A STAT1-GBP1 axis modulates epithelial proliferation in postpartum breast tissue by repressing CDKI expression

Abstract

Background

Postpartum breast cancer, diagnosed within five years of childbirth, is associated with heightened mortality compared to breast cancers in nulliparous women. Although the postpartum breast undergoes extensive involution and remodeling, the molecular drivers that promote subsequent tumor development remain incompletely understood. We investigated whether signal transducer and activator of transcription 1 (STAT1) and guanylate binding protein 1 (GBP1) contribute to epithelial proliferation through suppression of cyclin-dependent kinase inhibitors (CDKIs).

Methods

Formalin-fixed, paraffin-embedded postpartum (n = 5) and nulliparous (n = 5) benign breast tissues were profiled using transcriptomic panels targeting oncogenic and immunologic pathways. Protein expression of STAT1, GBP1, and the proliferation marker Ki67 was examined by immunohistochemistry. Functional studies were performed in human mammary epithelial cells (HMECs) derived from the same postpartum and nulliparous tissues. Small interfering RNA (siRNA) and lentiviral knockdown strategies were used to reduce STAT1 and GBP1, followed by assessment of CDKI expression, cell cycle distribution, and cell proliferation.

Results

Transcriptomic profiling revealed a postpartumspecific interferon signature (STAT1, GBP1), elevated Ki67, and reduced CDKIs; immunofluorescence across > 200 lobules confirmed these increases and suggested that GBP1 fully mediates the STAT1–Ki67 link. In HMECs, knockdown of STAT1 or GBP1 induced a marked rise in p21 and p57 (CDKN1A and CDKN1C), accompanied by G1 cell cycle arrest and reduced proliferation. Combined knockdown had an additive effect on suppressing epithelial proliferation, suggesting a cooperative role for STAT1 and GBP1 in modulating cell cycle progression.

Conclusions

These findings identify a STAT1–GBP1 axis that enhances postpartum epithelial proliferation by repressing CDKI expression. This mechanism may help to explain the heightened vulnerability observed after childbirth and highlights potential biomarkers or early intervention targets in postpartum breast tissues.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13058-025-02117-3.

Introduction

Postpartum-related breast cancer (PPBC), defined as breast cancer diagnosed within 5 to 10 years after childbirth [1, 2], represents a clinically distinct and challenging context in breast oncology. Compared to nulliparous women, those who have recently given birth experience a heightened risk of aggressive disease and worse clinical outcomes [3–5]. Notably, nearly 50% of breast cancers diagnosed in women under 45 years of age occur within 10 years of their last birth, and PPBC is associated with a 2-to-3-fold higher risk of mortality for both estrogen receptor-positive (ER+) and estrogen receptor-negative (ER−) subtypes [6, 7], which disproportionately affect young African American women, contributing to the racial disparities in patient survival outcomes [8–10]. Unraveling the biological processes that contribute to PPBC is therefore critical for understanding its differential impact and developing strategies to improve clinical outcomes across diverse populations.

The postpartum breast undergoes a complex remodeling process, including post-lactational involution, during which lactation-associated alveolar epithelium undergoes widespread cell death (primarily apoptosis), followed by regenerative phase that restores the non-lactating state. This transition involves extensive immune cell infiltration to clear dying cells, as well as dynamic extracellular matrix reorganization [11, 12]. While these physiologic events are well characterized, their molecular consequences for the epithelial compartment—especially those that might predispose cells to neoplastic transformation—remain incomplete and poorly understood. An inflammatory milieu, enriched with cytokines and growth factors, may inadvertently create fertile ground for tumor-promoting signals, thereby contributing to the increased risk in PPBC [13–16].

Cytokine-driven JAK/STAT signaling pathways, including those involving interferon-stimulated genes (ISGs), are central to immune responses but can also influence epithelial cell fate [17–20]. STAT1 and GBP1, in particular, have drawn attention for their dual roles: while classically associated with antiviral and antiproliferative functions, emerging evidence suggests that under certain contexts, these ISGs support tumorigenic processes by modulating cell cycle regulators [21–24]. Recent tumor profiling studies extend the relevance of STAT1 to the pregnancy setting. Peña Enríquez et al. analyzed 33 PABC tumors with the NanoString BC360 panel and reported STAT1 among the top immune proliferative genes enriched in postpartum cases [25]. Complementing these human data, Schreiber’s group showed that Stat1 null mice spontaneously develop ERα⁺ luminal mammary carcinomas, implicating STAT1 as a context dependent tumour suppressor [26]. Conversely, tumor intrinsic STAT1 activity can enhance growth and immune evasion via myeloid derived suppressor cells (MDSCs) [27]. These diverse observations raise the unresolved question of how STAT1 functions in normal postpartum epithelium, a gap our present study addresses by coupling tissue level profiling with functional interrogation of primary HMECs.

Mammary epithelial cell proliferation is tightly controlled by cyclin-dependent kinases (CDKs), cyclins, and cyclin-dependent kinase inhibitors (CDKIs), which together regulate cell cycle progression [28–30]. Progression from the G1 to S phase of the cell cycle is driven by cyclin-CDK complexes (e.g., CCND1/CDK4/6) that phosphorylate and inactivate retinoblastoma (RB), a key tumor suppressor. This process is normally inhibited by CDKIs such as CDKN1A (p21), CDKN1B (p27), and CDKN1C (p57), which induce G1 arrest and prevent excessive cell proliferation [31, 32]. Disruption of this balance, through upregulation of cyclins/CDKs or suppression of CDKIs, can drive abnormal cell cycle progression and increase the likelihood of tumorigenesis.

In postpartum tissues, the JAK/STAT pathway has been implicated in mammary gland remodeling, alveolar cell clearance, and immune cell infiltration during involution. Specifically, STAT1 is recognized as a key mediator of cytokine-driven signaling that regulates epithelial homeostasis, although its postpartum-specific role remains less fully explored [33, 34]. By contrast, GBP1 has not been extensively characterized in the postpartum context, despite emerging evidence for its involvement in cell cycle regulation and tumor progression [22, 23].

