Reconstruction of epithelial transcriptional trajectories reveals heterogeneous progression and early therapy-resistance programs in high-grade serous carcinoma precursors

Background: Serous-tubal intraepithelial-carcinoma (STIC) is considered the principal precursor of tubo-ovarian high-grade serous-carcinoma (HGSC), yet its biological uniformity, progression risk and potential response to poly(ADP-ribose) polymerase-inhibitors (PARPi) remain poorly defined. Methods: We performed trajectory-based transcriptomic reconstruction of Fallopian-tube epithelial regions spanning normal epithelium, precursor lesions, STIC, and invasive carcinoma using pseudotime inference and publicly available spatial transcriptomic data. Gene expression and pathway dynamics were defined along pseudotime, and interpatient heterogeneity examined at both lesion and patient levels. Transcriptional signatures associated with PARPi-resistance were quantified across STICs, BRCA-mutant and BRCA-wild type subtypes. Results: Trajectory inference captured a continuous transcriptional progression from normal epithelium to HGSC, with STICs occupying heterogeneous evolutionary positions rather than a single precursor state. Incidental isolated STICs(STICi) spanned early to later pseudotime states and frequently aligned with loss-of-cilium organisation and less advanced epithelial phenotypes. STICs associated with concurrent cancer(STICc) exhibited more advanced malignant progression signatures, including cell-cycle activation, epithelial-to-mesenchymal transition (EMT), interferon signaling, and DNA-repair. Histologically similar lesions occupied divergent pseudotime positions with marked interpatient heterogeneity. PARPi resistance–associated signatures were variably enriched across precursor and precancer-stage lesions, with persistence into invasive disease. Conclusions: STICs are heterogeneous and occupy distinct evolutionary positions along a continuum, highlighting potentially different progression risks from normal epithelium to HGSC. STICi differ from STICc which harbour signatures of more advanced malignant 4 progression. Heterogeneity in PARPi resistance–associated programs in STICs cautions against uniform (non-stratified) use of PARPi-based primary prevention strategies. Future research should explore evolutionary-trajectory informed biomarkers for risk stratification and early interception strategies. Keywords High grade serous ovarian cancer, serous tubal intraepithelial carcinoma, STIC, incidental, malignant progression, trajectory inference, pseudotime, PARPi, resistance, spacial transcriptomics 5

Serous tubal intraepithelial carcinoma (STIC) is an established precursor of tubo-ovarian high‑grade serous carcinoma (HGSC).(1, 2) Evolutionary genomic analyses support a clonal continuum in which malignant progression initiates in the Fallopian tube(1, 3) and indicate an estimated window of 6-7 years between development of a STIC and initiation of HGSC.(1, 4) Nevertheless, our understanding of the biological processes underlying early disease evolution remains limited. Isolated incidental STICs (STICi) may represent biologically indolent lesions or more advanced precursors comparable to STICs associated with concurrent cancer (STICc). Whether these lesions occupt distinct of overlapping evolutionary states remains unclear. These distinctions have direct implications for progression risk and therapeutic management, including sensitivity to poly(ADP-ribose) polymerase inhibitors (PARPi). Data from HGSC cell lines implicate transcriptional programs associated with attenuated replication stress as markers of PARPi resistance.(5) The recently published multimodal atlas of Fallopian tube precursors and HGSC(6) encompasses spatial transcriptomic profiling of 407 epithelial regions of interest (ROIs), including normal epithelium (Fallopian tube and/or fimbriae), precancer lesions, and regions of invasive cancer (HGSC), derived from 34 patients. We applied trajectory inference to this dataset to reconstruct the disease progression continuum, define transcriptomic changes across pseudotime, and characterise transcriptional heterogeneity among Fallopian tube precursor lesions. Finally, we quantified enrichment of a PARPi resistance-associated transcriptional signaturse in STICs and related these to evolutionary position and malignant transformation phenoptypes. 6

