HE4 drives PD-L1 expression in myeloid cells via IFN-γR-JAK-STAT3 signaling to promote tumor immune evasion

Summary

Immune checkpoint inhibitors (ICIs) targeting PD-1/PD-L1 have achieved clinical success, yet most patients fail to respond and many develop immune-related adverse events (irAEs). Although interferon gamma (IFN-γ) is considered the canonical driver of PD-L1 expression, regulation of PD-L1 in myeloid cells within the tumor microenvironment (TME) remains poorly defined. Here, we identify human epididymis protein 4 (HE4), a tumor-secreted glycoprotein overexpressed in multiple cancers, as an unrecognized inducer of myeloid PD-L1 transcription. HE4 directly binds IFN-γ receptors, activates JAK-STAT3 signaling, and upregulates PD-L1. Neutralization of mouse or human HE4 with monoclonal antibodies reduced myeloid PD-L1 expression, restored CD8⁺ T cell activity, and suppressed tumor growth in syngeneic and humanized models, while inducing fewer irAEs than PD-1 blockade. Clinically, high HE4 expression predicts poor prognosis but correlates with improved response to PD-1 inhibitors in lung adenocarcinoma, highlighting HE4 as both a therapeutic target and predictive biomarker.

Graphical abstract

Highlights

• •Tumor-derived HE4 suppresses CD8⁺ T cell cytotoxicity via myeloid PD-L1 induction
• •HE4 binds IFN-γ receptors to activate JAK-STAT3 and drive PD-L1 transcription
• •HE4 neutralization restores antitumor immunity with fewer irAEs than PD-1 blockade
• •HE4 predicts response to PD-1/PD-L1 therapy and serves as a therapeutic target

Zeng et al. report HE4, a tumor-secreted glycoprotein overexpressed in multiple cancers, induces PD-L1 transcription in myeloid cells by directly engaging IFN-γR and activating JAK-STAT3 signaling. Targeting HE4 reduces myeloid PD-L1 expression, restores CD8⁺ T cell function, and suppresses tumor growth with fewer immune-related adverse events.

Introduction

The tumor microenvironment (TME), composed of malignant, immune, and stromal cells and the extracellular matrix, critically determines tumor progression and therapeutic response.¹ Immune cells, including cytotoxic T lymphocytes and natural killer cells, restrain tumor growth by eliminating malignant cells, whereas tumors evade immune surveillance by engaging immune checkpoints, most notably the programmed cell death protein 1 (PD-1)/programmed death-ligand 1 (PD-L1) axis.^2,3 Immune checkpoint inhibitors (ICIs) targeting PD-1/PD-L1 have achieved clinical success and are now frontline therapies for multiple malignancies.^4,5 However, their efficacy is limited by low response rates, tumor-type-specific resistance, and immune-related adverse events.^6,7,8,9,10 Although PD-L1 expression is a key determinant of therapeutic response, the mechanisms regulating PD-L1 expression within the TME remain incompletely understood.

The expression of PD-L1 is regulated at transcriptional and post-translational levels in the TME.^5,11,12,13 Transcriptionally, PD-L1 is induced by cytokines such as IFN-γ via JAK-STAT3 and PI3K-AKT signaling in cancer and myeloid cells^11,14,15 and by type I IFN in myeloid-derived suppressor cells.^16,17 Hypoxia also enhances PD-L1 transcription through HIF-1α stabilization and promoter binding.¹⁸ Post-translationally, PD-L1 stability is regulated by N-linked glycosylation, which limits ubiquitin-proteasome degradation,^19,20,21 and by CMTM6-mediated protection from lysosomal degradation,^22,23 with additional modifications such as phosphorylation further modulating PD-L1 function.^24,25 PD-L1 is broadly expressed in the TME, particularly on myeloid antigen-presenting cells, where it dominantly suppresses antitumor T cell immunity; accordingly, host/myeloid and DC-intrinsic PD-L1 critically determine responses to PD-(L)1 blockade.^26,27 However, the upstream extracellular ligands, receptors, and cell-type-specific signaling pathways driving PD-L1 induction in non-tumor compartments remain poorly defined, limiting mechanism-based strategies beyond direct PD-(L)1 targeting.^26,27,28,29 Because PD-1/PD-L1 signaling also maintains peripheral immune tolerance, defining cancer-specific regulators of PD-L1 may enable safer immunotherapeutic interventions with reduced immune-related adverse events (irAEs).¹⁰

HE4 (human epididymis protein 4) is a secreted glycoprotein originally identified in the human epididymis³⁰ and is thought to contribute to tissue homeostasis by regulating protease activity and cell adhesion.^31,32 In pathological settings, HE4 expression is dysregulated, particularly in fibrotic and inflammatory conditions, implicating it in aberrant tissue repair and fibrosis progression.^33,34 Importantly, HE4 has emerged as a key cancer biomarker, particularly in ovarian cancer, where increased serum levels correlate with tumor progression, poor prognosis, and chemotherapy resistance.³⁰ Beyond its clinical use as a biomarker, HE4 (WFDC2) has been implicated in context-dependent regulation of protease-related processes, suggesting broader extracellular functions.³⁵ Elevated HE4 expression has been linked to enhanced cancer cell proliferation, migration, and invasion in various cancers, such as ovarian carcinoma.³⁶ HE4 promotes immune evasion by reshaping cytokine profiles and limiting immune-cell infiltration, thereby fostering an immunosuppressive TME.^36,37 However, the mechanisms underlying HE4-mediated immunosuppression have remained unclear. Here, we identify HE4 as a key regulator of PD-L1 expression in the TME. HE4 engages IFN-γ receptors on myeloid cells to activate JAK-STAT3 signaling and drive PD-L1 transcription. Neutralization of mouse or human HE4 reduced PD-L1 expression on myeloid cells, enhanced CD8⁺ T cell activation, and suppressed tumor growth in multiple models, while inducing fewer irAEs than PD-1 blockade. Clinically, elevated HE4 is associated with poor prognosis yet predicts improved responsiveness to PD-1 inhibitors in non-small cell lung cancer (NSCLC), highlighting its dual role as a therapeutic target and predictive biomarker.

Results

Extracellular HE4 promotes tumor immune evasion by suppressing CD8⁺ T-cell-mediated cytotoxicity

To identify potential immunotherapy targets in the TME, we performed single-cell RNA sequencing (scRNA-seq) on paired tumor and paratumor tissues from treatment-naïve lung adenocarcinoma (LUAD) patients. Uniform manifold approximation and projection (UMAP) analysis identified 23 cell clusters (Figure 1A), revealing significantly higher WFDC2 (HE4) expression in epithelial/malignant cells (EPCAM⁺KRT8⁺KRT18⁺KRT19⁺) from tumors than from paratumor tissues (Figures 1B and 1C). Furthermore, epithelial cells (clusters 7, 10, 14, 17, and 20) were stratified into malignant and non-malignant populations using inferCNV analysis, and WFDC2 was expressed at significantly higher levels in malignant epithelial cells than in their non-malignant counterparts (Figures S1A and S1B). Consistently, analysis of NSCLC scRNA-seq datasets from TISCH2 confirmed predominant WFDC2 expression in malignant populations (Figure S1C), and GEPIA analysis validated higher WFDC2 mRNA levels in LUAD tumors than in normal tissues (Figure 1D). As WFDC2 is a known biomarker in ovarian cancer,³⁰ its elevated expression in ovarian tumors was also confirmed (Figure S1D).

Figure 1.

HE4 promotes tumor immune evasion by upregulating PD-L1 expression within TME

(A–C) scRNA-seq analysis of two paired tumor and paratumor samples from LUAD patients showing cell clustering (A), WFDC2/HE4-expressing cell populations (B), and quantification of WFDC2/HE4 expression in epithelial/malignant clusters (C).

(D) Online dataset analysis of WFDC2 mRNA expression in tumor versus normal LUAD tissues.

(E and F) HE4 overexpression (E) or knockout (F) respectively promoted or suppressed the growth of subcutaneous LLC tumors in mice.

(G) Survival of mice intraperitoneally inoculated with HE4-overexpressing or control ID8 cells.

(H) HE4 knockout attenuated ID8 peritoneal tumor progression, shown by representative abdominal images and ascites volume.

(I) SDS-PAGE with Coomassie blue staining showing the purity of Fc and mouse HE4-Fc (mHE4-Fc) recombinant proteins.

(J) Administration of mHE4-Fc promoted MC38 subcutaneous tumor growth.

(K) mHE4-Fc administration reversed the growth suppression of HE4-KO LLC subcutaneous tumors.

(L) mHE4-Fc failed to reverse tumor growth suppression in HE4-KO LLC tumors implanted in Rag1-deficient mice.

(M) Extracellular HE4 did not promote LLC cell proliferation in vitro, as assessed by CCK8 assay.

(N) mHE4-Fc administration suppressed CD8⁺ T cell activation in the TME of HE4-KO LLC tumors, assessed by IFN-γ⁺ and granzyme B⁺ CD8⁺ T cells.

(O) HE4 upregulated PD-L1 expression on macrophages in the microenvironment of LLC, MC38, and HE4-KO LLC tumors.

(P–R) mHE4-Fc administration failed to reverse tumor suppression of HE4-KO LLC subcutaneous tumors in Cd274-deficient mice, with treatment scheme (P), tumor growth (Q), and tumor weight (R).

Statistical analyses were performed using Wilcoxon rank-sum test (D), two-way ANOVA (E, F, J–L, and Q), log rank (Mantel-Cox) test (G), and two-tailed paired (C) or unpaired t tests (H, M, N, O, and R). Data represent two independent experiments (E and M).

Based on these observations, we examined HE4 expression in mouse tumor cell lines (Figure S1E) and generated HE4-overexpressing or knockout Lewis lung carcinoma (LLC) and ID8 cells (Figures S1F and S1G). By using an LLC subcutaneous tumor model, we assessed the role of HE4 in tumor progression in vivo. The efficiency of HE4 overexpression and knockout was independently validated by measuring serum HE4 levels, as shown in Figure S1H. Compared with that in control cells, ectopic overexpression of HE4 markedly potentiated tumor growth (Figures 1E and S1I). In contrast, HE4 knockout significantly attenuated tumor growth in vivo (Figures 1F and S1J). In an ID8 intraperitoneal ovarian cancer model, HE4 overexpression significantly reduced survival (Figure 1G), while HE4 knockout—validated by ascites HE4 levels (Figure S1K)—nearly abolished ascites formation (Figure 1H) and peritoneal tumor deposits (Figure S1L).

As HE4 is a tumor-secreted glycoprotein, we hypothesized that its protumorigenic activity is mediated, at least in part, by its extracellular form. To test this, we purified Fc-tagged recombinant mouse HE4 (mHE4-Fc) (Figure 1I) and administered it intraperitoneally to mice bearing MC38 subcutaneous tumors (Figure S1M), which express low endogenous HE4 (Figure S1E). mHE4-Fc administration significantly accelerated tumor growth in vivo (Figures 1J and S1N). Notably, the administration of mHE4-Fc protein also effectively reversed the growth suppression of HE4-knockout LLC tumors (Figures 1K, S1O, and S1P), indicating that extracellular HE4 promotes tumor progression in vivo. Although HE4 has been implicated in tumor cell proliferation and invasion, mHE4-Fc failed to restore tumor growth in Rag1-deficient mice lacking mature T and B cells (Figures 1L, S1Q, and S1R) and did not affect LLC or ID8 cell proliferation in vitro (Figures 1M and S1S). These results indicate that extracellular HE4 exerts its protumorigenic effects primarily through adaptive immunity-dependent microenvironmental mechanisms rather than direct tumor cell-intrinsic proliferation.

Given that IFN-γ and granzyme B are key CD8⁺ T cell effector molecules^38,39 and were elevated in mice bearing HE4-knockout tumors (Figures S1T and S1U), we examined whether HE4 suppresses antitumor immunity in the TME. Flow cytometric analysis revealed increased frequencies of IFN-γ⁺ CD8⁺ T cells in HE4-knockout ID8 and LLC tumors (Figures S2A–S2C), whereas HE4 overexpression reduced IFN-γ⁺ CD8⁺ T cells (Figure S2D). Consistently, administration of mHE4-Fc suppressed CD8⁺ T cell effector function, as reflected by reduced IFN-γ⁺ and granzyme B⁺ CD8⁺ T cells in HE4-knockout LLC tumors (Figures 1N and S2E). Finally, depletion of CD8⁺ T cells using anti-CD8α antibody substantially reversed the tumor-suppressive effect of HE4 knockout (Figures S2F–S2H) and reduced IFN-γ and tumor necrosis factor alpha (TNF-α) levels in the TME (Figures S2I and S2J), demonstrating that HE4 promotes tumor immune evasion primarily by impairing CD8⁺ T cell function.

Extracellular HE4 promotes tumor immune evasion via myeloid PD-L1 in the TME

PD-L1 is a key immune checkpoint that suppresses cytotoxic T cell function. HE4 has been reported to stabilize PD-L1 protein via post-transcriptional mechanisms,⁴⁰ analysis of LUAD-patient-derived single-cell suspensions cultured ex vivo revealed a significant positive correlation between extracellular HE4 levels and PD-L1 expression on CD45⁺ immune cells (Figure S3A). Consistently, analysis of public scRNA-seq datasets (GSA: CRA001160; GEO: GSE155446, GSE146771) showed positive correlations between tumor-cell-expressed WFDC2 and macrophage CD274 expression in pancreatic adenocarcinoma, medulloblastoma, and colorectal cancer (Figures S3B–S3D), supporting a functional link between tumor-derived HE4 and immune PD-L1 regulation. In mouse LLC and MC38 tumor models, recombinant mHE4-Fc selectively increased PD-L1 expression on CD45⁺ immune cells but not CD45⁻ tumor cells (Figures S3E–S3G). Moreover, mHE4 failed to promote tumor growth in Cd274-deficient mice (Figures S3H and 1P–1R), indicating that HE4-driven tumor promotion depends on immune PD-L1.

Beyond tumor cells, PD-L1 expression on myeloid populations critically regulates tumor progression in vivo.^29,41,42 We analyzed PD-L1 expression across myeloid subsets in LLC tumors, including macrophages (Figure S4A), monocytes (Figure S4B), myeloid-derived suppressor cells (MDSCs) (Figure S4C), and dendritic cells (DCs) (Figure S4D). HE4 knockout significantly reduced PD-L1 expression on macrophages, DCs, and MDSCs (Figures 1O and S4E), whereas intraperitoneal administration of recombinant mHE4-Fc restored PD-L1 expression in these populations (Figures 1O and S4E). In contrast, PD-L1 expression on monocytes was not significantly affected by either HE4 knockout or mHE4-Fc treatment (Figures S4E and S4F).

