Drug responses are conserved across patient-derived xenograft models of melanoma leading to identification of novel drug combination therapies

Background Patient-derived xenograft (PDX) mouse tumour models can predict response to therapy in patients. Predictions made from PDX cultures (PDXC) would allow for more rapid and comprehensive evaluation of potential treatment options for patients, including drug combinations. Methods We developed a PDX library of BRAF-mutant metastatic melanoma, and a high-throughput drug-screening (HTDS) platform utilising clinically relevant drug exposures. We then evaluated 34 antitumor agents across eight melanoma PDXCs, compared drug response to BRAF and MEK inhibitors alone or in combination with PDXC and the corresponding PDX, and investigated novel drug combinations targeting BRAF inhibitor-resistant melanoma. Results The concordance of cancer-driving mutations across patient, matched PDX and subsequent PDX generations increases as variant allele frequency (VAF) increases. There was a high correlation in the magnitude of response to BRAF and MEK inhibitors between PDXCs and corresponding PDXs. PDXCs and corresponding PDXs from metastatic melanoma patients that progressed on standard-of-care therapy demonstrated similar resistance patterns to BRAF and MEK inhibitor therapy. Importantly, HTDS identified novel drug combinations to target BRAF-resistant melanoma. Conclusions The biological consistency observed between PDXCs and PDXs suggests that PDXCs may allow for a rapid and comprehensive identification of treatments for aggressive cancers, including combination therapies.

This source plate was used to dose 384-well microplates containing PDXC samples. Before dosing, the general health of the cells in a control lane are visually assessed. Using a simple .csv well map, the acoustic liquid handler automatically transferred each drug at each concentration from the source plate to the appropriate location on each of the cell-containing 384-well microplates. The drugged cell lines were then placed in an incubator for 72 hours, after which the CellTiter-Glo (Promega, San Luis Obispo, CA) assay was used to quantify cell viability/proliferation. Data from luminescence reads were subsequently rendered graphically in Transcriptic's web interface and made available for download in .csv format for further data analysis.

Culturing conditions for PDXC in HTDS
Blood serum albumin (BSA) in fetal bovine serum (FBS) is primarily responsible for binding and reducing activity of many drugs (11). Unlike standard cell lines, media used to propagate PDXCs does not contain FBS in order to more readily preserve the genomic and biological characteristics of the original tumor (12,13). Therefore, 0.1% purified BSA was added to PDC and PDXC growth media for HTDS in order that results could be more readily cross-compared to past and future studies using standard cell culturing techniques (14). Higher concentrations of BSA were not used because they significantly increase the cleaning maintenance of the automated cell dispenser cassettes.

Drug storage and preparation
Drugs were obtained from Selleckchem (Houston, TX) and master stocks were primarily made at half maximum solubility with DMSO or water diluent per manufacturer specifications for molecular weight, diluent and solubility. Drug stocks were then diluted to clinically relevant 4 maximum concentrations and transferred to 384-well stock drug plates at 65 µL. To prevent multiple freeze-thaw cycles of the master stocks, drugs are first aliquoted to a master plate, and then multiple working daughter 384-well plates were made, which were discarded monthly.
For drug dosing, drugs from working passage plates were transferred at 25 nL to individual wells containing a 25 µL cell suspension media pre-seeded with tumorspheres. All drugs and drug plates were stored at -80 ℃. All drugs and drug plates were set to have one-year shelf life and discarded upon passing this point.

Next Generation Sequencing
Isolation of genomic DNA from fresh human and mouse PDX tissue samples was performed using DNeasy tissue kit (Qiagen, Redwood City, CA) after tissue disruption using ruptor disposable probes, and DNA was quantified using PicoGreen (Thermo Fisher Scientific, South San Francisco, CA). The integrity was determined using agarose gels.
GM-CSF: Granulocyte-macrophage colony-stimulating factor. All samples were collect after the specified therapy.

Supplementary Table 3. Comparison of single-nucleotide variant (SNV) and insertion/deletion (ins/del) variants of DNA mutational hotspots from patient tumors and
first (X) and second (X1) generation PDX derivatives. Patient tumors and corresponding first generation (X) PDXs as well as X and X1 PDXs where compared. The match pairs are grouped using colored columns. The mutations identified are listed by row. Allele frequencies are also listed by row (for example 0.14 = 14%). Allele frequencies of SNV or ins/del that are present in matched samples are highlighted in yellow. Concordance (%) = ( ÷ × 100). Concordance at ≥ 10% AF and ≥ 25% AF was calculated for each matched pair and for all samples (Total Concordance).

Supplementary Table 4. Drug score and AUC for HTDS in 8 BRAF mutant melanoma PDXCs.
Drug scores and AUC values were calculated for single agents and vemurafenib + combimetinib 8 when assessing melanoma cell viability/proliferation. Drug score of 0 equals no effect and 100 represents killing all the cells. Drugs that stimulate cell growth produce a score < 0. An AUC value of 0 represents killing all the cells, a value of 270 represents no cell kill, values > 270 represent stimulation of cell growth. Table 5