Integrated Spatial and Bulk Untargeted Metabolomics Characterize Location- and Histology-Associated Metabolic Heterogeneity in Colorectal Cancer

Abstract

Colorectal cancer (CRC) exhibits profound molecular heterogeneity according to primary tumor location, but spatially resolved assessment is limited by interindividual variability. Here, to minimize interindividual genetic confounding, we performed an exploratory single-case spatial metabolomics study (MALDI-MSI) on a rare case of synchronous bilateral tumors from one patient. This was complemented by untargeted metabolomics (UHPLC-HRMS/MS) on tumor and matched normal tissues from a cohort of 30 CRC patients with different anatomical sites and histologies. As an observational finding in this single case, the left-sided colon cancer (LCC) showed distinct metabolite distribution features from the tumor core to the invasive margin, whereas in right-sided colon cancer (RCC), metabolic differences were more closely associated with a histological subtype. UHPLC-HRMS/MS showed that, among tubular adenocarcinomas, LCC exhibited enrichments in ether lipid and phosphatidylcholine metabolism, whereas RCC preferred nucleotide and fatty acid metabolism. Furthermore, right-sided mucinous adenocarcinomas displayed unique sphingolipid alterations, consistent with a possible desaturase-associated shift from apoptosis-promoting pathways toward structural functions compared with the tubular subtype. Collectively, these findings underscore the potential influence of primary tumor location and histology on a microenvironmental metabolic architecture, offering exploratory insights for future hypothesis testing on location- and histology-dependent tumor adaptation.

1. Introduction

Colorectal cancer (CRC) ranks as the third most prevalent malignancy globally, accounting for approximately 10% of all new cancer cases and a leading contributor to cancer-related mortality. Despite substantial advancements in multimodal therapeutic regimens, encompassing surgery, chemotherapy, and radiotherapy, achieving satisfactory long-term survival outcomes in patients with advanced CRC remains a formidable challenge, with the 5 year overall survival rate for stage IV disease still being below 20%. The prognosis and clinical management of CRC are governed by multiple factors, with tumor location emerging as a particularly critical determinant. − Specifically, left-sided colon cancer (LCC) and right-sided colon cancer (RCC) exhibit distinct clinicopathological and molecular profiles, underscoring the imperative need for subtype-specific research and tailored interventions.

Embryologically, the colon is divided into two functionally and anatomically distinct segments: the right colon, derived from the midgut and comprising the cecum, ascending colon, and hepatic flexure, and the left colon, originating from the hindgut and encompassing the descending colon and sigmoid colon. , At the molecular level, RCC is more commonly diploid and characterized by mucinous histology, high microsatellite instability, CpG island methylation, and BRAF mutations. Conversely, LCCs more frequently harbor TP53 and KRAS mutations. Clinically, a study integrating overall survival data from 1437 846 CRC patients across 66 published studies revealed that patients with LCC had a significantly 19% lower risk of death compared to those with right-sided tumors. Beyond genetic alterations, metabolic reprogramming has emerged as a core hallmark of cancer progression, enabling tumor cells to rewire metabolic pathways to meet the elevated demands for energy, biosynthetic precursors, and redox homeostasis during malignant transformation. Key metabolic adaptations include the Warburg effect, enhanced de novo lipid synthesis, and altered amino acid metabolism, all of which contribute to tumor proliferation, invasion, and chemoresistance. For CRC, accumulating evidence points to location-specific metabolic signatures: RCC is characterized by striking perturbations in bile acid metabolism, while LCC exhibits markedly elevated carnitine levels, a hallmark of enhanced fatty acid oxidation (FAO). Reflecting these inherent differences, the NCCN guidelines recommend EGFR inhibitors (e.g., cetuximab and panitumumab) primarily for LCC, highlighting the clinical relevance of tumor laterality. However, how these location-specific metabolic programs are organized within the spatial architecture of the tumor microenvironment remains surprisingly understudied, and their potential correlation with localized histopathological features has not been elucidated.

Spatial metabolomics, a pivotal subfield of metabolomics, has advanced rapidly and furnished transformative insights into biomedical research. , As a cutting-edge approach, it integrates high-resolution mass spectrometry imaging (e.g., MALDI-2 imaging) and spatially resolved sequencing to map metabolite distributions across tissue microenvironments, achieving a pixel resolution as low as 10 μm. This addresses key limitations of conventional metabolomics, which inherently lacks spatial information and fails to resolve regional metabolic heterogeneity. In contrast, spatial metabolomics enables micron-scale metabolic profiling by pinpointing metabolic hotspots and correlating spatiotemporal metabolic phenotypes with histopathological features. Moreover, it supports the tracking of dynamic metabolic reprogramming (e.g., progression of the Warburg effect) via time-series sampling and deciphers the functional roles and interactomes of metabolites in situ. − These capabilities are invaluable for unraveling tumor metabolic heterogeneity.