Flow cytometric analysis of macrophage subsets revealed that HE4 preferentially regulated PD-L1 expression in M0- and M2-like macrophages, with a weaker effect in M1-like macrophages (Figures S4G and S4H). Conversely, mHE4-Fc increased the proportion of PD-L1⁺ cells mainly in M0- and M2-like macrophages, whereas the response in M1-like macrophages was attenuated (Figure S4I), potentially reflecting near-saturating PD-L1 induction driven by strong IFN-γ-IFNGR-STAT1 signaling in M1-polarized cells.

HE4 transcriptionally upregulates PD-L1 expression via the IFN-γR-JAK-STAT3 pathway

Next, we examined the effects of HE4 on PD-L1 expression in vitro across myeloid cell types. Recombinant HE4 robustly increased PD-L1 expression in macrophages (Figure 2A) and DCs (Figure S5A). In polarization assays, HE4 markedly upregulated PD-L1 in undifferentiated and M2-polarized Raw264.7 cells, whereas this effect was attenuated in M1-polarized macrophages (Figure S5B). Notably, HE4 treatment did not affect the phagocytic capacity of Raw264.7 cells (Figure S5C), indicating a selective effect on PD-L1 regulation.

Figure 2.

HE4 transcriptionally upregulates PD-L1 via the IFN-γR-JAK1-STAT3 pathway

(A and B) Raw264.7 cells were pretreated with cycloheximide (CHX) for 2 h, followed by HE4-Flag treatment for 12 h; PD-L1 expression was analyzed by immunoblotting (A) and flow cytometry (B).

(C) Cd274 mRNA levels in Raw264.7 cells treated with control protein or recombinant HE4 for 3 h, assessed by qPCR.

(D) Raw264.7 cells were pretreated with Stattic (STA) or ruxolitinib (RUX) for 1 h, followed by HE4-Fc treatment for 3 h; Cd274 mRNA was quantified by qPCR.

(E) Binding of recombinant HE4-Flag to Raw264.7 cells assessed by flow cytometry using FITC-anti-Flag antibody.

(F) Activation of JAK/STAT signaling in BMDMs treated with Fc or HE4-Fc, analyzed by immunoblotting.

(G–I) Control or IFN-γR1-knockdown Raw264.7 cells were treated with HE4-Fc for 3 h (G) or 12 h (H and I), followed by analysis of Cd274 mRNA (G), PD-L1 protein (H), or STAT3 phosphorylation (I) by qPCR, flow cytometry, or immunoblotting, respectively.

(J–L) Recombinant IFN-γR1, IFN-γR2, and HE4 were incubated and subjected to immunoprecipitation using Ni-TED resin (J) or anti-Flag antibody (K and L), followed by immunoblot analysis.

(M) AlphaFold-3-predicted binding model of HE4 with IFN-γR1 and IFN-γR2.

(N) Immunoprecipitation analysis of recombinant IFN-γR1-His and IFN-γR2-His incubated with wild-type or mutant HE4-Flag.

(O–Q) Raw264.7 cells were treated with wild-type or mutant HE4, and Cd274 mRNA (O) and PD-L1 protein levels were analyzed by qPCR (O), immunoblotting (P), or flow cytometry (Q). Statistical analyses were performed using one-way ANOVA (G) or two-tailed unpaired t test (O). Data represent three independent experiments.

Although HE4 has been reported to stabilize PD-L1 post-translationally by inhibiting MMP9/13-mediated degradation,⁴⁰ blockade of protein synthesis with cycloheximide completely abolished HE4-induced PD-L1 upregulation (Figures 2A and 2B). Moreover, recombinant HE4 significantly increased Cd274 mRNA levels in Raw264.7 cells (Figure 2C), supporting a transcriptional mechanism underlying HE4-driven PD-L1 induction.

Multiple signaling pathways, including IFN-γ/JAK/STAT, HIF-1α, nuclear factor κB (NF-κB), and MAPK/ERK, have been implicated in transcriptional regulation of PD-L1.¹² To define the mechanism underlying HE4-induced PD-L1 expression, we performed a pharmacological inhibitor screen. HE4-mediated PD-L1 upregulation was selectively blocked by JAK1/2 and STAT3 inhibitors (Figure 2D) but not by inhibitors of p38, JNK, NF-κB, HIF-1α, EGFR, or STAT1 (Figures S6A and S6B), indicating a primary role for JAK-STAT3-dependent transcription. Consistently, recombinant HE4 directly bound macrophages (Figure 2E) and activated JAK-STAT3 signaling (Figure 2F), suggesting engagement of a surface receptor upstream of this pathway.

Because IFN receptors can activate JAK-STAT3 signaling,^43,44 we next interrogated their involvement. Knockdown of Ifnar1 did not impair HE4-induced Cd274 mRNA upregulation (Figures S6C–S6E), whereas knockdown of Ifngr1 markedly attenuated HE4-induced PD-L1 mRNA and protein expression as well as STAT3 phosphorylation (Figures 2G–2I). Direct interaction between HE4 and IFN-γR1/2 was confirmed by immunoprecipitation of recombinant proteins (Figures 2J–2L). AlphaFold 3 structural modeling identified key residues in mature HE4 mediating IFN-γR1/2 binding (Figure 2M); alanine substitution of these residues abolished receptor interaction (Figure 2N) and significantly reduced PD-L1 induction at both mRNA and protein levels (Figures 2O–2Q). Together, these data demonstrate that HE4 drives PD-L1 transcription through direct engagement of IFN-γR and activation of the JAK-STAT3 signaling axis.

Integrated scRNA-seq analysis of LUAD samples revealed broad expression of IFN-γR1 and IFN-γR2 across cell types, with enrichment in myeloid populations, particularly macrophages (Figures S6F and S6G). In these cells, IFN-γ-IFN-γR signaling promotes antimicrobial functions, antigen presentation, and inflammatory cytokine production.^45,46 To compare HE4- and IFN-γ-induced transcriptional programs, we performed RNA-seq in Raw264.7 cells. HE4 and IFN-γ induced overlapping yet distinct gene sets (Figures 3A and 3B), with 74 shared upregulated genes, including Cd274 (Figure 3C; Table S1). KEGG analysis showed convergence on pathways such as cytokine-cytokine receptor interaction and JAK-STAT signaling (Figures S7A and S7B), while HE4 uniquely enriched pathways including interleukin-17 (IL-17) signaling and lipid/atherosclerosis-related programs (Figure S7A). Consistent with these differences, IFN-γ triggered rapid but transient STAT3 phosphorylation, whereas HE4 induced delayed yet sustained STAT3 activation (Figures S7C and S7D), indicating distinct signaling kinetics and feedback regulation.

Figure 3.

HE4 competes with IFN-γ for IFN-γR binding and modulates downstream gene expression

(A and B) Raw264.7 cells were stimulated with HE4-Fc (20 μg/mL) or IFN-γ (100 ng/mL) for 3 h, followed by RNA-seq; volcano plots of HE4- (A) or IFN-γ-regulated genes (B) are shown.

(C) Genes commonly upregulated by HE4 and IFN-γ.

(D) AlphaFold-3-predicted interfaces of HE4-IFNGR1/2 and IFN-γ-IFNGR1/2 complexes, with shared receptor-contact residues highlighted.

(E) Competitive binding assay: His-tagged IFNGR1/2 was incubated with Flag-HE4 in the presence or absence of IFN-γ, followed by Ni-TED pull-down and immunoblotting.

(F and G) SPR sensorgrams showing binding of mHE4-Fc (F) or mIFN-γ-Fc (G) to mIFNGR1-His.

(H and I) ELISA quantification of HE4 and/or IFN-γ levels in ascites from ID8-tumor-bearing mice (H) and LLC-tumor-conditioned media (I).

(J) High concentrations of HE4 reduced IFN-γ binding to IFNGR1/2 in competitive pull-down assays.

(K) PCA of RNA-seq profiles from Raw264.7 cells treated with HE4-Fc, HE4-Fc plus IFN-γ, or IFN-γ for 12 h.

(L) Expression (TPM) of representative STAT1- or STAT3-associated genes following the indicated treatments.

Statistical analyses were performed using two-tailed paired Student’s t tests (H and I). Data in (E) and SPR sensorgrams (F and G) are representative of three independent experiments.

Furthermore, by using AlphaFold 3, we identified three identical sites on IFN-γR1/2 bound to both HE4 and IFN-γ: Ser51/Lys105 of IFN-γR1 and Tyr47 of IFN-γR2 (Figure 3D). This tripartite molecular mimicry positions HE4 as a potential antagonist of IFN-γ-mediated immune surveillance. Thus, we conducted a pull-down assay to test whether HE4 and IFN-γ compete for binding to IFNγ-R1/2. IFN-γ could compete for the binding of HE4 and IFNγ-R1/2, as IFN-γ significantly attenuated HE4 binding to IFNγ-R1/2 at comparable concentrations (Figure 3E). Unexpectedly, HE4 could not compete for the binding of IFN-γ and IFN-γR1/2, as the amount of IFN-γ bound to IFNγ-R1/2 did not decrease even when the concentration of HE4 was 2-fold greater (Figure S7F), which could be due to the different affinities between HE4 and IFNγ-R1/2 and between IFN-γ and IFNγ-R1/2. Surface plasmon resonance (SPR) analysis revealed that the binding affinity between HE4-Fc and IFNGR1-His (KD ≈ 20.8 nM; Figure 3F) was approximately 10-fold lower than that between IFN-γ-Fc and IFNGR1-His (KD ≈ 1.83 nM; Figure 3G). Notably, the concentration of HE4 in the TME of ID8 ascites or LLC tumors was approximately 14.6- or 16.2-fold greater than that of IFN-γ, respectively (Figures 3H and 3I). Mechanistically, such a pronounced concentration disparity could compensate for the relatively weak receptor affinity of HE4, ultimately potentiating its competitive binding capacity against IFN-γ. Indeed, a 15-fold excess of HE4 significantly attenuated the binding of IFN-γ to IFNγR1/2 (Figure 3J). Collectively, our findings support a concentration-dependent competitive mechanism, whereby elevated TME HE4 levels may effectively outcompete IFN-γ for receptor binding, potentially modulating IFN-γ-mediated signaling pathways in the tumor milieu.

Given that stimulation of Raw264.7 cells with HE4 versus IFN-γ did not elicit fully overlapping transcriptional programs or KEGG enrichments, we next profiled macrophages under mHE4/mIFN-γ (1:1) competition by RNA-seq. Principal-component analysis (PCA) revealed clear separation of transcriptomic states across treatments (Figure 3K). As expected, macrophages treated with mHE4-Fc or mIFN-γ alone formed distinct clusters along PC1 and PC2, indicating that each stimulus drives a unique gene-expression signature. Notably, co-treatment with mHE4-Fc and IFN-γ did not cluster with either single-stimulus group; instead, these samples occupied an intermediate yet clearly discrete position in PCA space. Together, these results suggest that concurrent HE4 and IFN-γ exposure induces a distinct transcriptional state, rather than a simple additive response or dominance of either the HE4- or IFN-γ-driven program.

As expected, transcriptomic profiling showed that IFN-γ robustly induced canonical STAT1 target genes—including Irf1, Stat1, Cxcl10, and Socs1—consistent with strong activation of the IFN-γ-STAT1 axis. Notably, co-treatment with HE4 markedly blunted the induction of these STAT1-dependent transcripts, indicating that HE4 suppresses IFN-γ-driven STAT1 signaling at the transcriptional level (Figures 3L and S7G). In contrast, STAT3-associated genes such as Il10, Vegfa, Ccl2, and Arg2 remained expressed at HE4-like levels under HE4-IFN-γ co-stimulation, suggesting relative preservation of a STAT3-skewed program (Figures 3L and S7G). Together, these data support a model in which HE4 preferentially antagonizes IFN-γ-mediated STAT1 activation, thereby reprogramming macrophages toward a state with reduced inflammatory and antitumor potential. KEGG enrichment analysis further revealed that, relative to IFN-γ alone, co-stimulation with HE4 and IFN-γ increased pathways related to oxidative phosphorylation, non-alcoholic fatty liver disease, and cytokine-cytokine receptor interaction, while downregulating olfactory transduction (Figure S7H). In addition, compared with HE4 alone, HE4-IFN-γ co-stimulation enriched multiple pathogen-sensing/antiviral response pathways—including HSV-1 infection, Epstein-Barr virus (EBV) infection, COVID-19, and cytosolic DNA-sensing—whereas Hippo signaling, inflammatory bowel disease, and several cancer-related pathways were reduced (Figure S7H). We next assessed antigen-presentation-related readouts in macrophages under conditions where HE4 may interfere with IFN-γ signaling. Flow cytometric analysis showed that IFN-γ stimulation markedly increased MHC-II expression on RAW264.7 macrophages, whereas HE4 alone did not induce MHC-II upregulation. Importantly, co-treatment with HE4 did not markedly attenuate IFN-γ-induced major histocompatibility complex (MHC) class II expression (Figure S7I). These data suggest that HE4-mediated immunosuppression is exerted through more selective mechanisms.

Blocking HE4 with a monoclonal antibody has a therapeutic effect on multiple mouse tumors

As HE4 promoted the transcriptional upregulation of PD-L1 (Figure 1O), a key immune checkpoint in the TME, we explored whether blocking HE4 with monoclonal antibody had any therapeutic effect on tumor progression. We screened multiple rats-anti-mHE4 monoclonal antibodies (mAbs) and identified one clone, #117 (IgG2a isotype), that significantly blocked the HE4-induced upregulation of PD-L1 at both the mRNA (Figure S8A) and protein (Figure S8B) levels in Raw264.7 cells. Notably, clone #117 did not affect tumor cell proliferation in vitro (Figures S8C–S8E). Furthermore, we evaluated the specificity and functional activity of clone #117. Using AlphaFold 3, we predicted the key residues on HE4 that mediate interactions with the variable heavy (VH) and light (VL) chains of clone #117. Alanine substitution of these residues (Glu48, Lys53-Cys55, Asp57-Arg59, and Lys72) abolished antibody binding to mHE4, supporting the accuracy of the structural model and the specificity of the interaction (Figure S8F). In addition, to determine whether clone #117 recognizes native, properly folded, and glycosylated HE4, we established a sandwich ELISA using a commercial anti-HE4 antibody for capture and clone #117 for detection. Conditioned medium from wild-type LLC cells produced a robust signal, whereas medium from HE4-knockout cells was nearly undetectable (Figure S8G). Consistently, HE4 was readily detected in conditioned media from wild-type LLC tumors, compared with the corresponding knockout controls (Figure S8H), confirming that clone #117 effectively detects native, tumor-derived HE4. Surface plasmon resonance (SPR) further demonstrated high-affinity binding of clone #117 to HE4 (KD = 0.0462 nM; Figure S8I), with negligible binding to the Fc tag (Figure S8J), further supporting antibody specificity. Finally, a competitive ELISA showed that clone #117 potently blocked HE4-IFNGR1 binding, with an IC₅₀ of 3.79 nM (Figure S8K). Collectively, these data establish clone #117 of anti-HE4 mAb as a highly specific, high-affinity, and functionally blocking antibody suitable for detecting and therapeutically targeting tumor-derived HE4.

