Many roads lead to resistance: a multimodal genomic approach to identify drug resistance mechanisms in leukemia.

We use an integrative genomic strategy interrogating six different types of genomic/epigenetic variation in primary leukemia cells from newly diagnosed patients to reveal glucocorticoid resistance genes, coupled with genomewide CRISPR screening to prioritize genes for downstream mechanistic studies.
Many roads lead to resistance: a multimodal genomic approach to identify drug resistance mechanisms in leukemia.

Many clinical variables influence the behavior of medications in patients, some environmental and some genetic in nature (Figure 1). Our research focuses on discovering variables with significant influence on antileukemic agents in children with acute lymphoblastic leukemia (ALL), a common disseminated malignancy  that can only be cured with effective systemic chemotherapy. Cure rates in children with ALL have improved from ~20% in the 1960s to ~90% today, yet more children die of ALL than any other cancer, and the cure rate in adults is much lower (~70%).

Figure 1. Variables that can influence the effects of medications in patients (adapted with permission from Vessell)

The major cause of ALL treatment failure is drug resistance, and genetics can have predominant effects on this phenotype. Our recent work has focused on discovery of somatic variants that cause leukemia cells to be resistant to chemotherapy. We initially used a candidate gene strategy2 looking for variants in genes thought to alter the pharmacology of antileukemic agents.  But our first genomewide study  of gene expression in ALL cells3 revealed that >90% of genes associated with resistance to prednisolone, vincristine, asparaginase or daunorubicin had not been previously linked to drug resistance in any human tumor, highlighting the limitations of candidate gene strategies.

As genomic technology advanced, we expanded our scope to interrogate genomic and epigenetic differences in primary leukemia cells that are resistant to chemotherapy4. In the current study, we report the integrated analyses of six types of genome variants (mRNA and miRNA expression, SNPs, copy number variants [CNV], DNA methylation, and coding variants from whole exome sequencing) and their role in resistance to glucocorticoids, an essential component of contemporary treatment of ALL.

In collaboration with our biostatistics colleagues, we developed a novel, gene-centric method, TAP (Truncated Aggregation of P-values), that integrates multiple types of genomic variation to assess the overall significance of each gene’s association with drug resistance. This identified 118 genes that were significant by both the integrative polygenomic analyses and the TAP statistic.

Genomewide CRISPR-knockout screening in a human ALL cell line served as an orthogonal method to identify and prioritize glucocorticoid resistance genes  revealing 15 genes that were significant by all three methods (i.e., polygenomic analysis, TAP and CRISPR). Interestingly, only one of these 15 genes had been previously linked to glucocorticoid resistance (NLRP3)4. Had we interrogated only mRNAs we would have found only 53% of these genes, CpG-methylation alone revealed only 33%, whereas interrogating mRNA, miRNA, CpG-me, SNV and SNP identified all 15 genes.

To assess the robustness of our approach, we determined the extent to which we had “rediscovered” known glucocorticoid resistance genes in ALL. This revealed that we had directly rediscover 30 of 38 known genes (79%). When we assessed whether we had found other genes in the same pathway, our approach identified 100% of the 38 pathways.

The top novel under-expressed gene significant by all three methods was CELSR2, a G-protein coupled receptor involved in non-canonical WNT signaling. Glucocorticoid resistance was recapitulated in human ALL cell lines by knocking down CELSR2 expression, and low CELSR2 expression was associated with glucocorticoid resistance in a second validation cohort of patients. Single-cell RNA sequencing documented the clonality of low CELSR2 expression in glucocorticoid resistant patients, and network-based transcriptomic modeling (NetBID) further identified CELSR2 as a key regulator in glucocorticoid resistant ALL.

To elucidate the mechanism by which low CELSR2 expression confers resistance, we identified differentially expressed genes in CELSR2-knockdown ALL cells compared to controls. We observed marked down-regulation of the glucocorticoid receptor (NR3C1) sufficient to alter transcriptional response to glucocorticoids. After glucocorticoid treatment (24H), the expression of the antiapoptotic protein BCL2 was significantly up-regulated in CELSR2-knockdown cells, suggesting the potential mitigation of this mechanism of glucocorticoid resistance via the co-administration of venetoclax, a BCL2 inhibitor. Indeed, venetoclax was highly synergistic with glucocorticoids in ALL cells with low CELSR2 expression, and this drug combination was more effective in treating NSG mice with ALL expressing low levels of CELSR2.

We found low CELSR2 expression in approximately 50% of glucocorticoid-resistant ALL in both children and adults. Since ex vivo resistance to glucocorticoids has been linked to inferior event free survival in childhood ALL5 and was associated with the persistence of residual leukemia cells (MRD) after remission induction therapy in our patients, these findings point to the potential utility of combining venetoclax with contemporary remission induction therapy to mitigate glucocorticoid resistance and improve the efficacy of ALL chemotherapy in this subset of patients. In summary, our study illustrates the power of a broad genomewide strategy of interrogating multiple forms of genome variation in cancer cells to discover new mechanisms of drug resistance and therapeutic options to mitigate, an approach that should have utility in many types of human cancer.

Figure 2. The challenge of integrating multiple types of genome variants causing drug resistance

Co-written by: Robert J. Autry and William E. Evans


1              Vesell, E. S. Pharmacogenetic perspectives gained from twin and family studies. Pharmacol Ther 41, 535-552, doi:10.1016/0163-7258(89)90130-7 (1989).

2              Evans, W. E. & Relling, M. V. Pharmacogenomics: translating functional genomics into rational therapeutics. Science 286, 487-491, doi:10.1126/science.286.5439.487 (1999).

3              Holleman, A. et al. Gene-expression patterns in drug-resistant acute lymphoblastic leukemia cells and response to treatment. The New England journal of medicine 351, 533-542, doi:10.1056/NEJMoa033513 (2004).

4              Paugh, S. W. et al. NALP3 inflammasome upregulation and CASP1 cleavage of the glucocorticoid receptor cause glucocorticoid resistance in leukemia cells. Nat Genet 47, 607-614, doi:10.1038/ng.3283 (2015).

5              Pieters, R. et al. Relation of cellular drug resistance to long-term clinical outcome in childhood acute lymphoblastic leukaemia. Lancet 338, 399-403 (1991).

Please sign in or register for FREE

If you are a registered user on Research Communities by Springer Nature, please sign in

Subscribe to the Topic

Cancer Biology
Life Sciences > Biological Sciences > Cancer Biology
  • Nature Cancer Nature Cancer

    This journal aims to provide a unique forum through which the cancer community will learn about the latest, most significant cancer-related advances across the life, physical, applied and social sciences.