Spatial Integration of Protein and Chromosomal States Reveals Early Copy-Number Changes and Genotype-Associated Immune Neighborhoods in Serous Ovarian Cancer Evolution

ORION-FISH integrates multiplex protein imaging with DNA-FISH to map chromosome copy-number changes in tissues, revealing early genomic diversification and genotype-specific immune neighborhoods in ovarian cancer initiation.

Abstract

Detecting chromosomal copy-number alterations together with protein-defined cell states in intact tissue is critical for understanding early clonal evolution and microenvironmental interactions in cancer. We developed ORION-FISH, which integrates high-plex tissue imaging with a morphology-preserving DNA fluorescence in situ hybridization (DNA-FISH) workflow and single-cell registration, yielding measurements concordant with clinical FISH. In high-grade serous ovarian carcinoma (HGSOC), ORION-FISH recapitulated known chromosomal changes while revealing subclonal heterogeneity missed by targeted sequencing. Applied to serous tubal intraepithelial carcinomas, precursors of HGSOC, ORION-FISH identified intermixed epithelial cells with MYC or CCNE1 copy-number gains, as well as concurrent alterations associated with distinct immune microenvironments. In addition, epithelial cells with MYC and CCNE1 copy-number gains were detected in morphologically normal fallopian tube epithelium, along with rare MDM4 increases across epithelial lineages. Together, ORION-FISH provides a framework linking chromosomal copy-number states to protein-defined phenotypes within preserved tissue architecture, enabling context-aware interrogation of early copy-number diversification at single-cell resolution.

Significance:

We introduce ORION-FISH, a spatially resolved workflow integrating multiplexed protein imaging with DNA-FISH to map genomic alterations within intact tissues. Applying this approach to ovarian cancer precursors reveals early copy-number diversification and associations with the local immune context, providing a foundation for studying how genomic and microenvironmental states coevolve during tumor initiation.

Introduction

Cancer emerges from genetic alterations that drive changes in cell state, enabling neoplastic cells to survive and proliferate within complex tissue ecosystems that include immune and stromal components (1–3). Next-generation sequencing (NGS) technologies have transformed cancer biology by enabling comprehensive detection of somatic mutations and copy-number alterations. However, most sequencing-based approaches require tissue dissociation or average signals across spatially heterogeneous domains, resulting in a loss of spatial context. As a result, rare genomic events—particularly those confined to small cell populations, early lesions, or precancers—are often obscured, limiting direct interrogation of early clonal selection and tumor–immune interactions.

Efforts to recover spatial genomic information have taken several forms. Histology-guided laser capture microdissection followed by multiregion sequencing or single-cell sequencing, including whole-genome approaches (4–8), can partially resolve subclone structure. However, these methods require dissociation or mechanical subdivision of tissue, and spatial relationships must be reconstructed computationally rather than measured directly. By contrast, spatial genomics approaches such as DNA fluorescence in situ hybridization (DNA-FISH; refs. 9–11), mutation-specific padlock probes (12), or in situ sequencing (13) allow direct visualization of genomic alterations within tissue but are typically limited in their ability to simultaneously resolve protein-defined cell states within the same cells.

Multimodal spatial profiling approaches—including multiplexed tissue imaging (14–18) and spatial transcriptomics—have provided unprecedented insight into tissue architecture, cell states, and immune organization. In these approaches, however, genomic alterations are generally inferred indirectly (e.g., from RNA expression patterns) or obtained from serial tissue sections. Consequently, the direct relationship between chromosomal copy-number alterations and tumor cell state remains incompletely defined, particularly in settings where tumor, normal, and stromal cells are intermixed within shared microenvironments (19). Because early cancer evolution is shaped by selection acting on such intermixed populations (19), there remains a critical need for methods that directly measure chromosomal alterations together with protein-defined cell states in intact tissue sections at single-cell resolution.

Such integrated measurements are particularly important for understanding early copy-number diversification, as they enable direct interrogation of how chromosomal alterations arise, coexist, and are differentially tolerated within local tissue and immune microenvironments. This is especially relevant in malignancies such as high-grade serous ovarian cancer (HGSOC), which is characterized by profound chromosomal instability (20) and arises from precursor lesions such as serous tubal intraepithelial carcinomas (STIC; refs. 21–23). Recent single-cell and spatial studies, including analyses of morphologically normal breast epithelia, have demonstrated that ostensibly normal tissues can harbor rare somatic copy-number alterations and mosaic genomic states without immediate clonal expansion (24, 25). In parallel, genomic analyses of epithelial cancers, including HGSOC, have revealed polyclonal and branching evolutionary architectures in which multiple genomic trajectories coexist prior to clonal dominance (26–31). Notably, emerging evidence indicates that not all early copy-number alterations are equivalently retained during tumor evolution, raising the possibility that some genomic changes arise transiently in normal epithelium but are selectively constrained or eliminated prior to malignant transformation. Together, these observations suggest that early copy-number alterations can precede morphologically defined precursor lesions and shape subsequent evolutionary trajectories, including interactions with the local immune microenvironment.

Despite these advances, directly visualizing copy-number alterations in situ and linking them to epithelial identity and local tissue context remains technically challenging. In practice, conventional DNA-FISH typically relies on the analysis of limited numbers of isolated nuclei (e.g., 50–100) to ensure scoring accuracy because the permeabilization and hybridization conditions required for probe binding can compromise nuclear morphology in intact tissue sections. This limitation restricts single-cell scoring within preserved tissue architecture and generally necessitates manual or semimanual evaluation of small numbers of nuclei, limiting scalability for whole-section imaging. Consequently, although DNA-FISH remains a clinical diagnostic standard, its capacity to resolve spatially organized clonal heterogeneity within intact tissue is constrained, particularly in densely packed epithelia and early precursor lesions such as STIC (21, 22).

As a result, several fundamental questions remain unresolved in early epithelial cancer precursors. In ovarian cancer, particularly STICs and their antecedent lesions, these include whether oncogenic copy-number gains arise as uniform, lesion-wide events or instead emerge as spatially intermixed subclonal states; whether cells within individual lesions harbor distinct combinations of alterations (e.g., isolated vs. combinatorial oncogene gains) reflecting parallel or branching genomic trajectories; how these genomic states relate to epithelial identity and differentiation; and whether they are associated with distinct patterns of local immune engagement or relative immune exclusion prior to invasive transformation. Addressing these questions requires an approach that preserves tissue and nuclear morphology, detects rare events within fields of morphologically normal cells, and directly links chromosomal copy-number states, protein-defined phenotypes, and spatial microenvironments within the same intact tissue section.