Next, we tested the therapeutic efficacy of clone #117 across multiple mouse tumor models, with anti-mHE4 mAb clone #1 used as a control since it fails to block HE4-induced PD-L1 upregulation (Figures S8A and S8B). We found that clone #117 significantly inhibited the growth of subcutaneous LLC tumors in female mice, whereas clone #1 did not have any therapeutic effect (Figures 4A–4C). To exclude sex, we repeated this experiment in male mice and observed the same therapeutic effect of clone #117 (Figures S9A–S9D). By using a urethane-induced spontaneous lung cancer model (Figure 4D), we demonstrated that the anti-mHE4 mAb of clone #117 significantly attenuated tumorigenesis compared with that of the control, as evidenced by reduced lung tumor nodule formation (Figure 4E). Consistently, the anti-mHE4 mAb of clone #117 also had a significant therapeutic effect on the ID8 peritoneal model, which manifested as a significantly decreased ascites volume (Figures 4F and 4G). The HGS-1 cell line represents another well-established high-grade serous ovarian cancer (HGSOC) model that recapitulates key biomechanical, cellular, and molecular features of human HGSOC, including strong concordance in mRNA expression profiles, innate and adaptive immune responses, tissue stiffness, and matrisome composition.⁴⁷ Following orthotopic implantation of HGS-1 cells into the mouse ovary, the therapeutic efficacy of anti-HE4 antibody treatment was evaluated (Figure 4H). As shown in Figure 4I, administration of the anti-HE4 antibody significantly reduced both primary tumor size and overall tumor burden compared with the isotype control group. Interestingly, the anti-mHE4 mAb of clone #117 did not have a significant therapeutic effect on MC38 tumors (Figures S9E and S9F), which may be due to the low expression level of HE4 in MC38 cells (Figure S1E). Interestingly, although 4T1 cells presented higher expression levels of HE4 than LLC cells (Figure S1E), the inhibitory effect of HE4 neutralization on 4T1 tumors was weaker than observed in LLC tumors, despite remaining statistically significant (Figures S9G–S9I). Notably, the free HE4 levels were significantly decreased (Figure 4J), indicating that clone #117 acts by blocking HE4-receptor interactions, rather than accelerating the degradation of HE4 in vivo.

Figure 4.

HE4 blockade exerts therapeutic efficacy across multiple mouse tumor models

(A–C) Mice bearing subcutaneous LLC tumors were treated intraperitoneally with isotype immunoglobulin G (IgG) or anti-mHE4 mAbs (clones #1 or #117). Treatment scheme (A), tumor growth curves (B), and representative tumors with weights (C) are shown.

(D and E) Anti-mHE4 (clone #117) treatment suppressed urethane-induced lung tumorigenesis, shown by representative H&E staining and tumor quantification. Scale bar, 1,000 μm.

(F and G) HE4 blockade attenuated ID8 intraperitoneal tumor progression, shown by representative abdominal images and ascites volume.

(H and I) Therapeutic efficacy of anti-mHE4 (clone #117) in an orthotopic HGS-1 ovarian tumor model, shown by treatment scheme and representative tumors with weights.

(J) Serum-free HE4 levels measured by ELISA in LLC-tumor-bearing mice one day after the final antibody treatment.

(K and L) Anti-mHE4 (clone #117) reduced PD-L1 expression on tumor-associated macrophages in the LLC subcutaneous and ID8 intraperitoneal models.

(M and N) Anti-mHE4 (clone #117) failed to suppress LLC tumor growth in Cd274-deficient mice.

(O) Combination therapy with anti-mHE4 and anti-mCTLA-4 enhanced tumor suppression in the LLC subcutaneous model.

(P) Combination therapy with anti-mHE4 and paclitaxel (PTX) enhanced antitumor efficacy in the LLC subcutaneous model.

(Q) Serum cytokines and clinical chemistry parameters following treatment with isotype IgG, anti-mHE4 mAb, or anti-mPD-1 mAb.

(R and S) Safety assessment showing representative H&E images and incidence of inflammation in heart, liver, lung, and colon after anti-mHE4 or anti-mPD-1 treatment.

Statistical analyses were performed using two-way ANOVA (B and N–P), one-way ANOVA (C, K, and N–Q), two-tailed unpaired t tests (E, G–J, and L), and chi-squared test (S). Data in (A–P) are pooled from two independent experiments. Scale bars, 100 μm.

Consistent with the in vitro findings (Figures S8A and S8B), blockade of HE4 with mAb clone #117 significantly diminished PD-L1 expression on macrophages in the TME of subcutaneous (s.c.) LLC (Figure 4K), intraperitoneal (i.p.) ID8 (Figure 4L), and in situ 4T1 (Figure S9J). Similarly, administration of clone #117 also decreased PD-L1 expression on DCs and MDSCs in the TME of LLC tumors (Figure S9K). Furthermore, in the microenvironment of LLC tumors, PD-L1 expression was reduced across all macrophage subsets, including M1-like, M2-like, and non-M1/M2-like population, following administration of clone #117 (Figure S9K).

Next, we explored whether the therapeutic antitumor effect of HE4 blockade was completely dependent on PD-L1 downregulation. By using an ID8 peritoneal model and an LLC subcutaneous model, we found that simultaneously blocking HE4 and PD-L1 did not result in better antitumor effects than single-agent therapy, which indicates that the antitumor activity of the anti-mHE4 mAb is completely dependent on its regulation of PD-L1 expression (Figures S9L–S9O). Consistently, the antitumor effect of clone #117 was also abolished when the tumors were inoculated subcutaneously into Cd274-deficient mice, further confirming that HE4 upregulates PD-L1 expression mainly on CD45⁺ immune cells including macrophages but not cancer cells (Figures 4M and 4N). In summary, these data indicate that HE4 is a critical regulator of PD-L1 and may even be a major regulator in the TME of certain tumors.

These findings indicate that the pro-tumorigenic activity of extracellular HE4 is predominantly mediated through the PD-1/PD-L1 axis and is mechanistically independent of CTLA-4 signaling. Given that CTLA-4 and PD-1/PD-L1 represent distinct yet complementary immune checkpoint pathways and that dual blockade of these axes has been clinically approved with improved antitumor efficacy across multiple cancer types, we next examined the therapeutic benefit of combining anti-HE4 with anti-CTLA-4 in vivo (Figure S9P). Indeed, co-administration of anti-mHE4 and anti-mCTLA-4 antibodies more effectively suppressed LLC tumor growth than either monotherapy (Figures 4O and S9P). Moreover, compared with single-agent treatment, HE4 blockade also synergized with the chemotherapeutic agent paclitaxel, yielding enhanced antitumor activity in both the LLC model (Figures 4P and S9Q) and the ID8 model (Figures S9R and S9S). Collectively, these results support anti-HE4 as a rational combination partner for CTLA-4-targeting immunotherapy or chemotherapy, consistent with clinical experience that combination regimens integrating PD-1/PD-L1 blockade with anti-CTLA-4 or chemotherapy outperform monotherapy.

irAEs pose a significant challenge in the clinical use of ICIs. To assess the potential inflammatory risks of HE4 blockade, we followed a previously described protocol.⁴⁸ Notably, anti-PD1 treatment significantly elevated serum levels of IFN-γ, IL-6, cardiac troponin I (cTnI), C-reactive protein (CRP), and ALT, whereas anti-HE4 treatment did not significantly alter these parameters (Figure 4Q). Histological analysis further revealed markedly lower incidence and severity of inflammation in the heart, liver, lung, and colon following anti-HE4 treatment compared with anti-PD1 treatment (Figures 4R and 4S). Collectively, these findings demonstrate that anti-HE4 treatment is associated with a reduced risk of irAEs, underscoring its superior safety profile as an immunotherapeutic strategy.

Blockade of HE4 regulates the immune environment in the TME

Next, we examined the overall immune microenvironmental changes after HE4 blockade via scRNA-seq. By dimensionality reduction analysis, 26 groups of cells were identified in the TME of LLC tumors (Figure 5A), and the ratios of several groups of tumor cells were significantly lower in the anti-HE4 mAb treatment group than in the isotype control treatment group (Figures 5B and 5C), which was consistent with the observed tumor growth suppression phenotype. PD-L1 expression was significantly reduced across the six macrophage groups (Figures 5D and 5E). In addition to the reduced proportion of PD-L1⁺ macrophage subsets, flow cytometric analysis revealed a decreased M2/M1 macrophage proportion (Figure S9T), although the scRNA-seq data did not show a significant difference in the proportions of M1 and M2 macrophages (Figure 5C). Additionally, tumors treated with the anti-HE4 mAb presented a greater percentage of effector CD8⁺ T cells than did control tumors, although the difference did not reach statistical significance, potentially because of the small sample size (Figure 5F). Consistent with the results from scRNA-seq, the percentages of effector CD8⁺ T cells (CD69⁺ or IFN-γ⁺) were significantly greater in the TME of LLC tumors or ID8 ascites after anti-HE4 mAb treatment than in that of the control treatment, as shown by flow cytometry analysis (Figures 5G and 5H). Consistently, the antitumor effect of the anti-HE4 mAb was not demonstrated when the tumors were inoculated into Rag1-deficient mice (Figures 5I–5K). Furthermore, depletion of CD8⁺ T cells with an anti-CD8α mAb completely abrogated the antitumor effect of the anti-HE4 mAb, demonstrating that HE4-blockade-mediated tumor control is strictly dependent on CD8⁺ T cells (Figures 5L–5O). Notably, macrophage depletion via the anti-CSF1R mAb also abolished the antitumor effect of HE4 blockade but exhibited comparable antitumor activity to that of the anti-HE4 mAb treatment (Figures 5P–5S). Taken together, these data indicate that neutralization of HE4 significantly attenuates the expression of PD-L1 on macrophages to remodel the tumor immune microenvironment and inhibit tumor growth.

Figure 5.

HE4 blockade promotes antitumor immunity in the tumor microenvironment

(A–C) scRNA-seq analysis of LLC tumors from mice treated with control IgG or anti-HE4 antibody (n = 3 per group), showing UMAP clustering of 26 cell populations (A), relative abundance of each cluster (B), and aggregated cell types (C). ∗p < 0.05

(D and E) HE4 neutralization reduced Cd274 (PD-L1) expression in myeloid compartments. UMAP feature plots show Cd274 expression in macrophage/monocyte and epithelial/malignant populations (D), with paired comparison across macrophage clusters (E).

(F) UMAP visualization of nine intratumoral T cell subclusters in control IgG– and anti-HE4–treated tumors.

(G) In the LLC subcutaneous model, intratumoral IFN-γ⁺ and CD69⁺ CD8⁺ T cells were quantified by flow cytometry.

(H) In the ID8 intraperitoneal model, IFN-γ⁺ CD8⁺ T cells were quantified by flow cytometry.

(I–K) HE4 neutralization failed to suppress LLC tumor growth in Rag1^−/− mice, shown by treatment scheme and tumor growth/endpoint analyses.

(L–O) CD8⁺ T cell depletion abrogated the antitumor efficacy of HE4 blockade, with treatment scheme, tumor growth/endpoint measurements, and confirmation of depletion efficiency by flow cytometry.

(P–S) Macrophage depletion using anti-CSF1R diminished the antitumor efficacy of HE4 neutralization, with tumor growth/endpoint measurements and confirmation of depletion efficiency by flow cytometry.

Statistical analyses were performed using unpaired t tests (B, C, H, and K), paired t test (E), one-way ANOVA (G, N, O, R, and S), and two-way ANOVA (J, M, and Q). Data in (G, H, and L–S) are pooled from two independent experiments.

Neutralization of HE4 has therapeutic effects on human cancers

We then examined whether HE4 had any effect on the expression of PD-L1 on human macrophages. Recombinant human HE4 protein significantly increased both the mRNA and protein levels of PD-L1 in PMA-differentiated THP-1 macrophages (Figures 6A–6C) and strongly bound to these cells (Figure 6D). Consistent with our previous data (Figures 2D, S5A, and S5B), treatment with JAK and STAT3 inhibitors, but not STAT1 inhibitors, significantly attenuated hHE4-induced PD-L1 upregulation in PMA-differentiated THP-1 cells (Figures 6E and 6F). These data indicate that HE4 plays a highly conserved role in regulating PD-L1 expression on macrophages.

Figure 6.

HE4 neutralization exerts therapeutic activity in human cancer models

(A–C) PMA-differentiated THP-1 macrophages were stimulated with Fc or hHE4-Fc, and PD-L1 expression was assessed by flow cytometry (A), immunoblotting (B), and RT-qPCR (C); a commercial hHE4-Fc was included as an independent control.

(D) Binding of hHE4 to PMA-differentiated THP-1 cells assessed by flow cytometry.

(E and F) PMA-differentiated THP-1 cells were pretreated with ruxolitinib, fludarabine, or Stattic, followed by hHE4-Fc stimulation; PD-L1 was quantified by RT-qPCR (E) and flow cytometry (F).

(G and H) Anti-hHE4 monoclonal antibodies inhibited hHE4-induced PD-L1 upregulation in PMA-differentiated THP-1 cells, assessed by RT-qPCR (G) and flow cytometry (H).

(I) Anti-hHE4 mAb clone #10 blocked hHE4 binding to PMA-differentiated THP-1 cells.

(J) Binding of wild-type or epitope-mutant hHE4-Fc to anti-hHE4 mAb clone #10 was quantified by ELISA.

