Abstract
Small cell lung cancer (SCLC) has limited therapeutic options and an exceptionally poor prognosis. Understanding the oncogenic drivers of SCLC may help define novel therapeutic targets. Recurrent genomic rearrangements have been identified in SCLC, most notably an in-frame gene fusion between RLF and MYCL found in up to 7% of the predominant ASCL1-expressing subtype. To explore the role of this fusion in oncogenesis and tumor progression, we used CRISPR/Cas9 somatic editing to generate a Rlf–Mycl-driven mouse model of SCLC. RLF–MYCL fusion accelerated transformation and proliferation of murine SCLC and increased metastatic dissemination and the diversity of metastatic sites. Tumors from the RLF–MYCL genetically engineered mouse model displayed gene expression similarities with human RLF–MYCL SCLC. Together, our studies support RLF–MYCL as the first demonstrated fusion oncogenic driver in SCLC and provide a new preclinical mouse model for the study of this subtype of SCLC.
The biological and therapeutic implications of gene fusions in SCLC, an aggressive metastatic lung cancer, are unknown. Our study investigates the functional significance of the in-frame RLF–MYCL gene fusion by developing a Rlf–Mycl-driven genetically engineered mouse model and defining the impact on tumor growth and metastasis.
This article is highlighted in the In This Issue feature, p. 2945
Introduction
Small cell lung cancer (SCLC) is an exceptionally aggressive malignancy that comprises 13% of all lung cancer cases and causes an estimated 250,000 deaths globally per year (1, 2). Most patients have metastatic disease at the time of diagnosis. Although SCLC tumors are initially sensitive to standard cytotoxics, nearly all patients develop recurrent and chemoresistant disease, leading to a median survival of slightly more than 1 year (1). The recent addition of immune checkpoint inhibitors to standard chemotherapy has led to durable responses in a small minority of patients but has extended median survival by only 2 months (3), underscoring the need to deepen our understanding of this disease in pursuit of more effective treatment strategies.
SCLC is recognized histologically by a characteristic cellular morphology and frequently expresses markers of neuroendocrine differentiation (4). Data from both human tumors and mouse models of SCLC have defined distinct subtypes of SCLC based on differential expression of key transcriptional regulators, including achaete-scute homolog 1 (ASCL1), neurogenic differentiation factor 1 (NEUROD1), POU class 2 homeobox 3 (POU2F3), and yes-associated protein 1 (YAP1; ref. 5). These subtypes have been referred to respectively as SCLC-A, SCLC-N, SCLC-P, and SCLC-Y. Canonical “classic” and “variant” subtypes of SCLC are associated with upregulation of ASCL1 and NEUROD1, respectively (6, 7). Common genetic hallmarks of SCLC include near-universal biallelic loss of tumor suppressors TP53 and RB1, amplification of MYC family genes, and inactivating mutations of NOTCH genes (8, 9). Recent work has advanced our understanding of how these molecular features may define subtype- or genotype-specific therapeutic vulnerabilities (2, 7, 10, 11). Due in part to a paucity of primary tumor samples, such studies have relied heavily on established cell lines and genetically engineered mouse models (GEMM), the latter of which recapitulate many of the key mutations, markers, and metastatic patterns observed in human SCLC (7, 12–17).
A number of GEMMs have been developed to model SCLC biology. Most include Rb1 and Trp53 alleles flanked by loxP sites, with lung tumor growth initiated through intranasal or intratracheal delivery of adenoviruses expressing Cre recombinase (17). Although concomitant overexpression of MYCL has been shown to accelerate tumor growth in this context (18), long latency of these models has spurred the adoption of faster growing models generated through the additional deletion of Rbl2 or Pten or stabilization of MYC (7, 13–16). These and other double- and triple-knockout models have been used to interrogate SCLC tumorigenesis, providing insight to putative cell types of origin and defining a requirement for Ascl1 in SCLC tumorigenesis (6, 19, 20). A stabilized and overexpressed MYC allele elicits a rapidly growing, metastatic phenotype and can drive tumor evolution from Ascl1hi to Neurod1hi and then to Yap1hi SCLC (7, 21), suggesting distinct roles for these transcriptional regulators in tumor initiation and progression.
GEMMs have provided insight into the molecular underpinnings of SCLC metastasis. Multiple studies identified the neural transcription factor NFIB as a putative metastatic driver across different mouse models (20, 22, 23). Nfib overexpression and/or amplification accelerates oncogenesis in the context of Mycl amplification and enhances metastasis in GEMMs; expression of the human ortholog NFIB correlates with poorer prognosis in patients (12, 22, 23). Induction of Cre under different lung epithelial lineage-specific promotors has suggested that NFIB-driven metastasis is lineage- restricted, such that tumors that arise from distinct cells of origin metastasize either dependently or independently of NFIB (24). Clonal heterogeneity is likely to influence the modes of metastasis available to SCLC tumors (25). Continued delineation of the drivers of metastasis in SCLC may reveal distinct therapeutic opportunities.
Chromosomal translocations, insertions, and deletions resulting in gene fusions represent a common pathway of oncogenesis in solid tumors (26). Although SCLC demonstrates extensive genomic structural aberrancies, the role of gene fusions in driving SCLC development and metastasis has not been defined. Several intrachromosomal rearrangements have been identified in SCLC cell lines and tumors, most notably a recurrent in-frame gene fusion between RLF and MYCL (8, 27, 28). MYCL expression is associated with SCLC-A (6) and has been shown to support tumorigenesis and proliferation (18, 29). However, the pathologic significance of RLF–MYCL fusions in SCLC remains unknown. The functional interrogation of fusion genes as oncogenic drivers in SCLC GEMMs has been challenging due to the lack of tractable somatic engineering strategies. However, with the development of CRISPR/Cas9 methods, it is now possible to engineer complex chromosomal rearrangements in vivo (30, 31). Here, we report the first gene fusion GEMM of SCLC and demonstrate that RLF–MYCL fusion promotes tumorigenesis and enhances metastasis in SCLC.
