Unperturbed dormancy recording reveals stochastic awakening strategies in endocrine treated breast cancer cells

Abstract

Hormone dependent breast cancer (HDBC) is the most commonly diagnosed tumor type in women. Adjuvant endocrine therapies (ET) have been the cornerstone in the clinical management of HDBC patients for over forty years. A vast proportion of HDBC patients incur long periods of clinical dormancy following ET, with tumour awakening appearing at a steady pace for up to 25 years (Pan et al., 2017). Extensive genomic studies have demonstrated that 15-30% of clinical relapses develop recurrent genomic changes which contribute to drug resistance (i.e. ESR1 activating mutations) (Bertucci et al., 2019; Magnani et al., 2017; Razavi et al., 2018). However, even in these cases, there is no conclusive evidence around the pre-existence vs. de novo nature of these events. We previously showed that ETs can trigger and select for dormancy in subpopulations of breast cancer (Hong et al., 2019). In this work we took two novel approaches to investigate the dormancy and awakening roadmap of HDBC cells at unprecedented detail. Firstly, we leveraged a rare cohort of n=5 patients which were treated with primary adjuvant ETs in the absence of surgery (TRACING-HT) to dissect the contribution of genomic aberrations to tumor awakening. Next, we developed a first of its kind evolutionary study in vitro to systematically annotate cancer cells adaptive strategies at single cell level in unperturbed systems during a period of several months (TRADITIOM). Collectively our data suggest that ETs steer HDBC cells into an inherently unstable dormant state. Over time, routes to awakening emerge sporadically and spontaneously in single lineages. Each dormant cell retains an intrinsic awakening probability which we propose is a function of epigenetic decay. Awakening occurs without an external trigger and involves multiple apparent endpoint phenotypes that cannot be fully explained by conventional Darwinian genetic selection processes. Finally, our data show that common genetic hits associated with resistance happen downstream of awakening. Overall, our data have uncovered previously unsuspected roles for stochastic nongenetic events during dormancy with profound clinical implications.

Genetic and epigenetic driven variation in regulatory regions activity contribute to adaptation and evolution under endocrine treatment

“…Systematic integration of these experimental and computational strategies led to the conclusion that while CREs do not display the strong signature associated with coding drivers, changes in the context-specific regulatory activity of a defined set of CREs plays a crucial role during therapy-induced dormancy. Our results stand out considering the stochastic processes dominating dormancy entrance and exit (see companion manuscript 12 ). For example, our SIDP screens strongly suggest that signalling converging on NF-kB activation plays a central role in maintaining long-term dormancy.…”

“…We have shown that bona fide coding drivers (i.e., ESR1 mutations) might not be the actual cause triggering the exit from dormancy as they could emerge and be selected for after awakening, owing to the increased mutational burden associated with replication 12 . We then reasoned that the activity of specific CREs might contribute to the adaptive process occurring during the transition from growth to dormancy 1 entrance 13,44 . 2 To investigate this hypothesis, we performed SIDP in long-term oestrogen deprived conditions (-E2), measuring gRNA frequencies at 7, 14, 21 and 60 days after infection (Figure 2a).…”

“…In a synergistic lineage tracing study (TRADITIOM, see accompanying manuscript), we show that entrance into dormancy is largely stochastic, with persister dormant lineages selected by chance each time, leading to a significant divergence between replicates 12 . To test if this process also influences the readout of SIDP, we tracked lineages leveraging the non-targeting sgRNAs (n = 501) for up to 60 days of hormone deprivation (full dormancy 12 ). Surprisingly, 210/501 non-targeting sgRNAs (42%, compared to 0.9% in SIDP +E2) showed apparent non-neutral expansion or exhaustion at day 60 (Figure 2c).…”

“…However, the interpretation of the genomic data is difficult due to the stochastic processes influencing individual lineages during dormancy entrance (Figure 2c-d and 12 ). For instance, CREs loss of function conferring fitness advantage under treatment (i.e., TLR/NF-kB) could be explained by three alternative scenarios: increased plasticity (a larger subset of lineages become persister), early awakening and clonal expansion 12 or complete dormancy bypass (Figure 3a). To test these hypotheses, we tracked the behaviour of cells carrying individual sgRNAs (GFP-NLS) mixed with non-targeting controls during dormancy entrance with live-cell imaging or FACS (Figure 3a).…”

“…Despite HDBC cells being largely dependent on the activity of these TFs, previous perturbation screens focusing on ERa or FOXA1 bound CREs found that only a minority of binding sites appear to be essential for steady-state proliferation in vitro 28,29 . Yet, TF-centric perturbation has missed CREs driven by additional TFs (i.e., YY1 and GATA3 [30][31][32] ) and overlooked critical intermediate states in cancer evolution such as adaptive dormancy of persister cells 12,13 . To identify CREs contributing to the evolution and adaptation of HDBC tumours exposed to endocrine therapies we developed a prioritised CREs panel (termed Systematic Identification of epigenetically Defined loci, or SID) to investigate the role they play both in vitro and in vivo.…”