A study that combined gene activity, metabolite profiles and tissue imaging found a pattern in triple-negative breast cancer tumors from patients with poor metastatic outcomes. These tumors showed stronger epithelial-to-mesenchymal transition, or EMT, WNT and angiogenesis programs, shifts in several metabolites, fewer estimated CD8-positive T cells and B cells, and greater physical separation between malignant cells and those immune cells. The findings suggest that the tumor and its surrounding tissue differed together, rather than in a single molecular signal. They remain associations, not evidence that any one feature caused the cancer to spread.
The comparison behind the findings
The comparison involved 52 patients treated at a single cancer center. Thirty-nine were in the group that remained metastasis-free, while 13 were in the MetaPoor group, defined as patients whose disease metastasized and was followed by metastasis-related death within 36 months of the metastatic diagnosis. The cohort was assembled between November 2016 and August 2017, and median follow-up exceeded 10 years. Because the groups were assigned retrospectively by outcome, the design can show which features travelled with metastatic failure, not what caused it.
Researchers compared the groups using transcriptomic analysis, LC-MS/MS metabolomics, imaging mass cytometry and spatial proteomics. The aim was to see whether signals from tumor genes, metabolism and the tissue environment pointed in the same direction.
A signal in genes, then in metabolism
At the gene-expression level, the MetaPoor tumors were enriched for EMT, WNT and angiogenesis programs, with EMT the strongest hallmark signal in the study's gene-set analysis. The researchers also used xCell, a computational estimate of cell-type fractions, and found lower estimated B-cell and CD8-positive T-cell infiltration alongside higher estimated proportions of M2 macrophages, fibroblasts and endothelial cells. In plain terms, the poor-outcome tumors carried both a more EMT-oriented molecular profile and a different mix of immune and supporting cells.
The metabolite results added a second layer to the comparison. GPC and adenine were higher in MetaPoor tumors. Valine, isoleucine and leucine were lower, as were the arginine-related metabolites N-methyl-L-arginine and SDMA-ADMA, along with taurine. The result describes an outcome-associated metabolic pattern, not evidence that changing any of these metabolites would change the course of disease.
Where the cells sat
Imaging mass cytometry, or IMC, measured 35 protein markers while preserving their location in the tissue. Its analysis identified 11 tumor-microenvironment populations. Compared with the metastasis-free group, MetaPoor tumors had fewer CD8-positive T cells and B cells, more cancer-associated fibroblasts and M2 macrophages, and greater distances between malignant cells and CD8-positive T cells or B cells. Endothelial cells were closer to stromal cells but farther from neutrophils. These spatial differences give the molecular pattern a physical setting inside the tumor.
Another spatial analysis separated seven cellular populations and three marker-defined tumor-cell subpopulations. Stem-like and EMT-like epithelial cells made up a larger share of the MetaPoor tumors. The EMT-like cells had higher levels of Vimentin, CD44, HADHA, ACADM, CD98, p4EBP1, PS6, HIF1α, CA9 and GLUT1, and lower levels of ATP5A, CYTC, CS and IKB. That combination brought HADHA, CD98 and ATP5A into focus as candidate proteins for follow-up.
Some signals tracked survival
The two approaches used to estimate the tissue environment also broadly agreed. Immune-cell fractions from xCell and IMC were positively correlated for CD8-positive T cells, B cells, M2 macrophages, fibroblasts and endothelial cells. The reported correlation was strongest for fibroblasts, at 0.567, followed by B cells at 0.506; both had p-values below 0.001. For CD8-positive T cells, the correlation was 0.422 with p = 0.002. A correlation here measures whether two estimates move together, not whether one causes the other.
Several molecular signals also tracked overall survival. Isoleucine, valine-norvaline and N-methyl-L-arginine had positive correlations with survival, while GPC had a negative correlation. Among proteins, HADHA, CD98 and CD44 were negatively correlated with survival, while ATP5A was positive. The reported coefficients included -0.608 for HADHA and -0.511 for CD98, both with p-values below 0.001, and 0.415 for ATP5A with p = 0.002. The study did not provide an absolute risk estimate for an individual patient.
An external TNBC survival analysis provided a further check for two proteins. High HADHA expression was associated with shorter overall survival, with a hazard ratio of 1.40 and a log-rank p-value of 0.030. A hazard ratio compares the event rate between groups over follow-up. High CD98 showed a similar poorer-survival trend, but its association was not statistically significant, with a hazard ratio of 1.27 and p = 0.12. Immunohistochemistry on additional sections from the original cohort also found higher HADHA and CD98 and lower ATP5A in MetaPoor tumors.
What the evidence can support
The caution is the size and design. Only 13 patients were in the MetaPoor subgroup. The study discussion estimated statistical power at about 64% for HADHA and 44% for CD98, using a two-sided significance level of 0.05, and described the correlations as exploratory. The cohort was retrospective and single-center, and the metabolomic and IMC comparisons used exploratory p-values without multiple-testing correction. The IMC work also relied on targeted antibody panels and one 500 by 500 micrometre region per sample, limiting its spatial and marker coverage.
The external cohort details and IHC sample count were not reported in the reviewed analysis. That leaves the survival check useful as supporting evidence, but not a substitute for testing the findings in larger independent groups.
The results leave GPC, the branched-chain and arginine-related metabolites, HADHA, CD98 and ATP5A as candidates for independent validation and mechanistic study. Larger cohorts will be needed to test whether the combined molecular and spatial pattern can predict metastatic relapse prospectively. For now, the study offers leads about what metastatic TNBC tumors may look like, rather than a validated biomarker or treatment strategy.
Disclosure
Funding was reported from the Beijing Natural Science Foundation, the National Natural Science Foundation of China and the Heilongjiang Provincial Department of Science and Technology. The authors declared no conflicts of interest.
Paper data and sources
Original title: Multi‐Omics Integration Reveals Metabolic Reprogramming and EMT‐Driven Metastasis in Triple‐Negative Breast Cancer
Authors: Pang Y, Shang Y, Wang W et al.
Journal/Repository: Cancer science
Status: Peer-reviewed
First online: 2026-10-11
DOI: Not available
Original paper · Full text