Results
RLF–MYCL Is a Recurrent Gene Fusion in SCLC with High MYCL Expression
To assess the frequency of the RLF–MYCL gene fusion in human SCLC, we first analyzed RNA-sequencing (RNA-seq) data from 105 SCLC samples (8, 9) using the Arriba and FusionCatcher fusion transcript discovery algorithms (https://github.com/suhrig/arriba; ref. 32). We found that in-frame RLF–MYCL gene fusions had the highest frequency of fusion reads, observing recurrent fusion events in 4.8% of all SCLC cases (5/105) and 6.9% of those classified as SCLC-A (5/72). RLF and MYCL are encoded on opposing DNA strands in close proximity on chromosome 1p, approximately 300 kb apart. RLF–MYCL fusion events in SCLC primary tumors and cell lines demonstrate multiple intronic breakpoints, all of which result in an inversion event that, at the transcript level, leads to splicing of the first exon 1 of RLF to the last two exons of MYCL (Fig. 1A). The resulting mature transcript encodes the first 79 amino acids of RLF, followed by all but the first 27 amino acids of the MYCL protein, generating a 446-residue fusion protein (Fig. 1B).
We confirmed the presence of the in-frame RLF–MYCL fusion transcript and protein by PCR and immunoblot analysis in five human SCLC cell lines: NCI-H889, CORL47, NCI-H1092, NCI-H1963, and NCI-H1836 (Fig. 1C; Supplementary Table S1). In three of the cell lines harboring the RLF–MYCL fusion, immunoblotting suggests that the fusion protein is present at a substantially higher level than endogenous MYCL in any other line examined. Analysis of RNA-seq data from published data sets (8, 9) confirmed that the identified SCLC tumors and cell lines with RLF–MYCL fusion events all belong to the SCLC-A (ASCL1hi) subtype (Supplementary Fig. S1). The RLF–MYCL fusion was associated with higher levels of detectable MYCL transcripts relative to other SCLC-A lines (Fig. 1D). Although the fusion protein could have neomophic function(s), these data suggest that the RLF–MYCL gene fusion might primarily contribute to the cancers in which it is present in part by increasing functional MYCL expression, thereby increasing MYCL-driven oncogenic signaling in SCLC. To determine whether this fusion has a functional role in oncogenesis and tumor progression, we investigated the effects of this fusion event in mouse models of SCLC.
Rlf–Mycl Fusion Promotes Tumorigenesis of Preneoplastic Neuroendocrine Cells
In the mouse genome, Rlf and Mycl are located on chromosome 4 (qD2.2), in a region that is syntenic to human chromosome 1 (p34.2; Fig. 2A). The proximity and opposing orientation of the two genes in both mouse and human enable modeling the human translocation event in the mouse genome by Cas9-mediated editing: in both species, this fusion requires an inversion event without loss of intervening genetic material. We attempted to induce Rlf–Mycl fusion in mouse cells through the use of single-guide RNAs (sgRNA) promoting Cas9-mediated double-strand DNA breaks in the first introns of Rlf and Mycl (Fig. 2B). Plasmids expressing individual sgRNAs and Cas9 (ref. 33; Fig. 2B) were cotransfected into NIH/3T3 cells, and editing was confirmed by surveyor assays (Supplementary Fig. S2A), followed by PCR and immunoblot analyses demonstrating the presence of the demonstrating the presence of the Rlf–MYCL fusion (Fig. 2C). The presence of the desired Rlf–Mycl inversion was confirmed by sequencing the corresponding Rlf–Mycl fusion genomic DNA and transcript (Fig. 2C).
We initially sought to assess the role of the RLF–MYCL fusion in early stages of tumorigenesis using derivatives from the Rb1fl/fl;Trp53fl/fl;Rbl2fl/fl (RPR2) transgenic model of SCLC (16). It has been previously demonstrated that isolated cells from early neoplastic lesions in the lungs of these animals that have not fully transformed to malignancy contain precursors of SCLC (“preSC cells”; genotype Rb1−/−;Trp53−/−;Rbl2−/+), in which the effects of putative oncogenic drivers can be assessed (29). We found that preSC cells in which we induced the Rlf–Mycl fusion by introduction of Cas9 with sgRNAs targeting intronic sequences of both Rlf and Mycl formed larger and more colonies in soft agar than controls transfected with either single sgRNA construct (Fig. 2D). To determine the effect of the fusion on tumor initiation in vivo, we injected preSC cells transfected with each one or both sgRNA constructs into the flanks of immunodeficient mice (n = 5/group). Animals injected with preSC cells transfected with both sgRNAs to induce Rlf–Mycl rearrangement developed tumors earlier and faster than tumors with either single sgRNA control (Supplementary Fig. S2B–S2D).
To more readily support subsequent in vivo transfection experiments, we modified a lentiviral vector (USEC) to drive expression of both Rlf and Mycl sgRNAs, or control sgRNAs, from tandem U6 promoters along with Cre recombinase (refs. 34, 35; Fig. 2E). To test this system, we cotransduced preSC cells with lentiviruses containing Cas9 and either USEC with sgRNAs targeting Rlf and Mycl (herein sgRlfsgMycl) or USEC with sgRNAs against neomycin (hereafter sgNeosgNeo) and injected these transduced preSC cells into the flanks of immunodeficient mice (n = 5 and 4, with sgNeosgNeo and sgRlfsgMycl, respectively). Mice injected with sgRlfsgMycl preSC cells developed larger tumors (Fig. 2F) and had shorter survival (Fig. 2G) compared with sgNeosgNeo controls (P = 0.007, log-rank test). Droplet PCR of cDNA confirmed the presence of the Rlf–Mycl fusion transcript in tumors, albeit with lower expression in tumors of viral-transduced preSC cells than achieved with vector-transfected preSC cells (Supplementary Fig. S2E). Hematoxylin and eosin (H&E) staining and synaptophysin (SYP) IHC showed typical SCLC characteristics in tumors (Supplementary Fig. S2F and S2G). Together, these data confirm that we can successfully engineer the RLF–MYCL fusion in mouse cells and that the Rlf–Mycl fusion accelerates oncogenic transformation and tumor growth in pre-SCs.
