bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
1 ASCL1 represses a latent osteogenic program in small cell lung cancer
2 in multiple cells of origin
3
4 Rachelle R. Olsen1, David W. Kastner1, Abbie S. Ireland1, Sarah M. Groves2, Karine Pozo3,
5 Christopher P. Whitney1, Matthew R. Guthrie1, Sarah J. Wait1, Danny Soltero1, Benjamin L.
6 Witt4,5, Vito Quaranta2, Jane E. Johnson3,6, Trudy G. Oliver1,7*
7
8 1Department of Oncological Sciences, Huntsman Cancer Institute, University of Utah, Salt Lake
9 City, UT 84112, USA
10 2Department of Biochemistry, Vanderbilt University, Nashville, TN, 37212, USA
11 3Department of Neuroscience, University of Texas Southwestern Medical Center, Dallas, TX
12 75390, USA
13 4Department of Pathology, University of Utah, Salt Lake City, UT 84112, USA
14 5ARUP Laboratories at University of Utah, Salt Lake City, UT 84108, USA
15 6Hamon Center for Therapeutic Oncology Research, University of Texas Southwestern Medical
16 Center, Dallas, TX 75390, USA
17 7Corresponding author
18
19 *Correspondence:
20 Trudy G. Oliver, Department of Oncological Sciences, Huntsman Cancer Institute, 2000 Circle
21 of Hope, Salt Lake City, UT 84112, USA, Tel: +1-801-213-4221, Email:
23
24 Keywords: SCLC, ASCL1, mouse models, lung cancer, cell of origin, plasticity
25
1 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
26 Abstract 27 28 ASCL1 is a neuroendocrine-lineage-specific oncogenic driver of small cell lung cancer (SCLC),
29 highly expressed in a significant fraction of tumors. However, ~25% of human SCLC are
30 ASCL1-low and associated with low-neuroendocrine fate and high MYC expression. Using
31 genetically-engineered mouse models (GEMMs), we show that alterations in Rb1/Trp53/Myc in
32 the mouse lung induce an ASCL1+ state of SCLC in multiple cells of origin. Genetic depletion of
33 ASCL1 in MYC-driven SCLC dramatically inhibits tumor initiation and progression to the
34 NEUROD1+ subtype of SCLC. Surprisingly, ASCL1 loss converts tumors to a SOX9+
35 mesenchymal/neural-crest-stem-like state that can differentiate into RUNX2+ bone tumors.
36 ASCL1 represses SOX9 expression, as well as WNT and NOTCH developmental pathways,
37 consistent with human gene expression data. Together, SCLC demonstrates remarkable cell
38 fate plasticity with ASCL1 repressing the emergence of non-endodermal stem-like fates that
39 have the capacity for bone differentiation.
40
2 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
41 Small cell lung cancer (SCLC) is a highly aggressive neuroendocrine lung tumor with a median
42 survival time of only 10-12 months1,2. Almost all SCLCs exhibit inactivation of the tumor
43 suppressor genes RB1 and TP53, along with mutually exclusive gain of a MYC family member
44 including MYC, MYCL or MYCN3-5. While SCLC has historically been treated as a single
45 disease in the clinic, it is increasingly appreciated to be composed of distinct molecular
46 subtypes1. Approximately 70% of SCLCs are characterized by high expression of the
47 transcription factors ASCL1 and MYCL, along with high expression of neuroendocrine (NE)
48 markers such as Synaptophysin (SYP), Ubiquitin C-terminal Hydrolase L1 (UCHL1) and
49 Chromogranin A (CHGA)1,4-6. NE-high SCLCs typically display classic morphology with small
50 round blue cells and scant cytoplasm. In contrast, approximately 25% of SCLCs are associated
51 with an ASCL1-low/NE-low gene expression profile, and these tend to have high MYC
52 expression and variant morphology. We reported the first genetically-engineered mouse model
53 (GEMM) of MYC-driven SCLC6, and demonstrated that MYC promotes the variant, ASCL1-
54 low/NE-low subtype of SCLC. Importantly, the MYC-driven subtype of SCLC demonstrates
55 unique therapeutic vulnerabilities including sensitivity to inhibition of Aurora kinases (AURK),
56 CHK1, IMPDH, and arginine depletion6-11.
57 ASCL1 and NEUROD1 are lineage-specifying basic helix-loop-helix transcription factors
58 that activate neuroendocrine genes and are required for neural differentiation1,12. Genetic
59 knockout studies in the mouse demonstrate that ASCL1 is essential for the development of
60 pulmonary neuroendocrine cells (PNECs)13, and ASCL1 is expressed in the adult lung
61 specifically in PNECs14. Importantly, ASCL1 is necessary for the development of classic SCLC
62 as demonstrated in the Rb1fl/fl;Trp53fl/fl;Rbl2fl/fl (RPR2) mouse model where conditional Ascl1
63 deletion abolishes tumor formation12. NEUROD1 is not required for SCLC initiation or
64 progression in the RPR2 model, consistent with the observation that NEUROD1 is not
65 expressed in classic SCLC models6,12. NEUROD1 is expressed in the variant MYC-driven SCLC
66 model, and is associated with MYC expression in human SCLC6. NEUROD1 has recently been
3 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
67 suggested to temporally follow ASCL1 during MYC-driven subtype evolution15. However, the
68 function of ASCL1 and NEUROD1 in MYC-driven tumors in vivo has not yet been determined.
69 PNECs have been generally accepted to be the cell of origin for SCLC, although recent
70 studies have challenged whether they are the only relevant initiating cell population8,15-17. SCLC
71 shares molecular similarities with normal PNECs, including expression of ASCL1 and its
72 neuroendocrine target genes. Studies in mouse models have demonstrated that neuroendocrine
73 cells are the most permissive for SCLC development in the original classic Rb1fl/fl;Trp53fl/fl (RP)
74 model and in RPR2 mice18,19. Club cells and alveolar type II (AT2) cells in these studies were
75 relatively resistant to SCLC transformation18,19. More recently, a subset of SCLC was revealed
76 to harbor a tuft-cell-like signature with dependency on the tuft-cell transcription factor POU2F38,
77 suggesting that the tuft cell may serve as a cell of origin for SCLC. Basal cells have also been
78 described as cells of origin for SCLC16,20. Together, these studies suggest that lung cells may
79 have broad plasticity for SCLC transformation, but the cells of origin for MYC-driven SCLC have
80 not been determined.
81 Here, we use new GEM models to determine the function of ASCL1 in MYC-driven
82 SCLC derived from multiple cells of origin.
83
84 Results
85 MYC-driven SCLC can arise in multiple lung cell types
86 Our previous work demonstrated that Myc overexpression in the Rb1fl/fl;Trp53fl/fl;MycT58ALSL/LSL
87 (RPM) mouse model promotes the development of a variant, neuroendocrine-low subtype of
88 SCLC that recapitulates features of the human disease6. To determine which cell types in the
89 normal lung epithelium are capable of initiating tumorigenesis in the RPM model, we
90 intratracheally-infected mice with adenoviruses carrying cell-type-specific promoters driving
91 expression of Cre recombinase. We used a general Cmv promoter, a neuroendocrine Cgrp
92 promoter, a club cell Ccsp promoter, and an AT2 Spc promoter, whose specificity have been
4 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
93 previously described17-19,21. The combination of Myc overexpression with Rb1 and Trp53 loss
94 was sufficient to initiate tumorigenesis with all four Cre viruses (Fig. 1a and 1b). RPM-CMV
95 tumors had the shortest latency (43 days median survival), followed by RPM-CGRP tumors (55
96 days median survival). Compared to results from the RP model18, targeting of club cells with
97 CCSP-Cre led to tumors with a surprisingly short latency (77 days median survival), and was
98 almost as efficient at tumor initiation as targeting neuroendocrine cells, suggesting that club
99 cells are highly susceptible to MYC-mediated SCLC transformation. Consistent with the location
100 of club cells, in situ tumors in the CCSP-Cre RPM mice were often located within bronchioles
101 (Supplementary Fig. 1a). In contrast, AT2 cells were the most resistant to transformation as
102 RPM tumors initiated with SPC-Cre had the longest latency (184 days median survival) (Fig.
103 1a). Review by board-certified pathologists (A. Gazdar and B. Witt) confirmed that RPM mice
104 infected with different cell-type-specific Cre viruses predominantly developed SCLC, with a
105 mixture of classic and variant histopathologies and perilymphatic spread, similar to that
106 observed in the original RPM-CGRP model (Fig. 1b and Supplementary Fig. 1a). Combined with
107 previous studies in the RP and RPR2 mice17-19, these results suggest that genetic alterations
108 can dictate the susceptibility of cells to serve as a cell of origin for SCLC.
109
110 MYC-driven SCLC induces ASCL1 in multiple cells of origin
111 To further determine whether tumors in this model were neuroendocrine, we examined in situ
112 and invasive tumors for expression of ASCL1 and other neuroendocrine markers including
113 CGRP, UCHL1 and SYP. Interestingly, in situ tumors from each cohort were dominated by an
114 ASCL1high/neuroendocrinehigh (NEhigh) phenotype even when tumors were initiated in non-
115 neuroendocrine cells with CCSP-Cre or SPC-Cre (Fig. 1c). NEhigh in situ tumors did not express
116 CCSP or SPC (Supplementary Fig. 1b), suggesting that the original cell fate identities were
117 rapidly lost. Consistent with the classic NEhigh state of these in situ lesions, the majority of in situ
118 RPM tumors expressed low-to-undetectable levels of NEUROD1, YAP1, and POU2F3 (Fig. 1c),
5 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
119 key transcription factors that are associated with NElow subtypes of SCLC1. These results
120 suggest that MYC overexpression in the context of Rb1/Trp53 loss alters the original cellular
121 differentiation state and initially reprograms cells toward a neuroendocrine-like progenitor.
122 In contrast to the in situ tumors, invasive RPM tumors were dominated by an
123 ASCL1low/NElow phenotype largely independent of the cell of origin, and tended to express higher
124 levels of NEUROD1 and YAP1, but not POU2F3 (Fig. 1d and 1e). Interestingly, tumors initiated
125 in club cells exhibited significantly higher ASCL1 and less YAP1 (Supplementary Fig. 1c),
126 suggesting club cells may not progress as readily as tumors from other cells of origin. Although
127 POU2F3 levels were not significantly elevated in all of the models, we observed an increase in
128 POU2F3+ invasive tumors specifically in the RPM-CMV mice (Supplementary Fig. 1c),
129 consistent with recent studies15 suggesting that the Cmv promoter may occasionally be
130 targeting a POU2F3+ tuft cell8. Together, this suggests that RPM tumors induce ASCL1 and a
131 neuroendocrine phenotype in multiple cells of origin that progress to a NE-low state.
132
133 ASCL1 loss delays tumorigenesis and promotes bone differentiation in multiple cells of
134 origin
135 Since ASCL1 has been previously shown to be required for development of classic SCLC in the
136 RPR2 model12, we sought to determine the role of ASCL1 in MYC-driven variant SCLC. We
137 crossed RPM mice to Ascl1-floxed animals to generate Rb1fl/fl;Trp53fl/fl;MycT58ALSL/LSL;Ascl1fl/fl
138 (RPMA) mice and infected RPMA mice with cell type-specific adenoviruses as described above.
139 In contrast to prior findings in classic SCLC models12, RPMA mice developed tumors in the lung,
140 albeit with significantly delayed latencies compared to RPM mice (Fig. 2a). RPMA-CMV mice
141 developed tumors with the shortest median survival of 81 days (1.9-fold longer than RPM-CMV
142 mice), while RPMA mice infected with CGRP-Cre had a median survival of 139 days (2.5-fold
143 longer than RPM-CGRP mice). Median survival in the RPMA-CCSP mice was 256 days (3.3-
144 fold longer than RPM-CCSP mice). RPMA-SPC tumors had the longest latency (median survival
6 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
145 undetermined; at least 2.4-fold longer than RPM-SPC mice). Successful Cre-mediated
146 recombination of the Rb1, Trp53, and Ascl1 alleles in these mice was assessed by genomic
147 PCR on individual micro-dissected RPMA tumors (Supplementary Fig. 2a). We observed
148 recombined Rb1, Trp53 and Ascl1 alleles in ~94% of tumors. Unrecombined alleles were also
149 detected in the majority of samples, but we cannot rule out that this represents normal tissue
150 contamination in each tumor since it is difficult to cleanly dissect these tumors. We also
151 detected rare tumors that did not recombine the Ascl1 allele, and in these cases, the tumors
152 expressed ASCL1 protein and exhibited SCLC morphology (Supplementary Fig. 2b).
