AUTHORS
Arwa Bohra, MD, Joachim L. Petit2, Kelli Ann Gyotoku3, Matthew Smith4, Wilson Gonsalves, MD5, P. Leif Bergsagel, MD6, Shaji Kumar, MD7
AFFILIATIONS
1 Mayo Clinic, Rochester, United States
2 Mayo Clinic, Jacksonville, United States
3 Mayo Clinic, Rochester, United States
4 Mayo Clinic, Rochester, United States
5 Mayo Clinic, Rochester, United States
6 Mayo Clinic, Arizona, United States
7 Mayo Clinic, Rochester, United States
ABSTRACT
Background
Functional drug testing can capture therapeutic vulnerabilities not apparent from genomic profiling alone, offering a path toward precision medicine in guiding therapy selection.
Methods
We performed same-day ex vivo drug-sensitivity testing on freshly obtained CD138+ myeloma cells using the previously described “Direct-to-Drug” platform (Bonolo de Campos et al., 2020), with 24-hour CellTiter-Glo viability assessment across 7-point drug dilutions. Drug response was summarized using normalized AUC, with lower AUC indicating greater sensitivity, and measured OS from sampling to death or last follow-up using one observation per patient. Drug AUCs were used to predict survival kinetics and optimal thresholds were obtained for drugs with impact on survival. Corresponding RNA-sequencing data were also analyzed. (n=12).
Results
Among 169 MM encounters from 158 unique patients, 110 (65.1%) were previously treated and 59 (34.9%) were treatment-naive. Median follow-up from testing was 57.7 mo, 63.4 mo for NDMM and 50.6 mo for treated MM. Overall, 96 drugs were tested, with a median of 82 per encounter. The lowest median AUC values overall (<0.2) were observed for PMX-464 (thioredoxin (Trx1)/thioredoxin reductase (TrxR1) system inhibitor), S63845 (MCL1 inhibitor), auranofin, carfilzomib, and JWT-001. Among treatment-exposed encounters, AUC <0.2 was most frequent for PMX-464 (87.7%), auranofin (86.0%), S63845 (81.8%), AMG-176 (MCL-1 inhibitor) (60.6%), carfilzomib (59.1%), YM-155 (survivin inhibitor) (51.5%), bortezomib (44.9%), JWT-001 (41.2%), and dinaciclib (40.8%).
Among treatment-naive patients, each 0.1-unit increase in AUC was associated with inferior OS, with the hazard ratio (HR) highest for lenalidomide (2.7; 95% CI, 1.1–6.8; P<0.05), followed by ixazomib (2.5; 1.5–4.2; p<0.01), bortezomib (2.2; 1.4–3.4; p<0.01), and carfilzomib (1.7; 1.2–2.5; p<0.01). At an AUC cutoff of 0.2, carfilzomib sensitivity predicted superior OS (0.1; 0.0–0.6; p<0.05), while venetoclax sensitivity showed a nonsignificant favorable trend (0.2; 0.1–1.2; p=0.08) suggesting specific use cases. Maximally selected cutoffs further identified resistant subgroups with inferior OS: for bortezomib, AUC >0.3 was associated with median OS of 60.7 months versus not reached (M=4.7; p<0.01); for carfilzomib, AUC >0.2 was associated with median OS of 72.5 mo versus not reached (M=3.3). For lenalidomide, AUC ≥0.8 was associated with median OS of 95.5 mo versus not reached (9.2; 1.1–75.7; p<0.05, max-adjusted=0.3).
Alternative ex-vivo activity remained evident within resistant subgroups. Among carfilzomib-resistant cases, PMX-464 and auranofin each had AUC <0.2 in 57.1%; among bortezomib-resistant cases, both were sensitive in 61.9%. In lenalidomide-resistant MM, AUC <0.2 was observed for auranofin (83.7%), PMX-464 (81.6%), S63845 (79.4%), carfilzomib (55.3%), AMG-176 (50%), and bortezomib (48%). Similarly, venetoclax-resistant cases retained sensitivity to auranofin (83%), PMX-464 (78.7%), S63845 (71.4%), carfilzomib (55.4%), and JWT-001 (50%).
Hierarchical clustering of ex vivo drug-sensitivity profiles (130 of 169 encounters meeting ≥70% AUC completeness) identified two clusters in both exposed and non-exposed. Among treatment-naive patients, Cluster 1 (n=7) trended towards better outcomes (95% CI, NR–NR, 7 events), vs Cluster 2 (n=39) median OS also not reached (95.5–NR mo; P=0.3, 7 events). In this treatment-naive setting, both clusters showed sensitivity (median AUC <0.20) to carfilzomib and venetoclax, with Cluster 1 additionally sensitive to dinaciclib, bortezomib, and panobinostat, Cluster 2 additionally sensitive to auranofin, PMX-464, S63845, and JWT-001. Among treatment-exposed patients, Cluster 1 (n=38) had a significantly shorter median OS of 17.9 months (7.5–36.6) compared to Cluster 2 (n=35), which had a median OS of 44.1 months (P=0.04). In the treatment-exposed setting, both clusters shared sensitivity to carfilzomib, S63845, PMX-464, and auranofin, while Cluster 1 was uniquely sensitive to dinaciclib, AMG-176, YM-155, and JWT-001.
Conclusion
This is proof of concept that ex vivo sensitivities to various drugs correlate with survival and could inform therapy selection and adaptive clinical trial design. Auranofin and PMX-464 induced ≥90% killing (AUC ≤0.1) in a significant proportion of samples, marking them as strong candidates for clinical trials.
DISCLOSURES
Arwa Bohra, MD: No relevant conflicts of interest to declare.