response profiling to explore promising drug combinations to target high-risk leukemia subtypes in a protective context.
To specifically monitor the live ALL cells that interact with MSCs in coculture, we engineered hTERT-immortalized MSC cells4 to express a CD19 binding synthetic Notch receptor (CD19-SynNotch-TetRVP64)7 that controls expression of an integrated reporter cassette TetO-BFP (MSC.CD19sensor.BFP, Fig 1a). When coculture with CD19+ ALL cells, the membrane tethered transcription activator TetRVP64 is released and drives the expression of a BFP reporter in MSC cells (Fig. 1b). As the synthetic Notch receptor only responds to surface tethered ligands7, the reporter of the MSC.sensor can only be induced by CD19 bound beads but not soluble CD19 (Supplementary Figure S1a). Thus, measurement of the reporter signal from MSC.sensor provides a way to monitor the viable ALL cells in contact with cocultured stromal cells in situ.
While imaging-based analysis can collect many spatially and temporally resolved parameters, a bioluminescence-based readout is more convenient for high-throughput viability assays. Therefore, we engineered a bioluminescence reporter of MSC.sensor (MSC.CD19sensor.Lum) using the nano luciferase fused to a PEST domain8. This reporter yields signals with higher sensitivity and dynamic range than fluorescence signals. Single cell clones of MSC.sensor cells were generated and displayed a similar capacity to support the viability of ALL patient-derived xenografts (PDXs) in coculture as the parental MSC population (Supplementary Figure S1b). Coculturing of a fixed number of MSC.CD19sensor.Lum cells (2500 / well in 384 well plates) with an increasing number of ALL PDXs cells yielded a linear increase of bioluminescence signal up to a ratio of ALL/MSC of 4 (Fig. 1c). Thus, the bioluminescence signal from MSC.CD19sensor.Lum can be used as a surrogate readout for viable cell count to scale functional screening.
Although MSCs usually have a higher tolerance to chemotherapy than primary ALL cells, their physiological status should not be neglected while profiling with drugs in cocultures. To evaluate the impact of drugs on the MSC.CD19sensor.Lum cells, we measured the bioluminescence signal (driven by CD19 bound beads) and viability (ATP-based assay) of MSCs treated with a selection of anti-cancer agents that are routinely included in contemporary functional precision hematology programs (Supplementary Figure S 1c). We tested 20 approved and investigational compounds in preclinical development with non-overlapping mechanisms of action. While most of the tested compounds (15 out of 20) did not perturb the viability or the reporter of MSC.CD19sensor.Lum cells in a dose-range of 1nM – 1µM, 10nM of vincristine, bortezomib and panobinostat, as well as 100nM of cytarabine and idarubicin, provoked an elevated luminescence signal accompanied by a decrease of MSC viability (Supplementary Figure S1c). Thus, highly bioactive compounds targeting pan-essential cellular functions may perturb the homeostasis of MSC cells and confound the readout during drug profiling. To monitor the kinetics of the reporter signal after killing of ALL cells, we tested the BCL2 inhibitor Venetoclax in 2 PDXs. A progressive decay of the reporter signal matched the decrease of the PDX cell count detected by flow cytometry (7-AAD staining) upon 24h treatment (Fig. 1d).
Combinational drug screening is a valuable strategy to identify drug synergy to be further developed as safe and effective treatment options for cancer patients9. This approach usually requires high-throughput measurement of cell viability in large-scale combinational matrices. We explored candidate drug synergisms targeting high-risk ALL subtypes using the MSC.CD19sensor.Lum coculture with ALL PDXs. We considered Venetoclax as an ideal backbone for exploring synergies with other compounds because: (1) Venetoclax has favorable pharmaceutical properties and efficacy against several high-risk ALL cohorts including TCF3-HLF and MLL-rearranged ALL10-12; (2) Combination of Venetoclax with other therapeutics achieves sustainable remission in many tumor types13; and (3) Genome-wide CRISPR analysis of chemotherapy–gene interactions in ALL cells uncovered that sgRNA targeting of BCL2 scored broadly as a common target whose inhibition could enhance the response to many chemotherapy agents14.
We established a high-throughput pipeline for combinational drug screening of Venetoclax with 12 therapeutic agents in high-risk ALL cells (Venetoclaxplus screen, Methods) (Fig 2a, Supplementary Figure S2, and Supplementary Table S1). Venetoclax showed the strongest synergy with the MCL1 inhibitor S36845, which is in consistent with the functional redundancy of BCL2 and MCL1 in cell death control and endorses the clinical development of this approach in ALL therapy15. Our profiling also revealed potent synergistic combinations of Venetoclax with L-asparaginase and the XPO1 inhibitor Selinexor consistently across all tested PDXs (Fig 2a and 2b). Combinational treatment in a TCF3-HLF cell line (HAL-01) and a TCF3-PBX1 cell line (697) recapitulated these synergies in monoculture (Supplementary Figure S3). The first-in-class nuclear transporter inhibitor Selinexor may represent a promising therapy for ALL patients and is currently in phase I/II clinical trial for pediatric relapsed/refractory acute myeloid leukemia (AML) patients (Clinicaltrials.gov Identifier NCT04898894). To confirm the on-target effect on XPO1 we performed a CRISPR knockout of XPO1 and a sgRNA competition assay in HAL-01 and 697 cells. We found that cells expressing XPO1 sgRNAs were rapidly outcompeted by non-transduced cells shortly after infection, which was further pronounced under Venetoclax treatment (Fig 2c). Given the evidence of the synergy between Venetoclax and Selinexor in AML, diffuse large B-cell lymphoma and multiple myeloma16, this combination may represent a paradigm of synthetic lethality in many hematologic malignancies.
Although cell lines have been widely used in functional studies and preclinical drug screenings, cell lines have adapted to grow without niche support and do not represent the molecular complexity of the disease at presentation3. Experimental models that more faithfully reflect the primary disease may facilitate the development of next generation therapeutics. We and others have established coculture systems to support primary ALL ex vivo enabling functional profiling using imaging-based approaches5,6. While imaging records many morphological features of the cells, it generally requires complex analytical pipelines with elaborate software for quantification which require costly infrastructure. Furthermore, inter-sample heterogeneity of leukemia cell morphology confers additional complexity for data analysis by automated imaging. Engineering of biosensor MSCs, which detect ALL cells that interact through CD19 as a surrogate biomarker confers several advantages: (1) Specific detection of leukemic CD19 blasts, which is particularly valuable for bone marrow samples with low leukemia cell infiltration (detection of a small subpopulation of cells without additional cell sorting steps); (2) Objective readout that captures the total amount of leukemia cells in contact in coculture; (3) Label-free and fast assay, which is suitable for high content exploration studies. Although CD19 is dispensable for ALL cells survival17 and binding with CD19 antibody does not impact ALL cell viability (Supplementary Figure S1b)18, engagement with cell surface molecules may rewire downstream signaling and should be considered when designing sensors for other biomarkers in a new model.
We demonstrate the application of MSC.sensor platform to facilitate the exploration of clinically effective combination therapies in ALL. Using a 384-well plate holding 2,500 MSC / well, it is practical to detect 100 – 10,000 ALL cells / well. This format can be further miniaturized, for example by using a 1536 well plate or microfluidic systems to spare primary material. Since the MSC.sensor “reflect” the presence of ALL through contact, this system may also constitute a powerful platform for pooled genetic screens interrogating non-cell-autonomous determinants supporting tumor survival 19.