153 While imaging RPMA mice by microCT, we made the unexpected observation that mice
154 were developing lung tumors with a tissue density consistent with bone (Fig. 2b). We used bone
155 analysis imaging software on the Quantum GX2 microCT to compare lung tumor development
156 in RPM and RPMA mice. Bone analysis of RPM mice with high lung tumor burden identified
157 only the skeleton (Fig. 2b, top panel, and Supplementary Videos). In contrast, bone analysis of
158 RPMA-CMV, -CGRP, and -CCSP animals identified both the skeleton as well as multiple bone-
159 like tumors within the lung area. These lesions were clearly evident on multiple microCT cross-
160 sections (Fig. 2b, bottom panels, and Supplementary Videos). At necropsy, many RPMA lesions
161 were mineralized and required decalcification prior to sectioning (Fig. 2c). Analysis of H&E-
162 stained tissues by a board-certified pathologist (B.W.) confirmed that RPMA lungs contained
163 high-grade osteosarcoma with well-developed osteoid (Fig. 2c, right panels, and Supplementary
164 Fig. 2c). Bone-like lesions in RPMA tumors were also confirmed by Trichrome staining (Fig. 2d),
165 Periodic Acid Schiff (PAS) with Alcian Blue staining (PAB, Supplementary Fig. 2d), and
166 Toluidine blue staining (Supplementary Fig. 2e). Overall, osteosarcomas represented between
167 45-90% of the total tumor burden in RPMA-CMV animals compared to RPM animals that had no
168 osteosarcomas detected despite equivalent tumor burden (Fig. 2e and 2f). Small RPMA tumors
169 were primarily soft, while large invasive RPMA tumors displayed a mix of soft and osteosarcoma
170 phenotypes (Fig. 2g), suggesting that tumors may arise in a dedifferentiated state that later
7 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
171 adopts a bone phenotype. We verified that the vast majority of both in situ and invasive RPMA
172 tumors lacked ASCL1 expression (Supplementary Fig. 2f). Some RPMA mice had lymph node
173 metastases, including one animal that appeared to have an osteosarcoma metastasis
174 (Supplementary Fig. 2g). RPMA mice also had occasional non-calcified tumors resembling
175 adenocarcinoma or large cell carcinoma that lacked ASCL1 expression (Supplementary Fig.
176 2h). The osteosarcoma-like tumors were frequently observed by microCT imaging in RPMA-
177 CMV and RPMA-CGRP mice, less frequent in RPMA-CCSP mice, and not detectable in RPMA-
178 SPC mice (Fig. 2h). However, one RPMA-SPC mouse did have areas of focal bone formation
179 with early osteoid detected upon histopathological review (Fig. 2h). These data suggest that in
180 the context of MYC-driven SCLC, ASCL1 represses a latent osteogenic fate in multiple lung cell
181 types that may have differing propensities for bone differentiation.
182
183 RPMA tumors are transcriptionally distinct with loss of ASCL1 and NEUROD1 target
184 genes
185 Following the striking observation that loss of ASCL1 in MYC-driven SCLC promotes
186 osteosarcoma formation, we sought to better understand the transcriptional program of RPMA
187 tumors. We performed RNA-sequencing (RNA-seq) on variant RPM (n = 11), classic RPR2 (n =
188 6), and un-calcified RPMA tumors (n = 7) including one RPMA tumor that was heterozygous for
189 Ascl1 (Rb1fl/fl;Trp53fl/fl;MycT58ALSL/LSL;Ascl1fl/wt; RPMAhet). Consistent with their histological
190 differences, principal component analysis of RNA-seq data revealed distinct clustering of all
191 three tumor types (Fig. 3a). Notably, the one heterozygous RPMAhet tumor clustered apart from
192 homozygous Ascl1-null RPMA samples and closer in identity to the RPR2 and RPM samples.
193 Thus, complete loss of Ascl1 appears to be required for the osteoid gene expression signature,
194 consistent with histological observations.
195 We verified that Ascl1 mRNA was significantly depleted in RPMA tumors (Fig. 3b).
196 Consistently, established ASCL1 target genes including neuroendocrine genes were
8 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
197 significantly reduced in RPMA compared to RPM tumors (Fig. 3c and 3d). We next performed
198 differential gene expression analysis to identify additional genes and pathways that discriminate
199 the RPMA and RPM tumors. Interestingly, these data revealed Neurod1 as one of the most
200 significantly down-regulated genes in RPMA versus RPM tumors (Fig. 3e) with complete loss of
201 Neurod1 transcript counts (Fig. 3f). As assessed by IHC, RPMA tumors were devoid of
202 NEUROD1 protein, in contrast to high levels observed in RPM tumors (Fig. 3g). Gene-set
203 enrichment analysis (GSEA) using established human NEUROD1 target genes12 combined with
204 NEUROD1 ChIP-seq from mouse RPM tumors (n = 2) revealed a significant depletion of
205 NEUROD1 target genes in RPMA compared to RPM tumors (Fig. 3h). These data reveal that
206 deletion of ASCL1 in RPM tumors during tumor initiation abolishes the NEUROD1+ subtype of
207 SCLC, consistent with recent findings that Ascl1 expression temporally precedes Neurod1
208 expression during MYC-driven SCLC progression15.
209 Consistent with the observed osteoid formation, GSEA revealed a significant positive
210 enrichment for bone development genes in RPMA compared to RPM tumors (Fig. 3i). To
211 identify other biological processes altered following Ascl1 loss in RPM mice, we performed GO
212 Enrichment Analysis on all significantly enriched or depleted genes in RPMA compared to RPM
213 tumors (Fig. 3j and 3k). As expected, ossification-related processes were significantly
214 upregulated in RPMA, as well as immune-related signatures (Fig. 3j). Significantly-depleted
215 biological processes in RPMA tumors included those related to neuronal development (Fig. 3k),
216 consistent with the loss of ASCL1 and NEUROD1 in these tumors. Together, these data
217 highlight a critical role for ASCL1 in promoting neuroendocrine cell fate and repressing an
218 underlying osteosarcoma-like fate in MYC-driven SCLC.
219
220 Network analyses predict transcriptional regulators that drive osteosarcoma cell fate
221 upon ASCL1 loss
9 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
222 To identify the key transcription factors that are responsible for this dramatic change in cell fate
223 upon ASCL1 loss, we turned to Weighted Gene Coexpression Network Analysis (WGCNA).
224 First, we generated a coexpression network of all genes, which allowed the identification of
225 distinct gene modules across all RPM and RPMA samples. Using an ANOVA statistical test, we
226 identified the gene modules that significantly differed between RPM and RPMA tumors (Fig. 4a
227 and Supplementary Fig. 3a). Strikingly, ~1/3 of the transcriptome was altered upon ASCL1 loss.
228 To generate a gene regulatory network, we focused on transcription factors predicted by
229 WGCNA to be central to differentially-expressed gene modules, as well as known regulators of
230 lung cancer cell fate. Using BooleaBayes22, we determined rules of interaction between these
231 transcription factors, where the rules underlie the dynamics of cell identity in the presence or
232 absence of ASCL1 (Supplementary Fig. 3b). For example, ASCL1 is regulated by eight parent
233 nodes (AR, E2F1, HES1, KLF4, MITF, NR3C1, PHC1, and RUNX1). Each ON/OFF combination
234 of these parent nodes determine the expression of ASCL1. Likewise, ASCL1 regulates
235 expression of a number of downstream transcription factors. These regulations define how a cell
236 may change its identity or reach a stable phenotype (an attractor state). Dynamic simulations
237 identified two attractor states, each corresponding to either RPM or RPMA tumors (Fig. 4b). As
238 described in [22], we used a random walk simulation to predict regulators driving these steady
239 states. Satisfyingly, in silico silencing of the ASCL1 or NEUROD1 node destabilized the RPM
240 attractor, consistent with experimental results (Fig. 4b). Conversely, activation of ASCL1 or
241 NEUROD1 in the RPMA attractor destabilized that steady state. This is reminiscent of human
242 SCLC, in which ASCL1 is a destabilizer of the non-neuroendocrine subtype (SCLC-Y)22. By
243 comparing human SCLC cell lines to mouse tumors using PCA analysis (Fig. 4c), RPMA tumors
244 clustered with the non-neuroendocrine POU2F3 and YAP1 subtypes, suggesting a similarity
245 between RPMA tumors and human non-neuroendocrine subtypes.
246 In our network, we included other known neuroendocrine fate specifiers like INSM1 and
247 transcription factors known to be important in endodermal cell fate and lung adenocarcinoma
10 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
248 fate, such as NKX2-1 and FOXA2. While all of these genes were differentially expressed
249 between RPM and RPMA tumors (Fig. 4d), only FOXA2 significantly affected the dynamics of
250 the transcription factor network. IHC analyses confirmed loss of NKX2-1, FOXA2, and INSM1
251 protein in RPMA compared to RPM tumors, largely independent of initiating cell of origin (Fig.
252 4e and 4f). Interestingly, NKX2-1 levels remained high particularly in CGRP and SPC-initiated
253 tumors; however, it is noted that the SPC-initiated NKX2-1+ tumors appeared to be largely
254 adenocarcinoma. Consistently, ASCL1 expression is positively correlated with lineage-related
255 transcription factor genes including NKX2-1, FOXA1, FOXA2, and INSM1 in human SCLC cell
256 lines (Fig. 4g) from the SCLC-CellMiner database23, with similar results in human tumor data
257 (Fig. 4h). Thus, ASCL1 loss leads to coordinate loss of other key lineage-related transcription
258 factors, which are known to function in the same SCLC super-enhancers12,24,25.
259
260 ASCL1 represses non-endodermal cell fates and Notch/Wnt developmental pathways
261 We next focused on understanding the cell fate and predicted regulators gained upon ASCL1
262 loss. The lung epithelium is believed to derive largely from endoderm, whereas bone fates are
263 derived from mesoderm or ectoderm26,27. Mesenchymal stem cells (MSCs) from mesodermal
264 tissue and neural crest stem (NCS) cells from ectodermal tissue are known to have the potential
265 to become neurons or bone28,29. Thus, we examined whether gene signatures for these cell
266 types are enriched in RPM or RPMA tumors. GSEA revealed that both MSC and NC-cell
267 signatures were significantly enriched in RPMA vs RPM tumors (Fig. 5a). Consistent with these
268 results, GSEA revealed that both MSC and NC-cell signatures were significantly enriched in a
269 panel of 109 human SCLC cell lines when comparing ASCL1-low versus ASCL1-high samples
270 (Fig. 5b). These data suggest that ASCL1 facilitates an endodermal tumor cell fate and
271 represses an MSC/NC-stem-like fate.
272 During development, both MSC and NC-stem-like progenitors have the capacity for bone
273 development given the appropriate developmental signals. Multiple developmental pathways
11 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
274 are known to drive bone fate including Indian hedgehog (Ihh), Transforming growth factor-beta
275 (Tgf), Bone morphogenic protein (Bmp), Wnt, Notch and others30. Numerous components of
276 these pathways were transcriptionally upregulated in RPMA tumors including Ihh, Bmp family
277 members (Bmp3, 5, 2k, 8a, r2), Tgf receptors (Tgfbr2, 3), Smads (Smad3, 6, 7), Wnt ligands
278 (Wnt2, 2b, 3a, 5a, 6, 7b, 10a, 10b), and Notch receptors and target genes (Notch1, Notch2,
279 Hes1, and Rest) (Fig. 5c). Importantly, network analyses predicted that CTNNB1, HES1,
280 NOTCH1 and REST promote the RPMA fate (Fig. 4b). Thus, we determined Wnt and Notch
281 pathway activity by examining protein levels of targets for which we could identify good
282 antibodies, including CTNNB1 (β-catenin), HES1 and REST (Fig. 5d). Consistent with
283 bioinformatic predictions, CTNNB1, HES1, and REST levels were significantly increased in
284 RPMA compared to RPM tumors regardless of cell of origin (Fig. 5e). Similarly, gene expression
285 data from human SCLC cell lines showed an inverse correlation between ASCL1 and key Wnt
286 and Notch pathway genes (Fig. 5f). Together, these data suggest that ASCL1 represses an
287 MSC/NC-stem like state as well as Wnt and Notch pathway activity in MYC-driven SCLC.
288
289 ASCL1 represses SOX9 in mouse and human SCLC
290 In both the developing limb bud mesenchyme and neural crest, SOX9 marks osteoblast
291 progenitors and precedes RUNX2 expression and bone differentiation30. RUNX1 also precedes
292 RUNX2 expression in developing bone31 and has been suggested to function in chondrogenic
293 lineage commitment of mesenchymal progenitor cells32. RUNX1 has been shown to directly
294 regulate multiple genes important for bone development including Sox9 and Runx233,34. RUNX2
295 and SP7 (i.e. Osterix) are required for differentiation of bone-producing osteoblasts, and RUNX2
296 normally precedes SP7 expression during bone differentiation30,35,36. In previous network
297 analysis, RUNX1, SOX9, and RUNX2 were predicted drivers of the RPMA attractor state (Fig.