In this study, we harnessed multiple complementary approaches—transcriptomic profiling, multiplex immunofluorescence, and functional assays in patient-derived human normal breast tissue biopsies and human mammary epithelial cells (HMECs)—to identify molecular programs that may contribute to subsequent development of PPBC. Examining postpartum and nulliparous normal breast tissues from a carefully matched cohort, we uncovered elevated expression of STAT1 and GBP1, alongside increased proliferation markers and altered CDKI dynamics, in postpartum tissues. Using multiplex immunofluorescence and lobulelevel mediation analysis, we further tested whether GBP1 transmits STAT1 signalling to the proliferative marker Ki67 in vivo. Functional knockdowns of STAT1 or GBP1 in postpartum-derived HMECs induced G1 arrest and increased CDKI expression, providing direct mechanistic support for a STAT1-GBP1 axis that sustains epithelial proliferation.

By situating these findings within a human tissue-based model system, we provide a physiologically relevant framework for understanding how postpartum breast remodeling might interact with oncogenic pathways. These insights lay the groundwork for future investigations into the molecular interplay between immune signaling, epithelial plasticity, and health disparities in PPBC. Notably, our data raise the question of why certain postpartum remodeling pathways—normally transient—may remain activated or become dysregulated in some parous women, thereby contributing to malignant transformation. Clarifying these mechanisms could guide risk stratification, inform early intervention, and enable the development of targeted strategies to mitigate the excess risk carried by postpartum women.

Materials and methods

Study population

Women under 45 years of age undergoing mastectomies for invasive breast cancer or breast reduction procedures at Mayo Clinic, Jacksonville, Florida, between July 2019 and April 2024 were recruited for this study. Eligible participants included postpartum women (within five years of their last childbirth) and age- and BMI-matched nulliparous controls. Both European American and African American women were included in the study population. Women with prior breast cancer diagnoses or previous breast surgeries were excluded. This study was conducted under the approval of the Mayo Clinic Institutional Review Board (IRB #19-001672). All participants provided written informed consent prior to enrollment. Breast tissue samples were collected and processed for organoid preparation, formalin fixation and paraffin embedding (FFPE), and snap-frozen for RNA extraction. All tissue samples used for analysis were confirmed cancer-free by pathological diagnosis.

Transcriptomic profiling and gene set enrichment analysis

Total RNA was extracted from 10 μm FFPE breast tissue sections using the RNeasy FFPE kit (Qiagen, Cat# 73504) following the manufacturer’s protocol. RNA quality and concentration were measured using a NanoDrop 2000 spectrophotometer (ThermoFisher Scientific). Samples were profiled using the NanoString IO360 and BC360 panels, which target 1,500 distinct genes relevant to oncology and breast cancer biology. Hybridization reactions were processed using the nCounter Prep Station, and fluorescent counts were acquired with the nCounter Digital Analyzer. Data were normalized using nSolver 4.0 software with background correction and housekeeping gene normalization.

Gene set enrichment analysis (GSEA) was performed on the NanoString data to identify enriched biological pathways, following the method described by Subramanian et al. [35] Expression datasets were formatted along with phenotype files for postpartum versus nulliparous comparisons. GSEA software was used to assess enrichment using gene sets from the Molecular Signatures Database. Differential gene expression was identified by evaluating the distribution of gene sets across ranked lists.

Tissue processing for organoids and derivation of HMECs

Breast tissue samples were processed into organoids and human mammary epithelial cells (HMECs) were derived as described by LaBarge et al. 2013 [36]. Briefly, tissues were trimmed of adipose and stromal components, minced into 1–3 mm fragments, and digested overnight at 37 °C in a cocktail containing crude collagenase, hyaluronidase, DMEM/F-12, 10% FBS, penicillin/streptomycin, fungizone, and polymyxin B. Following digestion, the mixture was centrifuged, and the pellet was resuspended in media with antibiotics. Organoids were isolated using a 100 µM pore strainer, transferred into tubes with fibroblast media, and cryopreserved in CPMII medium (DME/F-12, 44% FCS, and 6% DMSO). A test aliquot was plated in M87A medium without cholera toxin to derive HMECs. HMECs emerged from organoids through differential trypsinization to remove fibroblasts. Cells were passaged to P4 before being used for experiments.

Knockdown of STAT1 and GBP1 in HMECs

siRNA Transfection. Knockdown of STAT1 and GBP1 in HMECs was achieved using small interfering RNA (siRNA) transfection. HMECs at 40% confluence in 6-well plates were transfected with 30 nM Silencer Select siRNA targeting STAT1 or GBP1 (ThermoFisher Scientific) using Lipofectamine 3000™ reagent according to the manufacturer’s protocol. After 24 h, the transfection medium was replaced with fresh M87A medium. Cells were harvested 48 h post-transfection for RNA extraction, RT-qPCR validation of knockdown efficiency, and assessment of CDKN1A and CDKN1C expression. Cell proliferation and cell cycle progression were analyzed using the Click-iT™ EdU assay and propidium iodide staining, respectively.

Lentiviral Transduction. For stable knockdown of STAT1 and GBP1, lentiviral constructs (Millipore Sigma) targeting STAT1 (KD1: NM_007315.2-1864s1c1; KD2: NM_007314.2-946s1c1) and GBP1 (KD1: NM_002053.1-1235s1c1; KD2: NM_002053.1-125s1c1) were used. Viral particles were produced in HEK293FT cells and collected from the conditioned medium. HMECs were transduced with lentivirus-containing medium supplemented with polybrene (3.5 µg/ml) for 24 h, followed by replacement with fresh medium. Knockdown efficiency was confirmed by RT-qPCR.

RNA extraction and gene expression analysis

Total RNA was extracted from HMECs using TRIzol reagent (Invitrogen) and quantified with a NanoDrop spectrophotometer. First-strand cDNA synthesis was performed using 1 µg of RNA and the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Quantitative real-time PCR (RT-qPCR) was conducted using the QuantStudio 7 Flex system with TaqMan assays specific for STAT1, GBP1, CDKN1A, and CDKN1C (Applied Biosystems). Relative gene expression was calculated using the 2^−ΔΔCT method, normalizing to GAPDH as the internal control.

