Multimodal Spatial Profiling Reveals Immune Suppression and Microenvironment Remodeling in Fallopian Tube Precursors to High-Grade Serous Ovarian Carcinoma

Authors

Kader T

Lin JR

Hug C

Coy S

Chen YA

de Bruijn I

Shih N

Jung E

Pelletier RJ

Lopez Leon M

Mingo G

Omran DK

Lee JS

Yapp C

Satravada BA

Kundra R

Xu Y

Chan S

Tefft JB

Muhlich J

Kim S

Gysler SM

Agudo J

Heath JR

Schultz N

Drescher C

Sorger PK

Drapkin R

Santagata S

Doi

10.1101/2024.09.25.615007

PMID: 39386723 · DOI: 10.1101/2024.09.25.615007 · Journal: bioRxiv (2024)

TL;DR

Kader and colleagues integrated cyclic immunofluorescence (CyCIF), 3D CyCIF, and GeoMx whole-transcriptome spatial transcriptomics across 44 fallopian tube (FT) FFPE specimens from 43 patients spanning the entire High-Grade Serous Ovarian Carcinoma (HGSOC) progression axis: incidental p53 signatures, incidental serous tubal intraepithelial carcinomas (STIC.I), STIC co-occurring with cancer (STIC.C), and invasive HGSOC. They show that localized type-I/II interferon (IFN) signaling, chromosomal instability with micronuclear rupture, and cGAS-STING activation are detectable as early as the p53-signature stage and progressively intensify. Early lesions display active immune surveillance (cDC1, NK cells, tissue-resident memory CD8+ T cells), but the NK–cDC1 axis collapses during the STIC.I→STIC.C transition while HLA-E+ epithelium, exhausted T cells (LAG3+, PD1+), and FoxP3+ Tregs accumulate, marking a transition to immune suppression that precedes invasion. Data are released as the cBioPortal study ovary_geomx_gray_foundation_2024.

Cohort & data

  • 44 FFPE fallopian tube specimens from 43 patients (one patient with bilateral STIC contributed two), assembled from the University of Pennsylvania and the Swedish Cancer Institute (Seattle) PMID:39386723.
  • Group 1 — STIC.C (n=24): invasive cancer with co-occurring STIC; BRCA WT 15/24, germline BRCA1 7/9, germline BRCA2 6/7, somatic 2/9; median age 65.5 (range 46–85); 8/24 stage I (6/8 gBRCA), 16/24 stage II–III (13/16 BRCA WT); only 3/24 received neoadjuvant chemotherapy.
  • Group 2 — incidental precursors (n=19): p53.I (n=10; 5 gBRCA1, 5 gBRCA2) and STIC.I (n=9; 5 gBRCA, 1 gBRCA2, 4 BRCA WT) found at risk-reducing salpingo-oophorectomy or opportunistic salpingectomy; median age 47 (range 34–72).
  • Cancer type: High-Grade Serous Ovarian Carcinoma and its FT precursors.
  • Dataset: ovary_geomx_gray_foundation_2024 on cBioPortal — both CyCIF images and GeoMx data are publicly available there.
  • Assays/methods: CyCIF with a 31-antibody panel (2D), 3D CyCIF on a Zeiss LSM980 confocal for one STIC.C case, and GeoMx WTA spatial transcriptomics (NanoString); 35/44 specimens had sufficient material for GeoMx (initial 603 ROIs → 542 after QC).
  • Image processing: MCMICRO Nextflow pipeline (ASHLAR registration, UNMICST2 segmentation); single-cell quantification via the gator gating tool (labsyspharm/minerva_analysis).
  • Statistics: Bayesian ordinal regression (brms) for stage-wise gene-expression trends, binomial GLMMs (lme4, glmmTMB) with patient random intercepts for cell-proportion comparisons, GSEA against MSigDB Hallmark and Reactome, and latent Dirichlet allocation (LDA, MATLAB fitlda, 21 topics) for spatial neighborhood analysis on 4.22×10⁷ single cells.

