The research team, led by Noémie Brassard-Jollive, Yoann Atlas, and senior authors Laurent Muller and Catherine Monnot at the Center for Interdisciplinary Research in Biology at the Collège de France, together with collaborators across Paris institutions including the Hôpital Saint-Louis pathology network, began by examining actual human ccRCC specimens at high resolution. Clear cell renal cell carcinoma is the most common form of kidney cancer and is driven overwhelmingly by loss of the von Hippel-Lindau (VHL) tumor suppressor. VHL protein normally tags hypoxia-inducible factors for destruction; without it, hypoxia signaling runs constitutively, flooding the tumor with vascular endothelial growth factor (VEGF) and other pro-angiogenic programs. The consequence is the striking hypervascularization familiar to any pathologist. But VHL loss also promotes epithelial-to-mesenchymal transition (EMT), a phenotypic switch in which epithelial tumor cells acquire invasive, migratory, fibroblast-like traits, and this connection between invasiveness and vessel architecture turned out to be central to the new findings.

Within the patient samples, the investigators identified two vascular structures that are morphologically distinct from ordinary tumor capillaries. They named them ponds and sheets. High-resolution three-dimensional imaging of optically cleared patient-derived xenografts, essentially fragments of human tumor grown in mice and rendered transparent so that confocal microscopy could reconstruct the full vascular volume, revealed the striking geometry of these structures. Ponds appeared as large, irregular, reservoir-like formations with wide luminal cavities, vessels dilated far beyond anything resembling normal capillary caliber. Sheets, by contrast, were thin, elongated ribbons that appeared partially collapsed, as if the endothelial tubes had been flattened or incompletely organized. Neither resembles the narrow, regularly branching capillaries that pathologists count when they measure microvessel density, a metric whose prognostic value in ccRCC has proven disappointingly limited. The implication is that vessel quantity matters less than vessel architecture, and that the architecture has been largely invisible to conventional two-dimensional histology.

Capturing these structures in patients was only the first step. The team then engineered an in vitro model capable of reproducing them, because no existing laboratory system had recapitulated ccRCC vascular morphology faithfully enough to support drug testing. Their solution was a three-dimensional co-culture in which EMT-like tumor spheroids, compact spheres of tumor cells exhibiting the mesenchymal phenotype associated with VHL loss and invasion, are embedded in a dense collagen matrix alongside human endothelial cells. Collagen hydrogels of defined fibrillar density provide a physiologically relevant extracellular scaffold, and as the spheroids invade outward and the endothelial cells assemble into capillary-like networks, the two compartments interact. Crucially, the co-culture spontaneously generated structures mimicking pond architecture, whereas endothelial cells cultured alone formed conventional capillary networks. The tumor cells, in their invasive, EMT-like state, were actively instructing the formation of aberrant vascular geometry.

Time-lapse live imaging made it possible to watch this process unfold and revealed a temporal coupling that the authors describe as a link between tumor invasion and pond morphogenesis. The aberrant dilated structures did not arise independently of tumor behavior; they emerged in concert with the outward movement of invasive tumor cells, suggesting that the mechanical and biochemical traffic between invading spheroids and adjacent endothelium drives vessels into these pathological configurations. This observation ties together two features of ccRCC biology that are usually studied separately: the EMT-driven invasive program of the tumor cells and the angiogenic program of the vasculature. In this model they are two faces of a single coordinated behavior, orchestrated in part by the VHL-deficient tumor microenvironment.

The pharmacological payoff came when the team exposed the system to clinically relevant targeted therapies: temsirolimus, an mTOR inhibitor; crizotinib, a tyrosine kinase inhibitor; and sunitinib, the multi-target VEGF receptor blocker that has long anchored first-line treatment of metastatic ccRCC. In monoculture, meaning endothelial cells alone forming capillaries, and in tumor spheroid monoculture, the drugs impaired capillary morphogenesis and tumor invasion to varying degrees, consistent with known clinical activity. But when the co-culture was treated, overall drug efficacy dropped. The presence of both cell types together created a microenvironment in which the same compounds that worked well on isolated compartments became substantially less effective, an in vitro echo of the resistance that frustrates oncologists in the clinic.