Multiplex immunofluorescence on FFPE breast tissue

Formalin-fixed, paraffin-embedded (FFPE) blocks of benign breast parenchyma obtained from five postpartum and five agematched nulliparous women were sectioned at 5 μm onto MAS slides (Matsunami) to minimize tissue loss during repeated heating and stripping cycles. Prior to multiplexing, each primary antibody was validated on serial sections by conventional chromogenic immunohistochemistry and by singleplex immunofluorescence to confirm both staining pattern and optimal working dilution, first in reference tissues with known abundance of the target antigen and then in normal breast tissue. Optimized conditions were carried forward into a fivecolor Opal protocol (Akoya Biosciences) adapted from Parra et al., 2017 [37]. Briefly, slides underwent sequential rounds of antigen retrieval, primary antibody incubation, HRPconjugated secondary amplification and Opal fluorophore deposition, with microwave stripping (AR6 or 9 buffer, Akoya) between cycles. A test slide containing all five antibodies was used to finetune marker order, Opal–antibody pairing and to serve as a positive control in every staining batch. An autofluorescenceonly slide received the full HRP cycle without Opal fluorophore and was included in spectral unmixing to characterize and isolate tissue autofluorescence.

The final panel was applied in the following sequence: STAT1 (Cell Signaling #14994, 1:2.13, Opal-520), GBP1 (Abcam #ab131255, 1:2.6 ≈ 0.39, Opal-690), Ki67 (Dako M7240, 1:9 ≈ 0.11, Opal-480), and cytokeratin AE1/AE3 (Dako M3515, 1:6.8 ≈ 0.15, Opal-780); all antibody dilutions are given as reciprocal values to allow direct replication. DAPI counterstain was applied in the final step.

Slides were imaged on a PhenoImager HT 2.0 platform (Akoya Biosciences) using manufacturerrecommended exposure times, spectral libraries and autofocus settings. Multispectral images were unmixed with inForm 2.6; the autofluorescence library generated from the control slide was applied to all samples. Epithelial regions were delineated by cytokeratin signal, nuclei were segmented by DAPI and fine-tuned with assisting components, and markerpositive cells were identified with phenotype thresholds derived from singleplex controls. For every morphologically intact, noninvoluting lobule present on the section, the percentage of STAT1, GBP1 or Ki67positive cells in the epithelial segment was exported for downstream statistical analysis.

Cell proliferation and cell cycle assays

EdU Incorporation Assay. HMEC proliferation was assessed using the Click-iT™ EdU Alexa Fluor 488 kit (ThermoFisher Scientific). HMECs were incubated with 10 µM EdU for 12 h to label newly synthesized DNA. Cells were then fixed, permeabilized, and stained according to the manufacturer’s instructions. Data were collected on an Attune NxT flow cytometer and analyzed using FlowJo V10 software.

Propidium Iodide Staining for Cell Cycle Analysis. Cell cycle progression was evaluated by propidium iodide staining using the Abcam propidium iodide flow cytometry kit. HMECs fixed overnight at 4 °C in 70% ethanol were stained with a staining solution composed of propidium iodide and RNase A, and fluorescence was measured on an Attune NxT flow cytometer. Cell cycle distribution was analyzed using FlowJo V10 software.

Statistical analysis

Statistical analyses were performed using GraphPad Prism 9.5.1. Categorical variables were compared using Fisher’s exact test, while continuous variables were analyzed with two-tailed unpaired t-tests with Welch’s correction or the Mann–Whitney U-test, as appropriate. Heatmaps and gene expression comparisons were generated using GraphPad Prism. NanoString data normalization and statistical analyses were performed with nSolver 4.0 software, while GSEA was performed using GSEA V4.3.2 software. A p-value of less than 0.05 was considered statistically significant.

Lobule‑level analysis of MxIF data with patient clustering and mediation analysis were performed in R version 4.4.0 using the packages readxl (1.4.3), dplyr (1.1.4), lmtest (0.9–40), sandwich (3.1-0), broom (1.0.5) and boot (1.3–30). For the initial parity check each marker (GBP1, STAT1, Ki67) was collapsed to the patient-level median, and the two groups (five postpartum versus five nulliparous breasts) were compared with a two-tailed Mann–Whitney U test, an appropriate non-parametric procedure for very small samples, was calculated using GraphPad Prism. We then retained lobule-level data and modelled marker intensity as a function of parity by ordinary least-squares regression. Point estimates were obtained with lm, whereas inference relied on a cluster-robust Huber–White variance estimator with patient ID as the clustering variable (vcovCL in sandwich), so that standard errors, t statistics and two-sided P-values were robust to intra-patient correlation (coeftest in lmtest). For these models, 95% confidence intervals were calculated using a standard Wald formula (β ± 1.96 × SE).

To examine whether STAT1 acts through GBP1 to influence proliferation, we applied the threeequation mediation framework proposed by Baron and Kenny (1986) [38], quantified the indirect effect with the productofcoefficients approach formalised by MacKinnon and Dwyer (1993) [39]. We denoted the lobulelevel intensities of STAT1, GBP1 and Ki67 as X, M and Y, respectively. First, we regressed the mediator on the predictor, M_i = α + a X_i + ε_1i​, which yields path a, the effect of STAT1 on GBP1. Second, we regressed the outcome on both mediator and predictor, Y_i = β + b M_i + c′ X_i + ε_2i​ ; in this equation b estimates the effect of GBP1 on Ki67 after accounting for STAT1, while c′ represents the remaining direct effect of STAT1 on Ki67. Third, we obtained the total effect of STAT1 on Ki67 from Y_i = γ + c X_i + ε_3i ​. All three ordinaryleastsquares models were fitted to the complete set of 217 lobules, and inference relied on Huber–White standard errors clustered on patient identity (ten clusters) to accommodate withinpatient correlation. The mediated, or indirect, effect was calculated as the product a×b ; its statistical significance was tested with (i) a clusterrobust Sobel ztest and (ii) a patientlevel percentile bootstrap based on 1 000 resamples in which entire patients, rather than individual lobules, were sampled with replacement. We repeated the analysis for postpartum lobules alone (n = 163), for nulliparous lobules alone (n = 54), and, as a negativedirection control, after reversing the presumed order of mediator and predictor (GBP1 → STAT1 → Ki67). R code for these procedures is provided in Supplementary Methods.