298 4b). Therefore, we sought to determine whether the MSC/NCS-like tumor cells express these
12 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
299 factors upon ASCL1 loss. Indeed, while RPM tumors rarely expressed RUNX1, SOX9, or
300 RUNX2, all three factors were dramatically increased in RPMA tumors (Fig. 6a, 6b, and
301 Supplementary Fig. 4a). Interestingly, RUNX1 and SOX9 expression were enriched in non-
302 calcified (soft) RPMA tumors compared to the more differentiated osteosarcoma-like tumors.
303 RUNX2 levels were high in non-calcified (soft) RPMA tumors and remained in more
304 differentiated osteosarcomas. In contrast, SP7 was predominantly expressed in the more
305 differentiated osteosarcoma-like tumors (Fig. 6a, 6b, and Supplementary Fig. 4b). High RUNX2
306 expression was not restricted to RPMA animals with bone-like lesions by microCT. Rather, we
307 identified multiple RPMA-CCSP and SPC animals with RUNX2+ tumors that lacked SP7
308 expression and did not produce osteoid, suggesting that these tumors represent earlier
309 developmental stages prior to formation of calcified bone. Together, these data are consistent
310 with RUNX1 and SOX9’s established roles preceding RUNX2+ and SP7+ differentiation during
311 bone development.
312 To determine whether ASCL1 repression of SOX9, RUNX1, and RUNX2 is conserved in
313 human SCLC cells, we knocked down ASCL1 in human SCLC cell lines (NCI-H2107 and NCI-
314 H889) using siRNAs and examined SOX9, RUNX1 and RUNX2 expression by immunoblot.
315 SOX9 was induced upon ASCL1 knockdown in both cell lines, while RUNX1 levels did not
316 change, and we did not detect induction of RUNX2 (Fig. 6c). Sox9/SOX9 is induced
317 transcriptionally in RPMA tumors and human SCLC cell lines, respectively, suggesting that
318 ASCL1 represses its expression (Fig. 6d and 6e). Analysis of ASCL1 ChIP-seq data did not
319 identify significant ASCL1 binding sites near SOX9 (Supplementary Fig. 4c and 4d), suggesting
320 that repression of SOX9 by ASCL1 is likely indirect. GSEA showed that established SOX9
321 target genes were significantly upregulated in RPMA compared to RPM tumors (Fig. 6f),
322 suggesting that SOX9 activity is increased in RPMA tumors. Furthermore, Ingenuity Pathway
323 Analysis (IPA) identified SOX9 as a top transcriptional regulator implicated in the transcriptional
324 differences between RPMA and RPM tumors (Fig. 6g and Supplementary Table 2). IPA also
13 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
325 predicted other key transcription factor regulators that distinguish RPMA from RPM such as
326 RUNX2, REST, NKX2-1, FOXA2, NOTCH1, and CTNNB1 (Supplementary Table 2), which we
327 verified herein. In human SCLC cell lines, SOX9 expression was enriched in samples that were
328 ASCL1-low (Fig. 6h), and high SOX9 expression also correlated with increased RUNX2
329 expression (Fig. 6i). GSEA revealed that known SOX9 target genes were significantly enriched
330 in ASCL1-low human cell lines (Fig. 6j), and in ASCL1-low human SCLC tumors (Fig. 6k).
331 Together these data reveal that ASCL1 represses a SOX9+ MSC or NCS-like state that has the
332 capacity for bone differentiation.
333
334 Discussion
335 The PNEC has been accepted as a major cell of origin for SCLC. This is consistent with multiple
336 similarities between SCLC and PNECs, and with the capacity of PNECs for SCLC
337 transformation in multiple mouse models18,19,37. Our data here suggest that specific genetic
338 alterations, namely MYC expression in the context of Rb1 and Trp53 loss, can promote SCLC in
339 other cells of origin, including club and alveolar cells, and potentially other cell types as well. It is
340 possible that Ccsp and Spc are targeting CCSP/SPC-double-positive bronchioalveolar stem
341 cells (BASCs)38. While club cells in RP and RPR2 mice were relatively refractory to SCLC
342 development18,19, targeting club cells in RPM mice led to SCLC with short latency, comparable
343 to tumor initiation in NE cells. This suggests that club cells may be particularly susceptible to
344 MYC-mediated SCLC transformation. Recent studies suggest that tuft and basal cells may also
345 serve as additional cells of origin in SCLC8,20. Together with recent findings15, these data are
346 consistent with the notion that cell of origin, genetics, and tumor cell plasticity can determine
347 SCLC phenotype. Moreover, SCLC diagnosis in the clinic does not necessarily mean that
348 tumors arose in NE cells, and this may need to be considered as we develop predictive models
349 for tumor evolution.
14 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
350 Regardless of the cell of origin, RPM tumors demonstrate NE features at the earliest stages
351 of development, suggesting that NE tumors can arise from non-neuroendocrine cells. This
352 suggests that multiple differentiated cells in the adult lung have the capacity to de-differentiate
353 into a neuroendocrine fate. It is notable, however, that some differentiated cell types like AT2
354 cells appear relatively refractory to transformation even in the context of profound oncogenic
355 changes like Rb1/Trp53/Myc; in contrast, AT2 cells are remarkably sensitive to transformation
356 by MAPK pathway activation, as observed upon expression of mutant EGFR or KRAS in
357 GEMMs. Tumor cell fate plasticity can also be appreciated in the clinic as it is known that
358 EGFR-driven lung adenocarcinomas (believed to arise in AT2 cells) can convert to SCLC upon
359 resistance to EGFR inhibitors39,40. Approximately 20-30% of prostate adenocarcinoma that
360 acquires resistance to androgen therapy also converts to a neuroendocrine fate through
361 transdifferentiation41,42, and these tumors frequently harbor RB1 and TP53 loss. Finally,
362 alteration of RB1, TP53, MYC, BCL2, and AKT in prostate or lung basal epithelial cells in vitro
363 can lead to neuroendocrine tumors in xenografts, resembling neuroendocrine prostate cancer
364 and SCLC20. These observations illustrate the potent capacity of oncogenic changes to alter cell
365 fate, and in particular, how RB1, TP53 and MYC cooperate to promote NE fate.
366 Given ASCL1’s role as a lineage-specific oncogene in SCLC that is highly expressed in a
367 significant fraction of tumors, this has prompted its consideration as a therapeutic target43. This
368 notion has been supported by studies showing that genetic deletion of Ascl1 in classic GEMMs
369 abolishes tumor formation12. However, here we find that ASCL1 is not required for MYC-driven
370 tumor development in the RPM model, even though it does appear to be required for
371 neuroendocrine cell fate. An important caveat in both studies is that Ascl1 is deleted at the time
372 of tumor initiation. It remains to be tested whether ASCL1 is necessary for the growth of
373 established tumors, with and without MYC expression, which will require the development of
374 more advanced conditional GEMM systems. ASCL1 is lowly expressed in MYC-driven human
375 SCLC6,10, together suggesting that ASCL1 inhibition may not be sufficient to block the growth of
15 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
376 MYC-driven SCLC. Our data suggest that loss of ASCL1 may simply convert SCLC to an
377 alternative cell fate. Indeed, chemotherapy-relapsed SCLC was found to exhibit significantly
378 reduced ASCL1 expression44, suggesting ASCL1 may not be required for SCLC progression.
379 Interestingly, chemotherapy-relapsed tumors had evidence for a gain in WNT pathway activity44,
380 consistent with our observations here that imply an antagonistic relationship between ASCL1
381 and WNT signaling in SCLC.
382 We were surprised to find that genetic disruption of Ascl1 led to conversion of SCLC to an
383 osteosarcoma-like fate. Both RB1 and TP53 alterations are remarkably common in
384 osteosarcomas (RB1, 70-90%; TP53, 50-70%)45,46. In addition, loss of Rb1 and Trp53 in
385 osteoblasts or MSCs in the mouse promotes tumors that highly resemble human osteosarcoma,
386 including high expression of RUNX245,47,48. During osteogenesis, RUNX2 has been shown to
387 bind pRB1 and HES149-51, and we observe upregulation of HES1 and other indicators of active
388 NOTCH signaling in RPMA tissue. ASCL1 and NOTCH/REST exhibit a mutually antagonistic
389 relationship in multiple contexts52,53. In SCLC, ASCL1 and Notch have an established
390 relationship whereby Notch signaling promotes non-neuroendocrine fate at least partially
391 through induction of the transcriptional co-repressor REST54. We observe induction of
392 NOTCH/REST in RPMA tumors, and we speculate that this event is key to inhibiting NE fate. As
393 NOTCH, WNT and other developmental pathways impinge upon RUNX2 to drive bone fate,
394 future studies will be required to elucidate the signals that promote bone differentiation in the
395 context of ASCL1 loss.
396 Our findings demonstrate that alterations in Rb1, Trp53 and Myc cooperate to
397 dedifferentiate tumor cells to a state that has the potential to be neuroendocrine in the presence
398 of ASCL1, and bone-like in its absence. One outstanding question prompted by these findings
399 regards the nature of this dedifferentiated cell state; during development, the lung epithelium is
400 believed to arise from endodermal progenitors, whereas bone is derived from mesoderm or
401 ectoderm. Therefore, is there a developmental cell type with NE and bone potential that could
16 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
402 be related to SCLC? At least two cell types are known to have the capacity for neural and bone
403 fates during development, MSCs (whose origin is not well understood) and neural crest (NC)
404 stem cells, which can give rise to both neurons and facial bone30,55,56. Gene expression
405 analyses of RPMA tumors suggest that they harbor MSC and NC stem cell-like gene expression
406 signatures. Other “small, round blue cell” tumors such as Ewing’s sarcoma have also been
407 shown to resemble MSCs57. While this issue requires further study, it suggests the surprising
408 notion that in the absence of ASCL1, SCLC can dedifferentiate to a SOX9+ stem-like cell that
409 precedes commitment to endodermal lineages.
410 These findings have led us to question whether SCLC can evolve to a bone-like state in
411 patients. A slow-growing bone phenotype would be under strong negative selective pressure
412 compared to rapidly-growing SCLC, and therefore likely to be an extremely rare event.
413 However, there are rare case reports of patients developing carcinoids with ossification58,59 and
414 extraskeletal osteosarcoma (ESOS) in the lung60, including one ESOS following chemotherapy
415 treatment of SCLC61. It is tempting to speculate that treatments that block ASCL1 potently in the
416 clinic, could push tumor evolution to a bone-like fate, or more likely, to a faster-growing
417 dedifferentiated MSC or NC stem cell-like state. Further study of chemotherapy-relapsed tissue
418 will be required to address this possibility.
419 Finally, we observed that Neurod1 was one of the most downregulated genes upon
420 ASCL1 loss in MYC-driven SCLC. This suggests the intriguing possibility that the NEUROD1+
421 subtype of SCLC requires ASCL1 for development. It has been debated whether the
422 NEUROD1+ SCLC subtype arises in the lung12; both our recent data15 and that shown here
423 suggest that the NEUROD1+ subtype is promoted by MYC but temporally follows ASCL1
424 expression. There are developmental contexts where ASCL1 expression precedes NEUROD1
425 such as in the olfactory epithelium and during adult hippocampal neurogenesis62,63. Thus, we
426 speculate that NEUROD1 expression follows ASCL1 during MYC-driven tumor progression in
17 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
427 SCLC. Together, our study suggests that cancer cells demonstrate remarkable plasticity and
428 capacity to evolve to early developmental programs.
429
430 Methods
431
432 Mice
433 Animal experiments were approved by the Institutional Animal Care and Use Committee at the
434 University of Utah and performed in compliance with all relevant ethical regulations.
435 Rb1fl/fl;Trp53fl/fl;MycT58ALSL/LSL (RPM, JAX #029971) mice were previously described6.
436 Rb1fl/fl;Trp53fl/fl; Rbl2fl/fl;Ascl1fl/fl mice were provided by J.E. Johnson12. Sperm was collected from
437 mice carrying the conditional Ascl1 allele64 and used to cross to our RPM strain through in vitro
438 fertilization. Anesthetized mice at 6-8 weeks of age were infected by intratracheal instillation65
439 with 1x108 plaque-forming units of adenovirus carrying Cre recombinase (University of Iowa).