Key findings

  • Early, persistent type-I/II IFN activation across the progression axis. GSEA shows IFN-α and IFN-γ Hallmark pathways enriched starting at p53.I vs FT.I and persisting through STIC.I, STIC.C, and cancer; Bayesian ordinal regression confirms significant upregulation of STAT1, IFITM1, IRF7, IRF9, ISG15, and TAP1 at every stage above matched normal epithelium PMID:39386723.
  • Stepwise rise in cells expressing IFN-pathway markers by CyCIF (p-TBK1, p-STATs, HLA-A, HLA-E): median 22% of epithelial cells in p53.I, 33% in STIC.I, 43% in STIC.C, 26% in cancer — independent of BRCA status. HLA-E+ epithelial cells alone rise 4% → 16% → 26% → 18% across the same stages.
  • Coordinated, localized IFN signaling. Pairwise CyCIF correlation analysis shows strong positive co-expression among p-STAT1, p-STAT3, p-TBK1, and HLA-E within individual cells (odds ratios 1.5–50) across all stages, suggesting clonal expansion under positive selection.
  • Late-stage IFN-ε downregulation. IFNE, IFNA2, and IFNA4 transcripts are significantly lower in STIC.C and cancer epithelium vs matched normal — a shift that is not seen in incidental precursors, implying loss of constitutive FT epithelial IFN-ε occurs after the initial IFN-α/γ response.
  • IRDS (IFN-related DNA damage resistance signature) emerges late. STAT1, MX1, and the anti-apoptotic MCL1 are upregulated in STIC.C and cancer (with a trend in STIC.I), consistent with chronic IFN signaling driving chemo/radio-resistance programs.
  • Micronuclear rupture and cGAS recruitment start as early as STIC.I. BANF1+ ruptured micronuclei (a sensitive marker of cytosolic DNA) increase in frequency from STIC.I through invasive cancer; a subset co-localize with CGAS and γ-H2Ax, validated by 3D CyCIF — implicating the cGAS–STING1 axis as the upstream trigger of the IFN response in HGSOC precursors.
  • HLA-E+ ROIs co-express a coordinated antigen-presentation/immune-modulation program. In STIC.I, HLA-E-positive epithelial regions are enriched for HLA-A, HLA-B, HLA-DRA, HLA-DQB1, HLA-F, TAP, TAPBP, complement (CFB, CFH, CFI, C1S, C2, C4B), MYC, and downstream IFN genes (MX1, IRF1, DDX60, OAS3, IFI44, PSMB8, XAF1).
  • Spatial topic modeling (LDA, 21 topics) identifies cancer-enriched neighborhoods: Topic 8 (CD163+ M2-like macrophages), Topic 10 (CD11c+ APCs), Topic 21 (HLA-DR+ activated APCs), Topic 14 (CD4+ T cells), and Topic 15 (CD8A+ T cells with CD103+ TRM and FOXP3+ Tregs). All five are significantly enriched in invasive cancer vs STIC.
  • NK–cDC1 axis collapses at STIC.I→STIC.C. HLA-DR+ cDC1 cells rise 12-fold in STIC.I stroma vs normal but drop 12-fold in STIC.C and 3-fold in cancer (vs STIC.I). NK cells fall from a median 0.1% in normal/p53.I to ~0.02% in STIC.I, STIC.C, and cancer. Spatial-transcriptomic gene sets corroborate: CLEC9A, BATF3, CLNK, XCL1, XCR1, IL-15, IL-12, KLRK1 (NKG2D), NCR1 (NKp46), CD226 (DNAM1) all decline; cytotoxic effectors PRF1, NKG7, GZMB, GZMK, GZMH, GZMA also decline in STIC.C and cancer.
  • Macrophages expand with progression. CD68+ M1-like and CD163+ M2-like macrophages reach peak abundance in invasive cancer; >50% of CD68+ cells co-express CD11c, classifying them as macrophage-derived APCs that frequently co-express HLA-DR and CD40. M1/M2 marker genes STAB1, AXL, IL10, CD163 increase in cancer-stage stroma.
  • CD4+ T cells infiltrate but show mixed activation/dysfunction. >35% of CD4+ T cells express HLA-DR or PD1 from STIC.I onward; ~14% are FoxP3+ Tregs and ~14% are LAG3+. Some HLA-DR+/CD40+ APCs co-express TIM3 (HAVCR2), implying a dampened antigen-presentation axis. Cell–cell interaction scores between activated CD4+ T cells, CD8+ T cells, and HLA-DR+ APCs decrease from STIC.I to STIC.C.
  • CD8+ T cell dynamics: surveillance then exhaustion. CD8+ T cells rise 2-fold in STIC.I epithelium and 10-fold in STIC.I stroma vs normal, then decline in STIC.C and cancer. Tissue-resident memory (CD8A+ CD103+ CD45RO+) cells rise 1.5-fold in p53.I and decline thereafter; conventional CTLs (CD8+ CD103− CD45RO−) peak in STIC.I and drop 4-fold in STIC.C, 2-fold in cancer (epithelium). Activation markers (Ki67+ or PDCD1+) rise from 11% in p53.I to 25–43% in STIC.I/STIC.C/cancer, while exhaustion (LAG3+ or PD1+LAG3+) increases 3- to 7-fold; CTLA4 and HAVCR2 RNA and GZMB protein/RNA also rise.
  • Stromal compartment mirrors epithelium early. Stromal ROIs in early stages are enriched for IFN response and IL6–JAK–STAT3 signaling, and MHC-class II (HLA-DRA, HLA-DMA, HLA-DRB1) increases in precursor and cancer stroma.
  • Late-stage tumor-promoting programs are HLA-E-independent. TGF-β, EMT (CLDN6, CDH3, COL4A1, MMP14, MMP2), hypoxia, hedgehog, angiogenesis, and TNF-α/NFκB hallmarks emerge primarily in STIC.C and cancer and are not restricted to HLA-E+ cells, indicating IFN-independent oncogenic mechanisms operating in parallel.