Results

Transcriptional profiling reveals a pro-proliferative and immune-modulated gene signature in postpartum breast tissues

To identify molecular pathways that might be relevant to the elevated breast‑cancer risk observed postpartum, we performed transcriptomic analysis of breast tissues from well-characterized, premenopausal postpartum women (n = 5, within five years of childbirth) and age- and BMI-matched nulliparous women (n = 5). Utilizing NanoString IO360 and BC360 gene panels, which encompass 1,500 oncologically and immunologically relevant genes, we achieved a high-resolution transcriptional portrait of these matched cohorts. Demographic characteristics of the participants are provided in Table 1.

Table 1.

Characteristics of subjects included in the study by parity category

	Parous (N = 5)	Nulliparous (N = 5)	P Value
	n(%)	n(%)
Age (Years)
Mean	36	36.6	0.834
Range	32–41	32–42
BMI kg/m2
Mean	23.22	23.42	0.899
Range	19.4–25.6	20.5–26.7
Menopausal Status (%)
Premenopausal	5(100)	5(100)	> 0.9999
Postmenopausal	0(0)	0(0)
Involution Status
None	2(40)	0(0)
Partial	3(60)	3(60)	0.1353
Complete	0(0)	2(40)
Parity Status
Parous	5(100)	0(0)	0.0079
Nulliparous	0(0)	5(100)
Breastfeeding Status
Yes	5(100)	0(0)
No	0(0)	0(0)
Number of children
1	1(20)	0(0)
2	3(60)	0(0)
3	1(20)	0(0)
Age at first birth
< 30	3(60)	0(0)
> 30	2(40)	0(0)
Years since last birth
1	2(40)	0(0)
2	1(20)	0(0)
3	0(0)	0(0)
4	0(0)	0(0)
5	2(40)	0(0)
Surgical procedure
Reduction	1(20)	3(60)	0.5238
Mastectomy	4(80)	2(40)

We identified 185 genes that were differentially expressed between postpartum and nulliparous tissues (p ≤ 0.05), including 141 genes elevated in postpartum samples. Hierarchical clustering (Fig. 1A) revealed a clear segregation of postpartum and nulliparous profiles, underscoring distinct biological states. Notably, postpartum tissues showed increased expression of interferon-stimulated genes associated with JAK/STAT signaling (e.g., STAT1, GBP1), as well as markers linked to cellular proliferation (MKI67) and cell cycle regulation (CCNB1). These transcriptional alterations were accompanied by increases in immune-related genes (CD8A, CD8B, CCL5, KLRK1), suggesting an immunomodulatory environment that may influence epithelial cell behavior. Additionally, we observed upregulation of PROM1, a stem cell-associated marker, consistent with a more dynamic epithelial state. Quantitative comparisons confirmed significantly higher levels of STAT1, GBP1, MKI67, and CCNB1, and a corresponding reduction in CDKN2B expression (Fig. 1B). Exploratory, hypothesis-generating analyses of clinical covariates (Supplementary Table S1 showed a positive trend between GBP1 expression and the number of prior births (Spearman ρ = 0.89, p = 0.041) and MKI67 with BMI (ρ = 0.9, p = 0.038). Given the sample size, we interpret this as a suggestive trend rather than a definitive association. Median STAT1 and GBP1 levels were modestly higher in Black women, but differences were not significant (p > 0.26) and are not presented. No other clinical variable showed a nominal association with STAT1, GBP1, or MKI67. Taken together, these findings indicate that postpartum breast tissues harbor a transcriptional program favoring proliferation and altered immune interactions, providing an early mechanistic clue that could be relevant to PPBC susceptibility.

Fig. 1.

Transcriptional profiling reveals a pro-proliferative gene signature in postpartum breast tissues. (A) Heatmap of differentially expressed genes in breast tissues from postpartum (n = 5) and nulliparous (n = 5) women. RNA was extracted from FFPE tissue sections and analyzed using NanoString IO360 and BC360 panels. Data were normalized to housekeeping genes, and differentially expressed genes (n = 185, p ≤ 0.05) were clustered by expression pattern. Genes in key pathways, including JAK/STAT signaling (STAT1, GBP1), proliferation (MKI67, CCNB1), and cell cycle regulation (CDKN2B), are displayed. (B) Quantitative comparison of STAT1, GBP1, MKI67 (Ki67), CCNB1, and CDKN2B mRNA levels between nulliparous and postpartum samples. Data are presented as mean ± SEM. Statistical significance was determined using unpaired t-tests; *p < 0.05

Gene set enrichment analysis highlights proliferation and cell cycle pathway activation in postpartum tissues

To gain deeper insight into the functional significance of the differentially expressed genes, we performed gene set enrichment analysis (GSEA) using curated pathways from the Molecular Signatures Database (MSigDB). This approach revealed that postpartum tissue show robust enrichment of signatures related to proliferation, cell cycle progression, and breast cancer biology. The invasive breast cancer pathway was prominently enriched (normalized enrichment score [NES] = 2.421, p < 0.0001), driven by upregulation of GBP1 and key cell cycle regulators such as CCNA2, CCNE2, and CDK1 (Fig. 2A).

Fig. 2.