Recent breakthroughs have revolutionized our understanding of tumor metabolic plasticity and therapy resistance. Spatial metabolomics integrated with multiomics has been successfully applied to head and neck tumors to uncover spatiotemporal drivers of progression and has also been highlighted as a powerful approach for dissecting metabolic reprogramming-related drug resistance. Notably, this spatial context-preserving capability makes it uniquely suited to dissect location-specific metabolic signatures in CRC research, addressing a longstanding gap in conventional metabolomic studies. However, a major yet underappreciated axis of CRC metabolic heterogeneity is its morphological diversity. While tumor laterality introduces a macroscopic gradient, distinct histological subtypessuch as tubular and mucinous adenocarcinomasfrequently coexist or exhibit vastly different clinical behaviors. How these microregional histological variances cross-talk with anatomical location to shape the local metabolic architecture remains fundamentally unclarified, largely because traditional bulk assays homogenize these discrete histological components.

Furthermore, extracting true location- and histology-dependent metabolic principles from standard multipatient cohorts remains profoundly challenged by interindividual confounding variables. Heterogeneity in gut microbiota composition, immune histories, and baseline genetic backgrounds often introduces massive noise that can obscure or distort genuine anatomically driven metabolic adaptations, masking the actual variations between LCC and RCC. ,

To control for host-specific genetic and environmental confounders, we performed high-resolution spatial metabolomics (MALDI-MSI) to spatially delineate the metabolic architecture of a rare case: a 47 year-old female patient presenting with synchronous bilateral primary CRC tumors exhibiting equivalent TNM staging. As both lesions coexisted within the same systemic environment, this paired design effectively controlled for confounding variables arising from host genetics, immune history, and gut microbiota composition, thereby facilitating exploratory interrogation of localized, space-dependent metabolic architecture at the microregional level. To mitigate the generalizability limitations inherent to a single-case study, complement these microregional findings, and robustly evaluate histology-associated variations, we sequentially expanded our framework to a macroscopic tissue level. Specifically, we performed bulk untargeted metabolomic profiling (UHPLC-HRMS/MS) on an independent validation cohort of 30 paired CRC patient samples rigorously stratified by both anatomical site and histopathological subtypes.

Conclusively, this study does not attempt to define a universal spatial atlas. Instead, by bridging the high-resolution spatial profiling of a rare bilateral case with a multipatient bulk validation cohort, we characterize the location- and histology-associated metabolic heterogeneity in CRC. This integrative approach allows us to identify a convergent set of metabolic reprogramming characteristics across pathway levels, thereby generating robust metabolic hypotheses regarding location- and histology-dependent tumor adaptation that can be rigorously tested in future large-scale spatial cohorts.

2. Materials and Methods

2.1. Patients

Specimens of synchronous bilateral primary tumors from the left and right colon of the same patient were collected from Tianjin Medical University Cancer Hospital and Institute. Over a 3 year period (2020–2023), a total of 1987 CRC patients who underwent surgical resection at our institution were retrospectively screened. Among these, a small number of cases of synchronous bilateral primary CRC (left and right colon) were identified via pathological and radiological assessments; however, only one patient met our strict inclusion criterianamely, no preoperative neoadjuvant therapy (chemotherapy, radiotherapy, or targeted therapy) and matched tumor-node-metastasis (TNM) stages for both left- and right-sided lesions, making this case suitable for our comparative spatial metabolomics analysis. Specifically, this patient was a 47 year-old female who underwent curative resection. Spatial metabolomics analysis was performed on the cancer tissues derived from both the left- and right-sided colon of this patient. In addition, 30 pairs of matched CRC samples were sourced from left-sided tubular adenocarcinomas, right-sided tubular adenocarcinomas, and right-sided mucinous adenocarcinomas, along with their corresponding normal colon tissues. Detailed pathological information for the synchronous bilateral case is provided in Table S1, and sample-level pathological and molecular data for the UHPLC-HRMS/MS cohort are provided in Table S2.

This study was approved by the Ethics Committee of Tianjin Medical University Cancer Institute and Hospital, with all participants prior to sample collection.