Genes & alterations

  • TP53 — defining mutation of HGSOC precursors; mutant p53 protein elevated by CyCIF in p53.I, STIC.I, STIC.C, and cancer vs normal FT epithelium. p53 signatures and STICs share clonal TP53 mutations with concurrent HGSOC (paper cites refs 11–13).
  • BRCA1, BRCA2 — cohort stratifier; germline status (gBRCA1, gBRCA2, somatic, WT) tracked across all groups. IFN/HLA-E phenotypes were independent of BRCA status.
  • CCNE1 — discussed (in introduction/citations) as a key oncogene amplified by chromosomal-instability/breakage-fusion-bridge cycles in HGSOC and STICs; not directly assayed for copy number in this paper.
  • HLA-A, HLA-E — both upregulated in IFN-active epithelium across all precursor stages. HLA-E overexpression is proposed as the key NK-evasion mechanism via NKG2A engagement.
  • STAT1, IFITM1, IRF7, IRF9, ISG15, TAP1, MX1 — IFN-stimulated genes upregulated from p53.I onward; IFITM1 induced as early as the Fim.I (incidental fimbriae) stage.
  • MCL1 — anti-apoptotic IRDS member, upregulated in STIC.C and cancer.
  • IFNE — Type-I IFN constitutively expressed by FT epithelium; significantly downregulated only in STIC.C and cancer.
  • CGAS, STING1, TBK1 — cGAS–STING–TBK1 axis activated by micronuclear rupture, evidenced by p-TBK1+ epithelium and cGAS recruitment to BAF+ ruptured MN.
  • BANF1 — used as a sensitive marker of cytosolic DNA / micronuclear rupture in CyCIF and 3D CyCIF.
  • MYC — upregulated in HLA-E+ STIC.I epithelium alongside the IFN/antigen-presentation program.
  • KLRK1 (NKG2D) and other NK markers (NCR1, CD226, KLRD1, KLRC2, KLRG1, KIRs) — broadly downregulated at the RNA level in STIC.C and cancer, consistent with NK-axis collapse.
  • CD4, CD8A, FOXP3, CXCL9 — immune-cell phenotype markers used to define infiltrating populations and APC chemoattraction.
  • PDCD1 (PD1), LAG3, HAVCR2 (TIM3), CTLA4 — exhaustion/checkpoint markers progressively upregulated on CD8+ T cells (and TIM3 on a subset of HLA-DR+/CD40+ APCs) in STIC.I, STIC.C, and cancer.