Gene set enrichment analysis (GSEA) identifies activation of proliferation and cell cycle pathways in postpartum breast tissues. (A) Enrichment plot for the invasive breast cancer pathway, showing significant enrichment in postpartum tissues (NES = 2.421, p < 0.0001). Key upregulated contributors include GBP1, CCNA2, CCNE2, and CDK1. (B) Enrichment plot for the cell proliferation pathway, demonstrating positive enrichment of cyclins and CDKs (CCNA1, CCNB1, CCND1, CDK1, CDK4) and negative enrichment of CDKIs (CDKN1A, CDKN1C). (C) Enrichment plot for the G2/M checkpoint pathway, highlighting activation of cyclins (CCNA2, CCND1) and repression of CDKIs (CDKN1B, CDKN2C) in postpartum tissues (NES = 2.419, p < 0.0001)

Similarly, a generalized cell proliferation pathway demonstrated strong enrichment of cyclins (CCNA1, CCNB1, CCND1, CCNE2) and CDKs (CDK1, CDK4), coupled with the relative depletion of cyclin-dependent kinase inhibitors (CDKN1A, CDKN1C, CDKN2B) (NES = 1.884, p < 0.0001) (Fig. 2B). The G2/M checkpoint pathway also showed heightened activity, with increased cyclin expression (CCNA2, CCND1) and diminished CDK inhibitors (CDKN1B, CDKN2C) (NES = 2.419, p < 0.0001) (Fig. 2C). These pathway-level analyses reinforce the conclusion that postpartum breast tissues adopt a regulatory milieu that favors cell cycle progression and may prime the epithelium for growth signals characteristic of neoplastic transformation.

Postpartum enrichment of STAT1/GBP1/Ki67 and GBP1-mediated link between STAT1 and proliferation

To confirm that the transcriptional changes observed were reflected at the protein level, we stained all ten breast specimens with a fivecolor multiplex immunofluorescence panel that detected STAT1, GBP1, Ki67, pancytokeratin and DAPI in a single optical cycle, thereby eliminating sectiontosection variability and allowing the three functional markers to be quantified within exactly the same histologic context. Every morphologically intact, noninvoluting lobule—163 in five postpartum breasts and 54 in five nulliparous breasts—was annotated (Fig. 3A) and compared between postpartum and nulliparous patients (Fig. 3B-G). The fraction of markerpositive epithelial cells was recorded for each lobule. Reducing these measurements to patient medians revealed that GBP1 and Ki67 remained significantly higher in postpartum tissue (p = 0.0079; p = 0.036), whereas STAT1 showed a positive trend (p = 0.094) (Fig. 3H-J). Ki67 transcript and protein correlated strongly across patients (Spearman ρ = 0.92, p = 0.0004), while GBP1 and STAT1 displayed modest but positive correlations (ρ = 0.61 and 0.32, respectively; Supplementary Fig. S1

Fig. 3.

Postpartum enrichment of STAT1/GBP1/Ki67 and GBP1-mediated link between STAT1 and proliferation. (A) Representative whole-slide view of a postpartum breast section stained in a single five-channel cycle (STAT1 = green, GBP1 = yellow, Ki67 = cyan, pan-cytokeratin = red, DAPI = blue). Outlines denote the manually annotated, non-involuting lobules that were analyzed. (B–G) Representative markers staining for STAT1 (B, E), GBP1 (C, F), and Ki67 (D, G), in postpartum (B-D) and nulliparous (E-G) lobular tissue. (H–J). Fraction of lobules positive for GBP1 (H), STAT1 (I) and Ki67 (J) after collapsing each sample to its lobule-level median (n = 5 nulliparous, blue; n = 5 postpartum, red). Brackets show two-tailed Mann–Whitney U test results (** p < 0.01; ns = not significant). (K) Cluster-robust ordinary least-squares estimates of the parity effect (β ± 95% bootstrapped confidence interval) for the same markers, retaining every lobule (n = 217) but computing Huber–White standard errors clustered on patient identity (N = 10). Positive β values indicate higher marker prevalence in postpartum tissue. All three markers remain significant after clustering (STAT1, p = 0.0016; GBP1, p < 0.0001; Ki67, p = 0.0211). (L) Conceptual path diagram depicting path a: STAT1 → GBP1, path b: GBP1 → Ki67 (controlling STAT1), and direct path c′: STAT1 → Ki67 (controlling GBP1). (M-O) Forest-style plots show cluster-robust coefficients (β ± 95% CI) for path a (M), path b (N), and direct path c’ (O). Red symbols = post-partum lobules (n = 163, 5 patients); blue symbols = nulliparous lobules (n = 54, 5 patients). The corresponding indirect effects (a × b) and full numerical output are in Supplementary Table S3

Comparisons of medians satisfies independence assumptions, but doesn’t account for intra-patient variability in per-lobule expression patterns, which are critical because lobules constitute the anatomic units where postpartum proliferation and interferon signaling coincide. Accordingly, we next analyzed the full lobule dataset but safeguarded against pseudoreplication by fitting a simple ordinary leastsquares model with a parity indicator while computing Huber–White standard errors clustered on patient identity. This clusteraware approach preserved every lobule yet treated the cohort as ten independent subjects. The resulting coefficients represented the mean difference in the marker expression between postpartum and nulliparous lobules, with robust inference that accounts for intra-patient variability. The estimated parity effects (STAT1 β = +0.0138, p = 0.0126; GBP1 β = +0.0792, p = < 0.001; Ki67 β = +0.0221, p = 0.0211) showed significance (Fig. 3K Supplementary Table S2), demonstrating that the elevated postpartum signal is consistent across both patient-level and lobule-level analyses.

The large number of anatomically discrete lobules in our multiplex data set, more than two hundred independent analytical units, allowed us to assess whether the correlations suggest whether the proliferative signal associated with STAT1 reaches Ki67 directly or must first pass through GBP1 using the clustered dataset and a Baron-Kenny mediation framework (Fig. 3L). In postpartum tissue STAT1 strongly predicted GBP1 (Path a, β = 2.327 (95% CI 1.812–2.810); Fig. 3M) and GBP1, in turn, independently predicted Ki67 while controlling for STAT1 (Path b, β = 0.154 (95% CI 0.094–0.225); Fig. 3N). The direct path from STAT1 to Ki67 (controlling for GBP1) became negative (Path c’, β=–0.431 (95% CI–0.686 – − 0.293); Fig. 3O), consistent with full mediation. In contrast, none of the paths reached significance in nulliparous lobules (Figs. 3M-O supplementary Table S3). Taken together, these findings support a model in which postpartum inflammatory signaling leads to elevated STAT1, STAT1 then drives GBP1 expression, and GBP1 serves as the immediate catalyst that propels epithelial cells into proliferation, a regulatory cascade that operates only in the postpartum breast and is entirely quiescent in nulliparous tissue.