2.2. Sample Preparation

For spatial metabolomes, fresh tumor tissues were fixed in 10% neutral buffered formalin, paraffin-embedded (FFPE), and stored at room temperature. For frozen sectioning, tissues were hydrated with three drops of distilled water prior to cryostat mounting. Tissue sections (12 μm thickness) were prepared using a Leica CM1950 cryostat at −20 °C (Leica Microsystems GmbH, Wetzlar, Germany). Afterward, sections were mounted onto indium tin oxide (ITO)-coated conductive slides and vacuum-desiccated for 30 min.

2.3. Matrix Coating

Desiccated tissue sections on ITO slides were coated with 15 mg/mL 2,5-dihydroxybenzoic acid dissolved in 90% acetonitrile/10% water (v/v) using a matrix sprayer. The sprayer was operated at 60 °C with a flow rate of 0.12 mL/min and a nebulizer pressure of 5 psi. Thirty coating cycles were applied, with a 5 s drying interval between cycles.

2.4. Mass Spectrometry Imaging

MALDI timsTOF metabolomics standards initiative (MSI) experiments were conducted on a timsTOF flex mass spectrometer (Bruker Daltonics, Bremen, Germany) equipped with a 10 kHz SmartBeam 3D laser. Laser power was calibrated to 80% irradiance and maintained constant throughout the entire data acquisition. Data were acquired in the positive ion mode across an m/z range of 50–1300. The spatial resolution was set to 50 μm, with 400 laser shots summed per pixel.

To communicate the confidence level of our molecular assignments, metabolite annotations for the MALDI-MSI data were rigorously classified according to the MSI guidelines. Given the inherent absence of chromatographic retention time separation in direct tissue imaging, all assigned spatial features were classified at level 2 (putatively annotated compounds).

Specifically, tentative identities were established via high-resolution accurate mass matching (m/z tolerance of <5 ppm) against reference patterns within the Human Metabolome Database (HMDB) and METLIN. To further improve the annotation confidence of significantly altered metabolites, structural assignment was validated by on-tissue collision-induced dissociation MS/MS fragmentation on the timsTOF flex platform or cross-verified against fragmentation patterns from our in-house spectral library. All tissue sections were analyzed under the same instrumental settings. Detailed metabolite identification results, including feature tables, metabolite annotations, and Supporting Information for metabolite assignment, are provided in Table S3.

2.5. Untargeted Metabolomics

Colon tissue samples were homogenized in 80% aqueous methanol, followed by centrifugation. The supernatant was evaporated and reconstituted in 50% acetonitrile with internal standards for UHPLC-HRMS/MS analysis. The chromatographic separation was performed on an Ultimate 3000 UHPLC system using either a C18 column for reversed-phase liquid chromatography or an Amide column for hydrophilic interaction liquid chromatography, both operating under optimized linear gradient elution profiles with water and acetonitrile (both containing 0.1% formic acid) as mobile phases. Metabolite detection was carried out using a Thermo Fisher Q Exactive mass spectrometer equipped with an electrospray ionization source, operating in both positive and negative ion modes via data-dependent target MS/MS (dd-MS²) acquisition. Internal standards such as decanoyl-l-carnitine-d3 and acetylcholine-d9 were sourced from C/D/N Isotopes, whereas reference compounds used for metabolite annotation and construction of an in-house library were purchased from various suppliers, including Sigma-Aldrich and Macklin Biochemical. LC–MS grade solvents were obtained from ANPEL Laboratory Technologies.

Prior to formal analysis, quality control evaluation was performed using six representative samples together with pooled QC samples. To ensure high reporting standards consistent with the MSI guidelines, metabolite features were classified into strict confidence tiers. In the positive ion mode (ESI⁺), a total of 7393 molecular features were initially detected (after excluding internal standards), among which 4397 features possessed MS/MS (MS²) spectral information. For the negative ion mode (ESI^–), 2373 features were detected, with 1657 retaining MS² data. To guarantee the fidelity of annotation across the data set, potential in-source fragments were algorithmically removed from the detected features. The untargeted metabolomics profiling was conducted using the standardized analytical platform of Shanghai ProfLeader Biotech Co., Ltd. Spectral matching was then executed against an integration of their rigorously validated, proprietary local in-house libraries alongside public databases, including mzCloud, HMDB, and MassBank. Mass tolerances for precursor (MS1) and product (MS2) ions were tightly restricted to 5 and 10 ppm, respectively, with a minimum MS/MS similarity threshold set at 70%. To facilitate metabolite traceability and external verification, all annotated metabolites and their corresponding HMDB identifiers have been systematically summarized in Table S4. Furthermore, putatively annotated candidates were manually curated based on their physicochemical properties, and any features with chromatographic retention times conflicting with known separation behaviors on the respective columns were strictly excluded.