The most provocative finding concerned sunitinib specifically. When the researchers analyzed the drug’s effects structure by structure within the co-culture, the pond-like formations showed markedly reduced sensitivity compared with the adjacent, more conventional capillaries. In other words, a single drug at a single concentration could silence one vascular compartment while leaving another largely intact, within the same three-dimensional culture. If ponds behave this way in patients, they would provide persistent vascular supply through the course of therapy, functioning as a structural reservoir of resistance even when standard capillaries regress. This offers a mechanistic candidate for the well-documented phenomenon of anti-angiogenic resistance in ccRCC, which has previously been attributed mostly to compensatory signaling pathways, hypoxic adaptation, and alternative pro-angiogenic factors rather than to the physical heterogeneity of the vessels themselves.

The study also carries a methodological message for the field. Two-dimensional assays, endothelial tube formation on flat Matrigel and microvessel density counts on tissue sections, dominate angiogenesis research, and the authors argue that these approaches fundamentally cannot detect structures like ponds and sheets, whose defining features are volumetric. The consensus is already shifting; recent volumetric imaging studies of ccRCC microvasculature have shown that three-dimensional vascular patterns correlate with tumor genomics and patient outcomes where density measurements fail. By combining optically cleared xenografts with a tractable in vitro co-culture that reproduces key pathological structures, the Paris team has supplied a platform on which candidate therapies can be evaluated not just for whether they reduce vessel number but for whether they eliminate the specific vascular geometries associated with invasion and drug failure.

Clinically, the identification of ponds as candidate contributors to treatment failure suggests several concrete directions. If pond formation is driven by EMT-like tumor cells, then targeting the invasive phenotype itself, or the tumor-endothelial signals that mediate pond morphogenesis, might sensitize these structures to VEGF blockade. Combination regimens pairing VEGF inhibitors with agents active against EMT-associated or mTOR-dependent pathways could be re-evaluated in the co-culture model before committing patients. And because the platform is built from human cells in a defined matrix, it could eventually support personalized testing, growing a patient’s own tumor cells in spheroid-endothelial co-culture to predict which targeted regimen will best dismantle the individual tumor’s vascular architecture.

For now, the immediate contribution is conceptual: ccRCC vasculature is not a uniform target but a heterogeneous collection of structures, some of which are intrinsically harder to drug than others, and the road to durable responses may run through the ones that pathologists’ slides have been flattening into invisibility. The study, published as volume 29, article 56 of Angiogenesis in August 2026, reorients attention from how many vessels a tumor has to what kind of vessels it has, and in doing so turns a laboratory curiosity, a dilated cavity in a collagen gel, into a plausible answer to one of kidney cancer’s most stubborn clinical questions.

The research integrates transcriptomic profiles and clinical data from over 1,500 gastric cancer patients spanning seven independent cohorts, creating a robust molecular classification system founded on glycogene expression patterns. This glycosylation-based molecular signature delineates distinct patient clusters associated with differential tumor stage, recurrence risk, and survival outcomes. Among the 12 critical glycogenes identified, GXYLT2 emerged as the most potent prognostic indicator, heralding its potential clinical utility.

Statistical analyses revealed a compelling correlation between elevated GXYLT2 expression and advanced tumor staging in gastric cancer, alongside significantly poorer overall survival and disease-free survival rates. Intriguingly, stratification by histological subtype demonstrates that GXYLT2 expression is markedly enriched in diffuse-type GC compared to the intestinal subtype, reinforcing the gene’s subtype-specific oncogenic relevance. Immunohistochemical staining substantiated these findings, showing strong GXYLT2 protein expression in diffuse gastric tumors, thus underscoring its viability as a translational biomarker.

Beyond correlative evidence, functional assays decisively illustrate that GXYLT2 actively promotes malignant phenotypes. Genetic silencing of GXYLT2 in diffuse-type gastric cancer cell lines led to a profound reduction in proliferative capacity, invasive potential, and sphere-forming ability—hallmarks of cancer aggressiveness and stemness. Conversely, overexpression experiments confirmed that while GXYLT2 alone is insufficient to induce oncogenic traits in intestinal-type GC models, its presence within the diffuse-type cellular context is essential to sustain malignancy, suggesting a context-dependent oncogenic network.