440 Viruses include: Ad-CMV-Cre (VVC-U of Iowa-5), Ad-CGRP-Cre (VVC-Berns-1160), Ad-
441 CCSP/CC10-Cre (VVC-Berns-1166), and Ad-SPC-Cre (VVC-Berns-1168). Viral infections were
442 performed in a Biosafety Level 2+ room following guidelines from the University of Utah
443 Institutional Biosafety Committee. Male and female mice were divided equally for all
444 experiments. For survival studies, mice were sacrificed due to tumor burden or visibly labored
445 breathing. Mice were euthanized by CO2 asphyxiation followed by necropsy.
446
447 MicroCT imaging
448 Mice were anesthetized with isoflurane and imaged using a small animal Quantum FX or
449 Quantum GX2 microCT (Perkin Elmer). Quantum FX images were acquired with 34 sec scans
450 at 90 kV, 160 µA current, and reconstructed at a 45 µm voxel size. Quantum GX2 images were
451 acquired with 18 sec scans at 90 kV, 88 µA current, and reconstructed at a 90 µm voxel size.
18 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
452 For improved imaging of bone, selected animals were imaged for 2 min on the Quantum GX2 at
453 90 kV, 88 µA current, and reconstructed at a 90 µm voxel size. Frequency of imaging ranged
454 from twice/week to once/month depending on the tumor growth rate. Resulting images were
455 processed with Analyze 11.0 or 12.0 software (Analyze Direct) as previously described6. For
456 bone analysis, acquired images were viewed with the Quantum GX2 Simple Viewer.
457
458 Immunohistochemistry
459 At necropsy, lungs were inflated with PBS, fixed overnight in 10% neutral buffered formalin, and
460 then transferred to 70% ethanol. Samples requiring demineralization were incubated in Rapid-
461 Cal Immuno Decal Solution (6089, BBC Chemical) prior to paraffin embedding. Tissue was cut
462 into 4 µm sections and stained with H&E to assess tumor pathology, or with specific antibodies.
463 Slides were deparaffinized in Citrisolv clearing agent (22-143-975, Fisher Scientific), and
464 rehydrated in an ethanol dilution series. Heat-mediated antigen retrieval was performed in a
465 pressure cooker for 15 min using either 0.01 M citrate buffer pH 6.0 or 0.01 M TE buffer pH 9.0,
466 as suggested by the antibody manufacturer. Endogenous peroxidases were blocked by 10 min
467 incubation with 3% H2O2, prior to 30 min blocking in 5% goat serum/PBST (PBS containing
468 0.1% Tween-20). Primary antibodies were diluted in either 5% goat serum/PBS-T or SignalStain
469 Boost IHC Detection Reagent (8114S, Cell Signaling Technology (CST)), as recommended by
470 the manufacturer. Slides were incubated in primary antibody overnight at 4°C using the
471 Sequenza coverslip staining system (ThermoFisher Scientific). The next day, slides were
472 incubated for 30 min at room temperature with VectaStain secondary antibody (PK-4001, BMK-
473 2202, Vector Laboratories), followed by 30 min incubation in ABC reagent and development
474 with DAB (SK-4100, Vector Laboratories). Slides were counterstained with hematoxylin and
475 mounted with a Shandon ClearVue coverslipper (Thermo Scientific). Primary antibodies include
476 the following: ASCL1 (556604, BD Pharmingen, 1:200), CCSP (AB07623, Millipore Sigma,
477 1:2000), CGRP (C8198, Sigma-Aldrich, 1:250), CTNNB1 (9587S, CST, 1:250), FOXA2
19 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
478 (ab108422, Abcam, 1:1000), HES1 (11988S, CST, 1:600), MYC (sc-764, Santa Cruz, 1:150),
479 NEUROD1 (ab109224, Abcam, 1:150), NKX2-1 (ab76013, Abcam, 1:250), POU2F3
480 (HPA019652-100UL, Sigma-Aldrich, 1:300), REST (ab202962, Abcam, 1:100), RUNX1 (25315-
481 1-AP, Proteintech, 1:250), RUNX2 (ab192256, Abcam, 1:1000), SOX9 (ab185966, Abcam,
482 1:1000), SP7 (ab209484, Abcam, 1:1000), SPC (AB3786, Millipore Sigma, 1:2000), SYP
483 (RB1461P1, Thermo Fisher Scientific, 1:200), UCHL1 (HPA005993-100UL, Sigma-Aldrich,
484 1:250), and YAP1 (14074S, CST, 1:400). Images were acquired on a Nikon Ci-L LED
485 Microscope with DS-Fi3 Camera. Selected slides were digitally scanned with a Pannoramic Midi
486 II (3DHISTECH) slidescanner. Staining was manually quantified by H-score on a scale of 0-300
487 taking into consideration percent positive cells and staining intensity as described66, where H
488 Score = % of positive cells multiplied by intensity score of 0-3. For example, a tumor with 80%
489 positive cells with high intensity of 3 = 240 H-Score. For SOX9 expression, scanned IHC slides
490 were quantified by H-score using CaseViewer software (3DHISTECH) and an automated
491 nuclear quantification program after manual identification of tumor regions.
492
493 Mouse tumor RNA-seq
494 RNA isolation from ~15 mg flash-frozen RPM (n=11), un-calcified RPMA (n=7), and RPR2 (n=6)
495 primary tumors was performed using RNeasy Mini Kit (Qiagen) with the standard protocol. RNA
496 from RPM (n=11) and RPR2 (n=2) tumors was subject to library construction with the Illumina
497 TruSeq Stranded mRNA Sample Preparation Kit (RS-122-2101, RS-122-2102) according to the
498 manufacturer’s protocol. RNA from RPR2 (n=4) and RPMA (n=7) tumors was subject to library
499 construction with the Illumina TruSeq Stranded Total RNA Library Ribo-Zero Gold Prep kit (RS-
500 122-2301) according to the manufacturer’s protocol. Chemically denatured sequencing libraries
501 (25 pM) from RPM (n=11) and RPR2 (n=6) tumors were applied to an Illumina HiSeq v4 single
502 read flow cell using an Illumina cBot. Hybridized molecules were clonally amplified and
503 annealed to sequencing primers with reagents from an Illumina HiSeq SR Cluster Kit v4-cBot
20 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
504 (GD-401-4001). Following transfer of the flowcell to an Illumina HiSeq 2500 instrument
505 (HCSv2.2.38 and RTA v1.18.61), a 50-cycle single-end sequence run was performed using
506 HiSeq SBS Kit v4 sequencing reagents (FC-401-4002). Chemically denatured sequencing
507 libraries (1.15 nM) from RPMA tumors (n=7) were applied to an Illumina NovaSeq S2 flow cell
508 using the standard workflow. Hybridized molecules were clonally amplified and annealed to
509 sequencing primers with reagents from an Illumina NovaSeq 6000 S2 Reagent Kit (20012861).
510 Following transfer of the flowcell to an Illumina NovaSeq 6000 instrument (20012850), a 50-
511 cycle paired-end sequence run was performed.
512
513 Bioinformatics
514 Fastq raw count files were aligned in the R statistical environment (v3.6). The mouse GRCm38
515 FASTA and GTF files were downloaded from Ensembl release 94 and the reference database
516 was created using STAR version 2.6.1b67 with splice junctions optimized for 50 base pair reads.
517 Optical duplicates were removed using clumpify v38.34 and adapters were trimmed using
518 cutadapt 1.1668. The trimmed reads were aligned to the reference database using STAR in two
519 pass mode to output a BAM file sorted by coordinates. Mapped reads were assigned to
520 annotated genes in the GTF file using featureCounts version 1.6.369. The output files from
521 cutadapt, FastQC, Picard CollectRnaSeqMetrics, STAR and featureCounts were summarized
522 using MultiQC70 to check for any sample outliers. To remove sources of unwanted variation from
523 tumor RNA-seq library preparation and sequencing platforms, all non-coding features, histones,
524 and ribosomal RNAs were removed for downstream analyses. The featureCount output files
525 were combined into a single raw count matrix. Differentially expressed genes (DEGs) were
526 identified using a 5% false discovery rate with DESeq2 version 1.24.071 (Supplementary Table
527 1). These data have been deposited in NCI GEO: GSE155692. PCA was performed on the first
528 two principle components using the regularized log count (rlog) values of the top 500 variable
529 genes. Log2(+1)-transformed, normalized intensity values were obtained to create heat maps
21 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
530 with unsupervised hierarchical clustering. A previously-described NE marker gene set6 was
531 used to compare RPMA (n=6) and RPM (n=11) tumor expression. Heat maps were generated
532 using Pheatmap version 1.0.12. Semantic similarity-based scatterplot visualizations of enriched
533 and depleted GO Biological Processes in RPMA vs RPM tumors were created with the REViGO
534 tool using GO Biological Processes input lists and corresponding p-values72. Scatterplot view
535 visualizes the GO terms in a “semantic space” where the more similar terms are positioned
536 closer together. The color of the bubble and size reflects the log 10 p-value obtained in the
537 Panther analysis. Input GO Biological Processes and log10 p-values were generated using
538 Panther v14 GO Enrichment Analysis tool for biological processes73 on enriched (>2.5 log2FC
539 by DESeq2) or depleted (<-2.5 log2FC by DESeq2) genes in RPMA vs RPM tumors. REViGO
540 and Panther GO Enrichment Analysis details and curated input gene lists are presented in
541 Supplementary Table 1.
542
543 GSEA
544 Gene set enrichment analysis (GSEA) was performed using GSEA version 3.0 software with
545 default parameters, classic enrichment, inclusion gene set size between 15 and 1000, and the
546 phenotype permutation at 1,000 times. GSEA was performed on a pre-ranked gene list
547 representing log2 fold change of RPMA vs RPM tumor bulk RNA-seq expression values
548 (Supplementary Table 1). Normalized enrichment scores and p values are shown below each
549 respective GSEA plot in the figures. A catalog of functional gene sets from Molecular Signature
550 Database (MSigDB, version 6.2, July 2018, www.broad.mit.edu/gsea/msigdb/
551 msigdb_index.html) was used for the ‘‘GO Bone Development”, “Mesenchymal stem cell” and
552 “Neural crest signature” gene sets. Gene sets for “ASCL1 Targets by ChIP-Seq” and
553 “NEUROD1 Targets by ChIP-Seq” represent conserved transcriptional targets identified by
554 ChIP-Seq from mouse SCLC tumor and human cell line SCLC models. The ASCL1 conserved
555 target gene list was previously published (GEO: GSE69398)12. The NEUROD1 conserved target
22 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
556 gene list was derived from published ChIP-seq data from human SCLC cell lines (GEO:
557 GSE69398)12 and NEUROD1 ChIP-seq data from RPM mouse tumors described in this study.
558 The “SOX9 target genes” list was derived from published mouse basal cell carcinoma ChIP-seq
559 data (GEO: GSE68755)74. Details of the curated gene sets are presented in Supplementary
560 Table 1.
561
562 ASCL1 and NEUROD1 ChIP-seq
563 Preparation of mouse lung tumors for ChIP-seq was performed as previously described7. Briefly,
564 20 million cells or ~250 g of mouse tumor per ChIP were crosslinked in 1% formaldehyde for
565 10 min at room temperature. Crosslinking was stopped with 125 mM glycine and nuclei were
566 extracted. Chromatin was sonicated using an Epishear Probe Sonicator (Active Motif) for 4 min
567 at 40% power. ASCL1 antibody (556604, BD Pharmingen), H3K27Ac antibody (39133, Active
568 Motif), or NEUROD1 antibody (sc-1084, Santa Cruz Biotechnology) were used for IP. An input
569 sample of each RPM tumor served as the control. Libraries were sequenced on an Illumina
570 HiSeq 2500 as single-end 50 bp reads to a minimum depth of 35 million reads per sample.
571 Reads were aligned to the mm10 build of the mouse genome with bowtie75 using the following
572 parameters: -m 1 -t --best -q -S -l 32 -e 80 -n 2. Peaks were called with MACS276 using a p
573 value cutoff of 10-10 and the mfold parameter bounded between 15 and 100. For visualization,
574 MACS2 produced bedgraphs with the –B and –SPMR options. Binding profiles were visualized
575 using Integrated Genome Viewer (IGV version 2.6.3) aligned to mm10 genome build. ASCL1
576 and NEUROD1 mouse ChIP-seq data are available on GEO: GSE155692. H3k27Ac mouse
577 ChIP-seq data are available on GEO: GSE1424967.
578 Human ChIP-seq for ASCL1 was performed as described in Borromeo et al., 2016 and
579 data are published in GEO with accession number GEO: GSE69398. Binding profiles of human
23 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
580 ASCL1 ChIP-seq were visualized using Integrated Genome Viewer (IGV version 2.6.3) aligned
581 to hg19 human genome build.