RLF–MYCL Endogenous Induction Accelerates SCLC Tumor Formation In Vivo
To further investigate the contribution of the Rlf–Mycl gene fusion in SCLC development in vivo, we employed an autochthonous SCLC model, that is, Rb1fl/fl;Trp53fl/fl;Rbl2fl/fl;Rosa26LSL-Cas9-GFP (hereafter RPR2C; ref. 31). Cre-mediated inactivation of Rb1, Trp53, and Rbl2 in this model has been previously reported to recapitulate morphologic characteristics and therapeutic vulnerabilities of human SCLC (16). Introduction of the LSL-Cas9-GFP cassette into the Rosa16 locus allows Cre-mediated induction of Cas9 expression together with deletion of these key tumor suppressors. First, we transduced a cohort of adult chimeric RPR2C mice via intratracheal instillation with USEC lentivirus expressing either sgRlfsgMycl or control sgNeosgNeo. Six months postinfection, we collected lungs for histologic analysis to assess tumor incidence and burden (Fig. 3A). Mice transduced with sgRlfsgMycl (n = 17) had nearly three times higher tumor burden (P = 0.012; Fig. 3B) and larger tumor areas (P = 0.041; Fig. 3C) than sgNeosgNeo controls (n = 20) estimated by quantitative histology (Fig. 3D). To analyze tumor progression over time with a higher potential penetrance of Rlf–Mycl rearrangement, we transduced a cohort of adult RPR2C mice by intratracheal delivery with 10 times higher viral titers of the USEC lentiviruses using both chimeric (n = 14 sgNeosgNeo/13 sgRlfsgMycl) and fully transgenic (n = 19 sgNeosgNeo/13 sgRlfsgMycl) mice and monitored these animals for development of respiratory distress or other adverse symptoms requiring euthanasia (Fig. 3E and F). We assessed tumor burden in living mice at 6 months following USEC induction by MRI. We detected significantly greater tumor volume in sgRlfsgMycl-transduced mice in both chimeric (P = 0.047) and fully transgenic cohorts (P = 0.016; Fig. 3G–J). A higher tumor burden was also evident upon histologic analysis of lungs of the sgRlfsgMycl as compared with sgNeosgNeo chimeric (P = 0.01) and fully transgenic cohorts (P = 0.03; Supplementary Fig. S3A–S3D). Overall survival of chimeric sgRlfsgMycl mice was significantly shorter than that of sgNeosgNeo mice (P = 0.001; Supplementary Fig. S3E). No significant difference in overall survival was observed in the fully transgenic cohort (P = 0.87; Supplementary Fig. S3F). Relative to sgRlfsgMycl mice, in sgNeosgNeo mice, we noted lower overall parenchymal tumor burden but also frequent occurrence of centrally located thoracic tumors leading to airway compromise, potentially explaining the similar duration of overall survival between these cohorts.
To evaluate whether a change in tumor spectrum might affect the acceleration of tumor progression in sgRlfsgMycl mice, histologic analyses were performed by a pathologist. In both sgRlfsgMycl cohorts, most tumors identified displayed typical histologic features of SCLC and stained positive for the neuroendocrine marker SYP (Supplementary Fig. S3G). Both 6-month and end-stage tumors showed a predominance of SCLC; in the chimeric cohort, we observed some cases of admixed SCLC with large-cell neuroendocrine carcinoma. The tumor architecture was classic and/or trabecular with lymphovascular invasion observed in most end-stage cases. We confirmed the presence of Rlf–Mycl fusion transcript by droplet PCR on cDNA from microdissected tumors from both chimeric (3/3) and fully transgenic (5/8) animals infected with sgRlfsgMycl (Supplementary Fig. S3H). Mycl1 transcripts by droplet PCR on cDNA generally matched Rlf–Mycl fusion transcript levels of Rlf–Mycl fusion mice (Supplementary Fig. S3I). Taken together, these data illustrate successful development of a novel SCLC GEMM harboring the Rlf–Mycl gene fusion and further support the oncogenic function of the RLF–MYCL fusion in SCLC.
RLF–MYCL Fusion Promotes Metastasis in SCLC
Human SCLC has a remarkable predilection for metastatic spread: approximately two thirds of patients have distant metastases evident at the time of diagnosis. In addition to analyzing the role of RLF–MYCL in promoting SCLC initiation and growth, we sought to clarify its potential contributions to metastasis. This was prompted in part by the observation that essentially all sgNeosgNeo-transduced mice ultimately required euthanasia due to respiratory distress from proximal airway compromise, whereas sgRlfsgMycl-transduced mice displayed a broader array of distress symptoms, including abdominal distension and lethargy. At necropsy, 93% of sgRlfsgMycl chimeric mice and 15% of sgNeosgNeo chimeric mice displayed overt metastasis (P < 0.0001; Fig. 4A; Supplementary S4A). A similar pattern was observed in the fully transgenic animals, in which 92% of sgRlfsgMycl mice displayed overt metastasis compared with 39% of sgNeosgNeo mice (P = 0.004; Fig. 4A). The organ distribution of observed metastases also differed between cohorts. Most notably, sgRlfsgMycl-transduced mice demonstrated widespread metastatic disease. Distant metastases to multiple organs in individual mice were observed in approximately 30% of chimeric and over 40% of fully transgenic sgRlfsgMycl mice (Fig. 4B), whereas only liver metastases were observed in sgNeosgNeo mice. In sgRlfsgMycl-transduced mice, other sites of metastasis included thoracic, paraspinal, and cervical lymph nodes; spleen; kidney; and mesentery (Fig. 4C; Supplementary S4A). Liver metastases of mice transduced with sgRlfsgMycl were substantially more extensive than those of control sgNeosgNeo mice, as assessed by both histologic sections and evaluation of total liver weight (P = 0.04; Fig. 4D and E). Histopathologic examination including staining for the neuroendocrine marker SYP confirmed SCLC phenotype in these metastases (Supplementary Fig. S4B). We also confirmed Rlf–Mycl fusion transcript by droplet PCR using cDNA from several metastatic nodules from both chimeric (three of three) and fully transgenic (five of six) sgRlfsgMycl-transduced animals (Supplementary Fig. S4C). To evaluate whether metastatic disease is an early event during SCLC tumor development in this Rlf–Mycl fusion model, we analyzed livers of chimeric RPR2C-transduced mice at an early time point 6 months after tumor initiation. We found that transduction with sgRlfsgMycl resulted in evident micrometastases to the liver in three of five animals versus zero of five control sgNeosgNeo-transduced mice (P = 0.038; Supplementary Fig. S4D and S4E). Collectively, these findings suggest that RLF–MYCL both accelerates tumorigenesis and promotes early metastatic dissemination in SCLC.