Dataset features The Kader et al. (6) GeoMx dataset, containing normalized gene expression values and clinical sample metadata, was obtained from the GEO repository (GSE281193). The dataset contained two groups of specimens: Group 1 encompassing specimens with HGSC and co-occurring STICs (hereforth referred to as “STICc”); and Group 2 lacking HGSC but containing isolated precursor lesions identified incidentally during risk-reducing salpingo-oophorectomy (RRSO) or opportunistic salpingectomy (hereforth referred to as “STICi”). ROI annotations followed those provided in the Kader et al. dataset (see Supplementary Table-S1). Twenty-five of the 34 specimens with epithelial ROIs had matched “normal” Fallopian tube and/or fimbriae epithelium within the same tissue section. Serous tubal intraepithelial lesions (STIL) were included. Cancer regions detached from the surrounding tissue were denoted as “floating cancer”. The number of ROIs per lesion type per group, along with relevant abbreviations, is listed in Supplementary Table-S1. Trajectory inference using slingshot Pseudotemporal ordering of the epithelial ROIs was performed using Slingshot.(7) Slingshot integrates clustering information with dimensionality reduction to infer lineage relationships using a minimum spanning tree and subsequently fits smooth principal curves to model continuous transcriptional progressions. Prior to trajectory inference, the 2,000 most variable genes in the dataset were selected. Principal component analysis (PCA) was performed on the normalized expression matrix, and the first four principal components were used as input for Slingshot. Trajectory inference was conducted using the lesion-type annotations as cluster labels to guide lineage construction. The analysis was performed in an unsupervised manner without pre-specifying start or end clusters, allowing slingshot to infer the trajectory structure and directionality based on the data topology. Slingshot constructed minimum spanning trees 7 connecting clusters and fitted principal curves to define smooth trajectories. Pseudotime values were calculated for each ROI along each identified lineage, representing relative transcriptional position along the trajectory. ROIs not assigned to a given lineage were assigned NA pseudotime values for that lineage. Gene set variation analysis Gene set variation analysis (GSV A) was conducted using the GSV A package, and heatmaps were generated using ComplexHeatmap. GSV A derives enrichment scores for each sample as a function of genes inside and outside a given gene set, analogous to a competitive gene set test.(8) This approach converts a gene–by-sample matrix, providing an estimate of pathway activity.(8) Gene Ontology Biological Processes (GOBP) and Hallmark gene sets were obtained from the human MSigDB collections. A transcriptional signature associated with PARPi resistance, defined by Tamura et al. (5) and focusing on upregulated genes (Supplementary Table-S6), was used to evaluate PARPi resistance-associated programs in the Kader dataset. We first assessed the performance of this signature in ovarian cancer cell lines from the Cancer Cell Line Encyclopedia (CCLE) dataset (9), confirming a significant positive correlation with olaparib IC50 values (mean of two experimental replicates). We then evaluated associations between this signature and GOBP GSV A scores in STIC ROIs and invasive HGSC using Pearson’s correlation, with Benjamini-Hochberg correction for multiple testing. Additional transcriptional signatures related to attenuated replication stress, paclitaxel resistance, and congression defects (defined in Tamura et al.) (5) were included to further examine processes linked to chromosomal instability (Supplementary Table S6). Word clouds were generated using tidytext and wordcloud. All analyses were performed in R version 4.4.2. 8

Trajectory analysis provides a valuable framework to investigate disease progression from transcriptomic data and capture gradual transitions underlying pathological processes, particularly in contexts where longitudinal sampling or explicit time-course measurements are infeasible.(7) Here, we applied trajectory inference using Slingshot(7) on GeoMx transcriptomes from 407 epithelial ROIs spanning Fallopian tube precursor lesions, invasive cancer, and normal tissue in the Kader dataset,(6) corresponding to 34 patients, to reconstruct HGSC progression. In principle, this approach estimates relative transcriptional ordering based on similarities in gene exprssion profiles, enabling assignment of pseudotime values to individual ROIs. Importantly, pseudotime reflects relative evolutionary position rather than chronological time. Trajectory inference reconstructed a continuous transcriptional evolution from normal Fallopian tube epithelium to HGSC. Principal component embedding revealed a smooth disease trajectory across epithelial ROIs (Figure 1A, Supplementary Figure S1A-Β). Pseudotime increased stepwise across lesion types, from normal Fallopian tube/fimbriae and p53 signatures lesions through STIL, STICi, STICc, and ultimately invasive and floating cancer, capturing known biological and histological phenotypes (Figure 1B, Supplementary Table-S1, Supplementary Table-S2). We observed a statistically significant difference in pseudotime distribution among p53 signature lesions and invasive cancer cells in BRCAmut vs BRCAwt cases (Figure S1C), consistent with the earlier clinical stage typically observed in BRCAmut specimens. Coordinated gene expression remodeling along pseudotime was evident (Figure 1C; Supplementary Table S3), with two distinct clusters of genes progressively downregulated or upregulated along the evolutionary trajectory to HGSC. Top genes positively