Through this rigorous multitier verification workflow, a total of 171 metabolites (in ESI⁺) and 174 metabolites (in ESI^–) were robustly annotated across both ionization modes. Level 1 confidence (confirmed by authentic physical standards, matching MS¹ and MS² spectra, and retention times) was achieved for 67 metabolites in ESI⁺ and 90 metabolites in ESI^–. Features lacking available physical standards but yielding consensus mirror-matches against validated databases were classified as Level 2 confidence (putatively annotated compounds), representing 104 metabolites in ESI⁺ and 84 metabolites in ESI^–. Detailed metabolite identification results, including feature tables, MSI confidence tiers, metabolite annotations, and Supporting Information for metabolite assignment, are provided in Table S4. Furthermore, to guarantee absolute transparency and reproducibility, the complete raw and processed data sets, along with detailed metadata, m/z tolerances, and data processing protocols, have been deposited and made publicly accessible in the MetaboLights repository (https://www.ebi.ac.uk/metabolights/MTBLS13729).

2.6. Hematoxylin and Eosin (H&E) Staining

Tissue specimens were fixed in 10% neutral buffered formalin (Sigma-Aldrich, HT501128) for 24–48 h at 4 °C before processing through a graded ethanol series (70%, 95%, and 100%) and xylene immersion. Paraffin-embedded blocks were sectioned at 4 μm thickness using a rotary microtome (Leica RM2235). For hematoxylin and eosin (H&E) staining, dewaxed sections underwent sequential staining in Mayer’s hematoxylin solution (Thermo Fisher, 72704) for 8 min, differentiated in 0.5% acid alcohol, and blued in Scott’s tap water substitute. Counterstaining was performed in eosin Y solution (0.5% aqueous with phloxine; Sigma-Aldrich, HT110316) for 90 s. Stained sections were dehydrated through ethanol gradients, cleared in xylene, and mounted with synthetic resin (Thermo Scientific, 8311-4). Histopathological evaluation was conducted by board-certified pathologists using bright-field microscopy (Olympus BX53) with NanoZoomer Digital Pathology imaging software (version 2.6.13).

2.7. Bioinformatics Analysis

The correlation between the expression of PPAT, GGT1, adenosine kinase (ADK), DAGLB, sphingosine kinase 1 (SPHK1), SLC22A5, and phospholipase C gamma 1 (PLCG1) and prognosis of colon cancer patients from TCGA COAD data was analyzed using the online in silico tool KM plotter (https://kmplot.com/analysis/). All data sets used in the analysis are indicated in the figures.

2.8. Statistical Analysis

For MALDI-MSI, hierarchical cluster analysis (HCA) results of samples and metabolites were presented as heatmaps with dendrograms, while Pearson correlation coefficients (PCC) between samples were calculated by the cor function in R (version 4.4.2) and visualized as heatmaps. Both HCA and PCC were carried out using the R package ComplexHeatmap. For HCA, normalized signal intensities of metabolites (unit variance scaling) were visualized as a color spectrum. For two-group microregional comparisons, due to the unique single-patient clinical context (n = 1), spatial differential features were determined using a stringent threshold combining an absolute |log₂ FC|≥1.0 (representing a strict minimum 2-fold reciprocal intensity shift) with a high-intensity signal-to-noise background suppression filter to exclude low-abundance matrix noise. These spatially resolved candidate features were subsequently subjected to pathway-level cross-verification using the multipatient bulk cohort. Identified metabolites were annotated using the KEGG Compound database (http://www.kegg.jp/kegg/compound/), and the annotated metabolites were then mapped to the KEGG Pathway database (http://www.kegg.jp/kegg/pathway.html). Pathways to which significantly regulated metabolites were mapped were then subjected to metabolomics set enrichment analysis, with their significance determined by the p-values from the hypergeometric test. In all the analyses of this metabolomics study, two main methods were used for data processing during the analysis, which are defined as follows:

• (1)Unit Variance Scaling (UV)

UV scaling (also termed Z-score normalization or autoscaling) standardizes the data set to a mean of 0 and a standard deviation of 1 according to the formula: x′ = (x – μ)/σ

where μ is the mean and σ is the standard deviation.