Rlf–Mycl SCLC Has an ASCL1hi Phenotype Consistent with SCLC-A Subtype
Human RLF–MYCL tumors and cell lines are all SCLC-A subtype (Supplementary Fig. S1). To determine the SCLC subtype of mouse tumors with Rlf–Mycl fusion, we compared lung tumors generated by sgNeosgNeo and sgRlfsgMycl with tumors from previously developed SCLC mouse models, including RPR2 mice and RPM (Rb1fl/fl;Trp53fl/fl;MycLSL/LSL) mice (7, 13, 16). Tumors in RPR2 mice demonstrate an exclusively ASCL1hi phenotype, whereas RPM mice develop a broader spectrum of phenotypes, including a predominance of NEUROD1hi tumors resembling SCLC-N and aggregates of POU2F3 and YAP1 expression. Tumors from both sgNeosgNeo- and sgRlfsgMycl-transduced animals demonstrated high expression of ASCL1 and a second neuroendocrine transcription factor INSM1; these findings were similar to and consistent with RLF–MYCL-positive human tumors (n = 5/group) and RPR2 mice (Fig. 5A and B; Supplementary Fig. S5A). In contrast to tumors from RPM mice, tumors from the sgNeosgNeo and sgRlfsgMycl models did not express detectable NEUROD1, POU2F3, or YAP1 by IHC (n = 5/group; P < 0.0001; Fig. 5C; Supplementary Fig. S5B–S5D). Human SCLC tumors have also been characterized into neuroendocrine (NE) and non-NE classes based on a 50-gene signature, identifying SCLC-A as NEhi tumors (36). Exploring expression across a panel of orthologous murine genes did not reveal consistent differences in NE state between sgRlfsgMycl and sgNeosgNeo lung tumors.
NFIB expression has been reported in multiple mouse models and demonstrates intertumoral and intratumoral heterogeneity in NEhi ASCL1hi models (7, 12, 22). We confirmed a similarly heterogeneous distribution of NFIB expression in lung tumors at 6 months and in end-stage lung tumors of both fully transgenic and chimeric mice (n = 5/group; Supplementary Fig. S5E). This pattern is reminiscent of what others have reported with amplification of chromosome 4, containing both Mycl and Nfib (12, 22). NFIB expression increased significantly at later stages of tumor development, potentially associated with progressive metastasis, although no significant differences were observed between sgRlfsgMycl and sgNeosgNeo mice (n = 5/group; Supplementary Fig. S5E). Taken together, these data support that RLF–MYCL fusion, in contrast to MYC, does not shift SCLC subtype but rather leads to more rapid progression of ASCL1hi SCLC, consistent with the observed phenotype of the corresponding fusion in human SCLC.
Expression Profiling of Murine RLF–MYCL and Human RLF–MYCL SCLC
Given that RLF–MYCL fusion promotes both protumorigenic and prometastatic phenotypes in mice, we sought to determine the underlying transcriptional programs driving Rlf–Mycl tumorigenesis, with reference to both sgNeosgNeo tumors and human RLF–MYCL-positive SCLC. We performed RNA-seq from transgenic sgRlfsgMycl (n = 4) and sgNeosgNeo (n = 7) tumors. We observed higher Mycl expression levels in sgRlfsgMycl tumors relative to sgNeosgNeo lung tumors (Fig. 6A), consistent with our earlier observation in human RLF–MYCL fusion-positive SCLC versus other SCLC-A samples (Fig. 1D). We reanalyzed RNA-seq data from our previously published data set of human SCLC (8) to identify genes differentially expressed between human SCLC carrying the RLF–MYCL fusion (n = 4) relative to other SCLC-A subtype SCLC (n = 29; Fig. 6B). The 10 most differentially expressed genes included PPT1, PPIE, RLF, SNX9, OTUD1, MYO6, NBPF, C1orf220, PLEKHG1, and HSD17B13 (Fig. 6B). Of these, neither NBPF nor C1orf220 has a known mouse ortholog. We sought to determine whether the mouse Rlf–Mycl primary lung tumors differentially expressed any orthologs of the remaining eight differentially expressed genes. We were able to detect nominally significant differences in the expression of three of the eight orthologous genes: similar to human RLF–MYCL samples, Pp1t, Ppie, and Rlf were overexpressed in RLF–MYCL-containing mouse lung tumors relative to sgNeosgNeo lung tumors (Fig. 6C). Differential expression was not confirmed in the other five candidates. Palmitoyl-protein thioesterase (PPT1) has been previously associated with tumor growth and metastasis (37), and elevated expression of PPT1 in tumors correlates with poor patient survival in a variety of cancers (38). Peptidyl-prolyl cis-trans isomerase E (PPIE, or CYP33) has been implicated in processes including metabolism, apoptosis, inflammation, and cancer (39).
Finally, to investigate pathways that discriminate fusion-associated from non–fusion-associated ASCL1 subtype SCLC, we performed gene set enrichment analysis (GSEA) on both the human and mouse RNA-seq data sets. GSEA of human SCLC revealed significant enrichment for six Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, all downregulated in RLF–MYCL fusion tumors relative to nonfusion SCLC-A tumors (Fig. 6D; Supplementary Table S2), notably including pathways downregulated in contexts of invasion and metastasis. Negatively enriched KEGG pathways included cell adhesion molecules, extracellular matrix (ECM)–receptor interaction and cytokine–cytokine receptor interaction, allograft rejection, complement and coagulation cascades, and focal adhesion. We then explored whether any of these six KEGG pathways could similarly distinguish murine sgRlfsgMycl versus sgNeosgNeo lung tumors. Strikingly, GSEA revealed that five of the six KEGG pathways identified as significantly downregulated in human RLF–MYCL SCLC were also significantly suppressed in mouse RLF–MYCL tumors (Fig. 6D and E). These included three of the top five differentially regulated pathways in the mouse data set: cell adhesion molecules, cytokine–cytokine receptor interaction, and ECM–receptor interaction (Supplementary Table S3). Additional KEGG pathways significantly associated with sgRlfsgMycl versus sgNeosgNeo lung tumors included upregulation of glycosaminoglycan biosynthesis heparan sulfate (40), consistent with the tumor growth phenotype we observed in Rlf–Mycl fusion mice (Fig. 6E). Both the individual gene data, particularly the KEGG pathway analyses, point to hallmarks shared between human RLF–MYCL SCLC and the corresponding mouse model with CRISPR/Cas9-engineered RLF–MYCL fusion.
