Infante, S. | Search bioRxiv

bioRxiv2026

ATF4 programs proline-dependent immune evasion in β-Catenin-driven hepatocellular carcinoma

Background & AimsHepatocellular carcinoma (HCC) frequently exhibits resistance to immune checkpoint inhibitors (ICIs), particularly in {beta} -catenin-driven tumors characterized by immune exclusion. While the Unfolded Protein Response (UPR) and the Integrated Stress Responses (ISR) enable tumor adaptation to metabolic stress their role in shaping tumor immunogenicity remains incompletely understood. We investigated whether ATF4, a central effector of the integrated stress response, couples metabolic reprogramming to suppression of anti-tumor immunity in HCC. MethodsWe combined transcriptomic analyses across three independent human HCC cohorts with mechanistic studies using an immunotherapy-resistant MYC/{beta}-catenin-driven murine HCC model. We integrated CRISPR/Cas9-mediated deletion of Atf4 with RNA-sequencing and targeted metabolomics. The impact of tumor-derived metabolites on macrophage differentiation and polarization was evaluated using primary bone marrow-derived cells. Therapeutic responses were evaluated in orthotopic and subcutaneous models treated with anti-PD-1 and anti-VEGFA. ResultsATF4 and XBP1 transcriptional signatures are selectively enriched in human HCC and associate with poor prognosis, vascular invasion, and an immunosuppressive myeloid-enriched tumor microenvironment. Genetic ablation of Atf4 markedly suppressed tumor growth in immunocompetent but not immunodeficient hosts, establishing a requirement for immune-mediated tumor control. Mechanistically, Atf4 loss downregulated Aldh18a1 and disrupted proline biosynthesis, resulting in extracellular proline depletion. This proline-deficient environment abrogated monocyte-to-macrophage differentiation and decreased M2 polarization, thereby reshaping the tumor microenvironment toward enhanced T cell infiltration and activation. Functionally, Atf4-deficient tumors exhibited restored sensitivity to anti-PD-1 monotherapy and showed pronounced responses to combined anti-PD-1/anti-VEGFA treatment in aggressive orthotopic models. ConclusionATF4 programs a proline-dependent metabolic axis that sustains macrophage-mediated immunosuppression and immune evasion in {beta}-catenin-driven HCC. Disruption of this pathway converts immune-excluded tumors into T cell-inflamed states and restores responsiveness to immunotherapy. By governing proline homeostasis and macrophage-mediated immunosuppression, ATF4 is a key metabolic checkpoint for immune evasion, linking stress adaptation to immune escape and a candidate therapeutic target in HCC. Impact and implicationsWe identify ATF4 as a crucial metabolic-immune orchestrator that sustains myeloid-driven immune evasion in {beta}-catenin-dependent HCC through proline-dependent circuitry. Disrupting the ATF4-proline axis converts immune-desert tumors into T cell-inflamed lesions by blocking macrophage differentiation, thereby sensitizing tumors to immune checkpoint therapy. This work positions ATF4 as a tractable therapeutic target to overcome immunotherapy resistance in HCC. Graphical abstract Highlights- ATF4 orchestrates an immunosuppressive tumor microenvironment in HCC by coupling metabolic stress adaptation to immune evasion. - Ablation of ATF4 disrupts proline biosynthesis, leading to a marked depletion of extracellular proline. - Cancer cell-derived proline availability contributes to macrophage differentiation and M2 polarization; its loss restores T cell-mediated anti-tumor surveillance and sensitizes beta-catenin-driven HCC to immune checkpoint blockade.

cancer biology↗

bioRxiv2026

Predicting Obstetric and Non-obstetric Diagnoses Co-occurrences during Pregnancy

Pregnancy care often involves simultaneous obstetric and other medical conditions, but their co-occurrence patterns are rarely modeled explicitly in a systematic, network-based approach. In this work, we formulate obstetric and non-obstetric diagnoses co-occurrences as a link prediction problem on a diagnosis-level homogeneous graph constructed from pregnancy encounters. Diagnoses are represented as nodes connected by co-occurrence edges, with node features capturing graph structure and demographic statistics3. We address this challenge by leveraging collected electronic health records data and study several standalone and hybrid graph neural network (GNN) architectures, including GCN, GAT, GraphSAGE, and three hybrid encoders that combine complementary aggregation mechanisms, namely GCN+GraphSAGE, GCN+GAT, and GAT+GraphSAGE. All models used consistent train-validation-test splits and are evaluated on 5- fold cross-validation sets. Among standalone models, GraphSAGE achieved the strongest performance, whereas hybrid GraphSAGE-based models (GCN+GraphSAGE and GAT+GraphSAGE) are best performers. The GCN+GraphSAGE hybrid, reaching an AUROC and AUPRC of approximately 0.90, consistently outperformed all other architectures. Further analysis of top-ranked predicted links revealed clinically plausible associations between pregnancy stage and risk-related diagnoses and common endocrine, metabolic, and hematological conditions. These findings indicate that graph-based link prediction may effectively prioritize obstetric and non-obstetric diagnosis pairs, providing a scalable framework for identifying clinically meaningful comorbidity patterns. They may further support hypothesis generation and downstream obstetric risk stratification efforts. AvailabilityAll codes including data preparation scripts, training and validation recipes, and experimental configurations are available at: https://github.com/kabir-ai2bio-lab/ob-nonob-diagnoses-cooccurrences.

bioinformatics↗