(K) Pharmacokinetic analysis of anti-hHE4 mAb clone #10 in C57BL/6 mice following intravenous administration.

(L–O) Fresh human LUAD tumor cell suspensions were treated with anti-hHE4 mAb clone #10, followed by flow cytometric analysis of PD-L1 and ELISA measurement of IFN-γ and granzyme B. (P) Recombinant HE4 suppressed IFN-γ production in human LUAD tumor cell suspensions.

(Q–T) HE4 blockade enhanced PBMC-mediated antitumor activity in humanized C-NKG mice bearing OVCAR3 or NCI-H358 tumors, shown by treatment scheme, representative tumors, and tumor weights.

Schematics (L and Q) were created using BioRender. Statistical analyses were performed using one-way ANOVA (C, E, and G), paired t tests (M–P), or unpaired t tests (S and T). Data in (A–J) are representative of three independent experiments; data in (Q–T) are pooled from two independent experiments.

Although human and mouse HE4 exhibit functional conservation, their sequence homology is relatively low (approximately 44.3%), and mAb raised against mouse HE4 (clone #117) does not recognize human HE4 (Figure S10A). Next, we screened mice-anti-human HE4 mAbs and found that one clone of the anti-hHE4 mAbs (clone #10, IgG2b) significantly inhibited hHE4-induced PD-L1 mRNA and protein expression in PMA-induced THP-1 cells (Figures 6G and 6H). This mAb also greatly reduced the binding of the hHE4 protein to THP-1 cells (Figure 6I). Furthermore, we evaluated the specificity of clone #10. Using AlphaFold 3, we predicted key residues on HE4 that mediate interactions with the variable heavy (VH) and light (VL) chains of clone #10. Alanine substitution of these residues (Asp57, Asn58, Lys76, Gln96, Ser101-Cys103, and Gln106) abolished antibody binding to hHE4, supporting the specificity of the interaction (Figure 6J). To further characterize the properties of this anti-hHE4 mAb, we assessed its pharmacokinetic behavior in vivo by intravenously administering clone #10 to C57BL/6 mice as a surrogate model for a fully humanized anti-hHE4 antibody. Following a single intravenous dose (10 mg/kg), serum concentration-time profiling showed a rapid distribution phase followed by a prolonged elimination phase (Figure 6K). The antibody achieved a peak serum concentration (C_max) of 36.15 μg/mL shortly after dosing and exhibited a long terminal half-life (t_1/2) of ∼137 h. The mean residence time (MRT) was 197.3 h, indicating sustained systemic exposure. Collectively, these pharmacokinetic parameters demonstrate prolonged in vivo retention and slow clearance of the anti-HE4 antibody.

Importantly, compared with the isotype control, treatment with the anti-hHE4 mAb clone #10 significantly diminished PD-L1 expression on CD45⁺ cells and increased the concentrations of IFN-γ and granzyme B in single-cell suspensions of human LUAD (Figures 6L‒6O). Conversely, compared with the control, the recombinant hHE4 protein significantly inhibited IFN-γ production (Figure 6P). Considering that the anti-hHE4 mAb clone #10 does not cross-react with mouse HE4 (Figure S10B), to further evaluate the antitumor efficiency of HE4 neutralization preclinically, we therefore employed a peripheral blood mononuclear cell (PBMC)-humanized xenograft models on C-NKG mice, which lack T, B, and natural killer (NK) cells (Figure 6Q). Administration of the anti-hHE4 mAb clone #10 significantly attenuated tumorigenesis in the human ovarian line OVCAR3 (Figures 6R and 6S) and NSCLC line NCI-H358 (Figure 6T), which suggests that blocking HE4 with specific mAbs holds promise for clinical application.

High HE4 expression in the TME predicted an improved response to immune checkpoint inhibitors

Furthermore, analysis of the TCGA and GTEx datasets revealed differential HE4 mRNA expression across additional solid tumors beyond LUAD and ovarian cancer. Specifically, beyond LUAD (Figure 1D) and ovarian cancers (Figure S1D), HE4 expression was also significantly upregulated in cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC), cholangiocarcinoma (CHOL), pancreatic adenocarcinoma (PAAD), uterine corpus endometrial carcinoma (UCEC), and uterine carcinosarcoma (UCS), whereas it was downregulated in colon adenocarcinoma (COAD), kidney renal clear cell carcinoma (KIRC), and sarcoma (SARC) (Figure 7A). Consistently, immunohistochemical analysis confirmed elevated HE4 protein expression in PAAD (Figure S11A), reduced expression in COAD (Figure S11B), and unchanged levels in bladder cancer relative to matched adjacent nontumor tissues (Figure S11C). Together, these data indicate that HE4 expression is high in certain human cancer types.

Figure 7.

Elevated HE4 expression in human cancers correlates with improved response to immune checkpoint inhibitors

(A) Pan-cancer transcriptional profiling of WFDC2 (HE4) expression across TCGA tumors and GTEx normal tissues (log₂[TPM+1]). ns, not significant; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001.

(B and C) HE4 protein levels in paired LUAD tumor (LT) and adjacent non-tumor (LN) tissues (n = 16) analyzed by immunoblotting, with densitometric quantification.

(D) Immunohistochemical analysis of HE4 expression in LUAD tumors (LT) and paracancerous tissues (LN) (n = 103), with representative images and quantitative scoring.

(E–G) HE4 expression in tumor samples from immunotherapy-sensitive versus resistant patients across two independent cohorts, assessed by immunohistochemistry.

(H) Correlation between HE4 expression and immunotherapy response in an online NSCLC cohort.

(I) Working model illustrating that tumor-derived HE4 engages IFN-γR on myeloid cells, biases signaling toward JAK-STAT3, induces PD-L1 transcription, and that HE4 neutralization restores antitumor immunity.

Statistical analyses were performed using two-tailed unpaired (A, E, G, and H) or paired (C and D) Student’s t tests. Scale bars, 100 μm.

By using western blotting and immunohistochemical staining, we observed significantly higher HE4 protein levels in LUAD samples than in matched paracancerous tissues (Figures 7B–7D), consistent with earlier findings (Figure 1D). Having established that HE4 regulates myeloid PD-L1 expression in the TME, we next examined its potential as a biomarker for predicting the therapeutic efficacy of anti-PD-1 immunotherapy. We collected one cohort of cancer samples from NSCLC patients who received anti-PD-1 immunotherapy plus chemotherapy and reported that HE4 expression was significantly higher in samples from patients who were sensitive to anti-PD-1 immunotherapy than in those from patients who were resistant to anti-PD-1 treatment (Figure 7E). Notably, HE4 outperformed PD-L1 expression as a predictive biomarker of immunotherapy, as evidenced by its greater AUC (0.7309 vs. 0.6133; Figure 7F). Analysis of other NSCLC cohorts that received anti-PD-1 immunotherapy plus (or not) chemotherapy yielded similar results (Figure 7G). Consistently, database analyses further showed that higher HE4 expression associated with a greater response rate to anti-PD-1 immunotherapy, which indicates that HE4 may be an appropriate biomarker for predicting the efficiency of anti-PD-1 immunotherapy (Figure 7H). Together, these data indicate that HE4 is highly expressed in certain human cancer types and may serve as a predictive biomarker for current anti-PD-1 immunotherapy.

Discussion

Immune checkpoint inhibitors targeting the PD-1/PD-L1 axis have improved cancer therapy but remain limited by incomplete responses and irAEs. Here, we identify tumor-secreted HE4 as a driver of myeloid PD-L1 transcription via IFN-γR-JAK–STAT3 signaling. HE4 neutralization suppresses PD-L1 and elicits robust antitumor activity in multiple models (Figure 7I), while inducing substantially fewer irAEs than PD-1 blockade (Figures 4Q–4S). Together, these findings establish HE4 as a key mediator of tumor-immune crosstalk and a promising immunotherapy target. Clinically, elevated HE4 may identify tumors with greater dependence on PD-1/PD-L1-mediated immune evasion, although this association may also reflect adaptive immune pressure and warrants further validation integrating T-cell-inflamed and IFN-γ signatures with immune-cell PD-L1.

Current ICIs, particularly anti-PD-1/PD-L1 antibodies, have transformed cancer therapy but often disrupt peripheral tolerance and cause irAEs, including dermatitis, colitis, hepatitis, and pneumonitis, thereby limiting broader clinical use.^{2,4,49,50,51,52} In contrast, HE4 shows minimal expression in normal tissues but is selectively upregulated in multiple malignancies, including ovarian, lung, gastric, breast, and pancreatic cancers.^30,53,54,55 Accordingly, HE4 blockade may preferentially target tumor-associated immune compartments while limiting systemic immune activation. Consistent with this notion, HE4 neutralization induced little multi-organ inflammation compared with anti-PD-1 therapy (Figures 4Q–4S), supporting a potentially improved safety profile that warrants further validation in autoimmune-prone and humanized models.

Rowswell-Turner et al. first linked extracellular HE4 to PD-L1 upregulation in myeloid cells through a proposed post-translational mechanism involving reduced MMP-mediated PD-L1 proteolysis.⁴⁰ In contrast, our data support a distinct model in which HE4 primarily regulates PD-L1 at the transcriptional level, as indicated by cycloheximide sensitivity and robust Cd274 mRNA induction. We further identify IFN-γR (IFNGR1/2) as the functional receptor for extracellular HE4 and define a downstream JAK–STAT3 pathway required for PD-L1 induction in myeloid cells. This receptor-defined mechanism enables a therapeutic strategy—neutralizing HE4 to block HE4-IFNGR1 engagement—that suppresses tumor growth in vivo and may offer an improved therapeutic window compared with direct PD-1/PD-L1 blockade. The apparent discrepancy with prior work may reflect context-dependent regulation of PD-L1 by HE4, with post-translational or transcriptional mechanisms predominating under different biological settings.

Consistent with prior reports identifying IFN-γ as the canonical driver of PD-L1 expression in tumor and myeloid cells,^14,56 our data show that HE4 can function as an alternative and in some contexts dominant, inducer of PD-L1. HE4 blockade markedly reduced macrophage PD-L1 and restored CD8⁺ T cell activity even in the presence of IFN-γ, indicating a non-redundant regulatory axis (Figures 4K–4L, 5G–5H, and S9J–S9K). Biochemical and structural analyses further demonstrate that HE4 competes with IFN-γ for IFN-γR binding (Figure 3J), suggesting that in HE4-enriched TMEs, HE4 may substantially contribute to IFN-γR-dependent PD-L1 induction. Together, these findings suggest that, beyond IFN-γ-STAT signaling, tumor-derived factors such as HE4 can dominate PD-L1 regulation and immune evasion in specific TMEs.^14,56

An important question concerns the relative contribution of PD-L1 expressed by myeloid versus tumor cells. Prior studies indicate that PD-L1 on tumor-associated macrophages, monocytes, and DCs—rather than on tumor cells—dominantly suppresses antitumor immunity in the TME.^{29,57,58,59,60,61} Consistent with this, our in vivo analyses show that HE4 regulates PD-L1 primarily within the CD45⁺ compartment: HE4 markedly increases PD-L1 on macrophages (Figure 1O) and also on DCs and MDSCs (Figures S4E–S4F), while exerting minimal effects on monocytes or CD45⁻ tumor cells (Figures S3F–S3G). Thus, HE4 selectively programs myeloid subsets toward a PD-L1^hi immunosuppressive state, explaining the dominant role of myeloid PD-L1 in immune evasion. Accordingly, HE4 blockade reduces myeloid PD-L1 and achieves antitumor efficacy comparable to PD-L1 blockade, suggesting an upstream, tumor-enriched target with potential to limit off-tumor toxicity. The basis of this myeloid selectivity remains unclear and may reflect cell-type-specific IFN-γR signaling thresholds or co-receptor contexts.

Limitations of the study

This study has several limitations. First, our in vivo efficacy and safety assessments relied mainly on syngeneic mouse models and short-term treatment settings, which may not fully capture the complexity and chronic immune effects present in human cancers. Second, while our data support myeloid PD-L1 as the dominant mediator of HE4-driven immune evasion, potential PD-L1-independent or non-immune functions of HE4 in the TME were not extensively explored. Finally, the clinical associations between HE4 expression and response to PD-1 blockade are retrospective and require prospective validation in larger, independent cohorts.

Resource availability

Lead contact

Requests for further information, resources, and reagents should be directed to and will be fulfilled by the lead contact, Chenhui Wang (wangch@uestc.edu.cn).

Materials availability

The materials and reagents used in this study are listed in the key resources table. Reagents generated in our laboratory in this study are available upon request.

Data and code availability

• •The human lung adenocarcinoma (LUAD) single-cell RNA sequencing (scRNA-seq) dataset has been deposited in the NCBI Gene Expression Omnibus (GEO) under accession number GEO: GSE298300. The mouse LLC tumor scRNA-seq dataset is available under accession number GEO: GSE304155. Bulk RNA sequencing data generated from RAW264.7 cells have been deposited under accession numbers GEO: GSE298587 and GSE313133. All datasets are publicly accessible through the NCBI GEO database.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request.

Acknowledgments

We thank Professor Zhou Shengtao (West China Second University Hospital of Sichuan University) for his assistance in establishing an orthotopic mouse model of ovarian cancer. This investigation was supported by grants from the National Science Fund for Distinguished Young Scholars, China (82225029, to C.W.); The Special Project of Natural Science Foundation of China (82441031, to C.W.); the Key Project of the National Natural Science Foundation of China (82430076, to C.W.); the Sichuan Science and Technology Program (2026NSFSC1929, to B.Z.; 2026NSFSC0699, to Y.W.); the General Program of National Natural Science Foundation of China (grant no. 82572023, to R.H.); the Youth Fund of the National Natural Science Foundation of China (82301987, to B.Z.; 82302628, to Y.D.; 82301989, to R.H.; and 82402704, to Y.L.); the National Key Research and Development Program of China (2020YFA0710700, to C.W.); the Postdoctoral Foundation of China (2024M750363, to B.Z.); the Sichuan Postdoctoral Innovation Plan (BX202202, to Y.D.); the Postdoctoral Foundation of Sichuan Provincial People’s Hospital (2022BH018, to L.F.; 2022BH07, to M.Y.); and the Postdoctoral Foundation of Sichuan Province (TB2022086, to R.H.; TB2023092, to L.F.).

