Interactions with a "Humanized" Mesenchymal Bone Marrow Stromal Niche In Vivo Modify the Patterns of Essential Genes for Myeloma Cells: Therapeutic Implications

Background: The biology and treatment response of human multiple myeloma (MM) cells in vivo is influenced by interactions with mesenchymal bone marrow stromal cells (BMSCs). For several key BMSC-derived cytokines (including IL-6) only the human, not murine, form optimally interacts with their respective receptor(s) on human MM cells. To better simulate the treatment responses of human MM cells in the BM, "humanized" BM-like niches in vivo have been engineered with biocompatible ceramic scaffolds "functionalized" via osteogenic differentiation of human mesenchymal BMSCs and implanted subcutaneously in immunocompromised mice.

Aim: To determine if the patterns of genetic dependencies elucidated through in vitro CRISPR-based functional genomic studies are recapitulated when human MM cells are grown in mice within BM-like scaffolds with "humanized" mesenchymal stromal compartment.

Methods: Cas9+ MM cell lines KMS11 and XG7 were transduced with a library of 1372 single-guide RNAs (sgRNAs) targeting 184 genes of interest (4 sgRNAs/gene), including 89 genes preferentially essential for MM cell lines compared to other neoplasms in vitro; broad-spectrum oncogenic targets (e.g. KRAS, MYC); tumor suppressor genes (e.g. PTEN); genes with limited in vitro essentiality in MM cells, but significantly higher expression vs. non-MM lines (e.g. ZFP91, ZBP1); and 155 olfactory receptor (OR) genes (typically not expressed or biologically inactive in tumor cells) as "DNA cutting" control sgRNAs. MM cells transduced with this focused CRISPR knockout (KO) library were injected into "humanized" scaffolds implanted subcutaneously (s.c.) in NSG mice. Tumors were collected when mice reached criteria for euthanasia and next-generation sequencing quantified the changes in sgRNA distribution at the end vs. start of experiment.

Results: A large majority of genes defined by in vitro CRISPR KO screens as MM-preferential dependencies were also essential for MM cells in BM-like scaffolds. Among 57 MM-preferential dependencies in vitro which were plausible dependencies for KMS11 cells (e.g. CERES scores <-0.4), 50 genes were essential for KMS11 cells in BM-like scaffolds in vivo (average log₂fold change<-1.0 and depletion of 3+ of 4 sgRNAs relative to the 99% confidence interval for control sgRNAs). These genes included key transcription factors/cofactors (e.g. IRF4, PRDM1, POU2AF1, RELB, MAF); epigenetic regulators (e.g. CARM1, MTA2, DOT1L); kinases upstream of NFkappaB (CHUK, IKBKB); ER regulators (e.g. SYVN1). Most "core-essential" or broad-spectrum oncogenic dependencies (e.g. MYC, CFLAR, CDK7 on both lines; KRAS in XG7) of this sgRNA library remained essential in vivo; while PTEN KO cells were enriched. Notably, several genes had more pronounced essentiality in vivo vs. invitro (e.g. BCL2, PIM2); or converted from non-essential in vitro to essential in vivo. For instance, among 95 genes of this library which are not likely dependencies in vitro (CERES scores >-0.4) for KMS11 cells, 29 genes exhibited in vivo essentiality for both KMS11 and XG7 cells: several of these latter "in vivo dependencies" are recurrently essential for other MM lines in vitro (e.g. ZBTB38, UBE2J1, TCF3, DNAJB11), while also others have limited if any in vitro essentiality (e.g. ZBP1, PYGO2) across MM despite significantly higher transcript levels vs. other neoplasias. Notably, several genes with increased essentiality in the "humanized" BM scaffolds vs. in vitro also had stronger in vivo dependency in the BM scaffolds vs. when growth of the same MM cells as s.c. plasmacytomas (e.g. BCL2, PIM2, UBE2J1, SYVN1, ALG9, AMFR). Co-culture with BMSCs or IL-6 treatment induces several of these transcripts in MM cells suggesting that increased dependency of MM on these genes in the "humanized" BM model is due at least partly to its human cytokines.

Conclusions: This study provides evidence for significant impact of the "humanized" BM-like niche on the patterns of genetic dependencies for human MM cells. Most genes preferentially essential for MM cells in vitro remain essential for their in vivo growth in the "humanized" BM model. Notably, several genes that do not meet criteria for dependency in vitro show such metrics in "humanized" BM scaffolds, but not sc plasmacytomas. These observations highlight important implications of the "humanized" BM-like in vivo model for current and future efforts to define and prioritize therapeutic targets for MM.