To address this need, we developed ORION-FISH, an integrated approach to clonal mapping that combines one-shot ORION multiplexed protein imaging (15) with a morphology-preserving DNA-FISH workflow and a single-cell registration and scoring pipeline. Applying ORION-FISH to HGSOC (32), its precursor lesions including STIC, and adjacent morphologically normal fallopian tube epithelium—the site of origin of HGSOC (21, 23)—we establish a quantitative spatial framework linking chromosomal copy-number states to epithelial identity and local tissue context in formalin-fixed, paraffin-embedded (FFPE) sections. Using this approach, we identify pronounced single-cell heterogeneity in MYC and CCNE1 copy-number states within STIC lesions, detect rare epithelial cells with elevated gene-specific signal counts in morphologically normal fallopian tube epithelium, and find genotype-associated differences in local immune organization. Together, these findings position ORION-FISH as a scalable strategy for resolving early cancer diversification and its tissue context in situ.

Results

Overview of the ORION-FISH Workflow

To enable simultaneous, spatially registered measurement of protein expression and chromosomal copy-number states within the same FFPE tissue section, we developed ORION-FISH, an integrated experimental and computational workflow (Fig. 1A). ORION-FISH integrates one-shot ORION multiplexed protein imaging (15) with a morphology-preserving DNA-FISH workflow and a single-cell registration and scoring pipeline, enabling quantitative analysis of gene-specific signal counts within intact tissue architecture. Using a commercially available immune-focused ORION antibody panel with minor modifications (Supplementary Table S1), we performed 16-plex protein imaging followed by four-plex DNA-FISH targeting MYC (8q24) and CCNE1 (19q12; refs. 10, 23, 28), together with corresponding centromeric controls (Fig. 1A–C; Supplementary Table S1).

Figure 1.

ORION-FISH enables integrated protein and chromosomal imaging in intact tissue. A, Schematic of the ORION-FISH workflow integrating one-shot multiplex immunofluorescence (mIF; ORION; 16–18 antibodies) with DNA-FISH on the same FFPE section. ORION staining and imaging (step 1; 20×, 0.8 NA) is followed by fluorophore deactivation (step 2), a four-plex DNA-FISH assay (step 3; 20×, 0.8 NA), and spatial registration with single-cell quantification (step 4), linking protein-defined cell states with chromosomal copy-number and spatial context. B, Study design and targets. Positive-control HGSOC tumors with MYC and/or CCNE1 copy-number alterations and incidental STIC lesions were profiled using an immune-enriched ORION antibody panel together with DNA-FISH probes targeting MYC (8q24) and CCNE1 (19q12) and matched centromeres. Antibodies (RareCyte) and probes (Empire Genomics) were commercially sourced. All positive control HGSOC tumors are mentioned in Supplementary Fig. S1A. C, Single-cell outputs including protein intensities, gene and centromere FISH counts, cell type/state assignments, and spatial coordinates. D and E, ORION-FISH imaging of CCNE1-amplified HGSOC (HGSOC-2). D, E-cadherin–positive tumor epithelium with multiple CCNE1 signals per nucleus and matched CEP19 signals. Multiplexed protein imaging shows proliferative tumor cells (Ki-67⁺ and E-cadherin⁺) and immune populations (E-cadherin⁻ and CD8^+/CD163⁺). Single-channel images are in Supplementary Figs. S1–S4; adjacent section conventional DNA-FISH in Supplementary Fig. S2B and S2C. E, Nuclear segmentation masks derived (ORION Hoechst, green) overlaid with DNA-FISH (red) in the same section, showing alignment across modalities. Arrows indicate rare nuclei lost or displaced during DNA-FISH. Additional single-color channels and segmentation examples are in Supplementary Fig. S1C and S1D. F and G, ORION-FISH mapping of MYC copy-number status in STIC epithelium. F, STIC from a BRCA1 mutation with p53-positive, E-cadherin–positive epithelium, and MYC copy-number gains. Dashed outlines indicate the STIC region; adjacent immune cells are present. G, Nuclear segmentation masks derived from Hoechst (green) overlaid with DNA-FISH DAPI (red) in STIC epithelium. The arrow indicates a displaced nucleus excluded during quality control (QC). Additional examples are in Fig. 2 and Supplementary Figs. S8–S10, S12, and S13. [A, Created with BioRender.com (https://app.biorender.com/portal/harvard)].

A key feature of ORION-FISH is the sequential integration of multiplexed protein imaging followed by DNA-FISH within the same tissue section. Protein imaging is performed first to preserve antigenicity and tissue architecture, after which fluorophores are chemically deactivated (14, 16), and an optimized DNA-FISH protocol [adapted from established traditional DNA-FISH methods (33, 34)] is applied. Protease digestion, hybridization, and wash conditions were optimized to preserve nuclear morphology while maintaining spatial correspondence with protein-defined cell phenotypes (Supplementary Figs. S1–S4). Preservation of nuclear morphology allows analysis of hundreds to thousands of cells from a sample, enabling scalable integration of protein and DNA measurements (Fig. 1D and E).

Validation in HGSOC with Known Genomic Alterations

We first validated ORION-FISH in four HGSOC cases with genomic alterations previously identified by clinical molecular pathology testing (OncoPanel NGS sequencing; Fig. 1B; Supplementary Figs. S1A, Supplementary Table S2; ref. 35). These included tumors with (i) MYC and CCNE1 coamplification, (ii) isolated CCNE1 amplification, (iii) isolated MYC gain, and (iv) MDM4 gain (1q32) with concurrent CBFB loss (16q22.1). Across all cases, ORION-FISH demonstrated strong concordance between nuclear segmentation and DNA-FISH signal localization at the single-cell level (Supplementary Figs. S1C and S1D, S5A and S5B). Signal intensities and spatial patterns were comparable with conventional DNA-FISH performed on adjacent sections (Supplementary Fig. S2B and S2C), indicating that integration with multiplex protein imaging does not compromise analytic performance.

ORION-FISH recapitulated expected copy-number patterns while enabling direct spatial visualization of heterogeneous copy-number states within intact tissue sections (Fig. 1D and E; Supplementary Figs. S4–S5). Independent validation using a clinically approved automated FISH scoring platform (BioView; refs. 36, 37) confirmed both the frequency and spatial distribution of alterations across >10,000 cells [e.g., (i) HGSOC-1, with ∼30% of cells harboring >5 copies of both MYC and CCNE1 and (ii) HGSOC-3, with ∼40% of cells harboring >5 copies of MYC alone; Supplementary Fig. S6; Supplementary Table S3]. Notably, ORION-FISH identified low-frequency subclonal events not reported by bulk sequencing, such as gains in CCNE1 in a subset of cells in HGSOC-3 (∼9%; Supplementary Fig. S6C–S6E). ORION-FISH also confirmed widespread low-level MDM4 gain without high-level amplification in HGSOC-4 (∼70% of cells with >2 copies without amplification; Supplementary Fig. S7).