Author contributions

B.Z. and Y.Z. performed the experiments with the assistance of P.S., T.P., Y.L., M.Y., R.H., L.F., Z.C., H.W., G.H., Y.D., and X.L. C.C. and S.Z. helped to establish the orthotopic mouse model of ovarian cancer. X.X., F.L., and Q.C. helped to obtain the human cancer samples and evaluated the tumor slides. B.Z. and C.W. designed the experiments and analyzed the data; C.W. wrote the manuscript and supervised the project with B.Z., Y.W., and X.X.

Declaration of interests

The authors have no competing financial interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
HE4 Rabbit mAb	Abcam	Cat# ab200828
Phospho-Jak1 (Tyr1034/1035) Antibody	Cell Signaling Technology	Cat# 3331; RRID: AB_2265057
Phospho-Jak2 (Tyr1007/1008) Rabbit mAb	Cell Signaling Technology	Cat# 3776; RRID: AB_2617123
Phospho-Stat1 (Tyr701) Rabbit mAb	Cell Signaling Technology	Cat# 9167; RRID: AB_561284
Phospho-Stat3 (Tyr705) Rabbit mAb	Cell Signaling Technology	Cat# 9145; RRID: AB_2491009
Mouse PD-L1 Rabbit mAb	Cell Signaling Technology	Cat# 60475; RRID: AB_2924680
Human PD-L1 Rabbit mAb	Cell Signaling Technology	Cat# 15165; RRID: AB_2798734
STAT3 Rabbit mAb	Abclonal	Cat# A19566; RRID: AB_2862671
IFNGR1 Rabbit pAb	Abclonal	Cat#A23652
IFNAR1 Rabbit pAb	Abclonal	Cat# A18594; RRID: AB_2862353
β-actin Rabbit mAb	Abclonal	Cat# AC026; RRID: AB_2768234
HRP-conjugated mice anti FLAG-tag mAb	Abclonal	Cat# AE005; RRID: AB_2770401
HRP-conjugated rabbit anti His-tag mAb	Abclonal	Cat# AE104
IFN-γ Rabbit pAb	Proteintech	Cat# 29788-1-AP; RRID: AB_3086161
HRP-conjugated goat anti-human IgG (H + L)	Proteintech	Cat# SA00001-17; RRID: AB_2890979
HRP-linked goat anti-mouse IgG (H + L) (L152)	ABP Biosciences	Cat# L152
HRP-linked goat anti-rabbit IgG (H + L) (L153)	ABP Biosciences	Cat# L153
Anti-mouse CD8α InVivo (Clone: 2.43)	Selleck Chemicals	Cat# A2102; RRID: AB_3099521
Anti-mouse CSF1R InVivo (Clone: AFS98)	Selleck Chemicals	Cat# A2159; RRID: AB_3722750
Anti-mouse PD-1 InVivo (Clone: RMP1-14)	Selleck Chemicals	Cat# A2122; RRID: AB_3644244
Anti-mouse CTLA-4 InVivo (Clone: 9H10)	Selleck Chemicals	Cat# A2103
Anti-mouse PD-L1 InVivo (Clone: 10 F.9G2)	Selleck Chemicals	Cat# A2115; RRID: AB_3675704
Rat control IgG2a InVivo	BioXcell	Cat# BE0089; RRID: AB_1107769
Mouse control IgG2b InVivo	BioXcell	Cat# BE0086; RRID: AB_1107791
Multi-rAb™ polymer HRP-goat anti-rabbit recombinant secondary antibody (H + L)	ProteinTech	Cat# RGAR011; RRID: AB_3094534
FITC-conjugated anti-human CD45	BioLegend	Cat# 982316; RRID: AB_2876779
FITC-conjugated anti-mouse CD45	BioLegend	Cat# 157214; RRID: AB_2894427
APC-Cy7-conjugated anti-mouse CD3	BioLegend	Cat# 100221; RRID: AB_2057374
PE-conjugated anti-mouse CD8	BioLegend	Cat# 126607; RRID: AB_961300
PerCP/Cyanine5.5-conjugated anti-mouse CD8b	BioLegend	Cat# 126609; RRID: AB_961304
APC-conjugated anti-mouse IFN-γ	BioLegend	Cat# 505809; RRID: AB_315403
PE-Cy7-conjugated anti-mouse CD11c	BioLegend	Cat# 117317; RRID: AB_493569
APC-conjugated anti-mouse MHC-II	BioLegend	Cat# 107614; RRID: AB_313329
PE-conjugated anti-mouse Granzyme B	BioLegend	Cat# 372207; RRID: AB_2687031
FITC-conjugated anti-mouse CD69	BioLegend	Cat# 104505; RRID: AB_313108
APC-Cy7-conjugated anti-mouse CD45	BioLegend	Cat# 103115; RRID: AB_312980
Pacific Blue-conjugated anti-human/mouse CD11b	BioLegend	Cat# 101223; RRID: AB_755985
APC-conjugated anti-mouse F4/80	BioLegend	Cat# 123115; RRID: AB_893493
FITC-conjugated anti-mouse CD206	BioLegend	Cat# 141703; RRID: AB_10900988
PE-conjugated anti-mouse CD86	BioLegend	Cat# 159203; RRID: AB_2832567
APC-conjugated anti-mouse Ly6C	BioLegend	Cat# 128015; RRID: AB_1732087
PE-conjugated anti-mouse F4/80	BioLegend	Cat# 111603; RRID: AB_3082990
Brilliant Violet 605-conjugated anti-mouse PD-L1	BioLegend	Cat# 124321; RRID: AB_2563635
PE-conjugated anti-mouse Ly6G/Ly6C (Gr1)	BioLegend	Cat# 108407; RRID: AB_313372
APC-conjugated anti-human PD-L1	BioLegend	Cat# 329707; RRID: AB_940358
PE-conjugated anti-human IgG Fc	BioLegend	Cat# 410707; RRID: AB_2565785
Alexa Fluor 488-conjugated anti-Flag tag	BioLegend	Cat# 637318; RRID: AB_2810690
Brilliant Violet 605™ Rat IgG2b, κ Isotype Ctrl Antibody	BioLegend	Cat# 400649; RRID: AB_2864282
APC Rat IgG1, κ Isotype Ctrl Antibody	BioLegend	Cat# 400411; RRID: AB_326517
PE Mouse IgG1, κ Isotype Ctrl Antibody	BioLegend	Cat# 400139; RRID: AB_493443
Human TruStain FcX™	BioLegend	Cat# 422301; RRID: AB_2818986
Mouse TruStain FcX™ (anti-mouse CD16/32)	BioLegend	Cat# 101319; RRID: AB_1574973
Bacterial and virus strains
DH5a competent cells	Vazyme	Cat# C502-02
Biological samples
Tissue microarrays of human lung adenocarcinoma	OUTDO BIOTECH	Cat# HLugA150CS04; HLugA060PG03
Tissue microarrays of human colon adenocarcinoma	OUTDO BIOTECH	Cat# HCol-Ade090PG01
Tissue microarrays of human pancreatic adenocarcinomas	OUTDO BIOTECH	Cat# HPanA120CS01
Tissue microarrays of human bladder cancer	OUTDO BIOTECH	Cat# HBlaU060CS02
Fresh LUAD tumor and paired paratumor tissues	Sichuan Provincial People’s Hospital	N/A
Paraffin-embedded NSCLC tumor specimens from patients treated with immuno-therapy, cohort #1	Sichuan Provincial People’s Hospital	N/A
Paraffin-embedded NSCLC tumor specimens from patients treated with immuno-therapy, cohort #2	Tongji Hospital affiliated to Huazhong University of Science and Technology	N/A
Human peripheral blood mononuclear cells	Quicell Biotech	Cat# P1885
Chemicals, peptides, and recombinant proteins
Linear polyethylenimine (25 kDa)	Beyotime	Cat# C0537
Polybrene	Beyotime	Cat# C0351
Penicillin‒streptomycin-amphotericin B solution (100×)	Beyotime	Cat# C0224
Puromycin dihydrochloride	Beyotime	Cat# ST551
Dimethyl sulfoxide	MP Biomedicals	Cat# D8418
Tween 20	BioFroxx	Cat# 1047ML500
DNase I	BioFroxx	Cat# 1121MG010
collagenase IV	BioFroxx	Cat#2091MG100
hyaluronidase Grade I	BioFroxx	Cat# 1140MG100
skim milk powder	BioFroxx	Cat# 1172GR500
Lipopolysaccharides, from E.coli 055:B5	MedChemExpress	Cat# HY-D1056
Phorbol 12-myristate 13-acetate	MedChemExpress	Cat# HY18739
Stattic	MedChemExpress	Cat# HY-13818
Fludarabine	MedChemExpress	Cat# HY-B0069
IMD0354	MedChemExpress	Cat# HY-0172
JNK-IN-8	MedChemExpress	Cat# HY-13319
PD169316	MedChemExpress	Cat# HY-10578
paclitaxel	MedChemExpress	Cat# HY-B0015
ruxolitinib	MedChemExpress	Cat# HY-50856
erlotinib	MedChemExpress	Cat# HY-50896
HIF-1α-IN	MedChemExpress	Cat# HY-115903
Recombinant Fc-tagged mouse IFN-γ protein	MedChemExpress	Cat# HY-P73252
no-tagged mouse IFN-γ protein	MedChemExpress	Cat# HY-P70667
Recombinant mouse IL-4	MedChemExpress	Cat# HY-P701093
recombinant mouse GM-CSF	MedChemExpress	Cat# HY-P7361
Recombinant Fc-tagged human HE4 protein	abinScience	Cat# HC595031
Critical commercial assays
E.Z.N.A.® Total RNA Kit I	Omega Biotek	Cat# D6915-04
Hiscript III reverse kit	Vazyme	Cat# R302-01
SYBR Green qPCR master mix	Vazyme	Cat# Q311
Enhanced chemiluminescent substrate	BioGround	Cat# BG0001
Endo-free plasmid medium kit	Omega Biotek	Cat# D6915-04
FectoPRO transfection reagent	Polyplus-transfection	Cat# 116–040
Mouse IFN-γ ELISA kit	Biolegend	Cat# 430804
Human IFN-γ ELISA kit	Biolegend	Cat# 430104
Mouse TNF-α ELISA kit	Biolegend	Cat# 430904
Human HE4 ELISA kit	Cusabio	Cat# CSB-E12923h
Mouse HE4 ELISA kit	Cusabio	Cat# CSB-EL026092MO
Human Granzyme B ELISA kit	Cusabio	Cat# CSB-E08718h
Mouse cTnI ELISA kit	Solarbio	Cat# SEKM-0153
Mouse Granzyme B ELISA kit	Elabscience	Cat# E-EL-M0594
Mouse FLT3L ELISA kit	Elabscience	Cat# E-EL-M0030
Mouse CRP ELISA kit	Beyotime	Cat# PC186
Mouse IL-6 ELISA kit	Elabscience	Cat# E-EL-M0044
Mouse ALT kit	Elabscience	Cat# E-BC-K235-S
Mouse AST kit	Elabscience	Cat# E-BC-K236-M
Amplex Red Creatinine Assay Kit	Beyotime	Cat# S0291S
Zombie Violet™ Fixable Viability Kit	BioLegend	Cat# 423114
Quick Genotyping Assay Kit for mouse tail samples	Beyotime	Cat# D7283
Cell counting kit-8	MedChemExpress	Cat# HY-K0301
Deposited data
scRNA-seq from primary LUAD patients	This study	GSE298300
scRNA-seq from mouse LLC tumors	This study	GSE304155
Bulk RNA-seq from Raw264.7 treated with Fc, HE4-Fc, or IFN-γ, respectively	This study	GSE298587
Bulk RNA-seq from Raw264.7 treated with HE4-Fc, IFN-γ, or HE4-Fc + IFN-γ	This study	GSE313133
Experimental models: Cell lines
Human: NCI-H358	Cyagen Biosciences	N/A
Human: OVCAR3	ATCC	Cat# HTB161
Human: HEK293T	ATCC	Cat# CRL-11268
Mouse: LLC	ATCC	Cat# CRL-1642
Mouse: ID8	Quicell Biotech	Cat# I451
Mouse: Raw264.7	ATCC	Cat# TIB-71
Human: THP-1	ATCC	Cat# TIB202
Mouse: 4T1	ATCC	Cat# CRL-2539
Mouse: MC38	ATCC	Cat# CRL-2640
Mouse: HGS-1	CancerTools. org	Cat# 160537
Bone marrow-derived macrophages	This study	N/A
Bone marrow-derived dendritic cells	This study	N/A
Experimental models: Organisms/strains
Wild-type C57BL/6JGpt	GemPharmatech	N000013
A/JGpt mice	GemPharmatech	N000018
BALB/cJGPT mice	GemPharmatech	N000020
FVB/NJGpt mice	GemPharmatech	N000026
Cd274+/− mice (C57BL/6JGpt-CD274em3Cd302d12132/Gpt)	GemPharmatech	T009724
Rag1−/− (C57BL/6NCya-Rag1em1/Cya)	Cyagen Biosciences	C001197
C-NKG (NOD. Cg-PrkdcscidIl2rgem1cya/Cya)	Cyagen Biosciences	C001316
Oligonucleotides
Human WFDC2 qPCR F: GCTGTGCCACCTTCTGCTC; R: TCAGAAATTGGGAGTGAC	Tsingke Biotech	N/A
Mouse Wfdc2 qPCR F: CTTGAACCAATTACGGACTG; R: ATTAGGCTTGGAGCAGACA	Tsingke Biotech	N/A
Human CD274 qPCR F: GCTGCACTAATTGTCTATTGGGA; R: AATTCGCTTGTAGTCGGCACC	Tsingke Biotech	N/A
Mouse Cd274 qPCR F: AGCCTGCTGTCACTTGCTAC; R: CCTGCCACAAACTGAATCAC	Tsingke Biotech	N/A
Human GAPDH qPCR F: ACGGGAAGCTTGTCATCA; R: GACTCCACGACGTACTCAGC	Tsingke Biotech	N/A
Mouse Actin qPCR F: GGTCATCACTATTGGCAACG; R: ACGGATGTCAACGTCACACT	Tsingke Biotech	N/A
gRNA-Wfdc2 (mouse)-F1: CACCGACTTTGGACAAGGACTGTG; R1: AAACCACAGTCCTTGTCCAAAGTC	Tsingke Biotech	N/A
gRNA-Wfdc2 (mouse)-F2: CACCGAATGCTGCCGCAATGGATGT; R2: AAACTCATCCATTGCGGCAGCATTC	Tsingke Biotech	N/A
gRNA-Wfdc2 (mouse)-F3: CACCGTTTCTCTGCATCGGTGCCTT; R3: AAACAAGGCACCGATGCAGAGAAAC	Tsingke Biotech	N/A
gRNA-Ifngr1 (mouse)-F1: CACCGATTAGAACATTCGTCGGTAC; R1: AAACGTACCGACGAATGTTCTAATC	Tsingke Biotech	N/A
gRNA-Ifngr1 (mouse)-F2: CACCGGGCTCGGAGAGATTACCCGA; R2: AAACTCGGGTAATCTCTCCGAGCCC	Tsingke Biotech	N/A
gRNA-Ifngr1 (mouse)-F3: CACCGACTTGAACCCTGTCGTATGC; R3: AAACGCATACGACAGGGTTCAAGTC	Tsingke Biotech	N/A
gRNA-Ifnar1 (mouse)-F1: CACCGATGTAGACGTCTATATTCTC; R1: AAACGAGAATATAGACGTCTACATC	Tsingke Biotech	N/A
gRNA-Ifnar1 (mouse)-F2: CACCGATGACAACTACACCCTAAAG; R2: AAACCTTTAGGGTGTAGTTGTCATC	Tsingke Biotech	N/A
gRNA-Ifnar1 (mouse)-F3: CACCGTTCAGCAGAATATCGAACGT; R3: AAACACGTTCGATATTCTGCTGAAC	Tsingke Biotech	N/A
Cd274-KO mice genotyping WT-F: CCACAGGAGACAGTTTGGTGAGAGG; R: GGCTTCCACCACCAAAGTGTTT	Tsingke Biotech	N/A
Cd274-KO mice genotyping KO-F: CTTACTGGGCAATCACTCCATCCC; R: CAGAACAACTGCTATCGAAGA GCC	Tsingke Biotech	N/A
Recombinant DNA
psPAX2	Addgene	Addgene_12260
plentiCRISPRv2	Addgene	Addgene_52961
pVSV-G	Addgene	Addgene_138479
pcDNA3.1-mHE4-Flag	Youbio BioTech	N/A
pcDNA3.1-hHE4-Flag	Youbio BioTech	N/A
pInfuse-hHE4-Fc	This study	N/A
pInfuse-mHE4-Fc	This study	N/A
pLVX-mHE4-Flag	This study	N/A
pInfuse-hHE4-Mut-Fc	This study	N/A
pInfuse-mHE4-Mut-Fc	This study	N/A
pInfuse-mIFNGR1-6×His	This study	N/A
pInfuse-mIFNGR2-6×His	This study	N/A
Software and algorithms
GraphPad Prism version 9.5.1	GraphPad Software	N/A
CFX ManagerTM version6.0	Bio-Rad	N/A
AlphaFold 3	Google DeepMind	N/A
FlowJoV10.8	BDBiosciences	N/A
NoveExpress software	Agilent	N/A

