Search bioRxiv⌕ Search
Home/Archive/bioRxiv

bioRxiv · 10.1101/2025.09.30.679631

Cancer-associated fibroblasts drive metabolic heterogeneity in KRAS-mutant colorectal cancer cells

Elton, E.·Tavakoli, N.·Cetin, H.·Finley, S. D.
View original source ↗Export metadata ↓

Abstract

KRAS-mutant colorectal cancer (CRC) is characterized by metabolic reprogramming that can lead to tumor progression and drug resistance. The tumor microenvironment (TME) plays a pivotal role in modulating these metabolic adaptations. In particular, cancer-associated fibroblasts (CAFs), which make up a large portion of the TME, have been shown to strongly contribute to metabolic reprogramming in CRC. This study applies flux sampling, a computational method that explores the full range of feasible metabolic states, combined with representation learning and hierarchical clustering, to a computational model of central carbon metabolism to understand how CAFs influence metabolic adaptations of KRAS-mutant CRC cells following targeted enzyme knockdowns. Focusing on twelve key enzymes involved in glycolysis and the pentose phosphate pathway, knockdowns were simulated under both normal CRC media and CAF-conditioned media (CCM) conditions. Analysis revealed that CCM induces greater metabolic heterogeneity, with knockdown models exhibiting more variable and distinct metabolic states compared to those cultured in normal CRC media. While some enzyme knockdowns produced similar metabolic states, this overlap was less frequent in CCM, indicating that CAF-derived factors diversify the metabolic responses of CRC cells to enzyme perturbations. Pathway-level flux analysis demonstrated media-specific shifts in central carbon metabolism pathways. Importantly, the predicted biomass flux showed that enzyme knockdowns reduced growth across both conditions, but models in the CCM condition indicated CAFs could offer a protective effect against metabolic perturbation. Overall, this study reveals that CCM significantly influences the metabolic state and adaptability of KRAS-mutant CRC cells to enzyme perturbations, emphasizing the importance of including TME components in metabolic modeling and therapeutic development. These findings provide valuable insights into the metabolic adaptability of CRC and suggest that targeting tumor-CAF metabolic interactions may improve treatment strategies. Graphical Abstract Overview of computational workflowModels of interest represent simulated enzyme knockdowns in central carbon metabolism. Flux sampling searches the entire metabolic solution space and results in a distribution of flux values for each reaction within each model. Samples can be organized by knockdown and condition into matrices for input into representation learning. Representation learning is applied to sampling data to identify shared and independent metabolic states. Metabolic states indicate a heterogeneous response to enzyme knockdowns. Overlap of dark and light blue flux distributions, sampling clusters, and metabolic responses exemplify a shared metabolic state separate from to the gray unperturbed state. This workflow provides a low-dimensional representation of metabolic state that captures both the pathway- and reaction-level differences that describe each simulated knockdown. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=172 SRC="FIGDIR/small/679631v1_ufig1.gif" ALT="Figure 1"> View larger version (34K): org.highwire.dtl.DTLVardef@cb6226org.highwire.dtl.DTLVardef@98d94eorg.highwire.dtl.DTLVardef@e2b20aorg.highwire.dtl.DTLVardef@116cfcd_HPS_FORMAT_FIGEXP M_FIG C_FIG

Explore related subjects

systems biologyPreprint

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Elton, E., Tavakoli, N., Cetin, H., Finley, S. D.. 2025-10-01. Cancer-associated fibroblasts drive metabolic heterogeneity in KRAS-mutant colorectal cancer cells. https://doi.org/10.1101/2025.09.30.679631

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

SpaReg: sparsity-based 3D reconstruction of tissue microenvironments at native resolution across morphological and spatial molecular modalities