Clinical implications

  • Biomarkers for early detection / risk stratification. Variability in proliferation (Ki67) and DNA damage (γ-H2Ax) within STIC.I lesions, combined with IFN-pathway and HLA-E status, may help identify STIC subsets with elevated potential for malignant transformation and dissemination — relevant to the “precursor escape” model linking STICs to peritoneal carcinomatosis.
  • NK-axis reactivation as an interception strategy. The authors propose that anti-NKG2A antibodies — explicitly naming monalizumab — could re-arm NK cells against HLA-E-high STICs, particularly for high-risk patients with incidental STICs. This is a hypothesis based on phenotypic data; no treatment was administered in this study.
  • Public spatial atlas as a resource. CyCIF images and GeoMx data are released through cBioPortal as study ovary_geomx_gray_foundation_2024 (dataset page) for community-wide hypothesis generation on HGSOC interception.

Limitations & open questions

  • Sample size. 44 specimens span the full progression spectrum but the authors call out that the incidental-STIC cohort in particular is small; more STIC.I cases are needed to dissect the heterogeneity that may stratify malignant potential.
  • STIC.C contamination. ~25% of STIC lesions associated with cancer have been reported elsewhere to actually be disseminated cancer cells (precursor-escape model, citing ref 12), confounding “precursor” vs “metastasis” assignment in STIC.C ROIs.
  • Cross-sectional, not longitudinal. The progression axis is reconstructed from snapshots in different patients; clonal evolution is inferred, not directly tracked.
  • No direct functional validation. Causal links from cGAS-STING activation → IFN signaling → HLA-E induction → NK suppression are model-consistent but not perturbed experimentally in the manuscript.
  • Drug claims are hypothetical. Monalizumab is suggested as a candidate but not tested; HLA-E expression as a predictive biomarker for NKG2A blockade in HGSOC interception is an open question.
  • Bulk-of-ROI confounders. GeoMx ROIs aggregate multiple cells; some inferred epithelial signal may reflect adjacent stromal/immune contributions despite pan-cytokeratin guidance.
  • Assay generality. A few NanoString WTA probes are not receptor-specific (e.g. KLRC2/KLRC1 collapsed to “KLRC2”), which limits resolution of NK inhibitory-receptor biology.

Citations from this paper used in the wiki

  • “We analyzed 44 FT specimens with precursor lesions collected from 43 individuals obtained from a multi-center collaboration” — defines the cohort size used throughout.
  • “GSEA revealed prominent activation of the IFN pathway (both IFN-α and IFN-ɣ) and cell cycle regulator pathways in the epithelium during early HGSOC development” — basis for the early-IFN claim.
  • “BAF+ MN ruptures were observed as early as STIC.I lesions, with increasing frequency in invasive cancer … a subset of ruptured MN contained cGAS, and some also contained the DNA damage marker ɣ-H2Ax” — supports the cGAS-STING/MN claim.
  • “the number of HLA-DR+ cDC1 cells … were 10-fold higher in p53.I epithelium and 15-fold higher in STIC.I epithelium compared to normal epithelium” and “cDC1 populations expressing HLA-DR increased 12-fold in STIC.I compared to normal tissues but decreased substantially in more advanced lesions (12-fold decrease in STIC.C and 3-fold in established cancers compared to STIC.I)” — quantifies the cDC1 collapse.
  • “NK cells, which were present at low levels in normal tissue and p53.I precursors (median 0.1%), became nearly undetectable in later stages of the disease (STIC.I, STIC.C, and cancer; median: 0.02%)” — NK collapse numbers.
  • “the proportion of exhausted CD8+ T cells rising 3- to 7-fold (~2% were LAG3+ T_RM CD8 T cells in STIC.I, STIC.C, and cancer; 1% were LAG3+ CTL CD8+ T cells in STIC.C and cancer)” — exhaustion quantification.
  • “targeting NK cell reactivation using therapies like humanized anti-NKG2A antibodies (e.g., Monalizumab58) could be a promising strategy for early intervention” — basis for the monalizumab clinical-implication claim.
  • “Both CyCIF images and GeoMx data will be available through https://cbioportal.org/study/summary?id=ovary_geomx_gray_foundation_2024” — confirms the dataset release.

This page was processed by crosslinker on 2026-05-04.