Experimental model and study participant details

Human specimen collection

The protocol for the collection of human LUAD samples was approved by the Ethics Review Committee of Sichuan Provincial People’s Hospital (approval number: 2022-383), and written informed consent was obtained from all patients. Fresh tumor and paired adjacent normal tissues were prospectively collected from patients who had not undergone radio-/chemotherapy or neoadjuvant therapy within the preceding 6 months. Sample collection was conducted immediately following surgical resection without interfering with the standard surgical procedures. Two paired tumor and adjacent normal tissues were dissociated into single-cell suspensions for single-cell RNA sequencing. An additional 16 tumor and paired paratumor tissues were used to assess HE4 expression. Furthermore, paraffin-embedded tumor specimens were retrospectively collected from 44 patients with NSCLC, including 21 lung adenocarcinoma (LUAD) and 23 lung squamous cell carcinoma (LUSC) cases, who were classified as immunotherapy-sensitive-sensitive (n = 27) or -resistant (n = 17). All patients received combined treatment with a PD-1 monoclonal antibody and platinum-based chemotherapy. The PD-1 inhibitors administered included camrelizumab (n = 20), pembrolizumab (n = 1), serplulimab (n = 1), tislelizumab (n = 9), and sintilimab (n = 12). Clinical stage at treatment initiation was as follows: Stage IIa (n = 8), Stage IIb (n = 10), Stage IIIa (n = 22), Stage IIIb (n = 3), and Stage IIIc (n = 1). For post-operative patients (n = 29), treatment sensitivity was defined as the absence of disease progression or recurrence for more than 2 years when immunotherapy was initiated at clinical Stage II, or for more than 1 year when initiated at Stage III; treatment resistance was defined as disease progression or recurrence within 6 months after surgery. For pre-operative patients (n = 55), treatment sensitivity was defined as achieving a partial or complete response, whereas resistance was classified as stable disease (SD) or progressive disease (PD), according to RECIST v1.1 criteria.

Additionally, an NSCLC microarray from patients who were either sensitive or resistant to anti-PD1 antibody plus (or not) chemotherapy was retrospectively provided by Dr. Qian Chu, Tongji Hospital. In these cohorts, a total of 57 patients with NSCLC were enrolled, including 43 cases of LUAD and 14 cases of LUSC. Among these patients, 52 received combination therapy with a PD-1 monoclonal antibody and platinum-based chemotherapy, 4 received PD-1 monoclonal antibody monotherapy, and 1 patient received combination therapy with a PD-L1 monoclonal antibody and platinum-based chemotherapy. Samples were obtained from patients treated in the pre-operative setting. Treatment sensitivity was defined as achieving a partial or complete response (PR; n = 37), whereas treatment resistance was defined as progressive disease (PD; n = 20), according to RECIST v1.1 criteria.⁶²

Animals

The Cd274^−/− and WT littermates used in this study were produced in-house by breeding heterozygous parents (5–9th generations). Genotyping was carried out using both wild-type (WT)–specific and knockout (KO)–specific primer sets designed by GemPharmatech. All the mice used in this study were 6–8 weeks of age and were housed under specific pathogen-free conditions at the animal facility of the School of Medicine of the University of Electronic Science and Technology of China. The mice were maintained at 24 ± 2°C on a 12/12-h light/dark cycle with free access to food and water. Animal housing and experimental procedures were performed in compliance with the guidelines for the care and use of animals, and the protocol was approved by the Animal Experimentation Ethics Committee of Sichuan Provincial People’s Hospital (approval number: 2022-404).

Cell lines and cell culture

HEK293T, LLC, ID8 and Raw264.7 cells were maintained in DMEM supplemented with 10% FBS, 100 U/mL penicillin and 100 μg/mL streptomycin. THP-1, 4T1, MC38, OVCAR3 and NCI-H358 cells were maintained in RPMI-1640 medium supplemented with 10% FBS, 100 U/mL penicillin and 100 μg/mL streptomycin. Bone marrow-derived macrophages (BMDMs) were differentiated as previously described.⁶³ Bone marrow-derived dendritic cells (BMDCs) were prepared by culturing bone marrow cells with GM-CSF (20 ng/mL) and IL-4 (10 ng/mL), as previously described.⁶⁴ All the cells were maintained and amplified in a CO₂ incubator at 37°C with 5% CO₂. The cells were passaged every 2 days 293 F cells were maintained in serum-free FreeStyle 293 Expression Medium and amplified in a shaking incubator (130 rpm) under 5% CO₂ at 37°C. Cell line authentication was performed by short tandem repeat (STR) profiling. All cell lines used in this study are free of mycoplasma contamination.

Method details

Preparation of murine and human recombinant protein

Plasmids containing the full-length complementary DNA of mouse and human HE4 cDNA, pcDNA3.1-mHE4-Flag and pcDNA3.1-hHE4-Flag, were purchased from Youbio BioTech Co. (Changsha, China). The plasmids were amplified in E. coli DH5α cells, purified via an endo-free plasmid medium kit and transfected into 293 F cells via FectoPRO transfection reagent according to the manufacturer’s protocol. After 7–10 days of culture, Flag-tagged recombinant proteins, mHE4-Flag or hHE4-Flag, were affinity purified from the supernatant via anti-Flag-conjugated agarose (#2058ES08, Yeasen BioTech, Shanghai, China). For Fc-tagged protein production, cDNA fragments encoding hHE4 (aa 31–123) or mHE4 (aa 26–174) were amplified via PCR, subsequently subcloned and inserted into the modified pINFUSE-hIgG-Fc vector with an IL-2 signal peptide. Following transfection into 293 F cells as described above, Fc-tagged HE4 protein (HE4-Fc) was affinity purified with protein G resin according to the manufacturer’s protocol (L10029, GenScript BioTech, Naning, China). In addition, mouse IFN-γR1 or IFN-γR2 extracellular fragments were inserted into the vector pcDNA3.1+C-His (pcDNA3.1-Ifngr1-His and pcDNA3.1-Ifngr2-His, respectively). The plasmids were subsequently transfected into 293 F cells. His-tagged IFN-γR1 or IFN-γR2 proteins were affinity purified with Ni-TED resin according to the manufacturer’s protocol (L00885, GenScript BioTech, Nanjing, China). The eluted products were subjected to four rounds of buffer exchange with ice-cold PBS via ultrafiltration and concentrated to the appropriate concentration. The purities of the recombinant proteins were analyzed by Western blot or SDS‒PAGE with Coomassie blue staining. Aliquots were stored at −20°C for subsequent experiments.

Cell stimulation

To determine whether HE4 triggers PD-L1 upregulation, Raw264.7 cells or BMDMs were seeded into 12-well plates at a density of 5 × 10⁵/well. Then, the cells were treated with the indicated concentration of HE4 recombinant protein in Opti-MEM for 3 h (for Western blot and real-time qPCR) or 12 h (for flow cytometry and immunoblotting). To evaluate the effect of HE4 on PD-L1 expression across distinct macrophage subsets, Raw264.7 cells were differentiated to M1-like macrophages with 300 ng/mL LPS plus 20 ng/mL IFN-γ for 24 h, or differentiated to M2-like macrophages with 50 ng/mL IL-4 for 48 h, followed by stimulation with HE4 recombinant protein for 12 h. Human monocyte THP-1 cells were stimulated with 100 ng/mL PMA for 48 h to allow them to differentiate into macrophages before treatment. Pretreatment with the inhibitors was performed for 1 h before HE4 stimulation.

Anti-HE4 antibody production and purification

Rat anti-mouse HE4 and mouse anti-human HE4 monoclonal antibodies were generated by AtaGenix Laboratories Co., Ltd. (Wuhan, China). Briefly, rats (or mice) were immunized with mHE4-Fc (aa26–174) or hHE4-Fc (aa31–123) four times. Single-cell suspensions of immunized splenocytes were subsequently fused with mouse myeloma sp2/0 cells via polyethylene glycol (p7181, Sigma‒Aldrich, St. Louis, MO, USA). The hybridomas were selected in hypoxanthine-aminopterin-thymidine (HAT) medium, followed by two rounds of limiting dilution cloning to establish monoclonal cell lines. The monoclonal hybridoma was maintained in serum-free RPMI-1640 medium for 5 days, and the supernatants were collected and tested via ELISA for antigen specificity (Table S2). HE4-specific hybridomas were maintained in serum-free specific medium (NC0301, YoCon, Beijing, China) and amplified in a CO₂ shaking incubator to produce monoclonal antibodies in large amounts. Following mass cultivation for 7–10 days, the culture media was centrifuged at 2000 rpm for 5 min to remove the cellular debris. The antibodies were isolated from the supernatant via protein G resin and then eluted with 0.1 M glycine-HCl buffer (pH = 2.7). The eluates were ultrafiltered four times with ice-cold 1× PBS and concentrated to the appropriate concentration. SDS‒PAGE and Coomassie blue staining were conducted to evaluate the purity of the eluted products. Aliquots were stored at −20°C for subsequent experiments.

To evaluate whether administering the anti-HE4 mAb caused immune-related adverse events, 10-week-old wild-type male A/J mice were injected with the anti-HE4 mAb (200 μg/mouse, n = 10) intraperitoneally every three days for 21 days. Additionally, as controls, the mice were injected with either a rat IgG2a isotype mAb (n = 10) or an anti-PD-1 mAb (n = 9). During the treatment period, serum was collected from each mouse for ELISA quantification of the levels of IL-6, IFN-γ and the liver enzymes ALT and AST. Tissues, including heart, lung, liver and colon, were harvested for histological evaluation.

Surface plasmon resonance (SPR) analysis

Surface plasmon resonance (SPR) experiments were conducted using a WeSPR 100 system equipped with carboxylated gold sensor chips (G10031), following the manufacturer’s instructions (XLement, Shanghai, China), to quantitatively assess the binding affinities of antigen–antibody and receptor–ligand interactions. Briefly, the sensor chips were activated using a freshly prepared NHS (0.1 M) and EDC (0.4 M) in MES buffer (pH = 6.0) at 37°C for 30 min, followed by thorough washing with ultrapure water and brief drying. For receptor–ligand interaction analyses, recombinant IFNGR1-His was diluted to 30 μg/mL in MES buffer (pH 4.5) and immobilized onto the activated chip surface (6 μL per well) by overnight incubation at 4°C. For antigen–antibody interaction analyses, recombinant HE4-Fc or Fc control proteins were similarly diluted and immobilized. On the following day, residual solution was removed by gentle tapping, and the chip surface was blocked with 150 μL of blocking buffer provided with the kit at 37°C for 1 h. After two washes with dilution buffer, binding analyses were performed using the Affinity Analysis module according to the standard acquisition and fitting workflow. Binding kinetics were analyzed using the WeSPR analysis software.

Lentivirus transduction for gene knockout or overexpression

The full-length complementary DNA of mouse HE4 (NM_026323.2) was inserted into the pCDH-CMV-MCS-EF1-GFP-puro (pCDH-HE4) or pLVX-CMV-puro (pLVX-HE4) vector to generate a lentivirus for ectopic expression. For lentivirus production, the pCDH-HE4 or pLVX-HE4 recombinant plasmid (6 μg), packaging plasmid pAX2 (3 μg), and envelope plasmid VSV-G (3 μg) were preincubated with polyethylenimine (30 μg) in Opti-MEM for 15 min and then added to HEK293T cells (80% density in 6 cm dishes). The culture supernatants were collected 48 h posttransfection, filtered through a 0.22 μm filter and supplemented with 8 μg/mL polybrene before being used to infect LLC and ID8 cells. Empty pCDH-puro or pLVX-puro plasmids served as controls. After 72 h of incubation, uninfected cells were depleted with puromycin (5 μg/mL) for 7 days. The surviving populations were expanded to validate HE4 expression via real-time qPCR and Western blotting.