correlated with pseudotime included the cytoskeletal organisation genes PDLIM7 and FLNA, the microtubule 9 regulator STMN1 – a known immunohistochemical marker of STIC lesions(10) – and ITGB1, which is involved in cell adhesion and migration (Figure 1D, Supplementary Table-S3). By contrast, genes negatively correlated with pseudotime included TMEM231, RSPH1, HYDIN, and CF AP43, all of which are related to ciliary structure and function (Figure 1D, Supplementary Table-S3). Each dot of the scatterplots corresponds to a ROI, demonstrating a continuum in expression trend across the pseudotime. Pathway level analysis demonstrated progressive loss of cilium organisation and upregulation of embryonic morphogenesis, proliferation, chromosome segregation, adhesion and epithelial-to-mesenchymal transition (EMT) (Figure 1E, Supplementary Table-S4). We further observed upregulation of TGF-β signaling accompanied by IL6/STAT3, MYC and TNF-α signalling via NF-κB (Figure 1E, Figure S1D, Supplementary Table-S5). Together, these changes indicate malignant transformation occurring alongside increasing pseudotime through coordinated disruption of epithelial structure and activation of oncogenic and inflammatory transcriptional programmes well known for promoting cancer growth and spread.(11-13) We next created a pseudotime position of each lesion at the patient level (Figure 2A) to assess interpatient heterogeneity cross the malignant transformation continuum. This analysis revealed substantial variability among lesions of the same histologic type. For example, STICi lesions from patient 43 occupied pseudotime positions similar to STICc lesions in patients 6 and 25, yet were positioned later than STICi lesions in patient 36. Enrichment of a published PARPi resistance-associated transcriptional signature correlated with increasing olaparib IC50 values of ovarian cancer cell lines from the Cancer Cell Line Encyclopedia (CCLE) dataset (Figure 2B) (5, 9). Within STICs, this signature was strongly associated with pathways linked to attenuated replication stress, neural tube development, TGF-β response, and embryonic morphogenesis (Figure 2C, Supplementary Table-S6 and Supplementary Table-S7). We further observed significant positive associations with WNT 10 signaling and EMT (adjusted p 0.6; Supplementary Table-S6). Similar associations were observed in HGSC; however, the relative ranking differed, with embryonic morphogenesis, WNT signaling, and TGF-β response emerging as the most strongly correlated processes (Figure S2A, Supplementary Table-S8). Patient-level pathway analyses (Figures 2D–F and S2B) revealed heterogeneous transcriptional profiles across EMT/morphogenesis, cell-cycle progression, interferon response, oncogenic signaling, DNA repair, congression delays, replication stress attenuation, and paclitaxel and PARPi resistance programs. In patients with concurrent HGSC (Figure 2D), some STICc lesions exhibited elevated EMT, proliferation, and DNA repair activity (e.g., patients 11, 17, 18, and 25), features that were largely preserved in matched invasive HGSC regions. By contrast, certain cases showed discordant DNA repair activity between STICc and invasive HGSC (e.g., patients 14, 16, and 25). PARPi resistance-associated activity was observed in a large number of STICc lesions (patients 11, 16, 17, 18, and 22), occassioanlly co-occurring with paclitaxel resistance. Notably, patient 18 displayed elevated PARPi resistance-associated programs in STICs but reduced activity in matched invasive regions. Several patients displayed strong interferon responses in HGSC (patients 9, 11, 12, 22, and 24), with corresponding STICc lesions showing variable interferon activity ranging from low (patient 11) to similarly elevated (patient 22 and 24). STICi lesions generally showed low EMT and proliferation activity in most patients, in keeping with their low DNA repair activity (Figure 2E; patients 36–41). These resistance-associated signatures had variable activity in STICi, albeit at lower levels compared to STICc (Figure 2E). Of particular interest, patient 36 harbored early-pseudotime STICs with minimal resistance-associated signaling, whereas patient 43 (Figure 2E) displayed transcriptionally advanced STICs enriched for DNA repair, cell cycle, and PARPi resistance programs while maintaining low EMT activity. p53 signature lesions in patients with HGSC showed low EMT activity in 11 patients 6, 9, 10, and 14 but reached STIC-like levels in patients 17, 20 and 25, consistent with their more advanced pseudotime positions (Figure 2D, Supplementary Figure S2B). By contrast, p53 signature lesions in patients without HGSC showed predominantly low pathway activity, with the exception of patient 34, who exhibited pathway enrichment also in matched normal fimbrial tissue (Figure 2F). Finally, unsupervised k-means clustering identified subgroups of STIC phenotypes characterized by differential pathway engagement (Figure 2G, Supplementary Table-S9), further highlighting heterogeneity across STICs. STIC clustering patterns were