582
583 Weighted Gene Coexpression Network Analysis
584 WGCNA77 was performed on RPM (n=11) and RPMA (n = 6) RNA seq data using the R
585 package ‘WGCNA’. Low variance genes, with coefficient of variation < 0.2 and median absolute
586 deviation < 1, were removed, resulting in 15,286 genes across 17 samples. A signed network of
587 genes, such that only positively correlated genes will be grouped into gene modules (negative
588 correlations are given a score of 0), was constructed as follows. WGCNA’s pickSoftThreshold
589 function was used to determine the exponent of the correlation coefficients matrix that best
590 produces a scale free network when the matrix is used as weights of network connections. A
591 threshold of 15 was chosen to produce this weights matrix, with the correlation function
592 “Pearson.” A topological overlap matrix was then generated based on the overlap in
593 connections for two genes. This gave a measure of distance between each pair of genes, which
594 was used in average-linkage hierarchical clustering. The ‘cutTreeDynamic’ function was used,
595 with minimum cluster size of 100 to generate clusters, or modules, of coexpressed genes.
596 WGCNA’s ‘sampledBlockwiseModules’ was then used to increase robustness of gene modules,
597 which merges or removes genes from modules that are not distinct enough in a majority of the
598 resampled module labels. Differentially-expressed gene modules were determined using an
599 ANOVA statistical test between RPM and RPMA tumors. Three gene modules were significant.
600
601 Network Structure Inference
602 Network structure was inferred as described in Wooten et al (2019)22. Briefly, transcription
603 factors were filtered to include those differentially expressed between RPM and RPMA tumors
24 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
604 with Log Fold Change > 1.5 and adjusted p value < 0.05. Genes were then filtered further to
605 include only those most central to each differentially-expressed gene module, calculated as a
606 module membership value (kME) from the WGCNA analysis. Lastly, expert knowledge was
607 used to add any transcription factors that were deemed central to RPM/RPMA cell identity,
608 including FOXA2, SOX9, and CTNNB1. Network connections were added using the Enrichr tool
609 from the Ma’ayan laboratory78,79, which mines various ChIP-seq databases for TF-gene
610 interactions including CHEA, ENCODE, and TRANSFAC. Gene nodes with no downstream
611 nodes (child nodes) were removed, as they do not affect the dynamics of the network, except
612 NKX2-1, INSM1, and SP7, which were kept as biomarkers.
613
614 Network Rule Inference and Attractors
615 Network rules of interaction were calculated using BooleaBayes as previously described22,
616 using the 11 RPM and 6 RPMA samples. For each TF node in the network, a rule of expression
617 was determined for any ON/OFF combination of parent nodes. BooleaBayes simulations
618 consisted of asynchronous directed random walks, with probabilities of transition between states
619 determined by TF rules. By simulating the network starting at each sampled state, we found
620 pseudo-attractors (here, referred to as “attractors) as previously defined22. Searches for
621 attractor states from each initial condition were cut-off at a TF-basin radius of 4 steps, and the
622 phenotypes of any attractors were determined by distance to initial condition. Two attractors
623 were found using this method, one for each phenotype. Driving transcription factors for each
624 attractor, or those whose silencing have a destabilizing effect on the attractor, were found using
625 in silico perturbations22.
626
627 ASCL1 knockdown and immunoblot
25 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
628 Human classic SCLC cell lines NCI-H2107 and NCI-H889 were obtained from ATCC. Cells
629 were transfected with human ASCL1 stealth siRNA (129901-HSS100744, Thermo-Fisher) using
630 X-tremeGENE siRNA Transfection Reagent (4476115001, Sigma) according to the
631 manufacturer’s protocol. Cells were harvested 72 hr after transfection for analysis by
632 immunoblot. Cell pellets were flash frozen and stored at -80°C until use. Total protein lysates
633 were prepared as previously described, separated via SDS-PAGE and transferred to a PVDF
634 membrane80. Membranes were blocked for 1 hr in 5% milk followed by overnight incubation with
635 primary antibodies at 4°C. Membranes were washed for 4 x 10 min at RT in TBS-T. Mouse and
636 rabbit HRP-conjugated secondary antibodies (711-035-152 and 115-035-205, Jackson
637 ImmunoResearch, 1:10,000) were incubated for 1 hr in 5% milk at RT followed by washing 4 x
638 10 min at RT in TBS-T. Membranes were exposed to WesternBright HRP Quantum substrate
639 (Advansta) and detected on Hyblot CL film (Denville Scientific Inc). Primary antibodies include:
640 ASCL1 (556604, BD Pharmingen, 1:300), SOX9 (ab185966, Abcam, 1:1000), RUNX1 (25315-
641 1-AP, Proteintech, 1:1000), RUNX2 (12556, CST, 1:1000), with HSP90 (4877, CST, 1:5000) as
642 a loading control.
643
644 PCR for Cre recombination efficiency
645 Approximately 15 mg of flash-frozen tumor tissue was cut on dry ice and processed with the
646 Qiagen DNeasy kit to isolate tumor genomic DNA. DNA concentrations were measured on a
647 BioTek Synergy HT plate reader. Equal quantities of tumor genomic DNA were amplified by
648 PCR with GoTaq (M7123, Promega) using the gene-specific primers listed. For Ascl1: Common
649 Fwd (MF1) 5’-CTACTGTCCAAACGCAAAGTGG-3’; WT Rev (MR1) 5’-
650 GCTCCCACAATCCTCGTAAAGA-3’; Flox Rev (VR2) 5’-TAGACGTTGTGGCTGTTGTAGT-3’;
651 and Recombined Fwd (5’ UTR Fwd) 5’-AACTTTCCTCCGGGGCTCGTTTC-3’. PCR conditions
652 were: 94 deg 5 min, 30 cycles of (94 deg 1 min, 64 deg 1.5 min, 72 deg 1 min), 72 deg 10 min,
26 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
653 hold at 4 deg. Expected band sizes are as follows: 5’ UTR Fwd-VR2 (recombined allele) = ~700
654 - 850 bp, MF1-VR2 (floxed allele) = 857 bp, MF1-MR2 (wild-type allele) = 411 bp.
655 Primers to detect Rb1 recombination include the following: D1 5'-
656 GCAGGAGGCAAAAATCCACATAAC-3', 1lox 5' 5’-CTCTAGATCCTCTCATTCTTCCC-3’, and 3'
657 lox 5’-CCTTGACCATAGCCCAGCAC-3’. PCR conditions used were 94 deg 3 min, 30 cycles of
658 (94 deg 30 sec, 55 deg 1 min, 72 deg 1.5 min), 72 deg 5 min, hold at 4 deg. Expected band
659 sized were ~500 bp for the recombined allele, and 310 bp for the floxed allele.
660 Primers to detect Trp53 recombination include the following: A 5'-
661 CACAAAAACAGGTTAAACCCAG-3', B 5'-AGCACATAGGAGGCAGAGAC-3', and D 5'-
662 GAAGACAGAAAAGGGGAGGG-3'. PCR conditions were 94 deg 2 min, (30 cycles of 94 deg 30
663 sec, 58 deg 30 sec, 72 deg 50 sec), 72 deg 5 min, hold at 4 deg. Expected band sizes were 612
664 bp for the recombined allele, and 370 bp for the floxed allele.
665 PCR products were run on 1-2% agarose/TAE gels containing ethidium bromide and images
666 were acquired using an Azure Biosystem C200 imager.
667 668 Statistics
669 Statistical analysis was performed with GraphPad Prism 8. Mouse survival studies were
670 analyzed with Mantel-Cox log-rank tests. Immunohistochemistry quantification and normalized
671 RNAseq counts were compared by non-parametric two-tailed Mann-Whitney t-tests.
672 Measurements were taken from distinct tumors in multiple mice for each experiment, as
673 indicated in the figure legend. P values < 0.05 were considered statistically significant. Error
674 bars represent mean ± standard deviation unless otherwise specified.
675
676 Acknowledgements
677 We appreciate Dr. Anton Berns for permission to use Ad-Cre viruses deposited at University of
678 Iowa. We are grateful to the late Dr. Adi Gazdar, who provided insightful pathological
27 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
679 observations for this study. We thank R. Dahlgren and L. Houston for administrative assistance
680 and members of the Oliver lab, especially D. Hansen, P. Ballieu, and A. Micinski for mouse
681 colony management. We acknowledge support from the National Cancer Institute (NCI) of the
682 National Institutes of Health (NIH) under award P30CA042014 to Huntsman Cancer Institute for
683 the use of core facilities and shared resources, including Biorepository and Molecular
684 Pathology, and High-Throughput Genomics and Bioinformatic Analysis. This work was
685 supported by the NIH NCI (R21-CA216504-01A1; U01-CA231844; U01-CA213338).
686
687 Author Contributions
688 Conceptualization: R.R. Olsen and T.G. Oliver
689 Data Acquisition: R.R. Olsen, D.W. Kastner, A.S. Ireland, K. Pozo, S. Wait, M.R. Guthrie, C.P.
690 Whitney, D. Soltero
691 Data Analysis and Interpretation: R.R. Olsen, D.W. Kastner, A.S. Ireland, S.M. Groves, B.L.
692 Witt, V. Quaranta, T.G. Oliver
693 Writing – Original Draft: R.R. Olsen, A.S. Ireland, T.G. Oliver
694 Writing – Review and Editing: R.R. Olsen, A.S. Ireland, S.M. Groves, V. Quaranta, J.E.
695 Johnson, T.G. Oliver
696 Funding Acquisition: T.G. Oliver
697 Resources: J.E. Johnson
698 Supervision: V. Quaranta, J.E. Johnson, T.G. Oliver
699
700 Competing Interests Statement
701 TGO has pending patent applications related to subtype stratification of SCLC: US16/335368;
702 JP2019522392, EP2017865057.
703
704 Resource availability
28 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
705 Corresponding author
706 Further information and requests for resources and reagents should be directed to and will be
707 fulfilled by the corresponding author, Trudy G. Oliver ([email protected]).
708
709 Materials availability
710 The RPM mice used in this study are deposited at The Jackson Laboratory, JAX #029971.
711 RPMA mice are available upon request with an MTA.
712
713 Data availability
714 All software is commercially available or cited in previous publications. Mouse RNA-seq and
715 ChIP-seq data unique to this study are deposited in NCBI GEO: GSE155692. Associated raw
716 data is available for Fig. 3-6 and is provided in Supplementary Tables 1-2.
717 718
719 References
720 1 Rudin, C. M. et al. Molecular subtypes of small cell lung cancer: a synthesis of 721 human and mouse model data. Nat. Rev. Cancer 19, 289-297 (2019).
722 2 Sabari, J. K., Lok, B. H., Laird, J. H., Poirier, J. T. & Rudin, C. M. Unravelling the 723 biology of SCLC: implications for therapy. Nat. Rev. Clin. Oncol. 14, 549-561 724 (2017).
725 3 Bragelmann, J. et al. Family matters: How MYC family oncogenes impact small 726 cell lung cancer. Cell Cycle 16, 1489-1498 (2017).
727 4 George, J. et al. Comprehensive genomic profiles of small cell lung cancer. 728 Nature 524, 47-53 (2015).
729 5 Gazdar, A. F., Bunn, P. A. & Minna, J. D. Small-cell lung cancer: what we know, 730 what we need to know and the path forward. Nat. Rev. Cancer 17, 725-737 731 (2017).
732 6 Mollaoglu, G. et al. MYC Drives Progression of Small Cell Lung Cancer to a 733 Variant Neuroendocrine Subtype with Vulnerability to Aurora Kinase Inhibition. 734 Cancer Cell 31, 270-285 (2017).
29 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
735 7 Chalishazar, M. D. et al. MYC-Driven Small-Cell Lung Cancer is Metabolically 736 Distinct and Vulnerable to Arginine Depletion. Clin. Cancer Res. 25, 5107-5121 737 (2019).
738 8 Huang, Y. H. et al. POU2F3 is a master regulator of a tuft cell-like variant of small 739 cell lung cancer. Genes Dev. 32, 915-928 (2018).
740 9 Sen, T. et al. CHK1 Inhibition in Small-Cell Lung Cancer Produces Single-Agent 741 Activity in Biomarker-Defined Disease Subsets and Combination Activity with 742 Cisplatin or Olaparib. Cancer Res. 77, 3870-3884 (2017).
