Weighted Gene Co-Expression Network Analysis (WGCNA) Identifies Highly Proliferative Myeloma Subgroup Responsive to CDK4/ARK5 Inhibition

Multiple myeloma (MM) is an incurable plasma cell malignancy accounting for more than 10,000 deaths in the US each year. Hence the pursuit for novel therapeutic agents remains critically important. Myeloma pathogenesis is associated in part with aberrant cell cycle progression. Inhibition of cyclin dependent kinases CDK4/6 results in cell cycle arrest and sensitization to Bortezomib and other active agents in MM (Huang, Blood 2012). Here, we show that ARK5, a novel member of the human AMPKfamily, is overexpressed in 70% of MM and helps promote proliferation and cell cycle progression via G1/S phase activation in an mTOR dependent manner.

We examined the role of ARK5 using loss of function studies by ARK5 siRNA transfection in MM1.S, NCI-H929 cells as well as treatment with ON 123300, a dual CDK4/ARK5 kinase inhibitor. ARK5 siRNA knockdown decreased MM cell viability and cell proliferation via G1/S arrest compared to control siRNA. ARK5 siRNA treatment significantly (~70%) induced apoptosis in MM cells as detected by Annexin V/PI staining. We observed that phosphorylation of Rb, a critical cell cycle protein was significantly reduced in ARK5 depleted cells. Moreover, mTOR pathway inhibition was confirmed by reduction of pS6K in ARK5depleted cells as compared to control siRNA treated cells.

ON 123300 decreased viability in MM cell lines and patient cells but was not lethal to normal PBMCs. A single treatment of 50nM drug stratified MM cell lines into 2 groups, 5 resistant (MM.1R, KMS11, U266, RPMI-8226 and ARP1) and 4 sensitive cell lines (>80% cell kill-MM.1S, EJM, JJN3, NCI-H929). ARK5 protein expression by western blot analysis was much higher in sensitive cell lines. ON 123300 triggered G0/G1 cell cycle arrest and induced apoptosis similar to the effect of ARK5 siRNA (80% vs 70%). ON 123300 treatment also reduced phosphorylation of pRb and pS6K downstream of mTOR pathway. These results confirm that cell inhibitory effects of ON 123300 in MM are mediated in a large part via inhibition of ARK5. Co-culture experiments with BMSCs showed that ON 123300 not only targets MM cells but also overcomes the cytoprotective effects of the MM-host BM microenvironment. 4/5 ARK5 positive primary samples with adverse cytogenetics including 1q amplification and CyclinD1 translocation were sensitive to ON 123300 (>80% cytotoxicity) at 50nM. Further, IP injection of ON 123300 (100mg/kg) in tumor xenograft models (MM1.S, NCI-H929) showed that ON 123300 is well tolerated and significantly inhibits tumor growth in vivo(p<0.001).

To study the mechanism of action for ON 123300, we performed geneset enrichment analysis (GSEA) on drug induced gene expression signature of RNA-Seq data from pre-post treated cell lines. We interrogated a wide array of geneset libraries, including MSigDB (Subramanian, PNAS 2005), drug induced transcriptional modules (Iskar, Mol Sys Bio. 2013) and disease signatures (Sirota, Sci Transl Med 2011). GSEA showed significant representation of genes that are enriched in normal plasma cells and rapamycin sensitive geneset.

Next, we developed a weighted co-expression network (WGCNA, Langfelder BMC Bioinformatics 2008) based classifier using 304 MM samples from MMRC collection and RNA-Seq from 28 MM patients from Mt. Sinai Hospital. WGCNA defines a network that continuously links all genes and then clusters the most highly co-expressed genes in defined modules. These network modules were associated with clinical traits and UAMS GEP classification (Zhan, Blood 2006) of each sample. There was significant overlap between highly proliferative “PR” and “Cyclin D1/2” patient subsets based on classification and sensitivity to CDK4/ARK5 inhibition. Our classifier accurately discriminated 4 sensitive primary samples from one resistant sample, all tested in vitro. All sensitive samples were either Cyclin D1/2 or PR as per UAMS classification.

Our preclinical studies provide the basis for clinical evaluation of CDK4/ARK5 inhibitor ON 123300 due to its selective cytotoxicity on MM cells in vitro and in vivo. Using WGCNA we establish a systematic framework by constructing for the first time, MM-associated gene co-expression networks contributing to tumorigenesis and progression. Thus, WGCNA modeling is a novel approach for identification of MM patient subgroups that have a higher likelihood of response in clinical trials with CDK4/ARK5 inhibitors.