Discussion
Recurrent structural gene rearrangements in SCLC have been only minimally catalogued (8, 9, 28), and the potential roles of fusion oncogenes as initiators and promoters of cancer growth and spread in SCLC have not been previously defined. In the work reported here, we sought to explore the role of the most commonly reported gene fusion in human SCLC, RLF–MYCL, primarily through analysis of GEMMs harboring this fusion event using CRISPR-mediated genome editing in vivo. The ability to effectively model this fusion was aided by the proximity of the relevant genes and by the fact that the local chromosome environments represent syntenic blocks in the mouse and human genomes. The opposite orientations of the Rlf and Mycl genes on mouse chromosome 4 and of the RLF and MYCL genes on human chromosome 1 both require a small inversion event to generate the relevant fusion gene. Using CRISPR/Cas9-based editing in an SCLC precursor model and GEMMs, we were able to successfully generate the Rlf–Mycl fusion and demonstrate its potential to accelerate oncogenesis and promote widespread metastasis of SCLC.
Chromosomal translocations resulting in gene fusions were first recognized as a mechanism of oncogenesis in hematologic malignancies and have been subsequently identified as drivers of a wide range of solid tumors (26). The identification and characterization of oncogenic fusions in non–small cell lung cancers (NSCLC) has been critical not only in understanding the biology of these tumors but also in defining tumor-specific therapeutic targets for intervention. ALK, RET, ROS1, and the NTRK family members are all tyrosine kinases, which has facilitated rapid development of highly selective inhibitors for tumors driven by fusions activating or stabilizing these kinases (41). MYC family members have been known as oncogenic drivers for a much longer period of time than any of these NSCLC targets, yet the development of selective targeted inhibitors for MYC family members remains an unmet challenge (42). Our data supporting that the RLF–MYCL fusion represents a driver of oncogenesis and metastasis provide a rationale for exploring this fusion as a therapeutic target. Although inhibiting MYCL activity remains a challenge, the fusion would represent a uniquely tumor-specific target for strategies such as a proteolysis-targeting chimera (43). The Rlf–Mycl fusion GEMMs described in this study can be used in future preclinical studies to test novel therapeutic strategies and to assess the relative sensitivity to standard chemotherapy of these tumors.
Recent profiling of both human and mouse SCLC suggests that most tumors can be categorized into predominant subtypes based on differential expression of key transcription factors (5). MYCL expression is primarily associated with tumors of the SCLC-A subtype, expressing high levels of ASCL1; in contrast, MYC expression is associated with the other major subtypes. All reported examples of human SCLC harboring the RLF–MYCL fusion are of the SCLC-A subtype, which is recapitulated in both primary and metastatic lesions from the Rlf–Mycl fusion GEMMs reported here. That RLF–MYCL fusion is associated with the same subtype of SCLC as tumors overexpressing wild-type MYCL raises the question of whether the fusion protein is oncogenic only by increasing MYCL signaling (a hypermorphic function) or also by altering the nature of MYCL signaling (a neomorphic function). These are not mutually exclusive, but we do believe our data support a neomorphic function—not only increasing the pace of oncogenesis but also changing the nature of progression, enhancing the frequency and distribution of metastases.
One distinction between human SCLC and most of the mouse models generated to date is that whereas the human disease is notorious for early and widespread metastases, most SCLC GEMMs generated to date develop primary tumors in the lung, with some demonstrating predominantly liver metastasis as recapitulated in our analysis of RPR2 and RPM models (14). The Rlf–Mycl GEMMs described here may more closely resemble the exceptionally broad metastatic tropism of human SCLC, including distant nodal and multiorgan disease, representing a valuable model to facilitate research into the metastatic drivers of SCLC. However, the distribution of metastatic sites in the Rlf–Mycl fusion GEMMs does not fully recapitulate that of the human disease, lacking evident predilection for some common metastatic sites in patients, most notably the central nervous system.
In summary, we present here the first mouse model of a recurrent gene fusion in SCLC, demonstrating that the RLF–MYCL fusion promotes oncogenic transformation in precursor cells, accelerates tumorigenesis in vivo, and facilitates frequent metastases to multiple organs. The resultant murine tumors consistently demonstrate an ASCL1hi state, consistent with human SCLC harboring the homologous RLF–MYCL fusion. Further analysis of selective vulnerabilities of this mouse model may yield targeted strategies for better treatment of the human disease.
Methods
Human Genomics
RNA-seq fastq files from primary tumors and cell lines were obtained from Rudin and colleagues (ref. 8; N = 51) and George and colleagues (ref. 9; N = 54). Sequencing files were aligned to the GRCh37 reference genome using the STAR aligner algorithm v2.7 (44). The resulting BAM files were interrogated to identify gene fusions by the Arriba fusion detection algorithm v1.1.0 (https://github.com/suhrig/arriba/) and FusionCatcher v1.10 (https://github.com/ndaniel/fusioncatcher). Gene expression was quantified using Kallisto v0.45.0 (45) to Ensembl v75. We determined the subtype of each sample based on the highest expression of four genes: ASCL1, NEUROD1, YAP1, and POU2F3. All samples with RLF–MYCL gene fusion were classified as SCLC-A (ASCL1hi), and further comparative analyses were performed within the 33 samples of subtype SCLC-A from Rudin and colleagues (8). We used Sleuth (46) and the following model to identify genes that were differentially expressed in the samples with RLF–MYCL fusions: Expression ∼ SampleType + Fusion, where Fusion is 1 or 0 depending on detection by either Arriba or FusionCatcher and SampleType is cell line or primary tumor. We identified 10 genes that were differentially expressed (q < 0.05). Heat maps of differentially expressed genes represent z scores calculated across samples. Clustering was performed using Manhattan distance with the Ward D method.