743 10 Cardnell, R. J. et al. Protein expression of TTF1 and cMYC define distinct 744 molecular subgroups of small cell lung cancer with unique vulnerabilities to 745 aurora kinase inhibition, DLL3 targeting, and other targeted therapies. 746 Oncotarget 8, 73419-73432 (2017).
747 11 Dammert, M. A. et al. MYC paralog-dependent apoptotic priming orchestrates a 748 spectrum of vulnerabilities in small cell lung cancer. Nat. Commun. 10, 3485 749 (2019).
750 12 Borromeo, M. D. et al. ASCL1 and NEUROD1 Reveal Heterogeneity in 751 Pulmonary Neuroendocrine Tumors and Regulate Distinct Genetic Programs. 752 Cell Rep. 16, 1259-1272 (2016).
753 13 Ito, T. et al. Basic helix-loop-helix transcription factors regulate the 754 neuroendocrine differentiation of fetal mouse pulmonary epithelium. 755 Development 127, 3913-3921 (2000).
756 14 Borges, M. et al. An achaete-scute homologue essential for neuroendocrine 757 differentiation in the lung. Nature 386, 852-855 (1997).
758 15 Ireland, A. S. et al. MYC Drives Temporal Evolution of Small Cell Lung Cancer 759 Subtypes by Reprogramming Neuroendocrine Fate. Cancer Cell 38, 60-78.e12 760 (2020).
761 16 Ferone, G. et al. FGFR1 Oncogenic Activation Reveals an Alternative Cell of 762 Origin of SCLC in Rb1/p53 Mice. Cell Rep. 30, 3837-3850.e3833 (2020).
763 17 Yang, D. et al. Intertumoral Heterogeneity in SCLC Is Influenced by the Cell Type 764 of Origin. Cancer Discov. 8, 1316-1331 (2018).
765 18 Sutherland, K. D. et al. Cell of origin of small cell lung cancer: inactivation of 766 Trp53 and Rb1 in distinct cell types of adult mouse lung. Cancer Cell 19, 754-764 767 (2011).
768 19 Park, K. S. et al. Characterization of the cell of origin for small cell lung cancer. 769 Cell Cycle 10, 2806-2815 (2011).
30 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
770 20 Park, J. W. et al. Reprogramming normal human epithelial tissues to a common, 771 lethal neuroendocrine cancer lineage. Science 362, 91-95 (2018).
772 21 Ferone, G. et al. SOX2 Is the Determining Oncogenic Switch in Promoting Lung 773 Squamous Cell Carcinoma from Different Cells of Origin. Cancer Cell 30, 519- 774 532 (2016).
775 22 Wooten, D. J. et al. Systems-level network modeling of Small Cell Lung Cancer 776 subtypes identifies master regulators and destabilizers. PLoS Comput. Biol. 15, 777 e1007343 (2019).
778 23 Tlemsani, C. et al. SCLC-CellMiner: A Resource for Small Cell Lung Cancer Cell 779 Line Genomics and Pharmacology Based on Genomic Signatures. Cell Rep. 33, 780 108296 (2020).
781 24 Christensen, C. L. et al. Targeting transcriptional addictions in small cell lung 782 cancer with a covalent CDK7 inhibitor. Cancer Cell 26, 909-922 (2014).
783 25 Pozo, K. et al. Lineage transcription factors co-regulate subtype-specific genes 784 providing a roadmap for systematic identification of small cell lung cancer 785 vulnerabilities. Preprint at 786 https://www.biorxiv.org/content/10.1101/2020.08.13.249029v1. (2020).
787 26 Zepp, J. A. & Morrisey, E. E. Cellular crosstalk in the development and 788 regeneration of the respiratory system. Nat. Rev. Mol. Cell Biol. 20, 551-566 789 (2019).
790 27 Serizawa, T. et al. Developmental analyses of mouse embryos and adults using 791 a non-overlapping tracing system for all three germ layers. Development 146 792 (2019).
793 28 Achilleos, A. & Trainor, P. A. Neural crest stem cells: discovery, properties and 794 potential for therapy. Cell Res. 22, 288-304 (2012).
795 29 Jiang, Y. et al. Pluripotency of mesenchymal stem cells derived from adult 796 marrow. Nature 418, 41-49 (2002).
797 30 Long, F. Building strong bones: molecular regulation of the osteoblast lineage. 798 Nat. Rev. Mol. Cell Biol. 13, 27-38 (2011).
799 31 Lian, J. B. et al. Runx1/AML1 hematopoietic transcription factor contributes to 800 skeletal development in vivo. J. Cell. Physiol. 196, 301-311 (2003).
801 32 Smith, N. et al. Overlapping expression of Runx1(Cbfa2) and Runx2(Cbfa1) 802 transcription factors supports cooperative induction of skeletal development. J. 803 Cell. Physiol. 203, 133-143 (2005).
31 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
804 33 Soung do, Y. et al. Runx1 dose-dependently regulates endochondral ossification 805 during skeletal development and fracture healing. J. Bone Miner. Res. 27, 1585- 806 1597 (2012).
807 34 Tang, J. et al. Runt-related transcription factor 1 is required for murine osteoblast 808 differentiation and bone formation. J. Biol. Chem. 295, 11669-11681 (2020).
809 35 Komori, T. et al. Targeted disruption of Cbfa1 results in a complete lack of bone 810 formation owing to maturational arrest of osteoblasts. Cell 89, 755-764 (1997).
811 36 Nakashima, K. et al. The novel zinc finger-containing transcription factor osterix 812 is required for osteoblast differentiation and bone formation. Cell 108, 17-29 813 (2002).
814 37 Song, H. et al. Functional characterization of pulmonary neuroendocrine cells in 815 lung development, injury, and tumorigenesis. Proc Natl Acad Sci U S A 109, 816 17531-17536 (2012).
817 38 Kim, C. F. et al. Identification of bronchioalveolar stem cells in normal lung and 818 lung cancer. Cell 121, 823-835 (2005).
819 39 Marcoux, N. et al. EGFR-Mutant Adenocarcinomas That Transform to Small-Cell 820 Lung Cancer and Other Neuroendocrine Carcinomas: Clinical Outcomes. J. Clin. 821 Oncol. 37, 278-285 (2019).
822 40 Oser, M. G., Niederst, M. J., Sequist, L. V. & Engelman, J. A. Transformation 823 from non-small-cell lung cancer to small-cell lung cancer: molecular drivers and 824 cells of origin. Lancet Oncol. 16, e165-172 (2015).
825 41 Oser, M. G. & Janne, P. A. Small-Cell Neuroendocrine Tumors: Cell State 826 Trumps the Oncogenic Driver. Clin. Cancer Res. 24, 1775-1776 (2018).
827 42 Terry, S. & Beltran, H. The many faces of neuroendocrine differentiation in 828 prostate cancer progression. Front. Oncol. 4, 60 (2014).
829 43 Augustyn, A. et al. ASCL1 is a lineage oncogene providing therapeutic targets for 830 high-grade neuroendocrine lung cancers. Proc Natl Acad Sci U S A 111, 14788- 831 14793 (2014).
832 44 Wagner, A. H. et al. Recurrent WNT pathway alterations are frequent in relapsed 833 small cell lung cancer. Nat. Commun. 9, 3787 (2018).
834 45 Berman, S. D. et al. Metastatic osteosarcoma induced by inactivation of Rb and 835 p53 in the osteoblast lineage. Proc Natl Acad Sci U S A 105, 11851-11856 836 (2008).
837 46 Velletri, T. et al. P53 functional abnormality in mesenchymal stem cells promotes 838 osteosarcoma development. Cell Death Dis. 7, e2015 (2016).
32 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
839 47 Walkley, C. R. et al. Conditional mouse osteosarcoma, dependent on p53 loss 840 and potentiated by loss of Rb, mimics the human disease. Genes Dev. 22, 1662- 841 1676 (2008).
842 48 Calo, E. et al. Rb regulates fate choice and lineage commitment in vivo. Nature 843 466, 1110-1114 (2010).
844 49 Berman, S. D. et al. The retinoblastoma protein tumor suppressor is important for 845 appropriate osteoblast differentiation and bone development. Mol. Cancer Res. 846 6, 1440-1451 (2008).
847 50 Thomas, D. M. et al. The retinoblastoma protein acts as a transcriptional 848 coactivator required for osteogenic differentiation. Mol. Cell 8, 303-316 (2001).
849 51 Lee, J. S. et al. HES1 cooperates with pRb to activate RUNX2-dependent 850 transcription. J. Bone Miner. Res. 21, 921-933 (2006).
851 52 Masserdotti, G. et al. Transcriptional Mechanisms of Proneural Factors and 852 REST in Regulating Neuronal Reprogramming of Astrocytes. Cell Stem Cell 17, 853 74-88 (2015).
854 53 Harris, L. & Guillemot, F. HES1, two programs: promoting the quiescence and 855 proliferation of adult neural stem cells. Genes Dev. 33, 479-481 (2019).
856 54 Lim, J. S. et al. Intratumoural heterogeneity generated by Notch signalling 857 promotes small-cell lung cancer. Nature 545, 360-364 (2017).
858 55 Bruder, S. P., Fink, D. J. & Caplan, A. I. Mesenchymal stem cells in bone 859 development, bone repair, and skeletal regeneration therapy. J. Cell. Biochem. 860 56, 283-294 (1994).
861 56 Hutchins, E. J. et al. Migration and diversification of the vagal neural crest. Dev. 862 Biol. 444 Suppl 1, S98-s109 (2018).
863 57 Tirode, F. et al. Mesenchymal stem cell features of Ewing tumors. Cancer Cell 864 11, 421-429 (2007).
865 58 Tsubochi, H., Endo, S., Oda, Y. & Dobashi, Y. Carcinoid tumor of the lung with 866 massive ossification: report of a case showing the evidence of osteomimicry and 867 review of the literature. Int. J. Clin. Exp. Pathol. 6, 957-961 (2013).
868 59 Özşen, M., Yalçinkaya, U., Akyildiz, E., Bayram, A. S. & Gökalp, G. Bronchial 869 Carcinoid Tumors with Massive Osseous Metaplasia: A Case Report and Review 870 of the Literature. Turk. Patoloji. Derg. 36, 159-163 (2020).
871 60 Qian, J. et al. Primary thoracic extraskeletal osteosarcoma: a case report and 872 literature review. J. Thorac. Dis. 9, E1088-e1095 (2017).
33 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
873 61 Takamura, K., Ogi, T., Yamamoto, M. & Kikuchi, K. Extraskeletal osteosarcoma 874 of the lung following treatment of primary small-cell lung carcinoma with 875 chemoradiotherapy: A case report. Mol. Clin. Oncol. 8, 99-102 (2018).
876 62 Schwob, J. E. et al. Stem and progenitor cells of the mammalian olfactory 877 epithelium: Taking poietic license. J. Comp. Neurol. 525, 1034-1054 (2017).
878 63 Kim, E. J., Ables, J. L., Dickel, L. K., Eisch, A. J. & Johnson, J. E. Ascl1 (Mash1) 879 defines cells with long-term neurogenic potential in subgranular and 880 subventricular zones in adult mouse brain. PLoS One 6, e18472 (2011).
881 64 Pacary, E. et al. Proneural transcription factors regulate different steps of cortical 882 neuron migration through Rnd-mediated inhibition of RhoA signaling. Neuron 69, 883 1069-1084 (2011).
884 65 DuPage, M., Dooley, A. L. & Jacks, T. Conditional mouse lung cancer models 885 using adenoviral or lentiviral delivery of Cre recombinase. Nat. Protoc. 4, 1064- 886 1072 (2009).
887 66 Flowers, J. L. et al. Use of monoclonal antiestrogen receptor antibody to evaluate 888 estrogen receptor content in fine needle aspiration breast biopsies. Ann. Surg. 889 203, 250-254 (1986).
890 67 Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15- 891 21 (2013).
892 68 Martin, M. Cutadapt removes adapter sequences from high-throughput 893 sequencing reads. EMBnet.journal 17 (2011).
894 69 Liao, Y., Smyth, G. K. & Shi, W. featureCounts: an efficient general purpose 895 program for assigning sequence reads to genomic features. Bioinformatics 30, 896 923-930 (2014).
897 70 Ewels, P., Magnusson, M., Lundin, S. & Kaller, M. MultiQC: summarize analysis 898 results for multiple tools and samples in a single report. Bioinformatics 32, 3047- 899 3048 (2016).
900 71 Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and 901 dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).
902 72 Supek, F., Bosnjak, M., Skunca, N. & Smuc, T. REVIGO summarizes and 903 visualizes long lists of gene ontology terms. PLoS One 6, e21800 (2011).
