Lymphoid malignancies range from indolent lymphomas to aggressive lymphomas and acute lymphoblastic leukemia (ALL). Aggressive B-cell and T-cell lymphomas are subdivided into several entities including diffuse large B-cell lymphoma (DLBCL) and peripheral T-cell lymphoma (PTCL). Indolent disease includes mantle cell lymphoma (MCL), chronic lymphocytic leukemia and follicular lymphoma1.
Treatment of aggressive lymphoid malignancies is based on intensive poly(immuno-)chemotherapy protocols, although specific regimens vary by diagnosis2. Prognosis differs substantially across subtypes: in pediatric ALL, cure rates exceed 90%3, whereas real-world data indicate a cure rate of ~60% for DLBCL4. PTCL generally carries a poor prognosis, with only one fourth of patients achieving durable remission5. MCL is characterized by recurrent relapses with progressively shorter remissions and long-term remission is expected only in a minority of patients6. Across all aggressive lymphoid malignancies, outcomes remain poor in relapsed/refractory (R/R) disease4,5,6. New advances in treatment of R/R aggressive B-cell lymphomas and B-cell ALL (B-ALL) including chimeric antigen receptor (CAR)-T cell therapy and bispecific antibodies have led to improved outcomes, but a substantial proportion of patients relapse again7. Disease control prior to CAR-T cell therapy has emerged as an important determinant of outcome, a finding also observed for patients undergoing autologous hematopoietic stem cell transplantation8,9. These observations underscore the continued relevance of both conventional chemotherapy and novel targeted therapies, while highlighting the unmet clinical need for rational treatment selection in heavily pretreated patients.
Precision medicine aims to improve therapeutic outcomes by aligning treatment strategies with individual patient characteristics10. This approach is particularly appealing for heavily pretreated (R/R) patients, where prior exposure to multiple and heterogeneous regimens increases biological variability within the cohort. Early precision medicine initiatives primarily relied on genomic profiling to guide therapy selection; however, this strategy failed to identify actionable targets for a substantial proportion of patients10. Consequently, functional precision medicine has emerged as a complementary approach, leveraging functional assays to inform treatment decisions. Ex vivo drug sensitivity and resistance testing (DSRT) offers an opportunity to both elucidate mechanisms of drug sensitivity and resistance and may support individualized treatment selection11,12.
Here we present our experience and initial findings from the Drug Sensitivity Assay – Lymphoma/Leukemia (DSA-LL) observational study where a DSRT is performed on cell samples from adult patients with R/R lymphoid malignancies.
To date, 26 patients have been enrolled, with one participant sampled at two different relapses. Clinical characteristics and diagnoses are summarized in Table 1. Samples were obtained from various anatomical sites (Fig. 1A). Up to 528 clinically approved and investigational drug conditions were tested in 5-point concentrations across a 104-fold concentration range (Fig. 1B). A full list of included drugs is available in supplementary table 1 and a target level breakdown of included kinase inhibitors are presented in Supplementary Fig. 1. Of 27 samples, three were excluded from analysis due to insufficient mononuclear cell (MNC) yield (median cell number 9.2 × 106 (range 0.01 × 106–1.39 × 1010), Fig. 1C): all core needle biopsies. Cell viability was not a primary cause for exclusion (99% (range 25–100), Fig. 1D). Average assay well density was 104 (range103–104) (Fig. 1E) and appears to contribute to exclusion as two samples failed assay quality control in all assay plates (median z’ 0.7 (range −33.53–0.88)) (Fig. 1F), both of which had an assay well density of 103 cells/well. The metabolic activity in the samples remained relatively stable during the assay period (fold growth 0.69 (range 0.18–0.88), Fig. 1G). A majority of successfully analyzed samples, 16/22, were tested against all 528 drugs across eight plates, 5/22 samples were tested against a smaller library of 3 plates due to a lack of MNCs. One sample was tested against 7 of 8 plates due to supply issues.
Fig. 1: Overview of sample characteristics, drug library and assay performance.
A Flowchart of included sample and excluded samples by tissue type. B Drug classes represented in the screening library; clinically relevant drugs highlighted. A full list of included drugs is presented in Supplementary Table 1. A target level breakdown of drugs in the kinase inhibitor group is presented in Supplementary Fig. 1. C Mononuclear cell counts per sample. D Pre-assay cell viability, red data points correspond to patients excluded due to failed quality control. E Cell density per well, red data points correspond to patients excluded due to failed quality control. F Assay quality assessed by Z’ factor, red data points correspond to patients excluded due to failed quality control. Values below 0 are set as 0. G Fold growth during 24-hour incubation, red data points correspond to patients excluded due to failed quality control. NSAID non-steroid anti-inflammatory drugs, HSP heat shock protein, MNC mononuclear cell.