distinct from those observed for HGSC lesions (Figure S2C, Supplementary Table-S10). Discussion Using trajectory-based transcriptomic analysis, we demonstrate that STICs are not a uniform precursor state but instead occupy distinct evolutionary positions along a continuum from normal epithelium to invasive HGSC. Although STICs are widely regarded as obligate precursors of HGSC, our data demonstrate marked interpatient heterogeneity. STICi span early to later pseudotime states, frequently presenting loss of cilium organisation. By contrast, STICc consistently exhibit transcriptional programs characteristic of more advanced malignant progression, including cell cycle and mitotic programs, EMT/morphogenesis, interferon response and DNA repair pathways. These findings suggest that STICs are heterogeneous and may have different progression risks. Biological processes related to cilia structure and function showed the strongest decrease alongside the malignant transformation pseudotime. Although loss of primary cilium is linked to carcinogenesis through Hedgehog and Wnt signaling(14), little is known about the role and significance of motile cilia loss in ovarian carcinogenesis(15-17). Our data indicate that 12 markers of cilium assembly and function could potentially be used to evaluate early carcinogenesis in HGSC. Importantly, we observed enrichment of PARPi resistance–associated transcriptional signatures that varied across STICs, including within the same lesion category (STICi/STICc) and across BRCA-status (STICc). While many STICi occupied early pseudotime states with limited therapeutic drug resistance-associated signatures (paclitaxel/ PARPi), a subset aligned with later pseudotime positions exhibited transcriptional profiles comparable to STICc, which more consistently showed activation of DNA repair and replication stress–attenuation programs. Notably, these resistance associated states were often established at the STIC stage and persisted with only modest amplification in HGSC. EMT is a process known to associate with poor survival in HGSC (18), and we observed variable activity of the EMT-associatd programs in early lesions.. In patients with concurrent cancer, transcriptional activity of EMT could be observed as early as p53 signature lesions. Heterogeneity was also evident with respect to DNA repair activity across STICs, with STICc potentially showing higher activity compared to STICi. Together, these findings raise caution against uniform PARPi-based primary medical prevention for all STICs. Not all STICs may be equally sensitive to PARPi, raising the possibility of toxicity risks, without clear benefit in biologically indolent or intrinsically resistant lesions. This highlights the need for further research in this area. As the first trajectory based transcriptomic analysis of STICs, this study provides an evolutionary framework but is limited by its cross-sectional design, relatively small number of cases, and lack of longitudinal clinical outcomes. Our data do not preclude that STICi could be sensitive to PARPi; however, this should be interpreted with caution given the limited sample size. In particular, expanding the number of STICi samples and capturing diverse ethnic backgrounds will be essential to more robustly define the spectrum of STIC phenotypes along 13 the malignant evolution timeline and also to better understand their relationship to PARPi sensitivity and resistance. Future studies in larger patient cohorts integrating multi-omics, longitudinal sampling, and outcome data will be required to refine risk stratification and prognosis, and to develop therapeutic as well as preventive strategies. Forjaz et al. recently reported the first multi-omics, whole-human Fallopian tube 3D imaging study, suggesting that STIC lesions may be substantially more common than indicated by existing protocols(19). Our findings of marked heterogeneity and differential progression among STICs are also consistent with this observation. Evolutionary trajectory-informed biomarkers may improve our ability to distinguish and identify the clinically significant lesions, enabling individualized surveillance, targeted interception, and prevention strategies using biology-driven approaches. Author Contributions Project conceptualisation by MS, EM, RM, FB. Data curation and formal analysis by MS and EM. Resources by RD, TK and SS. First manuscript draft by MS and EM. Manuscript supervision by RM and FB. Manuscript review and editing by RM, FB, RD, TK and SS. Acknowledgements FB and EM acknowledge support from UKRI Frontier Research grant EP/X028704/1 and City of London CRUK Core Award CTRQQR-2021\100004. 14