Application to Precursor Lesions and Normal Epithelium

We next applied ORION-FISH to three cases containing isolated STIC—the recognized precursor of HGSOC—and adjacent morphologically normal fallopian tube epithelium (Fig. 1B, F, and G; Supplementary Table S2). Morphologically normal fallopian tube epithelium is shown as a comparator to STIC-associated regions in the corresponding figures. The chromosomal alterations in these STIC lesions were not known a priori. These tissues present a technically challenging setting due to dense epithelial packing and overlapping nuclei, particularly in the fimbriae epithelium in which early lesions typically arise. Reliable single-cell analysis in this context requires preservation of nuclear morphology and accurate assignment of DNA and protein signals to segmented cells. ORION-FISH enabled robust single-cell measurements in these settings by leveraging segmentation masks derived from ORION-Hoechst imaging, which were consistently applied to FISH-DAPI channels. This approach allowed accurate quantification of gene-specific copy-number signals even in tightly packed epithelium (Figs. 1G, 2A–2D; Supplementary Figs. S8–S10).

Figure 2.

Single-cell segmentation and cross-modality registration enable quality control (QC) of ORION-FISH data. A and B, Representative incidental STIC [STIC-3; BRCA wild-type (WT)]. A, Full section ORION imaging including normal epithelium. B, Higher magnification view of STIC lesion showing p53 overexpression; E-cadherin–positive epithelium with adjacent CD8⁺ and CD4⁺ T cells. C, Region of interest from B with single-nucleus segmentation mask (derived from ORION Hoechst) overlaid on ORION (green) and DNA-FISH (red) nuclear staining channels. Nearly all nuclei show spatial correspondence across modalities. Arrows indicate nuclei lost or displaced during DNA-FISH and excluded during QC. Dashed outline denotes STIC epithelium. D, Additional region from B showing segmentation mask overlaid on ORION and DNA-FISH nuclear channels. Dashed outline denotes STIC epithelium. E–I, Quality control of nuclear alignment and signal consistency. E, Hexbin plot of ORION and DNA-FISH nuclear intensities across epithelial cells from STIC-3. F, Cook distance-based outlier detection. Red points indicate nuclei with large residuals. G and H, Identification of low-confidence nuclei with reduced DNA signal intensity in one or both channels. I, Example of a nucleus lost during DNA-FISH (cell ID indicated), excluded during QC. Additional STIC specimens are shown in Supplementary Figs. S8–S10. H&E, hematoxylin and eosin.

Computational Integration and Quality Control Enable Scalable Single-Cell Quantification

To support scalable and reproducible single-cell analysis, we implemented a computational pipeline for image registration, segmentation, signal detection, and quality control. Whole-slide ORION and DNA-FISH images were registered using Ashlar (38) to achieve precise alignment across imaging modalities (see “Methods”). Nuclear and cellular segmentation was performed using a deep learning–based approach (Cellpose; Pachitariu and colleagues, bioRxiv 2025) applied to the ORION-Hoechst channel. Segmentation masks were propagated across both protein and DNA channels to generate a unified single-cell feature table.

FISH signal detection was performed using a deep learning-based framework [Spotiflow (39)], which identifies gene-specific and centromeric signals within each segmented nucleus. This enabled per-cell quantification of signal counts and integration with protein-defined phenotypes across >150,000 cells (per specimen) analyzed in precursor lesions and adjacent normal epithelium (Supplementary Fig. S11).

Quality control procedures excluded nuclei with insufficient DNA signal or evidence of displacement during processing, and spatial inspection confirmed that these excluded nuclei corresponded to poorly resolved or lost nuclei rather than biologically meaningful states (Fig. 2C–I; Supplementary Fig. S10A–S10G). Together, these procedures ensure robust quantification of gene-specific signal counts while minimizing technical artifacts and establish a framework for downstream biological interpretation.

STIC Epithelium Exhibits Heterogeneous MYC and CCNE1 Copy-Number States

Having established ORION-FISH as a robust approach for integrated protein and chromosomal profiling, we applied it to three STIC specimens. These analyses, derived from a small number of precursor lesions, illustrate the types of spatial genotype–phenotype relationships that can be directly resolved using this approach in early-stage tissue contexts. Following quality control, ORION-FISH enabled quantification of an average of 26,130 epithelial cells across these cases, including morphologically normal epithelium adjacent to STIC lesions. Consistent with prior reports (10, 23, 28, 40), STIC lesions exhibited frequent copy-number alterations involving MYC and CCNE1 (Figs. 1F, 3; Supplementary Fig. S8D). Elevated MYC signal counts relative to centromeric control were observed in the majority of STIC epithelial cells (∼66–77%), whereas CCNE1-associated signal increases were present in a subset (∼25–46%; Fig. 3A). These findings are consistent with early copy-number diversification as a characteristic feature of these precursor lesions (10, 23, 28).

Figure 3.

ORION-FISH reveals copy-number (CN) heterogeneity and spatial organization of STIC lesions. A, Quantification of epithelial cells with MYC gain, CCNE1 gain, or concurrent MYC and CCNE1 gain across three incidental STIC specimens. CN states were determined by automated single-cell FISH signal quantification normalized to matched centromeric controls (“Methods”). Cells were classified as exhibiting locus-specific CN gain when the gene-to-centromere ratio exceeded 1.3. Number of STIC epithelial cells analyzed: STIC-3 (n = 7,326), STIC-1 (n = 108), and STIC-2 (n = 1,735). B, Representative ORION-FISH image of a STIC lesion (STIC-3; BRCA WT) showing spatially distinct epithelial regions with MYC-only gain (purple), concurrent MYC and CCNE1 gains (green), and mixed regions with predominantly MYC-gain cells and a minority of dual-gain cells (red). Image were acquired at 20× (0.8 NA) on the ORION platform. C, Higher-magnification view of the region outlined in red in B, showing epithelium with predominantly MYC gain and rare cells with concurrent CCNE1 gain (arrowheads). Adjacent immune cells are present. Single-channel images with segmentation masks are shown in Supplementary Fig. S13. D, Higher-magnification view of the region outlined in green in B, enriched in epithelial cells with concurrent MYC and CCNE1 gains (arrowheads). Insets show representative nuclei (right) and single-channel MYC signal with segmentation masks (left). Corresponding single-channel overlays are shown in Supplementary Fig. S12; centromeric controls and orthogonal validation are shown in Supplementary Fig. S14. E, Higher-magnification view of the region outlined in purple in B, showing epithelial cells with MYC gain alone (arrowheads). Insets show representative nuclei with segmentation masks. Additional examples are shown in Supplementary Fig. S12. F, Independent validation of concurrent MYC and CCNE1 gains using a clinically validated automated FISH platform (BioView). Representative nuclei show increased MYC and CCNE1 CNs relative to centromeric controls. Full fields of view are shown in Supplementary Fig. S14. scale bar = 20 μm.