• (2)Zero-Centered (Ctr)

Centering is achieved by subtracting the variable mean from the original data points: x′ = x – μ.

For UHPLC-HRMS/MS, data were processed with Compound Discoverer software (version 3.3, Thermo Fisher Scientific), and statistical analysis was conducted using both univariate and multivariate methods. For multivariate statistical analysis, the normalized data were preprocessed by Pareto scaling and mean centering before performing OPLS-DA. The model quality is described by the R ² Y and Q ² values. R ² Y is defined as the proportion of variance in the data explained by the model, indicating the goodness of fit. Q ² represents the predictive ability of the model, estimated via a 7-fold cross-validation. Additionally, a 200-times permutation test was performed to evaluate the risk of model overfitting.

For univariate statistical analysis, the normalized data were analyzed in R (version 4.4.2), and data normality was assessed using the Shapiro–Wilk test to accommodate potentially unequal variances. Parametric Welch’s t-test was reserved exclusively for features satisfying the Shapiro–Wilk normality criteria (p > 0.05) to accommodate potential variance inequalities without inflating error rates, while nonparametric tests were performed on non-normally distributed data using the Wilcoxon Mann–Whitney test. For pairwise comparisons, fold-change values were calculated using the latter group in each comparison as the reference group. Specifically, fold changes were calculated relative to LCC in the RCC versus LCC comparison, relative to Ltu in the Rtu (Right-sided colon tubular adenocarcinomas) versus Ltu (Left-sided colon tubular adenocarcinomas) comparison, and relative to Rmu (Right-sided colon mucinous adenocarcinomas) in the Rtu versus Rmu comparison.

To identify location- and histology-specific metabolic alterations, differential features with full MS² coverage were screened using a combined threshold of a p-value < 0.05, variable importance in projection (VIP) score >1.0, and |log₂FC|> 0.26. Distinct from the conservative threshold used in MALDI-MSI (|log₂FC|> 1.0) to mitigate spatial background noise, this threshold was customized for the UHPLC-HRMS/MS cohort to preserve subtle yet robust biological shifts, ensuring cross-platform compatibility through functional pathway-level concordance. This filtering pipeline yielded 2452 and 1588 significant differential features in ESI⁺ and ESI^– modes, respectively.

2.9. Ethics Declaration

This study was approved by the Institutional Review Board of Tianjin Medical University Cancer Hospital and Institute (Approval no. EK2022243) and performed in accordance with the Declaration of the Helsinki. Patients provided informed consent prior to surgery for storage of tissue samples in the Medical Biobank of Tianjin Medical University Cancer Hospital and Institute for clinical and research purposes. No additional informed consent was required for this noninterventional study, as approved by the institutional ethics board.

3. Results

3.1. Global Metabolic Features of Bilateral Primary Colon Tumors

To investigate the spatial metabolic profiles of LCC and RCC, we performed MALDI-MSI analysis based on a single rare case of synchronous bilateral colon cancer. This initial analysis was conducted on two consecutive, 12 μm-thick frozen sections obtained from one 47 year-old female patient with synchronous bilateral primary tumors. Subsequently, we performed untargeted bulk tissue metabolomics using UHPLC-HRMS/MS to provide cohort-level support for the pathway-level metabolic changes suggested by the spatial analysis, rather than direct validation of segment-level spatial architectures. This analysis was conducted on a cohort of 30 paired CRC and adjacent normal tissue specimens, comprising 10 cases each of left-sided tubular adenocarcinoma, right-sided tubular adenocarcinoma, and right-sided mucinous adenocarcinoma. A schematic of the experimental design is provided in Figure A.

1.