1. Labidi-Galy SI, Papp E, Hallberg D, Niknafs N, AdleZ V , Noe M, et al. High grade serous ovarian carcinomas originate in the fallopian tube. Nat Commun. 2017;8(1):1093. 2. van den Berg CB, Dasgupta S, Ewing-Graham PC, Bart J, Bulten J, Gaarenstroom KN, et al. Does serous tubal intraepithelial carcinoma (STIC) metastasize? The clonal relationship between STIC and subsequent high-grade serous carcinoma in BRCA1/2 mutation carriers several years after risk-reducing salpingo-oophorectomy. Gynecol Oncol. 2024;187:113-9. 3. Cheng Z, Ennis DP , Lu B, Mirza HB, Sokota C, Kaur B, et al. The genomic trajectory of ovarian high-grade serous carcinoma can be observed in STIC lesions. J Pathol. 2024;264(1):42-54. 4. Wu RC, Wang P , Lin SF , Zhang M, Song Q, Chu T, et al. Genomic landscape and evolutionary trajectories of ovarian cancer precursor lesions. J Pathol. 2019;248(1):41-50. 5. Tamura N, Shaikh N, Muliaditan D, Soliman TN, McGuinness JR, Maniati E, et al. Specific Mechanisms of Chromosomal Instability Indicate Therapeutic Sensitivities in High-Grade Serous Ovarian Carcinoma. Cancer Res. 2020;80(22):4946-59. 6. Kader T, Lin JR, Hug CB, Coy S, Chen YA, de Bruijn I, et al. Multimodal Spatial Profiling Reveals Immune Suppression and Microenvironment Remodeling in Fallopian Tube Precursors to High-Grade Serous Ovarian Carcinoma. Cancer Discov. 2025;15(6):1180-202. 7. Street K, Risso D, Fletcher RB, Das D, Ngai J, Yosef N, et al. Slingshot: cell lineage and pseudotime inference for single-cell transcriptomics. BMC Genomics. 2018;19(1):477. 8. Hanzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics. 2013;14:7. 9. Barretina J, Caponigro G, Stransky N, Venkatesan K, Margolin AA, Kim S, et al. The Cancer Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature. 2012;483(7391):603-7. 10. Karst AM, Levanon K, Duraisamy S, Liu JF , Hirsch MS, Hecht JL, et al. Stathmin 1, a marker of PI3K pathway activation and regulator of microtubule dynamics, is expressed in early pelvic serous carcinomas. Gynecol Oncol. 2011;123(1):5-12. 11. Balkwill FR, Mantovani A. Cancer-related inflammation: common themes and therapeutic opportunities. Semin Cancer Biol. 2012;22(1):33-40. 12. Bowtell DD, Bohm S, Ahmed AA, Aspuria PJ, Bast RC, Jr., Beral V , et al. Rethinking ovarian cancer II: reducing mortality from high-grade serous ovarian cancer. Nat Rev Cancer. 2015;15(11):668-79. 13. Naylor MS, Stamp GW, Foulkes WD, Eccles D, Balkwill FR. Tumor necrosis factor and its receptors in human ovarian cancer. Potential role in disease progression. J Clin Invest. 1993;91(5):2194-206. 14. Collinson R, Tanos B. Primary cilia and cancer: a tale of many faces. Oncogene. 2025;44(21):1551-66. 15. Richardson MT, Recouvreux MS, Karlan BY , Walts AE, Orsulic S. Ciliated Cells in Ovarian Cancer Decrease with Increasing Tumor Grade and Disease Progression. Cells. 2022;11(24). 15 16. Guo Y , He X, Liu J, Tan Y , Zhang C, Chen S, et al. The relationship between HYDIN and fallopian tubal cilia loss in patients with epithelial ovarian cancer. Front Oncol. 2024;14:1495753. 17. Tao T, Lin W, Wang Y , Zhang J, Chambers SK, Li B, et al. Loss of tubal ciliated cells as a risk for "ovarian" or pelvic serous carcinoma. Am J Cancer Res. 2020;10(11):3815-27. 18. Rai L, Ravaggi A, Bignotti E, Hollis RL, Garsed DW, Pandey A, et al. Oxford Classic-Defined EMT Risk Stratification of High-Grade Serous Ovarian Cancer for Guiding Treatment Decisions. Clin Cancer Res. 2026;32(1):188-202. 19. Forjaz A, Queiroga V , Li Y , Hernandez A, Crawford A, Qin X, et al. 3D multi-omic mapping of whole nondiseased human fallopian tubes at cellular resolution reveals a large incidence of ovarian cancer precursors. bioRxiv. 2025:2025.09.21.677628. Disease Evolution Trajectory nFim_c nFim_i nFT_c nFT_i p53_i p53_c STIL_c STIC_i STIC_c Inv_Cancer Cancer_floating 0 20 40 60 80 100 120 Pseudotime by Lesion Type Pseudotime Top 500 GOBPs Along Disease Trajectory Lesion_Type Pseudotime z-score −2 −1 0 1 2 Pseudotime 0 50 100 150 Top 60 Genes Along Disease Trajectory MEST MARCKSL1 STMN1 EIF4G1 ITGB1 PDLIM7 FLNA TUBB SNRPB H2AC17 DNAJB13 C4orf47 RABL2B CFAP46 ANAPC4 MOK TMEM231 CCDC170 MAP6 CES4A TMEM232 SAXO2 NME5 CFAP45 SNTN WDR38 ZMYND10 ROPN1L TEKT1 TPPP3 CDHR3 PRR29 PIFO RSPH1 CFAP157 SPATA18 FAM183A C9orf24 C20orf85 BAIAP3 TUBA4B LDLRAD1 CIBAR2 CCDC78 UBXN10 CCDC187 CFAP70 CCDC17 DLEC1 CFAP43 VWA3A HYDIN CAPSL ODAD1 ZBBX RASSF7 FYB2 WDR77 ARHGAP26 PKP2 Lesion_Type Pseudotime z-score −2 −1 0 1 2 Lesion_Type nFim_c nFim_i nFT_c nFT_i p53_i p53_c STIL_c STIC_i STIC_c Inv_Cancer Cancer_floating 0 20 40 60 80 100 120 1 2 3 4 5 6 7 8 PDLIM7 (r = 0.8 ) 0 20 40 60 80 100 120 3 4 5 6 7 8 FLNA (r = 0.8 ) 0 20 40 60 80 100 120 3 4 5 6 7 8 ITGB1 (r = 0.76 ) 0 20 40 60 80 100 120 2 3 4 5 6 7 TMEM231 (r = −0.85 ) 0 20 40 60 80 100 120 1 2 3 4 5 6 7 RSPH1 (r = −0.84 ) 0 20 40 60 80 100 120 1 2 3 4 5 6 HYDIN (r = −0.84 ) 0 20 40 60 80 100 120 1 2 3 4 5 6 CFAP43 (r = −0.84 ) Pseudotime 0 50 100 150 Pseudotime Expression Figure 1 A B C D E −40 −20 0 20 −20 −10 0 10 20 PC1 PC2 Lesion Type Cancer floating p53 signature STIC STIL inc. Fimbriae inc. FT inc. STIC inc. p53sig Fimbriae FT Inv. Cancer Lesion_Type nFim_c nFim_i nFT_c nFT_i p53_i p53_c STIL_c STIC_i STIC_c Inv_Cancer Cancer_floating 0 20 40 60 80 100 120 2 3 4 5 6 7 8 STMN1 (r = 0.76 ) Figure 1. Pseudotemporal reconstruction of transcriptional changes from normal to invasive ovarian cancer epithelium. A) Principal Component Analysis (PCA) of