At single-cell resolution, these alterations were not uniform. Instead, STIC lesions comprised spatially intermixed epithelial cell populations with distinct copy-number states, including cells harboring isolated MYC gains, isolated CCNE1 gains, or concurrent gains of both oncogenes, rather than a single dominant copy-number state (Fig. 3B–E; Supplementary Fig. S12 and S13).

Independent validation using a clinically approved automated FISH scoring platform (BioView) confirmed both the presence and spatial distribution of signal patterns, including regions with concurrent MYC and CCNE1 gain within STIC lesions (Fig. 3F; Supplementary Fig. S14). This validation was particularly relevant for CCNE1 gains detected in STICs arising in BRCA1 mutation carriers (Supplementary Fig. S15), given prior observations that CCNE1 amplification and BRCA1/2 mutation status are typically mutually exclusive in advanced HGSOC (41, 42). The direct detection of these alterations at single-cell resolution in intact precursor tissue highlights the value of spatially resolved profiling for capturing early heterogeneity that may not be retained in advanced disease.

Morphologically Normal Epithelium Contains Cells with Extra Copies of MYC and CCNE1 and Associated Immune Neighborhoods

Unexpectedly, ORION-FISH also identified epithelial cells with higher MYC and/or CCNE1 copy numbers relative to centromeric controls within morphologically normal fallopian tube epithelium adjacent to STIC lesions (Fig. 4A–D; Supplementary Fig. S16–S18). These cells were interspersed among otherwise normal-appearing epithelium and lacked overt architectural or cytologic abnormalities by conventional histopathologic criteria. Examination of ORION protein data demonstrated that these cells did not exhibit p53 overexpression. Independent automated FISH scoring (BioView) confirmed the presence of copy-number gains, including CCNE1 gains, in normal epithelium from both BRCA-wild-type and BRCA-mutant cases.

Figure 4.

ORION-FISH detects copy-number gains in morphologically normal epithelium and associated immune neighborhoods in STIC cases. A, Representative ORION-FISH image from a STIC case (STIC-2; BRCA1 mutant) showing morphologically normal epithelium (p53-negative) adjacent to STIC epithelium (p53-overexpressing). The red box highlights E-cadherin–positive normal epithelium adjacent to STIC. Arrowheads indicate p53-negative epithelial cells with MYC copy-number gain. Images were acquired at 20× (0.8 NA) on the ORION platform. Additional examples are shown in Supplementary Fig. S9. B, Independent validation of MYC copy-number gain in morphologically normal epithelium using a clinically validated automated DNA-FISH scoring platform (BioView; 60×, NA 1.42, oil immersion with z-stack acquisition). Representative nuclei show increased MYC signal relative to CEP8. Full fields are shown in Supplementary Fig. S18. CN, copy number. C, Higher magnification view of the magenta-boxed region outlined in A, showing morphologically normal, p53-negative epithelium with CCNE1 copy-number gain (arrowheads). In contrast to regions with MYC gain, this area shows sparse immune cells. Images were acquired at 20× (0.8 NA) on the ORION platform. D, Independent validation of CCNE1 copy-number gain using BioView (60×, NA 1.42, oil immersion with z-stack acquisition). Representative nuclei show increased CCNE1 signal relative to CEP19. Full fields are shown in Supplementary Fig. S17. E, Quantification of CD8⁺Ki-67⁺ T-cell neighborhoods in morphologically normal epithelium from STIC cases (n = 3), stratified by epithelial copy-number status. Bars represent the proportion of epithelial cells with >1% CD8⁺Ki-67⁺ cells within a 50-µm radius. Statistical significance was assessed with binomial GLMMs (**, P < 0.01). Total epithelial cells analyzed: n = 69,201. F, Immune neighborhood composition associated with MYC-only copy-number gain versus diploid epithelial cells in STIC (n = 3). Bars represent the proportion of STIC epithelial cells with >1% of the indicated immune population within a 50-µm radius (**, P < 0.01; ***, P < 0.001, binomial GLMMs; Supplementary Table S4). Total number of STIC epithelial cells analyzed: n = 9,169; n = 2,179 total diploid; and n = 4,540 MYC gain. G, Comparison of immune neighborhoods associated with CCNE1-only copy-number gain versus diploid epithelial cells in STIC (n = 3). Bars represent the proportion of epithelial cells with >1% of the indicated immune population within a 50-µm radius. Total number of STIC epithelial cells analyzed: n = 9,169; n = 2,179 total diploid; and n = 752 CCNE1 gain (**, P < 0.01; *, P < 0.05, binomial GLMMs; Supplementary Table S4).

Although normal fallopian tube epithelium is not globally enriched for immune cells (19), spatial analysis of ORION immunofluorescence data revealed that normal epithelial cells with MYC gains—either alone or in combination with CCNE1 gain—were associated with enriched immune neighborhoods within a 50-µm radius, including CD8⁺ T cells and proliferating CD8⁺ T cells (Fig. 4A and E; Supplementary Fig. S16, Supplementary Table S4; P < 0.01). Neighborhoods were defined using a 50-µm radius to capture intraepithelial and immediately adjacent stromal cells (see “Methods”). Together, these observations indicate that epithelial cells with copy-number gains involving canonical HGSOC-associated loci can be detected within morphologically normal fallopian tube epithelium and are associated with localized immune engagement prior to the emergence of histologically defined precursor lesions. These findings motivated subsequent analyses in STIC lesions, in which such copy-number states are more abundant, heterogeneous, and spatially organized.

ORION-FISH Maps Genotype-Specific Immune Interactions within STIC

Because these incidental STIC specimens were previously shown to exhibit coexisting immune activation and suppression in our Pre-Cancer Atlas study (19), they provide a well-defined context for examining how specific chromosomal alterations relate to local immune organization. We therefore used ORION-FISH to map epithelial chromosomal copy-number states together with protein-defined cell identity and spatial proximity to immune populations at single-cell resolution within intact STIC lesions. Neighborhood analyses were performed across all quality-controlled cells in the imaged sections, rather than selected fields, with neighborhoods defined per epithelial cell (50-µm radius).

Within STICs, epithelial cells harboring MYC copy-number gains were frequently embedded within immune-enriched microenvironments. Spatial proximity analysis revealed a significant enrichment of activated immune populations—including proliferating CD8⁺ and CD4⁺ T cells, antigen-presenting cells (CD11c⁺), and CD68⁺ macrophages—within local neighborhoods surrounding these epithelial cells [Figs. 1F and 4F; Supplementary Table S4; P < 0.01, binomial generalized linear mixed model (GLMM)].

