"Direct to Drug" Screening As a Route to Individualized Therapy in Multiple Myeloma

Although more than 2,000 multiple myeloma (MM) genomes are now available, the expected era of genomic-driven precision medicine has not emerged since actionable mutations are uncommon, mostly subclonal, and, for now, lack matched therapeutics. Advances in high throughput screening and the rapid expansion of FDA approved targeted oncology drugs now raise the prospect of "direct to drug" screening as an alternate route to genomic-driven individualized therapy and as a source for chemogenomic interrogation of drug sensitivity and resistance.

Using an in house bioassay of FDA approved or late stage clinical trial drugs, epigenetic and kinase inhibitors screened in MM and NHL cell lines and by collating clinical experience, we established a panel of ˜80 MM relevant FDA approved and late stage clinical trial drugs for our standardized screening platform. This custom drug panel was specifically designed for screening live CD138+ selected tumor cells from MM patients with immediately actionable results.

Feasibility studies were conducted in human MM cell lines over 72h using a real time 384 well format, non-lytic cellular viability assay (Real Time Glo, Promega). These studies revealed an optimal cellular concentration per well, and that drug efficacy can be detected as early as 24 hours for most compounds, with the notable exception of immunomodulatory drugs (IMiDs). Further optimization using a lytic cellular viability assay (Cell Titer Glo, Promega) confirmed that a 24h drug exposure time point could be fixed for clinical relevance (72h for IMiDs). We then derived consensus conditions to screen the MM drug panel in live CD138+ sorted tumor cells collected from MM patients. Baseline clinical and cytogenetic characteristics are recorded, and a library of baseline DNA and RNA is assembled for further genomic interrogation.

The MM ˜80 drug panel was first screened in 25 human MM and 10 lymphoma cell lines of known genetic subtype and drug dose dependency measured in duplicate for 7 drug doses covering a broad concentration range. All drugs exhibiting dose-dependent reduction of cellular viability were reported as hits in the screen and the data collated into a drug sensitivity map. At 24h, 10 drugs exhibited broad cytotoxicity across all cell lines, increasing to 22 drugs at 72h. These included all proteasome inhibitors, the CDK5 inhibitor dinaciclib, the CRM disruptor selinexor and the HDAC inhibitors panobinostat and romidepsin. Conversely, 16 drugs were completely inactive at 24h. As expected, IMiDs and other activities emerged at 72h and the number of inactives was reduced to 7 drugs, including erlotinib, vismodegib, enasidenib, nindetanib, plerixafor, tofacitinib, and the negative control of unmetabolized cyclophosphamide. Other drug classes, exemplified by venetoclax, presented differential activity profiles that correlated with genomic features, and novel target classes for MM were suggested (e.g. PIKfyve).

To date, 65 primary patient samples have been screened, of which 23 were newly diagnosed and 42 received previous treatment (median of prior therapies = 5). Eleven drugs were active in less than 5% of primary patient samples, including: lenvatinib, tofacitinib, paclitaxel, pemetrexed, vismodegib, decitabine, and plerixafor. In contrast, 8 drugs were active in more than 70% of patients samples at nanomolar concentrations, including dinaciclib, bortezomib, carfilzomib, ixazomib, panobinostat, selinexor, venetoclax, romidepsin, and belinostat. The most potent drugs, with the majority of patient samples presenting EC50s lower than 10nM, were dinaciclib, carfilzomib, panobinostat, and romidepsin.

Using bromodomain inhibitors (BETi) as an example, in 42 primary MM samples, 17 (40%) were responsive to all three BETis and 32 had dose-dependent response to at least 1 of the BETis. Of these 32 drug sensitive patients, 18 were genetically hyperdiploid and 5 had a FISH defined disruption at the Myc locus (71%), suggesting the expected correlation of Myc expression with responsiveness. Similar positive correlation was seen with venetoclax and presence of a t(11;14).

As numbers of tested patients increase, a powerful dataset emerges of ex vivo drug sensitivities linked with clinical outcomes and genomic biomarkers, which can be harnessed for clinical trial enrichment, MOA discovery, and potentially patient care, enlightened by clinical and genomic annotation.