Lineage plasticity is an important, and likely underappreciated, mechanism of treatment resistance in lung cancer. Here, Quintanal-Villalonga and colleagues integrate results from multiomic analyses to provide key new insights into the biology of lineage plasticity.
See related article by Quintanal-Villalonga et al., p. 3028.
Lineage Plasticity in Lung Cancer
Despite advancements in tumor genomics and molecular diagnostics, histology remains an important framework for therapeutic selection and prognosis in lung cancers. However, tumor histology is not always “clear-cut”—mixed histology tumors have been recognized for decades, and lineage transformation following lung cancer–directed therapy has been described since 2006 (1). Although the initial case of lung adenocarcinoma (LUAD) transformed to small cell lung cancer (T-SCLC) was described in an EGFR-mutant LUAD, histologic transformation has since been identified in other molecular contexts following multiple different tyrosine kinase inhibitors (TKI) and immune checkpoint inhibitors (2, 3).
The umbrella of lineage plasticity also extends beyond the LUAD–SCLC continuum and can be seen in the form of mixed adenosquamous histology and LUAD to lung squamous cell carcinoma (LUSC) transformation at the time of treatment resistance. Neuroendocrine (NE) transformation is also well documented in the prostate cancer literature, and transformed prostate cancers share common molecular features with T-SCLC (4). Despite this growing awareness of lineage plasticity in cancer, a full understanding of its underlying biology remains elusive. What is clear, however, is that the process of histologic transformation is far from linear. In this report, Quintanal-Villalonga and colleagues (5) analyze a cohort of lung tumors of mixed histology (LUAD/SCLC) and pre-/post-SCLC transformation, shedding light on the nuances of lineage plasticity along the NE versus non-NE spectrum.
Predisposition to SCLC Transformation
What constitutes a lung cancer that will undergo histologic transformation? One possibility is that NE histology exists at baseline in these “pretransformed” tumors, either as NE cell clones not microscopically detected or as more prominent areas of mixed histology that are missed due to sampling error in the biopsy specimen. The latter hypothesis is supported by the fact that mixed histology tumors are indeed identified at diagnosis, especially in surgical resection specimens. For example, in a series of 100 surgically resected cases, combined non–small cell lung cancer (NSCLC)/SCLC histology was found in 28% of cases with all subtypes of NSCLC represented (6). Importantly, in this report by Quintanal-Villalonga and colleagues (5), the authors establish clonality between the two histologic components in their transformed LUAD (T-LUAD)/T-SCLC samples, demonstrating shared genomic and epigenomic features. This clonality is critical for proposing that true lineage plasticity is occurring. However, as the authors state, it is important to note that directionality cannot be assumed.
Genomic characterization of transformed specimens reveals TP53/Rb1 alterations in most T-LUAD/T-SCLC (Rb1 mutation/loss in 79%; TP53 mutation in 93% of transformed tumors) and evidence for whole-genome duplication (WGD) events before transformation, consistent with our expectations based on previously published literature (7). Adding to this growing body of literature, the authors postulate that chromosomal 3p loss may represent a novel risk factor for NE transformation, after identifying this variant in 85% of LUAD components of mixed histology tumors. Loss of 3p is also associated with LUSC, suggesting that 3p loss may broadly be indicative of a lineage plasticity–enabled genotype. However, recent clinical data revealed that no 3p loss events were detected in a series of EGFR-mutant LUAD-to-LUSC transformation cases (8). The gene(s) resident on 3p that may play a causal role in lineage plasticity therefore remains unclear.
Complementing these genomic analyses, transcriptomic and methylation analyses of T-LUAD/T-SCLC specimens compared with control LUAD and de novo SCLC specimens revealed that T-LUAD and T-SCLC represent intermediate states that retain gene expression patterns of their respective histologies but also contain substantive overlap with each other. These findings further support the hypothesis that T-LUAD/T-SCLC samples represent adjacent phenotypes on the continuum of lineage plasticity and contain genomic and epigenomic features that may be “priming” them for histologic transformation (Fig. 1).
Mechanism of Transformation
Once a cell is “primed” for histologic transformation, how does this transformation actually occur? Previous studies of samples from EGFR-mutant T-SCLC demonstrated that clonal divergence of T-SCLC from its LUAD precursor occurs prior to EGFR-targeted therapy and that the APOBEC-induced mutation signature is frequently seen in the process of SCLC transformation (7, 9). Aligned with these studies, 5 of 11 of the mixed histology tumors showed the APOBEC signature in the mutation profile of the LUAD portion of the sample. Although APOBEC-mediated DNA hypermutation is not a unifying mechanism for all SCLC transformation, its recurrent identification in analyses of transformed tumors likely signifies biologic importance, at least in a subset of tumors. Building on these genomic analyses, the current studies further advance the field by demonstrating a clear evolution of transcriptional and methylation profiles between T-LUAD/T-SCLC. Furthermore, genes associated with the PRC2 complex and stemness, WNT, PI3K/AKT, and NOTCH signaling, as well as immunosuppressive gene signatures are expressed at higher levels in T-SCLC compared with T-LUAD. These data support the leading hypothesis in the field that SCLC transformation is driven by transcriptional reprogramming rather than by acquisition of new mutations.