Cell Lines and In Vitro Assays
Pre-SC cells were generously provided by Dr. Kwon Park and derived as previously described (29). Human SCLC cell lines were obtained from ATCC or Sigma and cultured in either RPMI (Corning #10–041-CV) supplemented with 10% FBS (Gemini Bio #900–108), 1% l-glutamine (Fisher #25030149), and 1% penicillin/streptomycin antibiotic cocktail (Fisher #15070063; H187, CORL88, H889, H209, H1963, H82, CORL47, H1092, pre-SC); modified HITES medium DMEM/F12 (ATCC #30–2006) supplemented with 0.005 mg/mL insulin (Fisher #12585–014), 0.01 mg/mL transferrin (Sigma #T2036), 30 nmol/L sodium selenite (Sigma #S5261), 10 nmol/L hydrocortisone (Sigma #H0888), 10 nmol/L β-estradiol (Sigma #E2758), 1% l-glutamine (Fisher #25030149), and 1% penicillin/streptomycin antibiotic cocktail (Fisher #15070063; H1882 and H1836); or DMEM (Corning #10–014-CV) with 10% FBS, 1% l-glutamine, and 0.1% penicillin/streptomycin antibiotic cocktail (HEK293T, NIH/3T3, and Green-Go cells). Soft agar assay was performed following standard protocols, seeding 1 × 105 preSCs per well in 6-well plates, resuspended in 1.5 mL RPMI growth media containing 0.35% agar (Invitrogen #16500500) on a bottom layer that contained 1.5 mL growth media containing 0.5% agar (Invitrogen #16500500). The media were regularly changed every 3 days for 6 weeks. After incubation, colonies were counted by using ImageJ (NIH) software.
RT-PCR and Droplet Digital PCR Analysis for Gene Fusion Detection
Total RNA was extracted from fresh-frozen cell pellets using RNeasy Plus Universal mini kit (QIAGEN #15596026) following the manufacturer's instructions. Extracted RNA samples were quantified using Qubit 2.0 Fluorometer (Life Technologies). cDNAs were prepared using the Superscript III Reverse Transcriptase kit (Thermo Fisher #18080093) following the manufacturer's instructions. For droplet digital PCR (ddPCR), total RNAs from cell pellets or tissue were extracted with TRIzol (Thermo Fisher #15596026) following the manufacturer's instructions. Quantity was assessed by PicoGreen (Thermo Fisher) and quality by BioAnalyzer (Agilent). Droplet generation was performed on a QX200 ddPCR system (Bio-Rad) using cDNA generated from 7 ng total RNA with the One-Step RT-ddPCR Advanced Kit for Probes (Bio-Rad #1864021) according to the manufacturer's protocol with reverse transcription at 42°C and annealing/extension at 55°C. Each sample was evaluated in technical duplicates. Reactions were partitioned into a median of approximately 30,000 droplets per well. Plates were read and analyzed with the QuantaSoft software to assess the number of droplets positive for the gene of interest and reference gene. PrimePCR ddPCR Expression Probe Assays were ordered through Bio-Rad for the following genes of interest: Rlf–Mycl, Mycl, and B2M. Primers, primer pairs, and probes used in the various PCR reactions (Supplementary Table S1).
Immunoblotting
Protein extraction and Western blot were performed as previously described (47). After quantification of protein extracts using the Bradford method (Bio-Rad #5000205), we loaded 20 to 30 μg total protein in the gels. Primary antibodies for MYCL (R&D Systems, #AF4050) and Vinculin (Cell Signaling Technology #13901S) were used at dilutions recommended by the manufacturer. Donkey anti-goat (Li-Cor Biosciences #925–32214) and donkey anti-rabbit (Li-Cor Biosciences #926–32213) secondary antibodies were used.
Genomic DNA Isolation and Surveyor Assay
Genomic DNA from cell lines or snap-frozen flank tumors of preSCs was isolated using the DNeasy Blood & Tissue Kit (Qiagen #69504) following the manufacturer's guidelines. PCR products for surveyor assay were amplified using Phusion High-Fidelity DNA Polymerase (NEB #M0530S), gel purified, and subsequently assayed with the Surveyor Mutation Detection Kit from Integrated DNA Technologies (IDT #706020). Primers for fusion detection and surveyor assay are as indicated in Supplementary Table S1.
sgRNA Cloning and Lentiviral Vector Cloning
Oligos containing the sgRNA sequences targeting Rlf, Mycl, or Neo were designed and purchased from IDT. The pX330 vector expressing Cas9 (Addgene #42230) was digested with BbsI (ThermoFisher #ER1011) and ligated to annealed and phosphorylated sgRNA oligos targeting Rlf and Mycl. To deliver both sgRNAs in the same lentiviral vector, we followed the cloning strategy described in Vidigal and Ventura (34). Briefly, sU6 promoter was inserted to 2 sgRNA ultramer through PCR and Gibson assembly with pDonor_sU6 (Addgene plasmid #69351), then digested with BbsI restriction enzyme and ligated to pUSEC backbone that was digested with BsmBI restriction enzyme (ThermoFisher #ER0451; ref. 35). Subsequently, colonies of transformed Stbl3 bacteria (ThermoFisher #C737303) were picked and sequenced by Sanger sequencing. Oligos sequences appear in Supplementary Table S1.
Lentiviral Production
Lentiviruses were produced by cotransfection of HEK293T cells with lentiviral backbone of LentiCas9-blast (Addgene #52962) or USEC constructs and packaging vectors (delta8.2 and VSV-G) using TransIT-LT1 Transfection Reagent (Mirus Bio #MIR2304). Supernatants were collected 48 and 72 hours posttransfection, concentrated by ultracentrifugation at 25,000 rpm for 120 minutes and resuspended in 50 μL PBS (Gibco). Green-Go cells (48) were used to determine lentiviruses titer.
Transfections and Transductions
NIH/3T3 and preSC cells were transfected in T75 flasks with 10 μg total plasmid DNA using Lipofectamine 3000 reagent (ThermoFisher #L3000001) following the manufacturer's instructions. To enrich for transfected cells, transfections included 3 μg of a plasmid expressing the Puro-resistance gene (pSico; Addgene #11578), and cells were incubated with 2 μg/mL Puromycin (ThermoFisher #A1113803) for 2 days. Transductions were performed by adding lentiviruses directly to each T75 flask.