For gene knockout, the guide RNAs (gRNAs) of target genes were designed via the CHOPCHOP online server (https://chopchop.cbu.uib.no/). Sense and antisense oligonucleotides were annealed by a thermal cycler, phosphorylated with T4 polynucleotide kinase, and ligated into a BsmBI-digested pLenti-CRISPRv2-GFP vector via T4 DNA ligase. Lentivirus production was performed according to the aforementioned protocol. The virus particles were enriched via ultrafiltration, suspended in fresh complete medium containing 8 μg/mL polybrene and added to the target cells. After incubation for 24 h, the medium was replaced with fresh complete medium and then incubated for 48 h. GFP-positive cells were individually sorted by fluorescence-activated cell sorting (FACS) into 96-well plates (i.e., single cells per well). Clonal populations were expanded, and the knockout efficiency of each gene was verified via Western blotting.

Murine syngeneic tumor models

All the animal experiments were performed in strict accordance with the relevant ethical guidelines and were approved by the Department of the Ethics Committee of Sichuan Provincial People’s Hospital, University of Electronic Science and Technology of China. Cancer cells were harvested at the logarithmic growth phase, washed with precooled 1 × PBS and resuspended in saline at an appropriate cell density. Subcutaneous tumor models were established by inoculating MC38 (1 × 10⁶/150 μL) or LLC (2 × 10⁶/150 μL) cells into the right flank of C57BL/6 mice. 4T1 cells (2 × 10⁶/100 μL) were implanted into the mammary fat pads of BALB/c mice to establish an orthotopic breast cancer model. HGS-1 cells (1 × 10⁶ cells in 10 μL per ovary) were bilaterally orthotopically implanted into the ovaries of C57BL/6 mice to establish an ovarian cancer model. The long diameter (L, mm) and short diameter (W, mm) of the tumor tissues were measured with Vernier calipers every two days, and the tumor volume was calculated via the following formula: (L∗W∗W)/2 (mm³). At the end of the experiments, the mice were subjected to humane euthanasia through a standardized procedure as follows: the mice were placed in a designated euthanasia chamber, and CO₂ euthanasia was performed by gradually introducing carbon dioxide into the chamber over 10 min to ensure peaceful transition. Following CO₂ euthanasia, each mouse underwent cervical dislocation to confirm irreversible loss of consciousness and death, according to accepted veterinary guidelines. The tumor tissues were resected and weighed. Tumor volume and weight served as the primary metrics of the subcutaneous and orthotopic tumor models. ID8 cells (2 × 10⁶/200 μL) were intraperitoneally injected into C57BL/6 mice to establish an ovarian cancer peritoneal metastasis model. At the experimental endpoint (approximately 40 days), ascites was harvested and quantified. The ascites volume served as the primary metric of the ID8 peritoneal metastasis model. In accordance with the ethical guidelines, the mice were sacrificed once the tumor volume reached 2 cm³ or if ulcers occurred.

To evaluate the impacts of HE4 expression on tumor growth in vivo, LLC and ID8 cells with HE4 knockout or overexpression were subcutaneously and intraperitoneally inoculated into C57BL/6 mice, respectively. In parallel, the same amounts of LLC and ID8 cells infected with empty vector-loaded lentivirus were incubated with control mice.

To mimic the role of extracellular HE4 fragments in tumor growth, MC38 (1 × 10⁶/150 μL) or HE4-knockout LLC cells (2 × 10⁶/150 μL) were inoculated into the right flank of C57BL/6 mice. HE4-Fc recombinant protein or Fc protein (2.5 mg/kg body weight) was intraperitoneally administered every other day until the experimental endpoint. Notably, HE4-Fc and Fc proteins were administered one day prior to tumor cell incubation in the LLC model or 3 days after inoculation in the MC38 model. The tumor sizes were recorded daily. At the experimental endpoint, intact tumors were excised, weighed, and enzymatically digested into single-cell suspensions for further experiments.

Tumor-bearing mice were administered HE4 monoclonal antibodies (anti-HE4 mAbs) to evaluate the antitumor efficacy of HE4 neutralization in vivo. Antibody administration was initiated beginning on day 4 following inoculation (ID8 or HGS-1 cells), or the volume of subcutaneous (LLC or MC38) or orthotopic (4T1) tumors reached 50–100 mm³. The anti-HE4 mAbs or isotype IgG control (10 mg/kg body weight) were administered via intraperitoneal injection every other day, six times in total. At the experimental endpoint, ascites or tumor tissues were harvested for further experiments.

To evaluate the coordinated antitumor effects of anti-HE4 and anti-CTLA-4 antibodies, LLC tumor–bearing mice were treated once tumor volumes reached approximately 50–100 mm³. Mice received intraperitoneal injections of anti-mHE4 mAb (10 mg/kg) and anti-CTLA-4 mAb (5 mg/kg), administered either alone or in combination.

The synergistic antitumor effects of HE4 neutralization combined with chemotherapy were evaluated. Specifically, when subcutaneous LLC tumors reached volumes ranging from 50 to 100 mm³, anti-HE4 mAbs (5 mg/kg) were intraperitoneally administered every other day for a total of 6 doses. Concurrently, paclitaxel (10 mg/kg) was administered intraperitoneally every 4 days until the experimental endpoint. For the ID8 peritoneal model, anti-HE4 mAbs were administered on day 4 after ID8 cell inoculation, six times in total. Paclitaxel administration was initiated on day 7, three times at weekly intervals. Isotype-matched IgG and vehicle were used as controls for anti-mHE4 mAbs and paclitaxel administration, respectively. Survival outcomes were monitored for each mouse throughout the study period, and Kaplan‒Meier survival curves were generated to analyze treatment efficacy.

CD8⁺ T cells and macrophage depletion in vivo

For the depletion of CD8⁺ T cells in vivo, anti-CD8α antibodies (5 mg/kg body weight) were intraperitoneally administered beginning on the day of LLC inoculation (day 0) and were delivered every other day until the animals were euthanized. Spleens were harvested and dissociated into single-cell suspensions, and the frequency of CD8⁺ T cells was quantified via flow cytometric analysis (CD3⁺/CD8α⁺). To deplete macrophages in vivo, anti-CSF1R antibodies (20 mg/kg body weight) were intraperitoneally administered prior to LLC inoculation for one week, with continued dosing every 3 days throughout the study. Tumor tissues were excised, enzymatically digested into single-cell suspensions and analyzed by flow cytometry to determine the proportions of macrophages (CD45⁺/CD11b⁺/F4/80⁺).

Urethane-induced lung cancer models

Primary non-small cell lung cancer was induced in FVB mice through weekly intraperitoneal injections of urethane (1 mg/g body weight, dissolved in sterile saline). Following six weeks of urethane treatment, the mice were subjected to intraperitoneal administration of either anti-HE4 mAbs or isotype-matched IgG control (10 mg/kg body weight, every other day) 6 times. Antibody administration was performed again from weeks 12–14. At the experimental endpoint (week 18), the mice were subjected to humane euthanasia through a standardized procedure, and the lung tissues were surgically excised. The largest lung lobe was processed for hematoxylin and eosin (H&E) staining, and the number of visible tumor nodules was quantified.

Immunohistochemical staining and scoring

The expression of HE4 in tumor tissues was evaluated via immunohistochemistry. In brief, paraffin-embedded tumor sections were deparaffinized, rehydrated and soaked in sodium citrate-citrate buffer solution (10 mM, pH = 6.0) at 95°C for 30 min. Before being incubated with primary antibody, the tissue sections were sequentially treated with 3% hydrogen peroxide in 1 × PBS for 20 min and blocked with PBS containing 5% goat serum and 0.1% Triton X-100 for 1 h. The slides were then incubated with the indicated antibodies at 4°C overnight, followed by incubation with the multirAb polymer HRP-goat anti-mouse/rabbit recombinant secondary antibody for 1 h at room temperature. Positive signals of the targets were visualized with DAB substrates, and the nuclei were counterstained with hematoxylin. In parallel, the specificity of the primary antibody was assessed by preincubating the samples with 50 μg of recombinant HE4-Fc protein at 4°C for 1 h before they were added to the sections. A semiquantitative scoring system was used to assess the expression level of HE4, as previously described.⁶⁵ Briefly, the stained images were scored by intensity (1: weak; 2, moderate; 3, strong) and percentage of positive cells (1: 0–25%; 2: 26–50%; 3: 51–75%; 4: 76–100%) via a double-blind protocol. The final immunohistochemistry score of each section was calculated as the product of intensity and percentage.

Western blot

For Western blotting, cultured cells and tissues were lysed in SDS‒PAGE sample loading buffer containing 100 μM dithiothreitol (DTT), and the resulting cell lysates were sonicated to reduce viscosity. The soluble proteins in the culture supernatants were precipitated with methanol and chloroform, as previously described.⁶⁶ The resulting pellets were dissolved in SDS‒PAGE sample loading buffer containing DTT. All the samples were boiled for 5 min, and equal amounts of protein were electrophoretically separated and transferred to polyvinylidene difluoride (PVDF) membranes. The membranes were blocked with 5% skim milk in PBS containing 0.1% Tween 20 (PBS-T) for 1 h. After being incubated with the indicated primary antibodies at 4°C overnight, the membranes were washed with PBS-T and incubated with HRP-conjugated secondary antibodies for 1 h at room temperature. Bands were visualized via enhanced chemiluminescent substrate (BG0001, BioGround, Chongqing, China). Images were captured via a chemiluminescence imaging system (Clinx, Shanghai, China) for further analysis.

RNA extraction, cDNA synthesis and real-time quantitative PCR

Following the indicated treatments, the cells were washed twice with ice-cold PBS, and total RNA was extracted via an E.Z.N.A. Total RNA Kit I (D6915-04, Omega Biotek, Norcross, Georgia, USA) according to the manufacturer’s protocol. Equal amounts of RNA were reverse transcribed into cDNA via Hiscript III reverse transcriptase (R302-01, Vazyme, Naning, China). The resulting cDNA was subjected to real-time PCR using SYBR Green qPCR master mix (Q311, Vazyme). Mouse β-actin and human GAPDH were used as internal controls, and the relative expression levels of genes were calculated via the 2^−ΔΔCT method.⁶⁷

Protein–cell binding assay

RAW264.7 cells and PMA-differentiated THP-1 cells (1 × 10⁵) were incubated with 5 μg of HE4-Flag or HE4-Fc protein at 37°C for 2 h, and equal amounts of 3 × Flag or Fc peptides served as negative controls. The cells were then fixed with 2% paraformaldehyde at room temperature for 15 min, followed by incubation with Alexa Fluor-488-conjugated anti-Flag or PE-conjugated anti-Fc secondary antibodies for 30 min on ice. The stained cells were washed once with FACS buffer (5% BSA in PBS) and then subjected to flow cytometry analysis.

In vitro protein–protein binding assay

To investigate whether HE4 directly interacts with IFN-γR1 and IFN-γR2, 2 μg of purified HE4-Flag, IFN-γR1-His or IFN-γR2-His recombinant proteins were diluted in binding buffer (20 mM phosphate, pH 7.4; 200 mM NaCl; 2 mM EDTA; 5% glycerol; 0.2% NP-40; and protease inhibitor cocktail) and incubated at 4°C for 4 h, followed by further incubation with 30 μL of Ni-TED or anti-Flag-conjugated agarose for 2 h. The beads were washed three times with high-salt buffer and then resuspended in SDS‒PAGE loading buffer for immunoblotting analysis.

To determine whether HE4 competes with IFN-γ for binding to IFN-γR1 and IFN-γR2, 2 μg of purified IFN-γR1-His and IFN-γR2-His were incubated with 30 μL of Ni-TED resin at 4°C for 2 h. The resulting mixture was washed three times with high-salt buffer and then incubated with the indicated amounts of recombinant HE4-Flag and IFN-γ proteins for an additional 2 h. After three washes, the resulting resin was resuspended in SDS‒PAGE loading buffer for immunoblotting analysis.

To determine the half-maximal inhibitory concentration (IC₅₀) of anti-mHE4 mAb of clone #117, we performed a competitive ELISA to assess its ability to block the interaction between mHE4-Fc and the receptor IFNGR1. Briefly, recombinant mIFNGR1-His protein (0.2 μg in 100 μL per well) was coated onto high-binding 96-well plates and incubated overnight at 4°C. mHE4-Fc (0.3 μg in 100 μL per well) was preincubated with increasing concentrations of the #117 anti-mHE4 mAb at 4°C for 1 h before being added to the plates, which had been blocked with 5% BSA. After incubation at room temperature for 2 h, an HRP-conjugated anti-Fc antibody was added and incubated for an additional 1 h. Plates were then washed four times with PBST, followed by the addition of TMB substrate. The reaction was stopped with stop solution, and absorbance was measured accordingly.

Flow cytometry

The abundances of T cells, myeloid-derived suppressor cells (MDSCs) and PD-L1-expressing cells within the tumor microenvironment were analyzed via flow cytometry. For subcutaneous or orthotopic tumors, tumor tissues were dissociated mechanically and digested in FACS buffer containing 2 mg/mL collagenase IV and 0.2 mg/mL DNase I at 37°C for 40 min. Digested cells were filtered through 70 μm cell strainers to obtain a single-cell suspension and then washed with ice-cold PBS. For the peritoneal ID8 model, ascites was centrifuged at 1500 rpm for 5 min, and the pellets were washed with ice-cold PBS and then incubated with red blood cell lysis buffer on ice for 5 min. The cells were washed with ice-cold PBS twice and resuspended in FACS buffer for further experiments. The live and dead cells were distinguished via a Zombie Violet Fixable Viability Kit according to the manufacturer’s protocol. The cells were preincubated with TruStain FcX at 4°C for 15 min to block Fc receptors, followed by staining with fluorochrome-conjugated monoclonal antibodies at 4°C for 30 min. Then, the cells were fixed with 2% paraformaldehyde (PFA) at room temperature for 15 min and resuspended in FACS buffer for analysis. For intracellular staining, the cells were incubated with a cell stimulation cocktail (1:1000) for 3 h at 37°C, followed by blocking with TruStain FcX and surface staining. Then, the PFA-fixed cells were permeabilized with Foxp3/transcription factor staining buffer (#00–5523–00, eBioscience) and incubated with the appropriate antibodies on ice for 30 min, followed by flow cytometric analysis.