the Kader dataset using the top 2,000 most variable genes, each dot is an epithelial region of interest (ROI, n = 407 ROIs), coloured by lesion type. B) Boxplot illustrating ROIs pseudotime values for each lesion type (Kruskal-Wallis p = 8.68e-56). C) Heatmap of normalised expression of the top sixty genes that significantly change in expression with pseudotime along the first slingshot trajectory of the Kader et al. dataset (Spearman correlation with Benjamini-Hochberg, BH, adjusted p < 0.05). D) Scatterplots illustrating top correlating genes (PDLIM7, FLNA, STMN1, ITGB1, TMEM231, RSPH1, HYDIN, CFAP43). Each dot corresponds to an ROI, ordered and coloured by pseudotime. E) Heatmap of gsva enrichment scores for top five hundred Gene Ontology Biological Processes (GOBPs) that significantly change in expression with pseudotime, along the first slingshot trajectory (Spearman BH adjp 0.5). Word cloud graph highlights common keywords of the up and down-regulated processes. Supplementary Figure 1 B C 28.2% 6.4% 5.0% 3.8% 0 10 20 30 PC1 PC2 PC3 PC4 Principal Component Variance Explained (%) Variance Explained by First 4 PCs 0.04680 0.01530 0.00049 0.01530 0 50 100 150Pseudotime BRCA status WT Mut Comparison of Pseudotime across lesions and BRCA status Significantly correlating Hallmark pathways Along Disease Trajectory BILE_ACID_METABOLISM ESTROGEN_RESPONSE_EARLY MYC_TARGETS_V2 E2F_TARGETS G2M_CHECKPOINT MITOTIC_SPINDLE MYC_TARGETS_V1 DNA_REPAIR GLYCOLYSIS PI3K_AKT_MTOR_SIGNALING UV_RESPONSE_UP UNFOLDED_PROTEIN_RESPONSE MTORC1_SIGNALING IL6_JAK_STAT3_SIGNALING ANGIOGENESIS HYPOXIA TNFA_SIGNALING_VIA_NFKB TGF_BETA_SIGNALING EPITHELIAL_MESENCHYMAL_TRANSITION APICAL_JUNCTION Lesion_Type Pseudotime z-score −2 −1 0 1 2 Pseudotime 0 50 100 150 D A −40 −20 0 20 −20 −10 0 10 20 PC1 PC2 Pseudotime 0 50 100 150 LateEarly Disease Evolution Trajectory nFim_c nFim_i nFT_c nFT_i p53_i p53_c STIL_c STIC_i STIC_c Inv_Cancer Cancer_floating Lesion_Type nFim_c nFim_i nFT_c nFT_i p53_i p53_c STIL_c STIC_i STIC_c Inv_Cancer Cancer_floating Supplementary Figure 1. A) PCA plot of Figure 1A, coloured by the slingshot’s pseudotime value. Pseudotime corresponds to a one-dimensional variable representing each ROI’s transcriptional progression towards terminal state. B) Scree plot showing variance captured by the first 4 PC components used in slingshot. C) Boxplot illustrating ROIs pseudotime values per BRCA status for each lesion type (p values correspond to BH adjusted Welch's t-test). D) Heatmap of gsva enrichment scores for Hallmark pathways that significantly change in expression with pseudotime, along the first slingshot trajectory (Spearman BH adjp 0.3) Figure 2 BRCAwt Α Β D Zscore −2 −1 0 1 2 BRCAstatus Mut WT Lesion type STIC_i STIC_c Pt_8 Pt_9 Primary cancer stage stage_IC stage_IIA stage_III stage_IIIA stage_IIIC stage_incidental Patient Pt_10 Pt_11 Pt_12 Pt_13 Pt_14 Pt_15 Pt_16 Pt_17 Pt_18 Pt_19 Pt_20 Pt_21 Pt_22 Pt_24 Pt_25 Pt_26 Pt_27 Pt_28 Pt_29 Pt_30 Pt_31 Pt_33 Pt_34 Pt_36 Pt_37 Pt_38 Pt_39 Pt_40 Pt_41 Pt_42 Pt_43 Pt_6 F Pt_10 Pt_11 Pt_12 Pt_13 Pt_14 Pt_15 Pt_16 Pt_17 Pt_18 Pt_19 Pt_20 Pt_21 Pt_22 Pt_24 Pt_25 Pt_26 Pt_27 Pt_28 Pt_29 Pt_30 Pt_31 Pt_33 Pt_34 Pt_36 Pt_37 Pt_38 Pt_39 Pt_40 Pt_41 Pt_42 Pt_43 Pt_6 Pt_8 Pt_9 0 50 100 Pseudotime Patient Pseudotime by Patient Pearson’s r = 0.73, p = 9.7e-06 A2780 CAOV4 EFO21 EFO27 ES2 FUOV1 IGROV1 JHOS2 JHOS4 KURAMOCHI OAW42 OC314 OV56 OV7 OV90 OVCAR4 OVCAR8 OVISE OVK18 OVKATE OVTOKO RMGI SKOV3 TOV112D TOV21G TYKNU OV17R SW626−0.2 0.0 0.2 0.4 0.6 0 100 200 300 400 500 Olaparib IC50 PARPi resistance sign. (gsva scores) CCLE Ovarian Cancer Cell lines Pt-36 Pt-40 Pt-41 Pt-43 Pt-37 Pt-38 Pt-39 Pt-6 Pt-9 Pt-12 Pt-24 Pt-25 Pt-10 Pt-11 Pt-14 PrSt IC PrSt ICPrSt IC PrSt IIIC IIIA EPITH._TO_MESENCHYMAL_TRANSITION EMBRYONIC_MORPHOGENESIS NEURAL_TUBE_DEVELOPMENT POSITIVE_REGULATION_OF_CELL_CYCLE MITOTIC_NUCLEAR_DIVISION REG._OF_TYPE_I_INTERFERON_MED._SIGN. PEPTIDE_ANTIG._ASS._WITH_MHC_CLASS_I. TYPE_I_INTERFERON_PRODUCTION RESPONSE_TO_INTERFERON_ALPHA TNFA_SIGNALING_VIA_NFKB TGF_BETA_SIGNALING MTORC1_SIGNALING POS._REG._DOUBLE_STR._BR._REP._VIA_HOMOL. REGULATION_OF_DNA_REPAIR DNA_DAMAGE_RESPONSE DOUBLE_STR._BR._REPAIR_VIA_NONHOM. BASE_EXCISION_REPAIR Congression Defects Attenuated Replication Stress Paclitaxel Resistance PARPi Resistance Pt-16 Pt-17 Pt-18 Pt-19 Pt-22 Lesion Type Cancer_floating EMT / Morphogenesis Cell Cycle / Mitosis IFN Response DNA Repair STIC_c RARPi Resistance Paclitaxel Resistance Congression Defects Attenuated Repl. Stress Tamura et al. GOBP / Hallmark PrSt III PrSt III PrSt IIIC Signaling Pt-26 Pt-27 Pt-28 Pt-29 Pt-31 Pt-33Pt-30 Pt-34 STIC_i p53_i BRCAwt BRCAmut BRCAmutBRCAwt BRCAmut E PrSt III PrSt III PrSt III PrSt III PrSt III STICs_2 STICs_1 STICs_3 STICs_4 STICs_10 STICs_8 STICs_9 STICs_5 STICs_6 STICs_7 Paclitaxel Resistance Congression Defects DOUBLE_STRAND_BREAK_REPAIR_VIA_NONHOMOL. POSITIVE_REGULATION_OF_CELL_CYCLE MITOTIC_NUCLEAR_DIVISION BASE_EXCISION_REPAIR REGULATION_OF_DNA_REPAIR DNA_DAMAGE_RESPONSE POS._REG._OF_DOUBLE_STRAND_BREAK_REPAIR_VIA_HOMOL. TGF_BETA_SIGNALING