By contrast, epithelial cells with CCNE1 gains—either alone or in combination with MYC gain—were associated with a different immune context characterized by reduced proximity to activated CD8⁺ T cells, fewer APC-rich neighborhoods, and decreased CD4⁺ T-cell association (Figs. 3B and D, 4G; Supplementary Table S4; P < 0.05, GLMM). For example, 41% of STIC epithelial cells with MYC gains had neighborhoods containing >1% CD11c⁺ cells compared with only 30% of cells with CCNE1 gains (Fig. 4G). Importantly, these differences were observed despite CCNE1 gains being present in a substantial fraction of epithelial cells within the same lesions, indicating that local immune organization tracked with copy-number state rather than cell abundance alone. Together, these findings demonstrate that distinct epithelial copy-number states are associated with spatially organized and genotype-specific immune microenvironments within the STIC lesions.

Distinct Classes of Early Copy-Number Alterations Exhibit Different Contextual Constraints in Benign Fallopian Tube Epithelium

Having established that MYC and CCNE1 copy-number gains can be detected in both STIC and morphologically normal fallopian tube epithelium, we asked whether early copy-number alterations are uniformly tolerated across genomic loci or instead subject to distinct biological constraints prior to overt transformation. We focused on alterations that may arise early but are not commonly retained in advanced HGSOC, reasoning that these may represent transient or context-dependent genomic states.

We examined copy-number changes at the MDM4 locus (1q32), motivated by reports of 1q gains in morphologically normal epithelia (24, 25, 43) and by the observation that MDM4 amplification is uncommon in HGSOC, in which TP53 mutation is nearly universal and typically mutually exclusive with MDM4 gain (28, 32, 44). As a dosage-sensitive negative regulator of p53 signaling, MDM4 provides a model for early alterations that may transiently modulate stress responses prior to TP53 inactivation by mutation (45). Using conventional DNA-FISH in benign fallopian tube specimens lacking STIC or HGSOC (n = 30; ref. 46), we detected rare epithelial cells with MDM4 copy-number gains (Supplementary Figs. S19–S20), with enrichment in BRCA1 mutation carriers (Supplementary Fig. S20D). These specimens were obtained from risk-reduction surgeries in BRCA mutation carriers or opportunistic salpingectomies in BRCA wild-type women, confirming representation of both genetic contexts.

To define the epithelial contexts of these events, we applied an expanded ORION-FISH panel incorporating lineage markers (PAX8⁺ and FOXJ1⁺) together with MDM4 probes across normal tissue, p53 signatures, and STIC lesions. MDM4 copy-number gains were confined to rare epithelial cells in morphologically normal tissue and were not observed in TP53-mutant epithelial populations of STIC and p53 signatures (Supplementary Fig. S20E; Supplementary Table S1), consistent with the possibility that some copy-number alterations arise in normal epithelium but are not retained in TP53-mutant precursor states (47). Notably, MDM4 copy-number gains were observed in both PAX8⁺ and FOXJ1⁺ epithelial populations and were not associated with local immune neighborhoods (Supplementary Figs. 20F and 20G). Together, these findings support a model in which distinct classes of early copy-number alterations exhibit different contextual constraints, while indicating that rare MDM4 gains can occur in benign fallopian tube epithelium; however, larger cohorts will be required to define their prevalence and biological significance.

Discussion

Understanding early tumor evolution requires approaches that directly link genomic alterations to protein-defined cell states within their native tissue context at single-cell resolution. This need is particularly acute for copy-number alterations, which are often assessed using dissociative or region-averaging methods that obscure spatial organization, or inferred indirectly from transcriptomic data rather than measured in situ (13, 48, 49). Sensitivity and accuracy are especially critical when studying early and rare events in cancer initiation, such as isolated epithelial cells harboring gene-specific copy-number alterations within morphologically normal tissue, in which interpretation depends on precise alignment of cell identity and microenvironmental context. The ability to identify such altered epithelial cells in situ raises the possibility that genomic alterations may arise prior to lesion formation and are subsequently shaped by local selective pressures. More broadly, these considerations suggest that early tumor evolution reflects not only the accumulation of genomic alterations within epithelial cells but also the interaction of these altered cells with the surrounding tissue ecosystem that influences their persistence or elimination.

Here, we introduce ORION-FISH, a spatially integrated workflow that combines one-shot multiplexed protein imaging with morphology-preserving DNA-FISH on the same FFPE tissue section, enabling single-cell registration of chromosomal copy-number states with protein phenotypes. By performing protein imaging first to preserve antigenicity and tissue architecture, followed by an optimized DNA-FISH protocol that maintains nuclear morphology, ORION-FISH enables quantitative copy-number scoring concordant with conventional clinical FISH while retaining the spatial context required for clonal mapping in densely packed epithelial tissues such as the fallopian tube. In this way, ORION-FISH extends multiplexed imaging frameworks (15), including our prior work in fallopian tube precursors (19), by directly measuring chromosomal alterations within the same cells in which immune and epithelial states are defined. More broadly, ORION-FISH complements emerging genotype–phenotype mapping approaches that link genomic alterations to transcriptional states in situ (50) and contributes to a growing class of integrated frameworks that directly couple genotype, phenotype, and tissue context at single-cell resolution.

Applying ORION-FISH across HGSOC, STIC, and morphologically normal fallopian tube epithelium reveals several features of early HGSOC evolution that are difficult to detect using existing approaches. Within STIC lesions, MYC and CCNE1 copy-number alterations are not uniform but instead exhibit pronounced cell-to-cell heterogeneity. Individual lesions comprise spatially intermixed epithelial subclones with isolated MYC gains, isolated CCNE1 gains, or concurrent gains, rather than uniformly converging on a single dominant copy-number state. This mosaic organization supports a model in which multiple genomic trajectories can coexist at the precursor stage, consistent with branching evolutionary architectures described in epithelial cancers, and provides a tissue-level framework for understanding how distinct clonal programs may emerge prior to invasive transformation (13, 24–27, 43). These findings further suggest that distinct copy-number states within STIC lesions may correspond to epithelial populations occupying different functional trajectories during early lesion evolution rather than representing static genomic differences. These observations are best interpreted as proof-of-principle examples of the types of genotype–phenotype–microenvironment relationships that can now be measured directly rather than as definitive statements of prevalence or clinical significance.