Animal Studies
Nude mice were purchased from Charles River Laboratories (stock #194). RPR2C embryonic stem cells (ESC) containing Rosa26-CAGGS-LSL-Cas9-GFP-Csy4 were generated by the Jacks laboratory at Massachusetts Institute of Technology as previously described (31). RPR2C chimeric animals were generated by the NYUMC Rodent Genetic Engineering Laboratory. Briefly, RPR2C ESCs were injected into C57BL/6N blastocysts (stock #005304; JAX) and cultured 2 to 3 hours in KSOM + AA medium (Sigma # MR-101-D) to blastocyst stage and implanted into pseudopregnant CD-1 (stock #022; Charles River Laboratories) fosters. To generate approximately 10 to 20 chimeras, we implanted 30 to 60 embryos per ESC injection session. Chimeric pups that were more than 6 weeks old displayed high degree of chimerism. To generate fully transgenic RPR2C animals, high-degree RPR2C chimeric animals (described above) were crossed to Rb1fl/fl;Trp53fl/fl;Rbl2fl/fl (RPR2; ref. 16; MMRRC: 043692-UCD) animals until RPR2 homozygous genes were acquired. Tumors in RPR2C mice were initiated by intratracheal instillation of 0.5 × 106 (chimeric) or 5 × 106 (chimeric and full transgenic) transduction units of lentivirus-expressing Cre recombinase, as previously described (49). Sections from previously described fl/fl;fl/fl;MycT58ALSL/LSL (RPM; JAX #029971; ref. 7) and RPR2 mice (n = 5/genotype) were placed on slides of IHC. For all animal studies, both male and female mice were equally divided between treatment groups. Investigators were not blinded with respect to which lentivirus was injected. All studies and procedures were approved by the Memorial Sloan-Kettering Cancer Center Institutional Animal Care and Use Committee or the NYU Langone Medical Center Institutional Animal Care and Use Committee.
MRI Analysis
Mouse scans were performed using a 9.4T 20-cm bore Bruker Biospec scanner (Bruker Biospin MRI GmbH) equipped with an ID 114-mm maximum strength of 530 mT/m Bruker gradient. An ID 40-mm Bruker volume resonator was used for RF excitation and MRI acquisition. The mice were anesthetized with 2% isoflurane (Baxter Healthcare Corp) gas in oxygen during MRI scanning. A small-animal physiologic monitoring system (SA Instruments) was used to monitor animal respiration during MRI scanning. Scout images along three orthogonal orientations were first acquired for animal positioning. For mouse lung imaging, respiratory gated T1-weighted axial images using the FLASH gradient echo sequence were acquired with TR of 170 ms, TE of 1.6 ms, slice thickness of 0.8 mm, FOV of 35 × 25 mm, in-plane resolution of 182 × 130 μm, and 10 averages. The tumors were measured from their Digital Imaging and Communications in Medicine (DICOM) files using ImageJ software. Tumor burden was calculated by outlining the region of interest (ROI) of tumor structures, taking the output of the ROIs in mm2 and multiplying each slice's ROI by its slice thickness.
IHC Analysis
Lung, liver, and other metastatic samples were fixed in 4% formalin and paraffin embedded. H&E staining was performed using standard methods. Formalin-fixed, paraffin-embedded sections at 4 to 5 μm were dewaxed, rehydrated, and subjected to high-temperature antigen retrieval by boiling 15 minutes in a pressure cooker in 0.01 mol/L citrate buffer at pH 6.0. Slides were quenched of endogenous peroxide in 3% H2O2 for 10 minutes, then blocked in 5% goat serum in PBS/0.1% Tween-20 (PBS-T) for 30 minutes, and then stained overnight with primary antibodies in blocking buffer (5% goat serum; Cell Signaling Technology #8112). For non–Cell Signaling Technology primary antibodies, a horseradish peroxidase–conjugated secondary antibody (Vector Laboratories) was used at 1:200 dilution in PBS-T and incubated for 30 minutes at room temperature followed by DAB staining (Vector Laboratories). Alternatively, Cell Signaling Technology primary antibodies were detected using 200 μL SignalStain Boost IHC Detection Reagent (Cell Signaling Technology #8114). All stainings were performed with Sequenza cover plate technology. For IHC, we used antibodies to NFIB (Sigma #HPA003956) 1:250, ASCL1 (BD #BD556604) 1:200, NEUROD1 (Abcam #109224) 1:150, INSM1 (Santa Cruz Biotechnology #sc-271408) 1:100, POU2F3 (Sigma #HPA019652) 1:300, YAP1 (Cell Signaling Technology #14074S) 1:400, and synaptophysin (Neuromics #MO20000) 1:250. For manual H-score quantification, images were acquired on a Nikon Ci-L LED Microscope with DS-Fi3 camera. H-score was quantified on a scale of 0 to 300, taking into consideration percent positive cells and staining intensity as described (50), where H-score equals the percentage of positive cells multiplied by an intensity score of 0 to 3. For example, a tumor with 80% positive cells with high intensity of 3 = 240 H-score. H&E- and IHC-stained slides were digitally scanned with the Zeiss Axio Scope A1 microscope using AxioVision SE64 software. Whole-slide images containing four to five lung lobes per animal were analyzed using CaseViewer software (3DHISTECH). Tumor regions were manually annotated.
Mouse RNA-Seq
RNA extraction of snap-frozen microdissected lung tumors and sequencing was done in collaboration with Genewiz and Integrated Genomics Operation (IGO) at the Memorial Sloan Kettering Cancer Center. RNA of full transgenic sgNeosgNeo mice (n = 7) and sgRlfsgMycl mice (n = 7) was subject to library construction prepared using the NEBNext Ultra RNA Library Prep Kit for Illumina (NEB) or TruSeq Stranded mRNA library Prep Kit (Illumina) following the manufacturer's instructions. The sequencing libraries were validated on the Agilent TapeStation (Agilent Technologies) and quantified by using Qubit 2.0 Fluorometer (Invitrogen) as well as by quantitative PCR (KAPA Biosystems). The sequencing libraries were clustered on two lanes of a flowcell. After clustering, the flowcell was loaded on the Illumina HiSeq instrument (4000) according to the manufacturer's instructions. The samples were sequenced using a 2 × 150-bp paired end configuration. Image analysis and base calling were conducted by the HiSeq Control Software. Raw sequence data (.bcl files) generated from Illumina HiSeq were converted into fastq files and demultiplexed using Illumina's bcl2fastq 2.17 software.