Silencing STAT1 or GBP1 in HMECs increases CDKI expression and alters the cell cycle profile

To directly test the functional role of STAT1 and GBP1 in regulating epithelial cell proliferation, we employed primary human mammary epithelial cells (HMECs) derived from postpartum breast tissues. Targeted knockdown of STAT1 or GBP1 using siRNA or lentiviral constructs led to substantial increases in key CDKIs: CDKN1A (p21) expression rose by 9- to 10-fold, and CDKN1C (p57) by 5- to 6-fold (Fig. 4A). These changes were accompanied by shifts in global gene expression patterns, confirmed by GSEA, showing reduced enrichment of cell cycle progression pathways and an increased emphasis on CDKI-related pathways (Figs. 4B–E). Principal component analysis (Fig. 4D) highlighted the distinct transcriptional states induced by knockdown of STAT1 or GBP1, indicating a profound molecular reprogramming of the postpartum-derived HMECs.

Fig. 4.

STAT1 and GBP1 regulate cyclin-dependent kinase inhibitors (CDKIs) in human mammary epithelial cells (HMECs). (A) Relative expression of CDKN1A (p21) and CDKN1C (p57) following knockdown of STAT1 or GBP1 in HMECs. Data are presented as mean ± SEM relative to control cells. (B-C) GSEA enrichment plots for the cell cycle progression pathway, comparing control HMECs to those with knockdown of GBP1 (B) or STAT1 (C). Knockdown of either gene results in negative enrichment of genes involved in cell cycle progression and positive enrichment of CDKI-related pathways. (D) Principal component analysis (PCA) of RNA-seq data from control, STAT1 knockdown, and GBP1 knockdown HMECs. Inset: (E) Enrichment plot for the cell cycle progression pathway following combined knockdown of STAT1 and GBP1, demonstrating a synergistic effect on CDKI expression

Knockdown of STAT1 and GBP1 induces G1 cell cycle arrest and inhibits proliferation

To assess the impact of STAT1 and GBP1 on cell cycle progression, we performed propidium iodide (PI) staining and EdU incorporation assays in HMECs following knockdown of STAT1, GBP1, or both genes. PI staining revealed significant G1 arrest, as the percentage of cells in G1 increased from 37% in controls to 74% in knockdown cells (Figs. 5A–C). Correspondingly, cells in the S and G2/M phases were reduced in knockdown conditions. EdU incorporation assays further confirmed the loss of proliferation. In control HMECs, 83.6% of cells were EdU-positive, indicating active DNA synthesis. However, knockdown of GBP1 reduced EdU-positive cells to 30.5% (Fig. 6). Knockdown of STAT1 produced a similar effect, and combined knockdown of both STAT1 and GBP1 produced a synergistic effect, leading to near-complete loss of proliferative activity (Fig. 7).

Fig. 5.

Knockdown of STAT1 and GBP1 induces G1 cell cycle arrest in HMECs. (A) Propidium iodide staining reveals a significant increase in the proportion of cells in G1 phase (blue) and a corresponding decrease in S (brown) and G2/M (green) phases following knockdown of GBP1. (B) Knockdown of STAT1 produces a similar shift in the distribution of cell cycle phases, with increased G1 arrest. (C) Combined knockdown of STAT1 and GBP1 has an additive effect on G1 arrest, reducing S-phase entry to minimal levels. Knockdown was performed using siRNA, and flow cytometry was used to measure cell cycle phase distribution

Fig. 6.

Knockdown of GBP1 suppresses cell proliferation in HMECs. (A) Representative flow cytometry plots showing EdU incorporation in control HMECs and GBP1 knockdown HMECs. Cells were incubated with Click-iT™ Plus EdU reagent for 12 h to label newly synthesized DNA. (B) Quantification of proliferating (EdU-positive) and non-proliferating cells. Knockdown of GBP1 reduced the proportion of proliferating cells from 83.6–30.5% while increasing non-proliferating cells from 15–68%. Data are presented as mean ± SEM from three biological replicates, and statistical significance was determined using unpaired t-tests: *p < 0.05

Fig. 7.

Combined knockdown of STAT1 and GBP1 reduces proliferation in HMECs. (A) Flow cytometry analysis of EdU incorporation in control HMECs. (B) Knockdown of GBP1 significantly reduces the proportion of EdU-positive cells. (C) Knockdown of STAT1 similarly decreases the proportion of proliferating cells. (D) Combined knockdown of STAT1 and GBP1 results in a further reduction in proliferation, with a synergistic effect on the inhibition of DNA synthesis. Data are presented as mean ± SEM from three biological replicates, and statistical significance was determined using unpaired t-tests: *p < 0.05

These results establish a mechanistic link between STAT1/GBP1 activity and the proliferative state of postpartum mammary epithelial cells. By suppressing CDKI expression, STAT1 and GBP1 appear to facilitate G1-S transition and sustained proliferation. Their cooperative action suggests a key regulatory axis that may be foundational to the pro-growth signaling milieu observed in postpartum breast tissues. Together, these data provide a physiologically relevant, human tissue-based framework for understanding how postpartum involution-associated signaling pathways modulate the epithelial cell cycle and potentially influence subsequent breast cancer risk.

Discussion

This study provides evidence that STAT1 and GBP1 cooperatively modulate mammary epithelial cell proliferation during the postpartum period by suppressing CDKIs and promoting cell cycle progression. Through an integrated approach—combining transcriptomic profiling of carefully matched postpartum and nulliparous human breast tissues, immunohistochemical validation, and functional manipulations in primary HMECs—we delineate a postpartum-specific network that may contribute to the heightened risk of PPBC.