An additional advance enabled by ORION-FISH is the ability to associate defined chromosomal copy-number states with local immune context within intact tissue. Although normal fallopian tube epithelium adjacent to STICs is not globally immune-enriched, epithelial cells with MYC- or CCNE1-associated copy-number gains in normal epithelium were frequently surrounded by locally enriched immune neighborhoods, suggesting that a subset of these early altered states may be immunologically visible prior to lesion formation. Within STIC lesions, epithelial cells with MYC-associated copy-number gains were preferentially embedded within immune-enriched neighborhoods containing proliferating T cells, antigen-presenting cells, and macrophages, whereas epithelial cells with isolated CCNE1 gains were associated with relative immune depletion despite comprising a substantial fraction of the lesion. These findings indicate that local immune organization tracks with epithelial copy-number state rather than cell abundance alone and suggest that early chromosomal alterations may differ in how epithelial cells are sensed or tolerated by the surrounding immune microenvironment. In this view, emerging epithelial subpopulations may not only mark clonal diversification but also correspond to distinct epithelial states that interact differently with—and are differentially constrained by—the surrounding tissue ecosystem during early tumor evolution. Defining the mechanisms underlying these associations will require larger cohorts and orthogonal measurements of antigen presentation, stress-response pathways, and immunomodulatory programs integrated with copy-number state.

Conceptually, ORION-FISH complements spatial transcriptomic and computational approaches that infer copy-number alterations from RNA or chromatin features. Inference-based methods offer genome-wide scope and scale but are indirect and may be less sensitive for rare or spatially isolated events, particularly in FFPE tissue (51). By contrast, ORION-FISH provides direct, quantitative measurement of targeted chromosomal states while preserving tissue architecture and linking these states to protein-defined cell identity and neighborhood features. This makes ORION-FISH well-suited for validating inferred alterations from atlas-scale datasets and for interrogating early clonal diversification in settings where rare events and spatial context are central. Although the present study used a 16- to 18-plex protein panel optimized for immune context, the ORION platform is inherently extensible through iterative imaging and cycling (15), providing a path toward deeper molecular phenotyping and integration of pathway-level features with chromosomal states.

This study has limitations. ORION-FISH is targeted to selected loci and does not provide genome-wide copy-number profiles, and the number of precursor lesions amenable to integrated analysis remains limited. Although the observed immune associations were consistent within lesions and supported by neighborhood-based analyses, larger cohorts will be required to refine effect sizes, establish population frequencies, and assess generality across clinical contexts. Future work expanding DNA probe panels, integrating additional epithelial, immune, and stress-response markers, and applying ORION-FISH to longitudinal or risk-stratified specimens will help clarify how early copy-number alterations arise, persist, or are selectively constrained. More broadly, by resolving rare subclonal populations together with their spatial contexts in intact tissue, ORION-FISH provides a practical framework for testing hypotheses about early genomic diversification, immune interactions, and selective constraints that shape the earliest phases of ovarian cancer evolution (22) and may ultimately inform strategies for early detection, interception, and prevention.

Methods

Ethics and Patient Specimens

All research described in this article was conducted in accordance with applicable ethical regulations and approved by Institutional Review Boards at Brigham and Women’s Hospital (BWH), Harvard Medical School, the University of Pennsylvania (UPenn), and the Swedish Cancer Institute. Four HGSOC cases were identified in the BWH Pathology department archives and used as positive controls for assay development and validation. Clinical molecular pathology reports for these cases document alterations involving MYC and CCNE1 (coamplification or isolated gains), or MDM4 gain with concurrent CBFB loss. These specimens were used to establish and optimize the ORION-FISH workflow.

To investigate early genomic alterations in ovarian cancer precursors and normal tissues, ORION-FISH was subsequently applied to fallopian tube epithelium and precursor lesions. Adjacent tissue sections from precursor lesions (n = 7), previously characterized as part of our Pre-Cancer Atlas study (19), were analyzed. In addition, 30 age-matched and BRCA status–matched benign fallopian tube specimens were identified at UPenn for analysis of morphologically normal epithelium as previously described (46). Following Institutional Review Board approval, serial 5-μm-thick sections were prepared for all cases. Hematoxylin and eosin staining was performed on adjacent sections to support histologic assessment. All specimens were FFPE, and diagnoses of STIC, p53 signatures, and benign epithelium were confirmed by board-certified pathologists using established criteria. Benign fallopian tube specimens were reviewed to confirm the absence of STIC or invasive carcinoma.

ORION Staining Protocol

ORION staining was performed largely as described by Lin and colleagues (15), with minor modifications to the antigen retrieval procedure. FFPE slides were baked at 55°C to 60°C for 55 minutes prior to shipment from different sites. Upon receipt, slides were processed on a BOND RX automated IHC/ISH stainer (Leica Biosystems). Slides were baked at 55°C for 60 minutes, dewaxed using Bond Dewax Solution at 72°C, and subjected to antigen retrieval using Epitope Retrieval Solution 2 (Leica Biosystems) at 100°C for 20 minutes, followed by Bond Wash Solution for 3 minutes at room temperature. Subsequent staining steps followed the standard one-shot ORION staining protocol using the v2 antibody panel (RareCyte Inc.). Multiplexed fluorescence images were acquired using a 20× objective (0.8 NA) on the ORION imaging platform (RareCyte Inc.).

Antibody Panels for ORION Staining

The commercially available RareCyte ORION v2 antibody panel was used with minor modifications, including the addition of a p53 antibody (panel 1: Fig. 1B). For panels incorporating additional fallopian tube–specific markers (panel 2: Supplementary Fig. S20E), antibodies were conjugated to ArgoFluor dyes according to RareCyte’s protocol and were used to construct the extraction matrix for spectral unmixing and downstream processing. Detailed information on all antibodies, conjugates, and panel compositions, including Research Resource Identifiers, is provided in Supplementary Table S1.

ORION Image Processing

Channel registration, illumination correction, and geometric distortion correction were performed with Artemis software on the ORION platform. Unstitched image files generated by Artemis (“pysed.ome.tif”) were used for downstream registration of both ORION and DNA-FISH images (see DNA-FISH and image registration below).

Development of a Morphology-Preserving DNA-FISH Protocol (ORION-FISH)

Following completion of ORION imaging, coverslips were removed by immersion in 1× PBS at room temperature until they detached. Slides were washed in PBS (2 minutes) and then incubated in a solution of 4.5% hydrogen peroxide (Thermo Fisher Scientific) and 24 mmol/L NaOH in PBS under white light illumination for 15 to 20 minutes to deactivate the fluorophores. Shorter deactivation times (5–10 minutes) could be used in the absence of strongly retained nuclear signals (e.g., p53), as prolonged exposure (>15 minutes) can compromise tissue integrity. After rinsing with PBS, slides were incubated in 0.2 N HCl (Sigma-Aldrich) for 20 minutes at room temperature, followed by washes in purified water (3 minutes) and 2× saline sodium citrate (SSC; 3 minutes; Thermo Fisher Scientific). Slides were then incubated in 10 mmol/L citrate buffer (pH 6.2; Sigma-Aldrich) at 80°C for 1 hour. Following rinses in 2× SSC and distilled water (5 minutes), excess liquid was removed, and ice-cold ISH pepsin (Dako Omnis, Agilent Technologies Inc.) was applied. Slides were incubated at 37°C for 6.5 to 7 minutes using a ThermoBrite system (Abbott), with digestion time adjusted based on tissue type.