All samples were processed with Kallisto (v0.45.0; ref. 45) to GRCm38 cDNA (Ensembl release 100). We identified Rlf–Mycl gene fusion in four of seven samples via either the FusionCatcher fusion detection algorithm (v1.10, http://code.google.com/p/fusioncatcher; n = 1) and/or ddPCR (n = 4). We used Sleuth (46) and the following model to study differential expression in the samples with Rlf–Mycl fusions: Expression ∼ Batch + Fusion, where Fusion is 1 or 0 depending on detection by either Arriba or FusionCatcher or ddPCR, and Batch denotes sequencing at IGO or Genewiz.
GSEA
GSEA was performed using KEGG pathways pulled from MSigDB using the R package msigdbr (51). When more than one Ensembl ID mapped to a given Hugo Symbol, the Ensembl ID with the lowest P value from differential expression analysis was used to represent the gene. Ranking of genes was determined by effect size (β) from Kallisto. GSEA was called from the clusterProfiler R package (52) using default parameters. Term was considered significant if the BH-adjusted P value was <0.05. GSEA plots were created by modifying gseaplot2 from the enrichplot R package (https://github.com/YuLab-SMU/biomedical-knowledge-mining-book). Human and mouse KEGG pathways appear in Supplementary Tables S2 and S3.
Statistical Analysis
Remaining statistical analyses were performed using GraphPad Prism 7 software (GraphPad Software). Error bars show mean ± SD. Significance was determined by the Student two-tailed unpaired t tests with 95% confidence intervals, and P values <0.05 were considered statistically significant. Survival analysis was performed by comparing two survival curves using the log-rank (Mantel–Cox) test. Statistical details are further described in respective figure legends.
Authors' Disclosures
D. Maddalo reports other support from Genentech outside the submitted work. E.H. Akama-Garren reports grants from NIH outside the submitted work. T. Jacks reports being a member of the Board of Directors of Amgen and Thermo Fisher Scientific and a cofounder of Dragonfly Therapeutics and T2 Biosystems; serves on the Scientific Advisory Board of Dragonfly Therapeutics, SQZ Biotech, and Skyhawk Therapeutics; and is also president of Break Through Cancer. These relationships did not impact the design or interpretation of experiments or results described in this manuscript. His laboratory currently receives funding from Johnson & Johnson and The Lustgarten Foundation. This funding did not support the work described in this manuscript. T.G. Oliver reports grants from the National Institutes of Health (U01CA231844; U24CA213274; R01CA251147) during the conduct of the study; other support from Known Medicine, personal fees from AbbVie, and nonfinancial support from Polaris outside the submitted work; in addition, T.G. Oliver has a patent for MYC Drives Progression of Small Cell Lung Cancer to a Variant Neuroendocrine Subtype with Vulnerability to Aurora Kinase Inhibition (nonprovisional) 16/335,368; 17865057.8; 2019-522392 pending. T. Papagiannakopoulos reports personal fees from Vividion Therapeutics, grants from Bristol Myers Squibb, grants from Dracen Pharmaceutical, grants from Agios Pharmaceuticals, personal fees from Calithera Biosciences, and grants from Kymera Therapeutics outside the submitted work. C.M. Rudin reports personal fees from AbbVie, personal fees from Amgen, personal fees from Astra Zeneca, personal fees from Epizyme, personal fees from Genentech/Roche, personal fees from Ipsen, personal fees from Jazz, personal fees from Lilly, personal fees from Syros, personal fees from Bridge Medicines, personal fees from Harpoon Therapeutics, and personal fees from Earli outside the submitted work. No disclosures were reported by the other authors.
Authors' Contributions
M. Ciampricotti: Conceptualization, data curation, formal analysis, investigation, writing–original draft, writing–review and editing. T. Karakousi: Investigation, writing–review and editing. A.L. Richards: Formal analysis, writing–review and editing. À. Quintanal-Villalonga: Investigation, writing–review and editing. A. Karatza: Investigation, writing–review and editing. R. Caeser: Investigation, writing–review and editing. E.A. Costa: Investigation, writing–review and editing. V. Allaj: Investigation, writing–review and editing. P. Manoj: Investigation, writing–review and editing. K.B. Spainhower: Investigation, writing–review and editing. F.E. Kombak: Methodology, writing–review and editing. F.J. Sanchez-Rivera: Investigation, writing–review and editing. J.E. Jaspers: Investigation, writing–review and editing. A. Zavitsanou: Investigation, writing–review and editing. D. Maddalo: Investigation, writing–review and editing. A. Ventura: Investigation, writing–review and editing. W.M. Rideout III: Investigation, writing–review and editing. E.H. Akama-Garren: Investigation, writing–review and editing. T. Jacks: Investigation, writing–review and editing. M.T. Donoghue: Investigation, writing–review and editing. T. Sen: Investigation, writing–review and editing. T.G. Oliver: Investigation, writing–review and editing. J.T. Poirier: Conceptualization, investigation, writing–review and editing. T. Papagiannakopoulos: Conceptualization, supervision, investigation, writing–original draft, writing–review and editing. C.M. Rudin: Conceptualization, supervision, funding acquisition, writing–original draft, writing–review and editing.
Acknowledgments
This work was supported by NCI R01 CA197936 and U24 CA213274 (to C.M. Rudin) and NCI U01 CA231844 (to T.G. Oliver). T. Papagiannakopoulos is supported by NCI R37CA222504, NCI R01CA227649, and American Cancer Society Research Scholar Grant RSG-17-200-01-TBE. We thank Dr. Kwon Park, University of Virginia, for the generous gift of preSC cells and Dr. S.Y. Kim, director of the Rodent Genetic Engineering Core at NYU Grossman School of Medicine, for assistance with ESC expansion and blastocyst injections.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.