Mechanistically, this study elucidates the molecular pathway through which GXYLT2 exerts its oncogenic function. GXYLT2 was found to potentiate the Wnt/β-catenin signaling pathway, a critical regulator of cellular proliferation and differentiation frequently deregulated in cancers. Loss of GXYLT2 function boosts phosphorylation of β-catenin—a modification that targets the protein for degradation—thereby diminishing its nuclear accumulation and decreasing transcription of downstream Wnt-responsive genes pivotal for tumor progression.

A key intermediary in this regulatory cascade is protein phosphatase 2A (PP2A), whose activity is suppressed by GXYLT2. The inhibition of PP2A ensures sustained activation of β-catenin signaling, facilitating tumor growth and invasion. Notably, restoring PP2A activity following GXYLT2 knockdown effectively counteracted Wnt pathway hyperactivation, providing a mechanistic rationale for targeting this axis therapeutically in diffuse-type gastric cancer.

The translational relevance of these molecular insights was confirmed through rigorous in vivo experimentation. Xenograft models demonstrated that GXYLT2 silencing markedly curtailed tumor growth and cell proliferation within diffuse GC contexts, cementing the gene’s role as a driver of tumor aggressiveness and a viable candidate for targeted intervention. These preclinical findings pave the way for further development of GXYLT2-focused diagnostic tools and therapeutic strategies.

Collectively, this study establishes GXYLT2 as a bifunctional molecule integral to both the clinical prognosis and the biological underpinnings of diffuse-type gastric cancer. By harnessing large-scale cohort data and cutting-edge molecular biology techniques, the researchers have defined a glycosylation-centric classification system that enhances patient stratification and opens new therapeutic avenues. Targeting GXYLT2 or its downstream effectors in the Wnt/β-catenin pathway may revolutionize treatment paradigms and improve outcomes for patients afflicted by this aggressive cancer subtype.

The discovery of GXYLT2’s role underscores the broader significance of glycosylation modifications in cancer biology, an area that has historically been underexplored. Aberrant glycosylation can profoundly influence cell signaling, adhesion, and immune evasion—factors instrumental in tumor development and metastasis. This study’s glycosylation-focused molecular classification thus not only serves as a prognostic tool but also generates new hypotheses regarding the molecular drivers of gastric cancer heterogeneity.

Moreover, the context-dependent effects of GXYLT2 highlight the complexity of oncogenic networks within tumor subtypes. The inability of GXYLT2 overexpression to drive aggressiveness in intestinal-type GC suggests that co-occurring molecular alterations or specific cellular environments are requisite for its oncogenic activity. This finding advocates for precision medicine approaches that consider tumor subtype and molecular context when designing targeted therapies.

Future research may explore combinatorial strategies incorporating GXYLT2 inhibition with agents modulating Wnt signaling or PP2A activity, capitalizing on the mechanistic insights uncovered. Additionally, the glycosylation signature identified here may serve as a foundation for the development of novel biomarkers to aid early diagnosis, prognostication, and treatment monitoring in gastric cancer.

In summary, this meticulously conducted and translationally relevant investigation firmly positions GXYLT2 as a critical biomarker and effector in the pathogenesis of diffuse-type gastric cancer. Advancing our understanding of glycosylation-mediated regulation and Wnt/β-catenin signaling in malignancy could ultimately lead to more effective, personalized interventions, offering hope to patients battling this formidable disease.

Subject of Research: Molecular characterization and functional analysis of GXYLT2 in diffuse-type gastric cancer

Article Title: Identification of glycogene-based molecular classification and correlations between the expression levels of 12-glycogene signature and molecular features in GC patients

News Publication Date: Not specified

Web References:

• Genes & Diseases Journal
• DOI: 10.1016/j.gendis.2025.101673

Image Credits: Jiale Yang, Jiajun Wu, Ziqiang Chen, Xiangyun Hou, Xiaojing Li, Zhaorui Liu, Kai Yin, Tao Pang, Ruimin Huang, Jun Yan

Keywords: Gastric cancer, GXYLT2, glycosylation, Wnt/β-catenin signaling, diffuse-type gastric cancer, prognostic biomarker, tumor aggressiveness, molecular classification