904 73 Mi, H., Muruganujan, A., Ebert, D., Huang, X. & Thomas, P. D. PANTHER 905 version 14: more genomes, a new PANTHER GO-slim and improvements in 906 enrichment analysis tools. Nucleic Acids Res. 47, D419-d426 (2019).
34 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
907 74 Larsimont, J. C. et al. Sox9 Controls Self-Renewal of Oncogene Targeted Cells 908 and Links Tumor Initiation and Invasion. Cell Stem Cell 17, 60-73 (2015).
909 75 Langmead, B., Trapnell, C., Pop, M. & Salzberg, S. L. Ultrafast and memory- 910 efficient alignment of short DNA sequences to the human genome. Genome Biol. 911 10, R25 (2009).
912 76 Zhang, Y. et al. Model-based analysis of ChIP-Seq (MACS). Genome Biol. 9, 913 R137 (2008).
914 77 Langfelder, P. & Horvath, S. WGCNA: an R package for weighted correlation 915 network analysis. BMC Bioinformatics 9, 559 (2008).
916 78 Kuleshov, M. V. et al. Enrichr: a comprehensive gene set enrichment analysis 917 web server 2016 update. Nucleic Acids Res. 44, W90-97 (2016).
918 79 Chen, E. Y. et al. Enrichr: interactive and collaborative HTML5 gene list 919 enrichment analysis tool. BMC Bioinformatics 14, 128 (2013).
920 80 Oliver, T. G. et al. Caspase-2-mediated cleavage of Mdm2 creates a p53- 921 induced positive feedback loop. Mol. Cell 43, 57-71 (2011). 922
35 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Fig. 1 a b 100 CMV (n=9) 80 CGRP (n=14) * CMV-Cre CGRP-Cre CCSP-Cre SPC-Cre CCSP (n=10) **** 60 SPC (n=19) 40
20 Percent surviva l 0 0 50 100 150 200 250 300 Days post-infection c NE non-NE ASCL1 CGRP UCHL1 SYP NEUROD1 YAP1 POU2F3 CMV-Cre
5
7 CGRP-Cre In situ tumors CCSP-Cre SPC-Cre d NE non-NE ASCL1 CGRP UCHL1 SYP NEUROD1 YAP1 POU2F3 CMV-Cre
5
7 CGRP-Cre Invasive tumors CCSP-Cre SPC-Cre e **** **** **** **** **** **** ** 300 300 300 300 300 300 300
200 200 200 200 200 200 200 -sc ore H-scor e H-scor e 100 100 100 100 100 100 100 1 H AP 1 SY P Y CGR P ASCL1 H-scor e UCHL1 H-scor e POU2F3 H-scor e NEUROD1 H-scor e 0 0 0 0 0 0 0 In situ Invasive In situ Invasive In situ Invasive In situ Invasive In situ Invasive In situ Invasive In situ Invasive bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
921 Main figure legends
922 Figure 1: MYC-driven SCLC can arise in multiple lung cell types and initially expresses
923 ASCL1
924 a) Survival of RPM mice infected with indicated cell-type-specific Ad-Cre viruses. Number
925 of mice indicated in the figure. Mantel-Cox log-rank test, * p < 0.05; **** p < 0.0001.
926 b) Representative H&E histology of SCLC from RPM mice initiated with indicated Ad-Cre
927 viruses. Scale bar: 25 µm.
928 c) Representative immunohistochemistry (IHC) of early in situ tumor lesions for indicated
929 antibodies (top) in RPM mice infected with indicated Ad-Cre viruses (left). Scale bar: 25
930 µm.
931 d) Representative IHC of large, invasive tumors for indicated antibodies (top) in RPM mice
932 infected with indicated Ad-Cre viruses (left). Scale bar: 25 µm.
933 e) H-score quantification of IHC in panels (c) and (d). H-Score = % of positive cells
934 multiplied by intensity score of 0-3; see Methods. Approximately 70-200 tumors from n ≥
935 3 mice per condition were quantified. Data is shown as mean ± standard deviation (SD).
936 Mann-Whitney two-tailed t-tests, **** p < 0.0001; ** p < 0.01.
937 See also Supplementary Figure 1.
938
36
bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Fig. 2 a b RPM RPMA 100 CGRP-Cre CMV-Cre CGRP-Cre CCSP-Cre 80 +
60 Bone
40
Percent surviva l 20 +
0 Axial 0 100 200 300 400 500 Days post-infection
RPM CMV (n=9) RPM CCSP (n=10) RPMA CMV (n=8) **** RPMA CCSP (n=8) **** RPM CGRP (n=14) RPM SPC (n=19) RPMA CGRP (n=8) **** RPMA SPC (n=10) **** Coronal
c Sagittal
MicroCT Necropsy H&E d RPM RPMA CMV-Cre
5
5 CGRP-Cre
7
7
CCSP-Cre e Soft Osteo 100
80
SPC-Cre 60
otal tumo r 40
% T % 20 ns *** f 0 100 100 RPM RPMA 80 80
g 6 60 60 7.5 10 Soft Osteo ) 2
40 40 m 5.0 106 μ umor burden ( 2.5 106
% T % 20 20 % Osteosarcoma
r s ize 5 0 0 1 10 1 105 RPM RPMA RPM RPMA Tumo
0 h RPM individual tumors 2.0 107
) Soft Osteo
Lung tumors 2
Ad5-Cre m by mCT Bone on mCT Bone on H&E 1.5 107 μ CMV 15/15 14/15 14/15 ( 1.0 107 CGRP 23/23 10/23 10/23 r s ize 5.0 106 CCSP 6/8 1/8 1/6 5 SPC 3/10 0/10 1/5 Tumo 1 10 1 105 0 RPMA individual tumors bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
939 Figure 2: ASCL1 loss delays tumorigenesis and promotes bone differentiation in multiple
940 cells of origin
941 a) Survival curve comparing RPM (solid lines, data from Figure 1a) vs RPMA mice (dashed
942 lines) infected with indicated cell-type-specific Cre viruses. Number of mice indicated in
943 the figure. + = mouse censored to end the cohort. Mantel-Cox log-rank test, **** p <
944 0.0001.
945 b) Representative bone analysis (top row) of microCT images in RPM vs RPMA mice with
946 advanced lung tumors infected with the indicated Ad-Cre viruses. Matched axial,
947 coronal, and sagittal cross-sections of RPM and RPMA tumors used for bone analysis
948 (bottom rows).
949 c) Representative microCT axial cross-sections, necropsy images, and H&E staining for
950 tumors from mice infected with indicated Ad-Cre viruses. In microCT panel, yellow
951 arrowheads indicate areas of focal bone formation. Necropsy image shows one whole
952 lung lobe. In H&E panels, scale bar indicates 50 µm in 10x image (left), and 25 µm in the
953 40x image (right).
954 d) Representative Trichrome staining in RPM-CMV vs RPMA-CMV tumors. Scale bar
955 indicates 50 μm in the 10x image (top) and 25 μm in the 40x image (bottom).
956 e) Quantification of the percent soft vs osteosarcoma tumor in RPM-CMV (n=3) vs RPMA-
957 CMV (n=6) mice with equivalent total tumor burden determined by PAS with Alcian Blue
958 (PAB) staining.
959 f) Average percent total tumor burden and total osteosarcoma in RPM vs RPMA mice
960 derived from Figure 2e. Mean ± SD. Unpaired two-tailed t-test, *** p < 0.001; ns = not
961 significant.
962 g) Size of individual soft tumors and osteosarcomas quantified from PAB-stained slides
963 from a representative RPM-CMV (top) or RPMA-CMV (bottom) mouse.
37
bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
964 h) Table indicating number of RPMA mice per cohort with lung tumors detected by microCT
965 imaging, bone detected by microCT, or bone detected by H&E review.
966 See also Supplementary Figure 2 and Supplementary Videos.
967
38
j g d a
RPMA vs RPM bioRxiv preprint Semantic space 2 SPC-Cre CCSP-Cre CGRP-Cre CMV-Cre Enrichment score PC2: 20.9% variance -0.25 -0.20 -0.15 -0.10 -0.05 -20 0.00 60 20 40 80 0 Significantly-enriched GObiologicalprocess RPM (n=11) P RPMA RPM es eltdMostdepleted Least depleted E 32 p<0.0001 NES -3.27 -40 (which wasnotcertifiedbypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. ASCL1 targetgenes doi: 2 20 0 -20 Semantic space 1 https://doi.org/10.1101/2020.11.11.362632 P2(=)RPMA (n=6) RPR2 (n=6) PC1: 39.5%variance
NEUROD1 H-score 100 200 300 0 e P RPMA RPM **** 060 40 RPMA het
Log10 (p-val) (n=1) h -20 -10
;
RPMA vs RPM this versionpostedNovember11,2020. Enrichment score b -0.30 -0.25 -0.20 -0.15 -0.10 -0.05 0.00 es eltdMostdepleted Least depleted k Ascl1
E 69 p<0.0001 NES -6.97 normalized counts 10000 10000 20000 30000 40000 NEUROD1 targetgenes 5000
Semantic space 2 0 P PARPR2 RPMA RPM Significantly-depleted GObiologicalprocess *** Runx2 Wnt10b Slit1 Lhx1 Mmp9 Sp7 Hmga2 Ibsp Sez6 Nhlh2 Rtn1 Pou6f2 Neurod1 Bmp8a ** RPMA RPM ** Semantic space 1
Log2(Normalized Counts+1) 10 15 5 0 i The copyrightholderforthispreprint f RPMA vs RPM
Enrichment score Neurod1 normalized counts c 0.00 0.05 0.10 0.15 0.20 1000 1500 2000 500 0 oterce Leastenriched Most enriched E .6p<0.0001 NES 3.66 GO BoneDevelopment P PARPR2 RPMA RPM *** * Dll3 Insm1 Chga Uchl1 Grp Ascl1 Syp Chgb Ncam1 Scg2 Calcb Calca ns RPMA RPM
Log10 (p-val) Fig. 3 -20 -10 Log2(Normalized Counts+1) 12 8 0 4 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
968 Figure 3: RPMA tumors are transcriptionally distinct with loss of ASCL1 and NEUROD1
969 target genes
970 a) Principal component (PC) analysis comparing gene expression by bulk RNA-seq in
971 RPM, RPMA, RPMAhet (Rb1fl/fl;Trp53fl/fl;MycLSL/LSL;Ascl1fl/wt), and RPR2 tumors.
972 b) Ascl1 expression shown as normalized counts by RNA-seq from lung tumors in indicated
973 GEMMs. Mean ± SD. Mann-Whitney two-tailed t-test, ** p < 0.01; *** p < 0.0003.
974 c) Heat map comparing log2 normalized counts for expression of select neuroendocrine
975 genes in RPM vs RPMA tumors.
976 d) Gene set enrichment analysis (GSEA) in RPMA vs RPM tumors using ASCL1 ChIP-seq
977 target genes from Borromeo et al, Cell Rep, 2016. Normalized enrichment score (NES)
978 and p value indicated in figure.
979 e) Heat map of top 200 (100-up and 100-down) differentially-expressed genes in RPMA vs
980 RPM tumors with select genes indicated.
981 f) Neurod1 expression as normalized counts by RNA-seq from lung tumors in indicated
982 GEMMs. Mean +/- SD. Mann-Whitney two-tailed t-test, * p < 0.05; *** p < 0.001; ns = not
983 significant.
984 g) Representative IHC and H-score quantification for NEUROD1 in RPM vs RPMA tumors.
985 Approximately 18-80 tumors were quantified from n ≥ 4 mice per condition. Scale bar: 25
986 µm. Mean +/- SD. Mann-Whitney two-tailed t-test, **** p < 0.0001.
987 h) GSEA comparing RPMA vs RPM tumors using NEUROD1 ChIP-seq target genes from
988 RPM tumors and human SCLC cell lines from Borromeo et al, Cell Rep, 2016.
989 Normalized enrichment score (NES) and p value indicated in the figure.
990 i) GSEA comparing RPMA vs RPM tumor gene expression to a known ossification
991 signature, “GO_Bone_Development.” NES and p values indicated in the figure.
992 j) Scatterplot visualizing semantic similarity of GO Biological Processes enriched in RPMA
993 vs RPM tumors.
39
bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
994 k) Scatterplot visualizing semantic similarity of GO Biological Processes depleted in RPMA
995 vs RPM tumors.