After digestion, slides were rinsed twice in 2× SSC and dehydrated through graded ethanol. After drying, probe mixtures (1 μL of probe; total volume 20 μL SwiftFISH hybridization buffer; Empire Genomics) were applied under a 24 × 50 mm coverslip, sealed, and denatured at 75°C for 7 minutes using the ThermoBrite system. Hybridization was performed overnight at 37°C on the ThermoBrite system. The following day, coverslips were removed, and slides were washed in 2× SSC containing 0.3% IGEPAL CA-630 (Sigma-Aldrich) at 72°C for 2 minutes, followed by a 2× SSC rinse (2 minutes) and ethanol dehydration. Slides were air-dried in the dark and mounted with VECTASHIELD Antifade Mounting Medium containing DAPI (12–14 μL; Vector Laboratories) and stored at 4°C in the dark for at least 15 minutes prior to imaging. Across samples, the proportion of nuclei lost or displaced during the DNA-FISH procedure was low and excluded during quality control. Commercially available DNA-FISH probes (Empire Genomics) were used for all experiments; probe details are provided in Supplementary Table S1.

ORION-FISH Imaging

To ensure optimal focus across both protein and DNA imaging modalities, additional focus points were specified during ORION imaging, particularly within regions of interest such as STIC lesions. These focus points were revisited and refined prior to whole-slide acquisition following DNA-FISH. Imaging parameters were held constant across samples within each experiment. Because DNA-FISH requires the removal of extracellular matrix proteins (34, 52, 53) to enable probe penetration, careful focusing was essential to minimize out-of-focus nuclei. Imaging parameters and channel settings are detailed in Supplementary Table S1. Unstitched imaging files (“pysed.ome.tif”) generated by Artemis Software were subsequently registered across both ORION and DNA-FISH modalities using Ashlar (38).

Image Processing and Quality Control

Image processing and analysis were performed with Ashlar (38), Cellpose (Pachitariu M and colleagues, bioRxiv 2025) for segmentation, and custom R scripts (v4.5.1; https://github.com/labsyspharm/cycif.importer), consistent with prior workflows (16, 19). Raw image tiles from both ORION and DNA-FISH acquisitions were stitched and registered using Ashlar (38) within the MCMICRO (54) pipeline. Following registration, OME.TIF files were segmented using Cellpose based on Hoechst signals from the ORION image round, and single-cell features were extracted to generate per-cell datasets (Supplementary Fig. S11A). Automated FISH spot detection and scoring were performed using Spotiflow (39) and integrated with the single-cell feature table to generate the final quantification dataset (Supplementary Fig. S11B and S11C). Quality control included visualization of FISH spot detection overlaid with nuclear segmentation masks to confirm accurate assignment of copy-number states at single-cell resolution (Fig. 2; Supplementary Figs. S8–S10, S12, and S13).

Quality control for copy-number analysis focused on epithelial cells, in which alterations were expected. Cells lacking detectable signals in either the ORION Hoechst or FISH DAPI channels were excluded. DNA signal intensity distributions were examined to identify low-quality nuclei, and outliers were detected using Cook distance thresholds (0.0001–0.001) to identify nuclei with abnormally low DNA signals (Fig. 2; Supplementary Fig. S10). To avoid overfiltering biologically valid cells, only low-intensity outliers (negative residuals) were removed for precursor lesions. Spatial inspection confirmed that excluded nuclei corresponded to lost or poorly resolved nuclei. Regions affected by tissue loss or folding were excluded from analysis, consistent with prior multiplex imaging workflows (14, 16, 19). For benign fallopian tube and p53 signature specimens (Supplementary Fig. S20E), which exhibit higher epithelial density, filtering was applied more conservatively, with all Cook distance outliers removed. One benign fallopian tube case was excluded due to the inability to reliably assign MDM4 signals at single-cell resolution. Copy-number status was defined using gene-to-centromere ratios (MYC/CEP8 and CCNE1/CEP19), with ratios >1.3 classified as copy-number gain. This framework enables reproducible, quantitative single-cell copy-number calls directly linked to protein-defined cellular phenotypes within intact tissues.

Cell-Type Identification

Cell types and states were defined using marker-based gating performed independently for each sample using the open-source Gater visualization and analysis tool, in combination with binary gating strategies as described previously (16, 19). Documentation for Gater is available at https://github.com/labsyspharm/minerva_analysis/wiki/Gating. Initial gating thresholds were defined based on marker intensity distributions, visually inspected, and refined prior to integration with the single-cell feature table. Cell type and state definitions were informed by established literature and prior ORION and CyCIF studies (16, 19).

Cell Population Proportions

Cell population proportions were calculated as the frequency of marker-defined phenotypes within each sample. Downstream quantification across disease stages and statistical analyses were performed in R (v4.5.1) using custom pipelines implemented in the cycif.importer package (https://github.com/labsyspharm/cycif.importer).

Proximity Analysis and Local Neighborhoods

To characterize the local immune microenvironment of individual epithelial cells, spatial proximity analysis was performed by quantifying neighboring cells within a 50-µm radius of each cell centroid. Euclidean distances were computed in two-dimensional (X and Y) coordinate space using the “st_is_within_distance” function from the “sf” package in R (v4.5.1). The focal cell was excluded from its own neighborhood. For each focal cell, neighborhood features were computed, including: (i) total neighbor count: number of cells within the 50-µm radius; (ii) counts of marker-positive neighbors (e.g., CD8⁺ and CD8⁺Ki-67+); and (iii) marker-positive neighbor percentages: marker-positive neighbors (calculated as count/total × 100).

A 50-µm radius was selected to capture local epithelial–immune interactions, including both intraepithelial immune cells and those in the immediately adjacent stromal compartment, consistent with prior spatial immunology studies of short-range cellular interactions in epithelial tissues (55).

Statistical Modeling of Immune–Genome Associations

Spatial neighborhood analyses were performed using whole-slide imaging data and included all cells passing quality control across each slide (n = 417,299 cells). Because neighborhoods were defined relative to each epithelial cell, immune-cell composition was quantified across all neighboring immune cells surrounding each epithelial cell, and the number of analyzed neighborhoods corresponds to the number of epithelial cells included.

To assess associations between epithelial genomic alterations and local immune composition, analyses were restricted to epithelial cells located in morphologically normal fallopian tube epithelium or STIC lesions (as defined by pathologist annotation) with at least one neighboring cell (total_neighbors >0). Associations between MYC and CCNE1 copy-number status (diploid vs. gain) and local immune composition were evaluated using binomial GLMMs implemented in the lme4 R package (Bates and colleagues, arXiv 2014). For each immune population (CD163⁺, CD8⁺Ki-67+, CD4⁺Ki-67+, CD11c⁺, CD68⁺, CD8⁺, and CD4⁺), the proportion of neighboring cells with the specified phenotype was modeled as a function of the focal cell’s MYC and CCNE1 copy-number status.