Table 1 Overview of patient characteristics
A 24-h incubation period was used for two reasons; 1) to maintain lymphoma cell viability ex vivo, as the cells exhibit a rapid decline in viability during prolonged culture13,14, 2) a short turnaround time is essential for future potential implementation as treatment of relapsed/refractory aggressive lymphoma and acute lymphoblastic leukemia must occur without delay.
Our findings show that ex vivo DSRT is feasible in R/R non-Hodgkin lymphoma and ALL, particularly in patients with leukemic disease or bone marrow involvement, and in those with nodal disease. Conversely, feasibility thus far has been limited in patients with exclusively solid organ involvement, although the sample size is small and further testing needs to be done.
Ex vivo drug profiling demonstrated low selective drug sensitivity scores (sDSS15) for commonly used chemotherapeutic agents, including platinum compounds (oxaliplatin median sDSS −6.08, range −11.37–3.42, carboplatin median sDSS −3.77, range −3.8–7.8), nucleoside analogues (cytarabine sDSS median −6.99, range −10–6.9, fludarabine sDSS median −8.45, range −11.83–−3.7), etoposide (sDSS median −2.01, range −6.26–11.94), and anthracyclines (doxorubicin sDSS median −6, range −7.62–1, epirubicin sDSS median −6.36, range −8.85–2.22) indicating substantially lower activity compared to healthy bone marrow controls in the majority of samples. This aligns well with expectations for a cohort with extensive prior treatment16. Glucocorticoids such as dexamethasone and BCL-2 inhibitors such as venetoclax exhibited a range of sensitivities in our cohort (median sDSS 2.8 (range −2.5–21.6), and 2.2 (range −5.1–24.2), respectively). The BCL-2/BCL-xL inhibitor navitoclax demonstrated the most consistent sensitivity (10.0 (range −8.5–22.7)). Clinical responses to BCL-2 inhibitors have previously been reported in relapsed ALL (particularly T-ALL) and MCL17,18, and the success of the venetoclax-based ViPOR regimen in relapsed DLBCL19 supports the clinical relevance of our findings. A heatmap demonstrating sDSS values for all tested drugs is shown in Fig. 2. A heatmap displaying DSS values for all tested drugs across all included patients can be found in Supplementary fig. 2.
Fig. 2: Heatmap showing the selective drug sensitivity score landscape of sampled patients.
Clustering was done using 1 minus Pearson correlation. Color scale spans from the 1st percentile (dark blue) to the 99th percentile (red) with 0 as neutral (white). Gray boxes indicate drugs not tested in the specific patient sample due to 1. Insufficient cell sample for testing or 2. Unavailability of drug at the time of testing. AITL Angioimmunoblastic T-cell Lymphoma, AML Acute Myeloid Leukemia, CLL Chronic Lymphocytic Leukemia, DLBCL Diffuse Large B-cell Lymphoma, HSTCL Hepatosplenic T-cell Lymphoma, LBCL Large B-cell Lymphoma, MCL Mantle Cell Lymphoma, Ph Philadelphia chromosome, ALL Acute Lymphoblastic Leukemia, PTCL Peripheral T-cell Lymphoma, T-LBL T-Lymphoblastic Lymphoma, sDSS Selective drug sensitivity score, WT Wild type.
Interestingly, hierarchical clustering analysis using 1 minus Pearson correlation (Fig. 2) demonstrates a grouping of p53-aberrant, defined as TP53 mutation, del(17p), or p53 overexpression, samples. Although few risk markers overlap across these diverse diagnoses, p53 aberrations are recognized as adverse prognostic factors in most hematologic malignancies. When comparing p53-aberrant disease, to p53 wild-type (WT) disease, we observed significantly higher sDSS for dasatinib in the aberrant group compared to the wild-type group (median sDSS 14.17 vs −2.6, P = 0.0003, q = 0.162784, Supplementary Data 1). All p53-aberrant disease cases demonstrated sensitivity to dasatinib compared to one p53-WT disease case (Fig. 3A–D). Further analysis of ABL and SRC inhibitors in our dataset revealed 3 additional ABL and/or SRC inhibitors with increased activity in p53-aberrant vs p53-WT samples, saracatinib an SRC and ABL inhibitor (median sDSS 5.4 vs −1.72, P = 0.025372, q = 0.115614), bosutinib both an ABL and SRC inhibitor (median sDSS 3.75 vs −3.94, P = 0.028617, q = 0.131555) and ponatinib a ABL inhibitor (median sDSS 3.6 vs −3.68, P = 0.043417, q = 0.131555).
Fig. 3: Drug sensitivity profiles and p53-associated effects.