EMBRYONIC_MORPHOGENESIS EPITHELIAL_TO_MESENCHYMAL_TRANSITION NEURAL_TUBE_DEVELOPMENT AttReplication Stress PARPi Resistance REG._OF_TYPE_I_INTERFERON_MEDIATED_SIGNAL. MTORC1_SIGNALING TYPE_I_INTERFERON_PRODUCTION PEPTIDE_ANTIGEN_ASSEMBLY_WITH_MHC_CLASS_I. RESPONSE_TO_INTERFERON_ALPHA TNFA_SIGNALING_VIA_NFKB Pt_8_9_STIC Pt_16_3_STIC Pt_16_STIC Pt_19_7_STIC Pt_19_6_STIC Pt_20_5_STIC Pt_17_9_STIC Pt_10_11_STIC Pt_16_2_STIC Pt_10_9_STIC Pt_19_8_STIC Pt_11_7_STIC Pt_11_11_STIC Pt_14_9_STIC Pt_9_6_STIC Pt_9_5_STIC Pt_36_7_incidental_STIC Pt_36_6_incidental_STIC Pt_14_10_STIC Pt_14_11_STIC Pt_37_8_incidental_STIC Pt_36_5_incidental_STIC Pt_9_16_STIC Pt_10_10_STIC Pt_10_12_STIC Pt_41_4_incidental_STIC Pt_41_6_incidental_STIC Pt_41_5_incidental_STIC Pt_22_12_STIC Pt_6_10_STIC Pt_6_3_STIC Pt_6_4_STIC Pt_6_5_STIC Pt_6_6_STIC Pt_38_2_incidental_STIC Pt_22_15_STIC Pt_9_18_STIC Pt_9_17_STIC Pt_36_8_incidental_STIC Pt_43_2_incidental_STIC Pt_9_13_STIC Pt_22_13_STIC Pt_43_7_incidental_STIC Pt_43_8_incidental_STIC Pt_40_5_incidental_STIC Pt_40_3_incidental_STIC Pt_37_6_incidental_STIC Pt_37_3_incidental_STIC Pt_37_4_incidental_STIC Pt_37_5_incidental_STIC Pt_37_7_incidental_STIC Pt_18_3_STIC Pt_18_STIC Pt_18_1_STIC Pt_18_5_STIC Pt_18_2_STIC Pt_17_8_STIC Pt_25_2_STIC Pt_16_1_STIC Pt_17_10_STIC Pt_8_11_STIC Pt_20_3_STIC Pt_20_6_STIC Pt_17_2_STIC Pt_17_STIC Pt_17_3_STIC Pt_17_11_STIC Pt_17_1_STIC Pt_20_7_STIC Pt_11_38_STIC Pt_11_39_STIC Pt_11_37_STIC Pt_11_25_STIC Pt_15_5_STIC Pt_15_7_STIC Pt_15_6_STIC Pt_16_7_STIC Pt_16_4_STIC Pt_16_6_STIC Pt_43_4_incidental_STIC Pt_25_7_STIC Pt_25_5_STIC Pt_25_6_STIC Pt_11_4_STIC Pt_8_10_STIC Pt_20_STIC Pt_15_10_STIC Pt_16_5_STIC Pt_15_9_STIC Pt_11_9_STIC Pt_25_1_STIC Pt_11_27_STIC Pt_11_5_STIC Pt_11_26_STIC Pt_11_24_STIC Pt_9_15_STIC Pt_22_9_STIC Pt_22_14_STIC Pt_9_14_STIC Pt_12_6_STIC Pt_18_4_STIC Pt_22_17_STIC Pt_17_7_STIC Pt_15_8_STIC Pt_11_8_STIC Pt_20_4_STIC Pt_11_10_STIC Pt_12_7_STIC Pt_43_10_incidental_STIC Pt_24_1_STIC Pt_24_4_STIC Pt_43_5_incidental_STIC Pt_43_6_incidental_STIC Pt_43_3_incidental_STIC Pt_43_9_incidental_STIC Pt_40_4_incidental_STIC Pt_22_19_STIC Pt_22_18_STIC Pt_22_16_STIC Pt_22_10_STIC Pt_22_11_STIC Pt_12_8_STIC Pt_36_9_incidental_STIC BRCAstatus Lesion type Patient Stage STIC_c CELL_SURF._RECEPTOR_PR._SERINE_THR._KINASE_SIGNAL. MORPHOGENESIS_OF_AN_EPITHELIUM EPITHELIAL_TUBE_FORMATION GROWTH TUBE_FORMATION EMBRYO_DEVELOPMENT CELL_PROJECTION_ORGANIZATION EMBRYO_DEVELOP._ENDING_IN_BIRTH_OR_EGG_HATCHING CELL_GROWTH REGULATION_OF_CELL_PROJECTION_ORGANIZATION MORPHOGENESIS_OF_EMBRYONIC_EPITHELIUM TGFb_RECEPTOR_SIGNALING_PATH. NEURAL_TUBE_FORMATION EPITHELIAL_TUBE_MORPHOGENESIS REG._OF_CELL._RESP._TO_TGFb_STIM. RESPONSE_TO_TRANSFORMING_GROWTH_FACTOR_BETA POS._REG._OF_CELL_PROJECTION_ORGANIZATION NEURAL_TUBE_DEVELOPMENT Attenuated Replication Stress 0.0 0.2 0.4 0.6 Pearson's r Top PARPi Resist. signature correlations in STICs adjp < 0.001 p53_c STIC_c STIL_c nFim_i nFT_ STIC_i p53_i nFim_c nFT_c Inv_Cancer C G STIC_c Zscore −2 −1 0 1 2 Lesion Type BRCA status CAOV4 FUOV1 JHOS2 JHOS4 OVCAR4 OVCAR8 OVKATE 0.0 0.2 0.4 0.6 0 100 200 300 400 500 Olaparib IC50 PARPi resistance sign. (gsva scores) Pearson’s r = 0.91, p = 0.0043 HGSC onlyAll Ovarian Figure 2. “Chronotope” of lesions per patient and transcriptional features of therapy resistance. A) Scatter plot illustrating the position of each lesion in pseudotime, per patient. Colours correspond to lesion type. B) Scatter plot illustrating correlation of PARPi resistance signature (Tamura et al) with Olaparib IC50 in ovarian cancer cell lines in the CCLE dataset (n = 28 ovarian cancer cell lines, n = 7 HGSC cells lines). C) Barplot of STIC ROI transcriptomes showing top 10 Biological Processes and transcriptional signatures related to Congression Defects, Attenuated Replication Stress, Paclitaxel Resistance significantly correlating with a PARP-inhibitor resistance signature from Tamura et al (pearson’s BH adjp < 0.05). D–F) Heatmaps of significant disease related GOBPs and Hallmarks as well as the Tamura et al transcriptional signatures for D) patients with invasive cancer E) patients with incidental STIC lesions and F) patients with incidental p53 lesions. Terms have been grouped in the following categories: EMT / morphogenesis (Epithelial to Mesenchymal Transition, Embryonic Morphogenesis, Neural Tube Development); cell cycle / mitosis (Positive Regulation of Cell Cycle, Mitotic Nuclear Division); IFN response (Regulation of Type I Interferon–Mediated Signaling Pathway, Peptide Antigen Assembly with MHC Class I Protein Complex, Type I Interferon Production, Response to Interferon-Alpha); Oncogenic Signaling ( TNFA signaling via NFkB, TGF-beta signaling, MTORC1 signaling); DNA repair (Positive Regulation of Double-Strand Break Repair via Homologous Recombination, Regulation of DNA Repair, DNA Damage Response, Double-Strand Break Repair via Nonhomologous End Joining, Base Excision Repair) as well as transcriptional signatures from Tamura et al. D-F) Lesion type colour coding as in A. G) Heatmap