The model was specified as follows:

cbind(n_success, n_failure) ∼ MYC_category + CCNE1_category + MYC_category:CCNE1_category + (1 | patient_id), where n_success is the number of neighboring cells with the given immune phenotype and n_failure is the number of all other neighboring cells. MYC_category and CCNE1_category were defined as categorical variables (diploid as the reference; gain indicating copy-number alteration), and patient ID was included as a random effect to account for interpatient variability. The interaction term (MYC_category:CCNE1_category) was used to test whether combined MYC and CCNE1 alterations on immune proximity differed from their individual effects (Supplementary Table S4: glm statistics summary).

Visualization of Immune–Genome Associations

For visualization, we quantified the proportion of epithelial cells with >1% neighboring cells of a given immune population. This threshold was selected to facilitate comparison across conditions, as immune cells are relatively sparse compared with epithelial cells, and full distribution can obscure differences. For every combination of MYC and CCNE1 status (both diploid, MYC gain only, CCNE1 gain only, or both gains), the proportion was calculated as the number of epithelial cells exceeding the 1% threshold divided by the total number of epithelial cells in that category. As an example, in STIC epithelium, there were 9,169 epithelial cells, of which 2,179 were diploid for both MYC and CCNE1. Of these diploid cells, 712 had >1% CD11c⁺ immune cells in their local neighborhood. The proportion of diploid epithelial cells exhibiting >1% CD11c⁺ neighbors was therefore 712/2,179 = 0.326, or approximately 33% (Fig. 4F).

Conventional DNA-FISH

Conventional DNA-FISH was performed according to established protocols (34). Slides were imaged either on the ORION platform (20×, 0.8 NA; for HGSOC cases) or using a VS200 slide scanner (Olympus; for benign fallopian tube cases) with 20× (0.8 NA) and/or 60× in oil immersion (1.42 NA) objectives using z-stack acquisition. The VS200 system was used for three-channel imaging (DAPI, FITC, and TRITC) of benign fallopian tube specimens probed for CEP1 (green) and MDM4 (red). Images were visualized and manually scored (50–70 cells per benign fallopian tube specimen) using QuPath (56). All probes were obtained from Empire Genomics.

BioView Imaging and Automated FISH Scoring

BioView is a clinically approved platform for automated scoring of conventional DNA-FISH (36, 37). To independently validate the DNA-FISH component of the ORION-FISH workflow, slides were submitted to BioView following completion of the DNA-FISH assay (n = 3 HGSOC positive controls; n = 3 incidental STIC cases). Image acquisition was performed using an Olympus BX63 microscope (60×, 1.42 NA). The BioView system employs dynamic z-stacking for each fluorescence channel, automatically determining the number of focal planes required to capture all FISH signals and generating composite images from the stacked planes. Due to the increased acquisition time associated with high-magnification z-stack imaging, regions of interest were manually selected, and fields of view were acquired using a 60× oil-immersion objective (NA 1.42) on FFPE tissue sections. Following the acquisition, nuclei were automatically segmented, and FISH signals were detected and assigned to individual nuclei by the BioView software. On average, more than 10,000 cells were quantified per tumor sample and more than 14,000 cells per STIC specimen. For visualization, representative fields of view are shown (Supplementary Figs. S6, S14, S15, S17, and S18; Supplementary Table S3). Cells were classified based on per-nucleus signal counts using standard automated criteria (e.g., ≤2, = 2, or >2 signals per cell). Signal counts and cell-level classifications were exported and used to quantify the copy-number alterations across samples. These data were used to validate copy-number states identified by ORION-FISH in positive-control HGSOC cases and to confirm the spatial distributions of alterations observed in STIC specimens.

Statistical Analysis

Statistical significance was defined as P < 0.05 unless otherwise stated. Statistical analyses were performed using R (v4.5.1) or GraphPad Prism (v10.0.2).

Schematic Diagrams

Schematic diagrams (Fig. 1) were created using BioRender.com.

Supplementary Material

Supplementary Table S1

Table S1 provides the details of antibodies (ORION) and FISH probes used in this manuscript.

Supplementary Table S2

Table S2 provides the cohort description.

Supplementary Table S3

Table S3 provides the FISH scores by BioView for HGSOC cases.

Supplementary Table S4

Table S4 provides the statistics summary.

Supplementary Figures S1-S20

Supplementary Figures provide additional evidence and validation of copy number changes from the ORION-FISH assay.

Data Availability

Custom code developed for this study is available at https://github.com/labsyspharm/orion-fish-ms-2026.

Authors’ Disclosures

Y.-A. Chen reports personal fees from RareCyte outside the submitted work. J.L. Muhlich reports grants from the Gray Foundation and Ludwig Cancer Research during the conduct of the study. C.W. Drescher reports grants from the Gray Foundation and Canary Foundation during the conduct of the study. P.K. Sorger reports personal fees from Merck and Danaher Corporation, grants and personal fees from RareCyte, and personal fees and other support from Glencoe Software during the conduct of the study, as well as personal fees from Montai Therapeutics outside the submitted work. R. Drapkin reports personal fees from Repare Therapeutics, Immunogen, GSK, and Light Horse Therapeutics outside the submitted work, as well as a patent for Methods of detecting ovarian cancer issued and a patent for HER3 inhibition in low-grade serous ovarian cancer issued. S. Santagata reports grants from the Gray Foundation and Ludwig Cancer Research during the conduct of the study, as well as personal fees and nonfinancial support from Roche Diagnostics Corporation and the San Antonio Breast Cancer Symposium outside the submitted work. No disclosures were reported by the other authors.

Authors’ Contributions

T. Kader: Conceptualization, formal analysis, investigation, visualization, methodology, writing–original draft. Y.-A. Chen: Data curation, formal analysis, visualization, methodology, writing–review and editing. C.B. Hug: Formal analysis, methodology, writing–review and editing. J.-R. Lin: Formal analysis, visualization, methodology, writing–review and editing. J.L. Muhlich: Data curation, software, visualization, methodology, writing–review and editing. S. Coy: Sample annotations. E. Jung: Provided and annotated reagents. L.E. Schwartz: Provided and annotated reagents. T. Fazio: Provided samples. C. Chiu: provided and annotated reagents. S.T. Ryall: Methodology. C.W. Drescher: Funding acquisition, writing–review and editing, provided and annotated reagents. P.K. Sorger: Resources, supervision, funding acquisition, writing–review and editing. R. Drapkin: Resources, funding acquisition, writing–review and editing. S. Santagata: Conceptualization, resources, supervision, funding acquisition, methodology, writing–review and editing.