Core Hypothesis

Integrating spatially resolved metabolomics with epigenomic profiling reveals that local accumulation of L‑glutamine at the invasive edge and dopamine in mesenchymal‑like zones directly modulates histone acetylation and DNA methylation patterns, creating a reversible epigenetic state that sustains therapy‑induced persistence. Targeting this metabolite‑epigenetic axis with enzyme inhibitors (e.g., glutaminase blockers) combined with epigenetic drugs (e.g., HDAC inhibitors) should prevent the emergence of resistance when administered adaptively guided by longitudinal multi‑omic VAE embeddings.

Mechanistic Rationale

• Metabolite‑Epigenetic Coupling: L‑glutamine fuels α‑ketoglutarate production, a cofactor for JumonjiC domain histone demethylases and TET DNA demethylases. Elevated glutamine therefore shifts the balance toward a more open chromatin state at stress‑response loci. Conversely, dopamine can inhibit histone deacetylases via catechol‑mediated oxidative modification, increasing acetylation of promoters governing epithelial‑to‑mesenchymal transition (EMT).
• Spatial Heterogeneity: Spatial multi‑omics has already mapped glutamine enrichment in SOX4+ edge cells and dopamine in GPNMB+ mesenchymal‑like glioblastoma subpopulations [Spatial multi-omics identifies locoregional metabolite regulators like L-glutamine in SOX4+ edge cells and DL-dopamine in GPNMB+ mesenchymal-like cells]. This creates distinct epigenetic niches that are invisible to bulk or single‑cell omics lacking spatial context.
• Dynamic Evolution: Longitudinal multi‑omic tracking shows that resistance-associated transcriptional programs emerge after metabolic shifts, suggesting a causal sequence where metabolite changes precede epigenetic remodeling [Longitudinal integration reveals therapy effects and biomarkers missed in cross-sectional single-omics].

Testable Predictions

1. Correlation: In patient‑derived xenograft models treated with standard chemotherapy, regions with high glutamine/dopamine (measured by MALDI‑imaging or spatial metabolomics) will show increased H3K27ac and reduced 5‑mC at promoters of EMT and drug‑efflux genes (measured by spatial CUT&Tag or ATAC‑seq).
2. Interference: Pharmacological reduction of glutamine (using CB‑839) or dopamine synthesis (using α‑methyl‑para‑tyrosine) will attenuate the epigenetic changes described above and delay the appearance of resistant clones.
3. VAE‑Guided Adaptive Dosing: A variational autoencoder trained on longitudinal spatial multi‑omics (metabolite, epigenome, transcriptome) will predict impending resistance with >80% accuracy when fed weekly liquid biopsy metabolites and circulating tumor DNA methylation signatures. Adjusting drug combinations based on VAE‑low‑dimensional alerts will significantly extend progression‑free survival compared to fixed‑schedule controls.

Experimental Design

• Model Systems: Orthotopic glioblastoma PDX and patient‑derived organoids with CRISPR‑tagged glutamine transporters and dopamine biosensors.
• Sampling Scheme: Baseline, then every 3 days for 4 weeks post‑treatment; collect tissue for spatial metabolomics (DESI‑MRI), epigenomics (spatial CUT&Tag for H3K27ac, 5‑mC), and transcriptomics (10x Visium). Parallel plasma draws for metabolomic and cfDNA methylation profiling.
• Computational Pipeline: Raw multimodal data fed into a multimodal VAE that imputes missing modalities, corrects batch effects, and yields a joint latent space. Change‑point detection on latent trajectories triggers adaptive therapy switches.
• Endpoints: Primary – time to radiographic progression; secondary – magnitude of metabolite‑epigenetic correlation, frequency of resistant clonal expansions measured by barcode sequencing.

Falsifiability

If spatial glutamine/dopamine levels do not predict local epigenetic states, or if manipulating these metabolites fails to alter the epigenetic landscape or delay resistance, the hypothesis is refuted. Likewise, if the VAE‑guided adaptive strategy shows no survival benefit over standard of care, the predictive utility claim is falsified.

Broader Impact

Confirming this hypothesis would establish a mechanistic bridge between microenvironmental metabolites and epigenetic plasticity, providing a rationale for spatially informed, temporally adaptive combination therapies that pre‑empt resistance rather than react to it.