996
40
bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Fig. 4 a c d *** ** 2.0 4000 2500
2000 1.5 3000 Sox9 Runx1 1.0 Runx2 1500 Hes1 2000 Rest 0.5 Notch1 1000 0.0 1000 500 -0.5
0 Foxa2 normalized counts 0 -1.0 Nkx2-1 normalized counts PC2: 14.7% variance RPM RPMA RPM RPMA Ascl1 Myc -1.5
Gene modules Mycn Foxa2 *** Nkx2-1 -2.0 4000 Insm1 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 Neurod1 PC1: 19.5% variance 3000 Human cell lines Mouse tumors A RPM-CGRP 2000 RPM RPMA A2 RPM-CMV N RPMA-CGRP 1000 Normalized expression Y RPMA-CMV -1 1 P RPMA-CCSP uncl RPMA-SPC Insm1 normalized counts 0 RPM RPMA b e Average NKX2-1 FOXA2 INSM1 Attractors states RPM RPMA RPM RPMA RPM RPMA AHR AR ASCL1 CEBPD RPM CMV-Cre CTNNB1 E2F1 ELF1 ELF5 RPMA ESR1
ETS2 CGRP-Cre FOXA2 FOXM1 FOXO1 FOXO3
FOXP2 CCSP-Cre GATA2 HES1 INSM1 IRF1 IRF8 SPC-Cre KLF2 KLF4 KLF5 f LMO2 **** **** ** ns **** **** **** **** **** ************ 300 300 300 LYL1 MECOM MITF 200 200 200 MYC MYCN 100 100 100 NEUROD1 INSM1 H-scor e FOXA2 H-scor e NFE2L2 NKX2-1 H-scor e NKX2-1 0 0 0 NOTCH1 SPC SPC SPC SPC SPC NR3C1 CMV CGRP CCSP SPC CMV CGRP CCSP CMV CGRP CCSP CMV CGRP CCSP CMV CGRP CCSP CMV CGRP CCSP PHC1 RPM RPMA RPM RPMA RPM RPMA POU3F2 PPARG g PRDM16 ASCL1 PRDM5 PROX1 RARG NKX2-1 1.00 REST RET 0.75 RUNX1 SOX1 0.50 RUNX2 SOX2 0.25 SMAD3 FOXA1 0.00 SOX11 FOXA2
SOX17 Expression (avg z-score) INSM1 SOX9 SP7 SCLC cell lines, n=106 SPI1 STAT6 h TBX3 ASCL1 TEAD4 INSM1 YAP1 10
SOX2 FPKM+1) 8 2 NKX2-1 6 4
SPC RPM CMV CMV FOXA1
CCSP 2
RPMA CGRP CGRP SOX1 0 Ad5-Cre PROX1 RET1 Expression (log SCLC tumors, n=81 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
997 Figure 4: Network analyses predict transcriptional regulators that drive osteosarcoma
998 cell fate upon ASCL1 loss
999 a) Weighted Gene Coexpression Network Analysis (WGCNA) reveals co-expressed gene
1000 modules in RPM and RPMA tumors.
1001 b) Binarized average states and found attractors in RPM (blue) and RPMA (purple) tumors
1002 initiated with the indicated Cre viruses. For each gene, colored squares are ON and
1003 white squares are OFF.
1004 c) Principal component analysis comparing bulk RNA-seq expression in human SCLC cell
1005 lines to mouse RPM or RPMA tumors initiated with the indicated viruses. Human cell
1006 lines were classified into subtypes based on high expression of Ascl1 (A and A2 variant),
1007 Neurod1 (N), Pou2f3 (P), or Yap1 (Y) or were unclassified (uncl).
1008 d) Normalized counts of indicated genes from RNA-seq. Data is shown as mean ± SD.
1009 Mann-Whitney two-tailed t-test, *** p < 0.001; ** p < 0.01.
1010 e) Representative IHC images for indicated antibodies in RPM and RPMA mice infected
1011 with cell-type-specific Cre viruses. Scale bar: 25 µm.
1012 f) H-Score IHC quantification for indicated proteins in RPM and RPMA mice.
1013 Approximately 5-33 tumors were quantified from n = 3-5 mice per condition. Data is
1014 shown as mean ± SD. Mann-Whitney two-tailed t-test, **** p < 0.0001; ** p < 0.01; ns =
1015 not significant.
1016 g) Heat map derived from “SCLC-CellMinerCDB” showing relative expression of SCLC
1017 transcription factor genes compared to ASCL1.
1018 h) Heat map showing relative expression of SCLC transcription factor genes in human
1019 tumors from George et al, Nature, 2015.
1020 See also Supplementary Figure 3.
1021
41
bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Fig. 5 a c
Neural crest signature Mesenchymal stem cell signature 0.25 0.40 0.20 0.30 0.15 0.20 0.10 0.10 0.05 RPM 0.00 0.00 RPMA
Enrichment score Ascl1 Enrichment score
RPMA vs RPM RPMA 12 Ihh 10 Bmp8a 8 Most enriched Least enriched Most enriched Least enriched Bmp5 Bmp2k 6 NES 2.48 p < 0.0001 NES 2.66 p = 0.008 Bmp3 Bmpr2 4 Tgfbr2 2 (Normalized counts+1)
Tgfbr3 2 b Smad3
Tgfb/Bmp 0
Smad6 Log Smad7 Neural crest signature Mesenchymal stem cell signature Wnt2 0.25 0.25 Wnt2b 0.20 Wnt3a 0.20 Wnt5a 0.15 0.15 Wnt6 0.10 0.10 Wnt Wnt7b 0.05 0.05 Wnt10a 0.00 0.00 Wnt10b -0.05 Notch1 Enrichment score Enrichment score Notch2 Hes1
Notch Rest
Most enriched Least enriched Most enriched Least enriched NES 2.65 p<0.0001 NES 1.91 p<0.015 ASCL1 -low vs -high cell lines d e
CTNNB1 HES1 REST **** **ns **** **** **** **** **** 300 300
RPM 200 200 CGRP-Cre
100 100 HES1 H-scor e CTNNB1 H-scor e
CMV-Cre 0 0 CGRP CMV CGRP CCSP SPC CGRP CMV CGRP CCSP SPC RPM RPMA RPM RPMA
**** **** **** **** 300 CGRP-Cre
RPMA 200
100 CCSP-Cre REST H-scor e
0 CGRP CMV CGRP CCSP SPC RPM RPMA SPC-Cre f ASCL1 WNT10A 1.00 0.75 WNT10B 0.50 NOTCH1 0.25 NOTCH2 0.00
REST Expression (avg z-score) SCLC cell lines, n = 116 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
1022 Figure 5: ASCL1 represses non-endodermal cell fates and Notch/Wnt developmental
1023 pathways
1024 a) GSEA in RPMA vs RPM tumors compared to neural crest and mesenchymal stem cell
1025 signatures. NES and p values indicated in the figure.
1026 b) GSEA in ASCL1-low vs ASCL1-high human cell lines (SCLC-CellMinerCDB) compared
1027 to neural crest and mesenchymal stem cell signatures. NES and p values indicated in
1028 the figure.
1029 c) Heat map showing expression of indicated genes in RPM vs RPMA tumors analyzed by
1030 RNA-seq for components of the Indian hedgehog (Ihh), TGFβ, BMP, WNT, and NOTCH
1031 pathways.
1032 d) Representative IHC for CTNNB1, HES1, and REST in RPM vs RPMA tumors initiated
1033 with the indicated cell-type-specific-Cre viruses. Scale bar: 25 μm.
1034 e) H-score IHC quantification for indicated antibodies in panel d. Data is shown as mean ±
1035 SD. Approximately 11-28 tumors were used for quantification from 3-5 animals per
1036 condition. Mann-Whitney two-tailed t-test, **** p < 0.0001; ** p < 0.01; ns = not
1037 significant.
1038 f) Heat map derived from “SCLC-CellMinerCDB” with expression of selected genes
1039 involved in WNT and NOTCH pathways relative to ASCL1 in human SCLC cell lines.
1040
42
bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Fig. 6 a RUNX1 SOX9 RUNX2 SP7 Soft Osteo Soft Osteo Soft Osteo Soft Osteo RPM CGRP-Cre CMV-Cre CGRP-Cre RPMA CCSP-Cre SPC-Cre b **** *** * **** **** **** **** **** ************ **** *********** ns 300 300 300 300
200 200 200 cor e 200 H- s 100 100 100 100 7 SP 7 SOX9 H-scor e RUNX2 H-scor e RUNX1 H-scor e 0 0 0 0
CMVCGRPCCSP SPC CMVCGRPCCSP SPC CMVCGRPCCSP SPC CMVCGRPCCSP SPC CMVCGRPCCSP SPC CMVCGRPCCSP SPC CMVCGRPCCSP SPC CMVCGRPCCSP SPC RPM RPMA RPM RPMA RPM RPMA RPM RPMA c d e h j SOX9 target genes * * 0.15 500 8 3 0.10 siASCL1 siCTRL siCTRL siASCL1 400 2
coun ts 6 0.05 ASCL1 300
(counts) 1 0.00 200 2 4 -0.05
(Avg. z-score) 0 SOX9 100 Enrichment score 2 0 -1 SOX9 log RUNX1 -100 0 SOX9 -2
Sox9 normalized RPM RPMA RUNX2 Most enriched Least enriched siCTRLsiASCL1 ASCL1-low
ASCL1-high ASCL1 -low vs -high cell lines NES 2.80 p < 0.001 HSP90 H2107 H889 f g i k
SOX9 target genes SOX9 target genes
0.30 *** 0.08 0.25 4 0.04 0.20 3 0.15 0.00 0.10 2 -0.04 0.05 (Avg. z-score) Enrichment score Enrichment score 0.00 1 -0.08
RPMA vs RPM RPMA 0
RUNX2 -1
Most enriched Least enriched Most enriched Least enriched ASCL1 -low vs -high tumors NES 5.19 p < 0.001 SOX9-lowSOX9-high NES 1.65 p < 0.03 bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
1041 Figure 6: ASCL1 represses SOX9 in mouse and human SCLC tumor cells
1042 a) Representative IHC for RUNX1, SOX9, RUNX2 and SP7 in RPM vs RPMA tumors
1043 initiated with the indicated cell-type-specific-Cre viruses. RPMA tumors were classified
1044 as non-calcified tumors (soft) or osteosarcomas (osteo). Scale bar: 25 μm.
1045 b) H-score quantification for the indicated proteins in RPM vs RPMA tumors from panel a.
1046 Data is shown as mean ± SD. Approximately 11-85 tumors from 3-6 mice per condition
1047 were quantified. Mann-Whitney two-tailed t-test, **** p < 0.0001; *** p < 0.001; * p <
1048 0.05; ns = not significant.
1049 c) Representative immunoblot following 72 hr treatment with control (CTRL) or ASCL1
1050 siRNAs in the indicated human SCLC cell lines. HSP90 serves as loading control.
1051 d) Sox9 expression as normalized counts by RNA-seq from lung tumors in indicated
1052 GEMMs. Mean ± SD. Two-tailed t-test, * p < 0.05.
1053 e) Expression of SOX9 by RNA-seq from human SCLC cell line NCI-H2107 treated for 72
1054 hr with ASCL1 or control (CTRL) siRNAs performed in biological duplicates. Results are
1055 reported as Log2 normalized counts with mean ± SD.
1056 f) GSEA for expression of SOX9 target genes from Larsimont et al, Cell Stem Cell, 2015,
1057 in RPMA vs RPM tumors. NES and p values indicated in the figure.
1058 g) Predicted SOX9 target genes enriched or depleted in RPMA vs RPM tumors by IPA.
1059 h) Expression of SOX9 in human SCLC cell lines grouped by ASCL1 expression levels in
1060 “SCLC-CellMinerCDB.” Data is shown as average z-score ± SD. Mann-Whitney two-
1061 tailed t-test, * p < 0.05.
1062 i) Expression of RUNX2 in human SCLC cell lines grouped by SOX9 expression levels in
1063 “SCLC-CellMinerCDB.” Data is shown as average z-score ± SD. Mann-Whitney two-
1064 tailed t-test, *** p = 0.005.
43
bioRxiv preprint doi: https://doi.org/10.1101/2020.11.11.362632; this version posted November 11, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
1065 j) GSEA for expression of SOX9 target genes from Larsimont et al, Cell Stem Cell, 2015,
1066 in ASCL1-low vs ASCL1-high human SCLC cell lines from “SCLC-CellMinerCDB.” NES
1067 and p values indicated in the figure.
1068 k) GSEA for expression of SOX9 target genes from Larsimont et al, Cell Stem Cell, 2015,
1069 in ASCL1-low vs ASCL1-high human tumors from George et al, Nature, 2015. NES and
1070 p values indicated in the figure.
1071 See also Supplementary Figure 4.
1072
44