In sum, although we still lack a fundamental mechanistic understanding of lineage plasticity and SCLC transformation in lung cancer, the addition of these data to previously published work allows us to conclude a few baseline assumptions for future studies (Fig. 1). First, TP53/Rb1 deficits, WGD, and 3p loss are all genomic findings associated with increased risk for SCLC transformation; however, the process of NE transformation itself is primarily driven by transcriptional reprogramming. The biologic link between these findings remains to be defined, and APOBEC-mediated hypermutation may play a role in lineage plasticity in some tumors. Second, T-SCLC demonstrates an NE phenotype, with acquired expression of NE markers such as INSM1 and SYP compared with T-LUAD counterparts. Relevant to recent advances in de novo SCLC subtyping by expression of lineage-defining transcription factors, T-SCLC can adopt all major subtypes of de novo SCLC (including the POU2F3-expressing SCLC-P subtype, suggesting a tuft-cell independent origin of SCLC-P). Finally, in the case of T-SCLC with a known oncogenic driver such as an EGFR mutation, transformed cells have uncoupled founder oncogenic signaling from key dependencies mediating tumor growth in the transformed state.
Clinical Approach to Transformed Lung Cancer
How do we treat patients whose tumors have undergone SCLC transformation? Standard of care for T-SCLC tumors is similar to treatment regimens for de novo SCLC, typically involving platinum/etoposide-based regimens. Retrospective analyses of patients with EGFR-mutant T-SCLC confirmed highest response rates to platinum/etoposide and taxane-based chemotherapy regimens and demonstrated limited utility of immune checkpoint inhibitors (10). Ongoing EGFR TKI therapy is typically recommended in cases of SCLC transformation from EGFR-mutant NSCLCs due to presumed intratumoral heterogeneity and some element of ongoing dependence on EGFR signaling. However, we lack prospective data to guide us on these clinical decisions, and the dynamic nature of transformed or mixed histology tumors in the face of varying treatments is likely underestimated by our current radiographic and static tissue biopsy monitoring of patients.
Encouragingly, the clinical trials landscape for de novo SCLC is rapidly expanding, with embedded hope for much-needed progress in survival outcomes. In the absence of widespread available clinical trial options for patients with T-SCLC, opportunities for enrollment on de novo SCLC trials, given the known biologic similarity (despite important differences regarding smoking exposure and other environmental factors), should be considered. As SCLC treatments improve, a deeper understanding of the similarities and differences between de novo SCLC and T-SCLC will become increasingly important. Another important effort moving forward will be developing clinical trial options specifically for patients with T-SCLC where possible, with the acknowledgment that small patient numbers may require multisite trials with innovative design schemas. One example of a potential target to explore in these tumors is AKT signaling, based on compelling preclinical data from Quintanal-Villalonga and colleagues (5) in this article (Fig. 1).
What is next for T-SCLC? We have already highlighted the need for more clinical trial options for patients, discussed above. In addition, this study and surrounding discussion of the data highlight the need to prioritize ongoing collection of biopsy samples from patients at the time of treatment resistance, in order to both identify and learn more about the occurrence of histologic transformation in lung cancer more broadly (Fig. 1). These ongoing efforts for tissue collection will allow more precise characterization of the biologic vulnerabilities associated with risk for transformation. In the future, we also envision a role for liquid biopsy–based plasma markers for histologic transformation, a phenomenon that currently evades our standard-of-care plasma DNA testing for known genomic mechanisms of treatment resistance.
Finally, in both T-SCLC and de novo SCLC, it is clear that we need to continue to move beyond single-gene markers to a broader understanding of genome-wide processes that are at play. In the data presented by these authors, we are able to appreciate the spectrum of changes during lineage transformation on the DNA, RNA, and protein levels, but what data are we still missing? Emerging platforms such as digital spatial profiling can be used moving forward to better understand not only the inner workings of the tumor cells themselves but also the details of how the location of each histologic compartment correlates with the immune infiltrate and the surrounding vasculature. The closer we are able to look, the more clearly we can see that lung tumors exist along a dynamic spatiotemporal spectrum of genomic, epigenomic, and histologic states that are likely driven by tumor-intrinsic as well as environmental factors. Our ability to develop more informed, targeted therapies for patients is reliant on our improved understanding of this biology.
C.M. Lovly reports grants from NIH/NCI during the conduct of the study, grants from Xcovery, and Novartis; other support from Pfizer, Roche, Genentech, AstraZeneca, Takeda, Blueprints Medicine, Cepheid, Foundation Medicine, Eli Lilly/Loxo, Amgen, Daiichi-Sankyo, D2G, and other support from Syros outside the submitted work. No disclosures were reported by the other author.