Our findings place ISGs, including STAT1 and GBP1, at the center of a regulatory hub that links the postpartum inflammatory milieu to the control of epithelial proliferation. By comparing postpartum and nulliparous breast tissues, we identified a distinct transcriptional signature characterized by elevated STAT1, GBP1, and multiple cyclins and CDKs, coupled with reduced expression of key CDKIs (e.g., CDKN1A/p21, CDKN1C/p57). Gene set enrichment analysis reinforced this connection, revealing significant enrichment of proliferation- and cell cycle–related pathways, as well as breast cancer–associated gene sets, in postpartum tissues [1, 40–42]. Parallel immunohistochemistry demonstrated that epithelial cells in postpartum lobules express higher levels of Ki67, STAT1, and GBP1, validating the transcriptomic signatures at the protein level. Notably, we found these upregulated genes in all of the postpartum samples, suggesting that the effect is not driven by outliers.

Our mediation analysis revealed a molecular chain that is active only in the postpartum breast: STAT1, a canonical interferonresponse transcription factor, transcriptionally upregulates GBP1, and GBP1 becomes the proximate driver of epithelial proliferation. The simplest view of the dataset, one median value per patient, already showed that STAT1, GBP1 and the proliferation marker Ki67 are all higher in the postpartum breast; the same pattern persisted when every lobule was analyzed with patientclustered inference. The mediation analysis clarified how these markers are linked. In postpartum lobules, nearly the whole association between STAT1 and Ki67 flowed through GBP1, whereas the direct effect of STAT1 on Ki67, after adjusting for GBP1, not only shrank but reversed sign, indicating that once GBP1 is considered, higher STAT1 is associated with lower Ki67. Such a sign change, termed inconsistent mediation, suggests that the pathway through GBP1 captures virtually all of the proliferative influence of STAT1. In nulliparous lobules neither the component paths nor the indirect effect was detectable, confirming that the cascade is dormant in resting breast tissue. This model is biologically plausible. STAT1 binding sites flank the human GBP1 promoter [43], and GBP1 has been linked to cellcycle progression and poor outcome in several cancer types [27, 44, 45]. Our findings extend this STAT1–GBP1–proliferation axis to the human postpartum breast and suggest that prolonged activation of the pathway could help explain the elevated risk for development of postpartum breast cancers.

In combination with our HMEC results, our findings refine the temporal and mechanistic landscape of STAT1 biology. Whereas the GEICAM EMBARCAM study linked STAT1 over expression to established pregnancy associated tumors [25], we demonstrate that STAT1, together with GBP1, is already elevated in histologically normal postpartum lobules and directly represses CDKN1A/CDKN1C. This repression distinguishes our model from STAT1 knock out mice, where loss of STAT1 accelerates tumor initiation [26], and from tumor-centric work showing STAT1 driven immune suppression [27]. By situating STAT1 activity prior to malignant conversion and identifying GBP1 as a co-regulator, our study reveals a postpartum specific STAT1–GBP1–CDKI axis that may prime the epithelium for subsequent oncogenic events.

The apparently favorable association of GBP1 and STAT1 with recurrencefree survival in bulktumor datasets [46] must be considered alongside our finding that the same genes, when activated within postpartum epithelium, repress CDKIs and sustain proliferation. In the Ascierto signature, STAT1/GBP1 expression primarily marks infiltrating immune cells, consistent with studies linking interferonresponsive immune transcripts to superior outcome [47]. By contrast, our multiplex segmentation and cell culture studies investigate STAT1/GBP1 specifically to cytokeratinpositive epithelial cells, and functional knockdown experiments demonstrate a cellintrinsic role in overriding G1 arrest. These observations highlight a central principle: the prognostic impact of interferonstimulated genes is highly celltype and stage specific. Transient postpartum epithelial activation may prime the tissue for transformation, whereas robust immunecellderived expression in an established tumor can facilitate immunosurveillance and limit recurrence.

We found that silencing STAT1 and GBP1 in postpartum-derived HMECs curtailed proliferative pathways (Figs. 4, 5, 6 and 7), whereas efforts to force expression of either gene in HMECs from nulliparous donors inhibited proliferation (data not shown). This dichotomy suggests that pregnancy may do more than raise transcriptionfactor abundance, but instead license a STAT1–GBP1 proliferative circuit by reshaping the enhancer landscape rather than merely increasing transcription-factor abundance. STAT1 was originally defined as an interferon-activated tumor suppressor, yet many advanced cancers show elevated STAT1 activity correlates with tumor progression [27]. Whether STAT1 arrests or drives cell growth appears to hinge on chromatin context: ectopic STAT1 halts hepatocellularcarcinoma cells [48], whereas in embryonic kidney progenitors IFNγ–activated STAT1 stimulates proliferation but blocks differentiation, and STAT1 inhibition releases differentiation while abolishing the proliferative signal [49]. Genome‑wide studies demonstrate that cytokine‑activated STATs create or reactivate lineage‑specific enhancers that cannot be reinstated by master regulators once STAT binding is lost [50]. Such enhancers can arise de novo as latent enhancers after a first stimulus and persist as an epigenetic memory that accelerates re-induction [51]. For IFN-γ, STAT1 is essential to establish, but not to maintain, long-term chromatin priming at clustered GBP genes in cervical epithelial (HeLa) cells, enabling a heightened secondary response [43]. Context matters for GBP1 as well: it suppresses tumour growth in colorectal epithelium but promotes aggressiveness in glioblastoma, underscoring the importance of epigenomic state in determining functional output [52, 53]. Our data extend this paradigm to postpartum mammary epithelium: parity-associated cytokine cues, which include IFN-γ [14], may configure enhancers that recruit STAT1/GBP1 to pro-proliferative targets; silencing either gene disrupts those interactions and limits growth, whereas overexpression in unprimed nulliparous cells cannot recreate the requisite enhancer topology and instead triggers the canonical antiproliferative interferon response.

Critically, functional studies in postpartum-derived HMECs provided direct mechanistic support for the role of STAT1 and GBP1 in regulating CDKI levels. Knocking down either STAT1 or GBP1 not only increased CDKN1A and CDKN1C expression but also induced G1 cell cycle arrest and suppressed proliferation. These perturbations were additive rather than purely overlapping, underscoring a cooperative relationship. These observations suggest that, in the postpartum setting, STAT1 and GBP1 can override normal proliferative checkpoints. Whether such changes facilitate early oncogenic events and influence PPBC risk remains to be determined in longitudinal studies.