Tissue microenvironments comprise cellular and acellular components whose three-dimensional (3D) architecture guides disease fate. Direct imaging of intact specimens by light-sheet and multiphoton microscopy, and computational reconstruction from serial sections, have established that 3D spatial context reveals cell and tissue organization inaccessible at single planes. Computational reconstruction in particular can leverage archived human tissue, benefiting from the cost-effectiveness, robustness, scalable storage, workflow compatibility, and century-long pathobiology knowledge of histology, and can integrate multiple spatial modalities. However, sectioning can introduce tears and folds, and computational alignment can further distort tissue integrity. Here we introduce SpaReg, a sparsity-based 3D reconstruction method spanning histology, spatial proteomics and spatial transcriptomics. Across multiple organs, SpaReg robustly reconstructs large tissue volumes with preserved subcellular morphology despite sectioning artifacts. On a standardized histology benchmark, SpaReg achieves the best balance between 3D reconstruction accuracy and tissue integrity, and on spatial transcriptomics benchmarks it ranks among the leading methods while scaling to hundreds of sections and millions of cells in a dataset that several existing methods fail to process. Preservation of subcellular morphology by SpaReg also enables training of a Hematoxylin and Eosin (H&E)-based epithelial, T and B cell classifier, generating single-cell-resolved 3D maps directly from H&E. Applied to pancreatic tissue containing pancreatic ductal adenocarcinoma arising from an intraductal papillary mucinous neoplasm, these maps reveal that 2D sections overestimate immune exclusion, and resolve lymphoid aggregates in 3D. SpaReg, therefore, provides a scalable foundation for morphologically faithful, multimodal 3D atlases and spatially informed disease modeling

systems biology↗
bioRxiv2026

TxCyto: A machine learning framework for estimating cytokine activity from whole transcriptome

Cytokines are critical mediators of intercellular communication, and a comprehensive characterization of their activity is essential for understanding health and disease. Existing tools to infer cytokine activity rely on experimental measurements. However, such measurements are available only for a small minority (43) of cytokines, and moreover, cytokine activity and response are highly context-specific, making a comprehensive experimental profiling across tissues, disease states, and biological contexts impractical. To address this gap, we developed TxCyto - a deep learning-based framework that infers the activity of cytokines, and more broadly of the tumor secretome, directly from the whole transcriptome profile of a sample. Trained on pan-cancer TCGA tumor transcriptomes, TxCyto was extensively validated in multiple independent datasets, including cytokine perturbation experiments. Across multiple cancer immunotherapy cohorts, TxCyto identified cytokines whose predicted activity was associated with therapeutic response. Furthermore, in spatial transcriptomic data for Liver cancer, TxCyto discovered spatial niches associated with response to immunotherapy. Overall, we develop a machine learning tool -TxCyto, for predicting the activity of 645 cytokines and tumor secretome from readily available whole transcriptomes. The TxCyto framework is generally applicable to other classes of regulatory molecules and TxCyto code base, and the tools are provided at https://github.com/Rahulncbs/TxCyto.

systems biology↗
bioRxiv2026

Interpretable machine learning coupled to gene regulatory networks uncovers subcircuits underlying cell fate decisions

Gene regulatory networks (GRNs) model causal linkages that control cell fate decisions and differentiation transitions. Prioritizing regulatory subnetworks underlying cell state differences is of critical importance, but current methods including those reliant on topological metrics introduce circularity as the metrics prioritizing TFs are computed from the same networks whose assumptions they inherit. Separately, interpretable machine learning methods can identify latent factors (LFs) that discriminate cellular states with formal statistical guarantees but do not model regulatory linkages. Here, we present FOCAL (Factor-Outcome Coupling for Assessment of Linkages), a paradigm to prioritize regulatory subnetworks by coupling state-specific and dynamic GRNs with outcome-supervised LFs learned using interpretable machine learning without reference to network topology. This shifts GRN focus from macroscopic TF nodes to state-specific and dynamic TF-gene linkages. In B and T cells, FOCAL identified GIFs (GRNs coupled to Interpretable latent Factors), prioritized regulatory subnetworks underlying established states as well as transient regulatory episodes preceding them. By coupling LFs learnt from perturbation experiments of lineage-defining TFs, FOCAL identified transcriptional predisposition to alternative fates within progenitor cell populations before overt differentiation. This uncovered a novel NFATC2-IRF8 interplay in activated B cells, that was validated by in-vitro and in-vivo genetic perturbations. The two transcription factors act cooperatively to restrain extrafollicular plasmablast differentiation and promote germinal center B cell fate.

systems biology↗