A Differential drug sensitivity between p53-aberrant and p53-wild-type samples using multiple Mann–Whitney U-tests (FDR = 1%); top performers based on q-value are highlighted as red dots and sDSS values are included figure 3D. B Comparison of dasatinib sDSS between p53-aberrant and p53-wild-type groups using Mann–Whitney U test. Blue dots correspond to sDSS-values from p53-wild type samples while red dots correspond to sDSS values from p53-aberrant samples. C Individual dose-response curves of dasatinib in representative p53-aberrant and p53-wild-type samples. Mean response curve fit across each group of patients is shown in bold. Blue curves correspond to p53-wild type samples and red curves correspond to p53-aberrant samples. D Heatmap showing selective drug sensitivity scores (sDSS) for a subset of 50 compounds across selected samples. Compounds were chosen based on 1. Common clinical use in relapsed/refractory disease 2. Top performance regarding sDSS (highest mean sDSS values) 3. Lowest q-values in multiple comparison testing or 4. BCR-ABL inhibitors and SRC-family kinase inhibitors to demonstrate differences compared to dasatinib. Color scale spans from the 1st percentile of sDSS values for all tested drugs (dark blue) to the 99th percentile of sDSS values in all samples (red) with 0 as neutral (white). Gray boxes indicate drugs not tested in the specific patient sample due to 1. Insufficient cell sample for testing or 2. Unavailability of drug at the time of testing. A table showing sDSS results in detail is available as Supplementary Data 2. AITL Angioimmunoblastic T-cell Lymphoma, AML Acute Myeloid Leukemia, CLL Chronic Lymphocytic Leukemia, DLBCL Diffuse Large B-cell Lymphoma, HSTCL Hepatosplenic T-cell Lymphoma, LBCL Large B-cell Lymphoma, MCL Mantle Cell Lymphoma, Ph Philadelphia chromosome, ALL Acute Lymphoblastic Leukemia, PTCL Peripheral T-cell Lymphoma, T-LBL T-Lymphoblastic Lymphoma, DSRT Drug screening and resistance testing, sDSS Selective drug sensitivity score, WT Wild type.
Sensitivity to dasatinib in p53-aberrant disease has been previously reported in CLL20,21. However, we observed increased dasatinib sensitivity may across multiple diagnosis in p53-aberrant R/R lymphoid malignancies. To explore this further, we leveraged publicly available data from the BEAT-AML trial22 to further investigate p53-aberrancy and dasatinib sensitivity in myeloid malignancies. In BEAT-AML p53-mutation was associated with decreased sensitivity to dasatinib compared to WT-p53 in AML patients (Supplementary Fig. 3A). Using DepMap23 to look at drug sensitivity in cancer cell lines there was no significant association between p53-mutation status and dasatinib sensitivity based on IC50 across all cancers (mutated-p53 vs unmutated-p53 median 5.535 vs 4.278, P = 0.3710) or in specifically lymphoid cell lines (mutated-p53 vs unmutated-p53 median 3.527 vs 2.895, P = 0.8274) (Supplementary Fig. 3B, C).
Furthermore, we observed a trend towards increased sensitivity to PI3K inhibitors, including Idelalisib (median sDSS 13.62 vs −2.16, P = 0.0009, q = 0.162784), Serabelisib (median sDSS 4 vs −1.54, P = 0.006, q = 0.300380), and Omipalisib (median sDSS 9.4 vs −5.75, P = 0.026, q = 0.368484). However, this was not consistent across all PI3K inhibitors.
Analysis with the present method required a large amount of MNCs limiting the applicability of our approach in Hodgkin’s lymphoma (characterized by a low proportion of Hodgkin–Reed–Sternberg cells24), primary CNS lymphoma and patients with limited disease burden. Reducing the number of compounds tested could improve feasibility in Hodgkin lymphoma, primary CNS lymphoma, and other cases with limited material. However, this approach restricts discovery to drugs expected to work in lymphoid malignancies, limiting repurposing opportunities. Moreover, patients with advanced disease and multiple prior therapies (median four) lack a clear standard of care, complicating design of disease-specific panels.
The limited efficacy of conventional chemotherapeutic agents in the assay requires further investigation, extensive resistance is reasonable in a heavily pre-treated cohort however other studies have shown mixed results for conventional chemotherapeutics using shorter incubations12,25 potentially introducing a disproportionate disadvantage for these agents.
In summary, we demonstrate that ex vivo DSRT using our approach is feasible in patients with R/R non-Hodgkin lymphoma and ALL, provided there is liquid or nodal involvement. Our findings are consistent with expectations for a heavily pretreated cohort, showing resistance to conventional chemotherapy and sensitivity to agents known to be active in R/R lymphoid malignancies, including BCL-2 inhibitors. Clinical outcome and treatment response data are being prospectively collected; future studies will evaluate their correlation with ex vivo DSRT profiles. Comprehensive mutational profiling, beyond p53 status, was not available for the current cohort. Access to broader genomic data in subsequent studies will facilitate the investigation of molecular determinants of drug resistance. The mechanistic basis for the observed dasatinib sensitivity in p53-aberrant disease remains unclear; ABL inhibition, SRC inhibition, combined target engagement, or off-target effects cannot be distinguished within this dataset, as these mechanisms are represented across multiple sensitive compounds. Larger, prospectively designed studies in heavily pretreated relapsed/refractory patients will be required to define the contributing mechanisms and establish the therapeutic relevance of this association.