These findings align with a growing body of literature indicating that the postpartum breast, undergoing involution and substantial remodeling, is a dynamic environment in which immune-derived signals can influence epithelial cell fate. The infiltration of immune cells, release of cytokines, and shifts in the extracellular matrix are well-documented hallmarks of involution [17–19]. Although we did not perform a comparison of all inflammatory cytokines between our postpartum and nulliparous cohorts, our data do show consistent upregulation of certain interferon-stimulated genes (e.g., STAT1, GBP1) in postpartum tissues. This mechanistic link adds a critical piece to the puzzle of how the postpartum microenvironment might influence epithelial behavior with possible implications for carcinogenesis. Moreover, since ISG overexpression is associated with poor outcomes in multiple cancers [22, 23, 54–62], the STAT1-GBP1 axis could be explored as a marker of PPBC risk or as a potential target for early intervention strategies aimed at mitigating postpartum-associated vulnerability. It is important to note that while GBP1 has been associated with poor prognosis in some contexts, other studies (including a recent comprehensive review in breast cancer subtypes [63] report mixed or even favorable prognostic correlations depending on tumor subtype and cytokine milieu.

Our functional studies implicate STAT1 and GBP1 as key upstream regulators of epithelial cell proliferation. Knockdown of either STAT1 or GBP1 in HMECs led to robust upregulation of CDKN1A (p21) and CDKN1C (p57), resulting in cell cycle arrest at G1 and decreased proliferation. These data provide direct experimental evidence supporting the regulatory role of STAT1 and GBP1 in controlling epithelial cell proliferation. Previous studies have identified STAT1 as a critical mediator of interferon responses and an enhancer of p53-dependent growth arrest [37–43]. However, our findings suggest that, in the postpartum context, STAT1 and GBP1 play cooperative role in suppressing CDKI expression. This context-dependent activity may reflect differences in post-transcriptional regulation, JAK/STAT pathway dynamics, or epigenetic modifications during postpartum involution.

While GBP1 is often discussed in the context of viral defense, it is also implicated in immunity to various intracellular pathogens (including Chlamydia) [64], underscoring its broader immunomodulatory function. The identification of GBP1 as a functional regulator of cell proliferation represents a significant advancement. While GBP1 has been implicated in immune response and viral defense, its role in modulating the epithelial cell cycle has been largely unexplored. Our findings suggest that GBP1 may act as a molecular switch, simultaneously driving cell cycle progression (via cyclin upregulation) and repressing CDKIs (via STAT1-mediated repression). Given the well-documented role of GBP1 in ovarian and prostate cancer progression [22, 23] its involvement in PPBC introduces an opportunity to explore it as either a therapeutic target or a potential target for early interception.

Our approach, leveraging patient-derived tissues and primary HMECs, offers a physiologically relevant perspective. The careful matching of postpartum and nulliparous cohorts and use of NanoString panels encompassing 1,500 cancer-related genes yielded a robust dataset that facilitated the identification of 185 significantly altered genes. While we acknowledge that the sample size (n = 5 postpartum, n = 5 nulliparous) is not large, it is substantial for an in-depth mechanistic analysis with primary human tissues. Critically, we observed consistent transcriptional shifts and pathway enrichment across the postpartum samples, and these data were reinforced by protein-level findings and functional assays in HMECs, strengthening confidence that these findings are biologically meaningful. Additional studies with larger cohorts and broader cytokine profiling will help clarify how widespread these changes are and whether they correlate with clinical risk factors for PPBC.

Future efforts can build on this foundation. Larger cohorts and longitudinal sampling may help determine whether STAT1/GBP1-driven changes precede malignant transformation or reflect an intermediate stage in cancer evolution. Similarly, dissecting the interplay between STAT1/GBP1 and other postpartum involution signals, such as specific cytokines or extracellular matrix components, will provide a more comprehensive understanding of the postpartum niche. Complementary approaches, including spatial transcriptomics and high-dimensional imaging, could elucidate the cellular architecture of postpartum breast tissue and define how immune cells, fibroblasts, and epithelial clusters interact to sustain JAK/STAT activation.

In addition, exploring targeted modulation of STAT1 or GBP1 in experimental models may illuminate their therapeutic potential. If inhibition of these factors can restore CDKI levels and normalize epithelial proliferation, then targeting the STAT1‑GBP1 axis could be explored as a strategy to mitigate postpartum‑associated vulnerability. Such strategies could be especially relevant for individuals at elevated risk due to parity-related factors or those who face health disparities, as postpartum breast cancer risk is known to vary by race and ethnicity [8–10].

In conclusion, this study characterizes a STAT1-GBP1 axis that directs epithelial proliferation by regulating CDKI expression in postpartum breast tissues. The alignment of transcriptomic, proteomic, and functional data provides a cohesive mechanistic model that advances our understanding of molecular changes that accompany the postpartum state. While further research is warranted to fully chart the postpartum regulatory landscape and translate these insights to clinical interventions, our work lays groundwork for developing strategies to identify and monitor postpartum molecular signals that could inform breast‑cancer risk assessment.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

J.W.O. performed experiments, wrote and edited the manuscript. J.V.C., S.A.T., and S.A. performed experiments. L.M.P.-S. and A.A. contributed to clinical trial documentation, patient consent, and sample acquisition. S.M., A.C.D., M.E.S., and D.C.R. provided guidance on study design and execution. All authors read and approved the final manuscript.

Funding

This study was supported by the National Cancer Institute through grants R01 CA229811, R01 CA237602, R01 CA262393, and an administrative supplement to R01 CA237602. The funding sources had no role in the design of the study; the collection, analysis, and interpretation of data; or in writing the manuscript.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

This study was conducted under the approval of the Mayo Clinic Institutional Review Board (IRB #19-001672). All participants provided written informed consent prior to enrollment. Breast tissue samples were obtained from women undergoing mastectomy or reduction procedures, and all tissues used in this study were confirmed to be free of malignancy by pathological evaluation.

Consent for publication

This manuscript does not contain any individual person’s data (including images or videos) that would require explicit consent.

Competing interests

The authors declare no competing interests.