Epigenetic modulation reveals differentiation state specificity of oncogene addiction

Abstract

Hyperactivation of the MAPK signaling pathway motivates the clinical use of MAPK inhibitors for BRAF-mutant melanomas. Heterogeneity in differentiation state due to epigenetic plasticity, however, results in cell-to-cell variability in the state of MAPK dependency, diminishing the efficacy of MAPK inhibitors. To identify key regulators of such variability, we screen 276 epigenetic-modifying compounds, individually or combined with MAPK inhibitors, across genetically diverse and isogenic populations of melanoma cells. Following single-cell analysis and multivariate modeling, we identify three classes of epigenetic inhibitors that target distinct epigenetic states associated with either one of the lysine-specific histone demethylases Kdm1a or Kdm4b, or BET bromodomain proteins. While melanocytes remain insensitive, the anti-tumor efficacy of each inhibitor is predicted based on melanoma cells’ differentiation state and MAPK activity. Our systems pharmacology approach highlights a path toward identifying actionable epigenetic factors that extend the BRAF oncogene addiction paradigm on the basis of tumor cell differentiation state.

AP-1 transcription factor network explains diverse patterns of cellular plasticity in melanoma cells

“…Treatment with MAPK inhibitors, for example, may induce the dedifferentiation of melanocytic cells toward the neural crest-like phenotype. Such adaptive phenotype switching occurs as early as 2-3 days of exposure to MAPK inhibitors, concomitantly with loss of MITF and acquisition of NGFR (Khaliq et al, 2021). In addition to drug-induced dedifferentiation, adaptive reactivation of the ERK pathway following a transient period of ERK inhibition is known as a common mechanism of adaptive resistance that helps tumor cells escape the effect of drug (Gerosa et al, 2020; Lito et al, 2012).…”

“…To classify the differentiation state of cells based on image-based protein measurements, we generated histograms of single-cell data on each of the previously validated melanoma differentiation state markers (MITF, SOX10, NGFR and AXL) (Khaliq et al, 2021; Tsoi et al, 2018). For each protein (X), we identified an appropriate binary gate, based on which individual melanoma cells were divided into two groups of X High and X Low cells.…”

“…The measurements included total levels of eleven AP-1 transcription factors (including cFOS, FRA1, FRA2, cJUN, JUNB, JUND, ATF2, ATF3, ATF4, ATF5 and ATF6), six AP-1 phosphorylation states (including p-cFOS S32 , p-FRA1 S265 , p-cJUN S73 , p-ATF1 S63 , p-ATF2 T71 and p-ATF4 S245 ), and four differentiation state markers MITF, SOX10, NGFR and AXL. Importantly, these four differentiation state markers were previously reported to represent transcriptionally distinct melanoma differentiation states (Khaliq et al, 2021;Tsoi et al, 2018). The panel of 19 melanoma cell lines tested represented a broad spectrum of differentiation states, including populations of melanocytic (MITF High /SOX10 High /NGFR Low /AXL Low ), transitory (MITF High /SOX10 High /NGFR High /AXL Low ), neural crest-like (MITF Low /SOX10 High /NGFR High /AXL High ) and undifferentiated (MITF Low /SOX10 Low /NGFR Low /AXL High ) cells (Figures 1C, S1A).…”

“…An example of cell-to-cell transcriptional heterogeneity with phenotypic consequences for therapy resistance is observed in melanomas (Emert et al, 2021;Fallahi Sichani et al, 2017;Shaffer et al, 2017). Numerous studies have associated fluctuations in the state of MAPK inhibitor sensitivity across BRAF-mutant melanoma cells to intrinsic variations in their differentiation state (Baron et al, 2020;Belote et al, 2021;Khaliq et al, 2021;Rambow et al, 2018;Tsoi et al, 2018;Wouters et al, 2020).…”

Epigenetic Regulation in Melanoma: Facts and Hopes

“…Recently, a library of small molecules acting as epigenetic modulators was used to identify the regulators of the abovementioned plasticity in BRAF-mutant melanoma cell lines in order to explain cell-to-cell variability, despite MAPK dependency. Results from this work have allowed differentiation among three further states: a lysine demethylase 1A (KDM1A)-dependent state, that could be efficiently inhibited by SP2509, a reversible KDM1A inhibitor mainly found in undifferentiated cells; a lysine demethylase 4B (KDM4B)-dependent state that is sensitive to JIB-04, a pan-inhibitor of Jumonji histone demethylases observed in neural crest-like cells; and a state induced by birabresib, which is a BET bromodomain inhibitor [113].…”