of gsva enrichment scores showing clustering patterns of STIC ROIs using the processes and pathways described in the previous panel (D-F). Clustering of ROIs was performed by k-means clustering. Supplementary Figure 2 A Pt-21 Pt-8 Pt-13 Pt-15Pt-42 Pt-20 Zscore −2 −1 0 1 2 BRCAstatus Mut WT Lesion type Inv_Cancer Pt_8 Pt_9 Primary cancer stage stage_IA stage_IC stage_IIA stage_III stage_IIIA stage_IIIB stage_IIIC BRCAstatus Lesion type Patient Stage Patient Pt_10 Pt_11 Pt_12 Pt_13 Pt_14 Pt_15 Pt_16 Pt_17 Pt_18 Pt_19 Pt_20 Pt_21 Pt_22 Pt_24 Pt_25 Pt_26 Pt_27 Pt_28 Pt_29 Pt_30 Pt_31 Pt_33 Pt_34 Pt_36 Pt_37 Pt_38 Pt_39 Pt_40 Pt_41 Pt_42 Pt_43 Pt_6 MUSCLE_STRUCTURE_DEVELOPMENT GASTRULATION REGULATION_OF_CELLULAR_RESPONSE_TO_GROWTH_FACTOR_STIMULUS EMBRYONIC_MORPHOGENESIS CELL_MORPHOGENESIS EMBRYO_DEVELOPMENT_ENDING_IN_BIRTH_OR_EGG_HATCHING REGULATION_OF_CELLULAR_COMPONENT_BIOGENESIS REGULATION_OF_CANONICAL_WNT_SIGNALING_PATHWAY REGULATION_OF_CELL_PROJECTION_ORGANIZATION RESPONSE_TO_GROWTH_FACTOR CANONICAL_WNT_SIGNALING_PATHWAY RESPONSE_TO_TRANSFORMING_GROWTH_FACTOR_BETA GROWTH CELL_SURFACE_RECEPTOR_PROTEIN_SERINE_THREONINE_KINASE_SIGNAL. ENZYME_LINKED_RECEPTOR_PROTEIN_SIGNALING_PATHWAY CELL_PROJECTION_ORGANIZATION REGULATION_OF_WNT_SIGNALING_PATHWAY WNT_SIGNALING_PATHWAY EMBRYO_DEVELOPMENT 0.0 0.2 0.4 0.6 Pearson's r Top PARPi Resist. signature correlations in Invasive Cancer B C INV_3 INV_4 INV_1 INV_2 INV_5 INV_7 INV_6 INV_8 INV_9 INV_10 POSITIVE_REGULATION_OF_CELL_CYCLE MITOTIC_NUCLEAR_DIVISION REGULATION_OF_DNA_REPAIR DNA_DAMAGE_RESPONSE DOUBLE_STRAND_BREAK_REPAIR_VIA_NONHOMOL. POS._REG._OF_DOUBLE_STR._BREAK_REP._VIA_HOMOL. BASE_EXCISION_REPAIR EPITHELIAL_TO_MESENCHYMAL_TRANSITION EMBRYONIC_MORPHOGENESIS TGF_BETA_SIGNALING NEURAL_TUBE_DEVELOPMENT Congression Defects Attenuated Replication Stress Paclitaxel Resistance PARPi Resistance RESPONSE_TO_INTERFERON_ALPHA MTORC1_SIGNALING REG._OF_TYPE_I_INTERFERON_MEDIATED_SIGNAL. TYPE_I_INTERFERON_PRODUCTION PEPTIDE_ANTIGEN_ASSEMBLY_WITH_MHC_CLASS_I. TNFA_SIGNALING_VIA_NFKB Pt_25_9_Inv_Cancer Pt_18_8_Inv_Cancer Pt_19_3_Inv_Cancer Pt_19_4_Inv_Cancer Pt_14_2_Inv_Cancer Pt_14_15_Inv_Cancer Pt_14_13_Inv_Cancer Pt_16_10_Inv_Cancer Pt_10_4_Inv_Cancer Pt_10_5_Inv_Cancer Pt_42_5_Inv_Cancer Pt_42_6_Inv_Cancer Pt_19_1_Inv_Cancer Pt_12_13_Inv_Cancer Pt_12_15_Inv_Cancer Pt_12_14_Inv_Cancer Pt_11_17_Inv_Cancer Pt_8_3_Inv_Cancer Pt_11_22_Inv_Cancer Pt_42_3_Inv_Cancer Pt_42_4_Inv_Cancer Pt_15_2_Inv_Cancer Pt_16_13_Inv_Cancer Pt_15_4_Inv_Cancer Pt_15_3_Inv_Cancer Pt_25_4_Inv_Cancer Pt_25_3_Inv_Cancer Pt_19_5_Inv_Cancer Pt_21_11_Inv_Cancer Pt_19_2_Inv_Cancer Pt_19_Inv_Cancer Pt_8_1_Inv_Cancer Pt_8_2_Inv_Cancer Pt_16_9_Inv_Cancer Pt_16_8_Inv_Cancer Pt_16_11_Inv_Cancer Pt_42_7_Inv_Cancer Pt_13_Inv_Cancer Pt_13_4_Inv_Cancer Pt_21_9_Inv_Cancer Pt_13_2_Inv_Cancer Pt_14_Inv_Cancer Pt_14_7_Inv_Cancer Pt_14_19_Inv_Cancer Pt_14_20_Inv_Cancer Pt_14_8_Inv_Cancer Pt_14_16_Inv_Cancer Pt_14_14_Inv_Cancer Pt_14_3_Inv_Cancer Pt_22_8_Inv_Cancer Pt_11_14_Inv_Cancer Pt_16_12_Inv_Cancer Pt_10_3_Inv_Cancer Pt_9_10_Inv_Cancer Pt_9_2_Inv_Cancer Pt_22_6_Inv_Cancer Pt_22_5_Inv_Cancer Pt_11_35_Inv_Cancer Pt_11_34_Inv_Cancer Pt_11_21_Inv_Cancer Pt_11_36_Inv_Cancer Pt_22_7_Inv_Cancer Pt_22_2_Inv_Cancer Pt_9_4_Inv_Cancer Pt_9_11_Inv_Cancer Pt_12_11_Inv_Cancer Pt_12_10_Inv_Cancer Pt_12_12_Inv_Cancer Pt_13_9_Inv_Cancer Pt_21_10_Inv_Cancer Pt_13_6_Inv_Cancer Pt_13_7_Inv_Cancer Pt_13_8_Inv_Cancer Pt_13_10_Inv_Cancer Pt_13_5_Inv_Cancer Pt_14_17_Inv_Cancer Pt_14_4_Inv_Cancer Pt_25_8_Inv_Cancer Pt_11_19_Inv_Cancer Pt_11_18_Inv_Cancer Pt_18_7_Inv_Cancer Pt_14_1_Inv_Cancer Pt_11_31_Inv_Cancer Pt_11_33_Inv_Cancer Pt_11_32_Inv_Cancer Pt_14_6_Inv_Cancer Pt_18_6_Inv_Cancer Pt_11_13_Inv_Cancer Pt_11_12_Inv_Cancer Pt_11_15_Inv_Cancer Pt_11_16_Inv_Cancer Pt_22_1_Inv_Cancer Pt_22_Inv_Cancer Pt_22_4_Inv_Cancer Pt_22_3_Inv_Cancer Pt_11_20_Inv_Cancer Pt_11_30_Inv_Cancer Pt_11_28_Inv_Cancer Pt_11_29_Inv_Cancer Pt_9_3_Inv_Cancer Pt_9_12_Inv_Cancer Pt_18_9_Inv_Cancer Pt_24_2_Inv_Cancer Pt_24_Inv_Cancer Pt_24_3_Inv_Cancer PrSt IIA PrSt IIIB PrSt IA PrSt IIIC PrSt III PrSt IA Lesion_Type nFim_c nFim_i nFT_c nFT_i p53_i p53_c STIL_c STIC_i STIC_c Inv_Cancer Cancer_floating Lesion Type BRCA status EMT / Morphogenesis Cell Cycle / Mitosis IFN Response DNA Repair RARPi Resistance Paclitaxel Resistance Congression Defects Attenuated Repl. Stress Signaling Pathways Lesion Type BRCA status BRCAstatus Mut WT Supplementary Figure 2. A) Barplot of Invasive Cancer ROI transcriptomes showing top 10 GOBPs significantly correlating with a PARP-inhibitor resistance signature from Tamura et al (pearson’s BH adjp < 0.05). B) Heatmaps of significant disease related GOBPs and Hallmarks as well as the Tamura et al transcriptional signatures for patients with invasive cancer. Primary cancer stage information is also noted on the heatmaps. C) Heatmap of gsva enrichment scores showing clustering patterns of invasive cancer ROIs using the processes and pathways described in Figure 2 panel (D-F). Clustering of ROIs was performed by k-means clustering.