To assess whether HE4 upregulates the expression of PD-L1. RAW264.7 or PMA-differentiated THP-1 cells were treated with recombinant HE4 protein as indicated. Then, the cells were harvested and washed with ice-cold 1×PBS, followed by staining to distinguish live/dead cells. Then, the cells were stained with a fluorochrome-conjugated anti-PD-L1 antibody on ice for 30 min and fixed with 2% PFA for flow cytometry. Data were acquired via an NovoCyte Quanteon Flow Cytometer System (Agilent) and analyzed with FlowJo V10 software (BD Biosciences).

Single-cell RNA sequencing

Fresh tumor and paired adjacent normal tissues were collected from two LUAD patients immediately after surgical resection. The tissues were rinsed with ice-cold PBS three times to remove blood clots and then minced on ice. The tissue block was incubated with a digestion solution composed of 0.25% trypsin and 10 μg/mL DNase I at 37°C with shaking at 50 rpm for 30 min. The cell suspensions were filtered through a 70-μm cell strainer and then through a 40-μm strainer again, and the filtrate was collected into a new centrifuge tube. The cell pellet was resuspended in 3 mL of red blood cell lysis solution and incubated on ice for 5 min. Then, 10 mL of precooled PBS was added, and the mixture was centrifuged at 250 × g for 5 min. The cell pellet was resuspended in PBS containing 0.5% FBS. The cells were stained with 0.4% AO/PI to evaluate cell viability. 10 × Genomics single-cell capture, Illumina library construction, preprocessing of scRNA-seq data, integration, dimensionality reduction and clustering were performed by Capital Biotech (Beijing, China). In brief, CellRanger software (https://support.10xgenomics.com/single-cell-vdj) was used to align and annotate sequencing reads against the human reference genome assembly GRCh38.p13, with gene annotations derived from GENCODE v34, to generate gene–cell count matrices. Following alignment, barcode information, including cell-identifying barcodes and unique molecular identifiers (UMIs), was processed and analyzed in R (version 3.6.0) using the Seurat R package (version 4.0.0).⁶⁸ Cells whose gene number was less than 200, or gene number ranked in the top 1%, or mitochondrial gene ratio was more than 25% were regarded as abnormal and filtered out. After filtering, high-quality cells were retained for further analysis. PCA was used for dimensionality reduction, and data visualization was performed using t-distributed stochastic neighbor embedding (t-SNE) and uniform manifold approximation and projection (UMAP).

Clusters were annotated to different cell types on the basis of the expression of canonical cell markers (http://117.50.127.228/CellMarker/)⁶⁹ and differential gene expression (DEGs), as previously described.⁷⁰ Single-cell RNA data were visualized via Loupe Browser (v8.1.2), and the expression profiles of genes of interest in different cell clusters were analyzed. InferCNV analysis was performed to stratify epithelial cells (EPCAM⁺) into malignant and non-malignant epithelial subsets, with CD3⁺ T cells were used as the reference normal population, to further analyze the differential expression of genes between malignant and normal epithelial cells. The raw data were submitted to the Gene Expression Omnibus database under accession number GSE298300.

Single-cell RNA sequencing was also performed to assess the impacts of HE4 neutralization on the tumor immune microenvironment. Subcutaneous LLC tumors were harvested on the day after the last administration of the mAbs. Tissues were minced on ice and then digested at 37°C with a shaking speed of 50 rpm or 30 min. Cell suspensions were filtered through a 40-μm strainer, and the cell pellet was resuspended in red blood cell lysis solution and incubated on ice for 5 min. The cell pellet was washed and resuspended in PBS containing 0.5% FBS, followed by the evaluation of cell viability. The samples were then sent to Majorbio Biopharm Technology Co., Ltd. (Shanghai, China), for further sequencing and bioinformatic analysis. Briefly, library construction was performed according to the manufacturer’s protocol (DNBelab C TaiM4 V3.0) and quantified via a high-sensitivity DNA chip on a Bioanalyzer 2100 and a Qubit high-sensitivity DNA assay (Thermo Fisher Scientific). Libraries were sequenced on the MGI DNBSEQ-T7 platform (MGI-Tech, Shenzhen, China). Reads from MGI DNBSEQ-T7 platform were processed using the dnbc4tools v2.1.2 pipeline with default and recommended parameters. FASTQs generated from DNBSEQ sequencing output were aligned to the Mus_musculus genome assembly GRCm38.p6, release M_25, using the STAR algorithm. Next, Gene-barcode matrices were generated for each individual sample by counting UMIs and filtering non-cell associated barcodes. Finally, we generated a gene-barcode matrix containing the barcoded cells and gene expression counts. This output was then imported into the Seurat v4.1.1 R toolkit for quality control and downstream analysis. Particularly, Genes expressed in fewer than three cells, cells with fewer than fifty detected genes, and cells with mitochondrial percentages exceeding 5% were excluded from the further analysis. We annotated the clusters to different cell types on the basis of the expression of canonical cell markers (http://117.50.127.228/CellMarker/).⁶⁹ The raw data were submitted to the Gene Expression Omnibus database under accession number GSE304155.

Bulk RNA sequencing and analysis

Bulk transcriptomic analysis was conducted to compare gene transcription in macrophages after stimulation with HE4 and IFN-γ. In brief, Raw264.7 cells were seeded in 6-well plates at a density of 5 × 10⁵ cells/well and then stimulated with either HE4-Fc (20 μg/mL) or IFN-γ (100 ng/mL) for 3 h. After washing twice with ice-cold PBS, total RNA was extracted and reverse-transcribed into cDNA. Library construction and sequencing were conducted via the MGI DNBSEQ-T7 platform (MGI-Tech, Shenzhen, China) at Majorbio Biopharm Technology Co., Ltd. (Shanghai, China). The sequencing data were analyzed via an online platform (https://cloud.majorbio.com/). Genes upregulated by HE4-Fc (compared with Fc stimulation) or IFN-γ (compared with the vehicle) are listed. Furthermore, KEGG pathway enrichment analysis was performed to systematically compare the signaling pathways modulated by HE4 and IFN-γ. The raw data were submitted to the Gene Expression Omnibus database under accession number GSE298587.

Cell proliferation assay

To investigate whether the extracellular HE4 protein promotes cell proliferation, ID8 or LLC cells, with or without HE4 knockout, were seeded into 96-well plates at a density of 3,000 cells per well. Recombinant HE4 protein or anti-HE4 mAbs were added to the culture medium at the indicated concentrations. At 0, 24, 48, and 72 h posttreatment, CCK-8 reagent was added to each well, and the absorbance at 450 nm was measured following a 1 h incubation at 37°C. The relative proliferation rate of the cells was calculated by normalizing the absorbance values at each time point to the baseline values.

Cytokine detection

At the experimental endpoint, blood was collected from each mouse and centrifuged at 5000 rpm for 5 min to separate the serum. The tumor tissues of LLC, MC38 and 4T1 cells were weighed, minced and then cultured in 24-well plates with 1 mL of serum-free RPMI-1640 medium supplemented with 1% penicillin‒streptomycin for 24 h. Culture supernatants were collected after centrifugation at 5000 rpm for 5 min and immediately stored at −80°C. The measured concentrations of each cytokine in the tumor culture supernatants were normalized to the weight of the tumor tissue (pg/g tumor). For the peritoneal ID8 model, ascites was quantified and centrifuged at 1500 rpm for 5 min, after which the supernatants were collected. IFN-γ, IL-6, TNF-α and HE4 levels in the serum, culture supernatant and ascites were measured via commercial ELISA kits according to the manufacturer’s instructions. For human cytokine measurement, after a single-cell suspension of LUAD was incubated with HE4-Flag or anti-HE4 mAbs for 48 h, the culture supernatant was collected and diluted 5-fold for ELISA. Mouse IFN-γ (#430804), human IFN-γ (#430104), and mouse TNF-α (#430904) ELISA kits were purchased from BioLegend. Mouse HE4 (CSB-EL026092MO), human HE4 (CSB-E12923h), and human granzyme B (CSB-E08718h) ELISA kits were purchased from Cusabio (Wuhan, China). Mouse CTnI ELISA kit (SEKM-0153) was purchased from Solarbio. Mouse granzyme B ELISA kit (E-EL-M0594) and FMS-like tyrosine kinase 3 ligand (FLT3L) ELISA kit (E-EL-M0030) were purchased from Elabscience (Wuhan, China). Mouse CRP ELISA kit (PC186) and Amplex Red creatinine assay kit (S0291S) was purchased from Beyotime. Serum alanine aminotransferase (ALT/GPT) and aspartate aminotransferase (AST/GOT) levels were measured with activity assay kits (E-BC-K235-S, E-BC-K236-M) according to standard protocols (Elabscience).

Prediction of protein‒protein interactions via AlphaFold 3

Protein structure prediction was conducted via the AlphaFold 3 web server (https://alphafoldserver.com/). Briefly, to investigate the binding mode and specific interaction interfaces between HE4 and the receptors IFNGR1 and IFNGR2, the mature mouse HE4 protein (aa26–174), and the extracellular domains of IFNGR1 (aa26–254) and IFNGR2 (aa20–262) were used to initiate the prediction. In addition, to further examine the specificity of the anti-HE4 monoclonal antibody, the variable heavy (VH) and variable light (VL) regions of the anti-mHE4 mAb (clone #117) and anti-hHE4 mAb (clone #10) were sequenced, and the full-length HE4 protein together with the VH and VL sequences were subjected to structure prediction using AlphaFold3. A predicted template modeling (pTM) score greater than 0.5 indicated high confidence in the overall structural model, whereas an interface PTM (ipTM) score greater than 0.6 suggested high reliability of the predicted protein interaction interface. The prediction results were downloaded and subjected to further analysis via PyMOL software.

In PyMOL, the “model_0.cif” file was loaded, and protein‒protein interaction sites were analyzed through a series of commands. The analysis involved three main steps: labeling different protein chains, identifying potential interaction regions between proteins, and selecting residues within a 4 Å interaction distance for visualization.

Preparation of human cancer sample suspensions and treatment

Fresh tumor tissues from patients with LUAD were minced and digested in FACS buffer containing 2 mg/mL collagenase IV and 0.2 mg/mL DNase I at 37°C for 40 min. The digested cells were filtered through 70 μm cell strainers and then resuspended in 10% RPMI-1640 medium for further experiments. The cells were treated with the indicated concentrations of recombinant HE4 protein (hHE4-Flag) or anti-hHE4 mAbs at 37°C for 48 h. Then, the cells were stained with a fluorochrome-conjugated anti-PD-L1 antibody for flow cytometry analysis, and IFN-γ and granzyme B in the culture supernatants were detected with an ELISA kit.

Pharmacokinetic analysis of anti-hHE4 mAb

To model the pharmacokinetic behavior of fully-humanized anti-hHE4 antibody in vivo, C57BL/6 mice (female, 6–8 weeks old) were administered a single dose of anti-mHE4 mAb of clone #10 via tail vein injection at 10 mg/kg body weight. Blood samples were collected at the indicated time points post-injection (0, 0.08, 0.25, 0.5, 1, 2, 4, 8, 12, 24, 72, 120, 168 h, 336 h and 504 h) through retro-orbital bleeding under isoflurane anesthesia. Blood samples were allowed to clot at room temperature for 30 min and were then centrifuged at 3,000 × g for 10 min to obtain serum, which was aliquoted and stored at −80°C until analysis. Serum concentrations of the anti-hHE4 mAb were quantified by enzyme-linked immunosorbent assay (ELISA). Briefly, high-binding 96-well plates were coated overnight at 4°C with recombinant hHE4-Fc protein (0.2 μg per well). After blocking with 5% BSA in PBS, diluted serum samples (1:10000) were added and incubated for 2 h at room temperature. The bound antibodies were detected using an HRP-conjugated anti-mouse IgG secondary antibody, followed by TMB substrate development and absorbance measurement at 450 nm. Antibody concentrations were calculated based on a standard curve generated using purified anti-hHE4 mAb of clone #10.

Pharmacokinetic parameters, including maximum serum concentration (C_max), elimination half-life (t_1/2), and mean residence time (MRT), were estimated using non-compartmental analysis with GraphPad Prism or Phoenix WinNonlin software.

Immune system humanized mouse tumor models

The PBMC-humanized xenograft mouse models were established as previously described to evaluate the antitumor efficiency of HE4 neutralization preclinically.⁷¹ In brief, after one week of adaptation to the environment, c-NKG mice (female, 8 weeks old) were subcutaneously inoculated in the right flank with a tumor cell-PBMC suspension containing 5 × 10⁶ NCI-H358 (or OVCAR3) cells and 3.0 × 10⁵ PBMCs. Beginning on day 3 after inoculation, tumor-bearing c-NKG mice were peritoneally administered 10 mg/kg/mouse anti-hHE4 mAbs or the IgG2a isotype control every other day, four times in total. On the day after the last administration, the mice were subjected to humane euthanasia through a standardized procedure, and the tumor tissues were resected, weighed and processed for flow cytometric analysis.

Bioinformatics analysis

Here, we used Gene Expression Profiling Interactive Analysis (GEPIA), an interactive online server that integrates gene expression profiling data from the Cancer Genome Atlas (TCGA) and Genotype-Tissue Expression (GTEx) projects, to analyze and verify the gene expression of HE4 in LUAD, ovarian cancer, colon adenocarcinoma, pancreatic adenocarcinomas and bladder cancer (http://gepia2.cancer-pku.cn/index.html).⁷² The clinical response of LUAD patients with high or low HE4 expression to immune checkpoint inhibitors was predicted via the tumor immune dysfunction and exclusion (TIDE) score, a computational framework developed to evaluate the potential for tumor immune escape from the gene expression profiles of cancer samples (http://tide.dfci.harvard.edu/query/).⁷³

Quantification and statistical analysis

In vitro experiments were performed at least three times independently. The animal experiments were independently repeated at least two times. The data are expressed as the mean ± standard deviation (SD) and were analyzed via GraphPad Prism 9.0 (GraphPad Software Inc., San Diego, CA, USA). The statistical significance of multiple comparisons was analyzed via one-way analysis of variance (ANOVA) followed by the Bonferroni post hoc correction, and a two-tailed Student’s t test was used for two-group comparisons. When multiple groups with a second variable, e.g., over time, were compared, two-way ANOVA followed by the Sidak post hoc test was performed. Survival curves were compared via the log rank test (Mantel‒Cox test). p < 0.05 was considered statistically significant.

Published: March 19, 2026