Global metabolic features of bilateral primary colon tumors. (A) Schematic diagram of experimental design; (B) H&E staining of LCC and RCC tissue and ×20 magnified H&E stain image of different colon cancer tissue regions, scale bar = 2 mm, for whole tissue section, scale bar = 100 μm, for magnified images; (C) the mean spectrum of LCC and RCC analyzed by MALDI-MSI; (D) the proportion of total tissue metabolite composition in LCC and RCC analyzed by MALDI-MSI; and (E) the dynamic distribution of differences in metabolite levels based on MALDI-MSI data. The red color indicates metabolites that are upregulated in RCC compared to LCC, and the blue color indicates metabolites that are downregulated; (F) differential abundance score plot showing overall KEGG pathway changes in RCC vs LCC based on MALDI-MSI data. RCC, right-sided colon cancer. LCC, left-sided colon cancer; (G) OPLS-DA of LCC and RCC based on independent UHPLC-HRMS/MS data (LCC, n = 10; RCC, n = 20); and (H) enrichment analysis of differential metabolites identified by UHPLC-HRMS/MS, showing metabolites increased or decreased in LCC, with RCC used as the reference group for fold-change calculation. In the tree map, red labels indicate pathways enriched in LCC, and blue labels indicate pathways enriched in RCC. Linoleic acid metabolism: p = 0.039; Phenylalanine, tyrosine, and tryptophan biosynthesis: p = 0.003; Glycine, serine, and threonine metabolism: p = 0.056; Glutathione metabolism: p = 0.002; Histidine metabolism: p = 0.022; alpha-Linolenic acid metabolism: p = 0.004; Arachidonic acid metabolism: p = 0.011; Fructose and mannose metabolism: p = 0.038; Purine metabolism: p = 0.011; Starch and sucrose metabolism: p = 0.051; Amino sugar and nucleotide sugar metabolism: p = 0.343; Tyrosine metabolism: p = 0.003; One carbon pool by folate: p = 0.056; Pentose phosphate pathway: p = 0.038; and Glyoxylate and dicarboxylate metabolism: p = 0.004.

Both tumor tissues from the patient who underwent spatial metabolomics analysis were histologically diagnosed as moderately differentiated adenocarcinoma with a pathological stage II classification and no evidence of lymph node metastasis. Histopathological assessment further revealed marked spatial heterogeneity within this individual case, characterized by a complex mixture of histological subtypes and tumor microenvironmental components. Specifically, while both tumors contained regions of tubular adenocarcinoma, RCC tissue also exhibited distinct areas of mucinous adenocarcinoma. These malignant regions were interspersed with tumor-glandular mixtures, muscularis mucosa, peritumoral muscularis, lamina propria, and connective tissue (Figure B). After importing the raw imaging data into SCiLS Lab software 2023c for smoothing and root-mean-square normalization, we obtained the mean spectrum for each region (Figure C). Analysis of these processed spectra indicated distinct metabolic profile differences observed in this individual case between the left- and right-handed lesions. Building on these spectral differences, we next performed comprehensive annotation and classification of all detected metabolites to characterize the predominant metabolic features distinguishing the left- and right-sided lesions in this case. The overall classification of these annotated metabolites is detailed in Figure D, with the three highest proportions being amino acids and their metabolites (19.14%), aldehydes, ketones, esters (14.17%), and organic acids and their derivatives (14.11%).

In this bilateral case, the total metabolite abundance appeared to be higher in the LCC lesion than in the RCC lesion across all detected metabolite classes (Figure D). Interestingly, three of the four metabolites with the highest expression in RCC are derivatives of arachidonic acid (Figure E). In our previous study, we found an increased level of PLA2G4A-mediated arachidonic acid metabolism in RCC. Consistently, the present study further revealed enhanced arachidonic acid-related lipid metabolism in the RCC lesion. The enriched metabolites in LCC are mostly secondary metabolites of fatty acids and amino acids that have been hydrolyzed by certain enzymes, such as decanamide, 10-(methylsulfonyl)- (D10MS), which is a metabolite of the medium-chain fatty acid capric acid, and methyl 3-phenyl-2-(trifluoroacetamido)­propanoate, which may originate from the metabolism of phenylalanine or its derivatives (Figure E). Next, KEGG pathway analysis showed that these metabolites were mainly involved in purine metabolism and nucleotide metabolism processes, which displayed relatively stronger pathway-associated signal intensity linked to the LCC lesion of this bilateral case, especially purine metabolism, sulfur metabolism, and nucleotide metabolism, while only beta-alanine metabolism and neuroactive ligand–receptor interaction showed a downregulation trend (Figure F). All pathway analyses were conducted using data retrieved from the KEGG Pathway Database. −

To determine whether the location-associated metabolic differences observed in this index case represent recurrent pathway-level features across an independent cohort rather than observations confined to the index case, we next extended our investigation to the expanded cohort of 30 patients via parallel bulk tissue UHPLC-HRMS/MS. This approach was applied to a cohort of 30 patient samples, encompassing those with LCC (n = 10) and RCC (n = 20). Following data normalization and pareto scaling, orthogonal partial least-squares-discriminant analysis (OPLS-DA) of the m/z peak data demonstrated distinct, nonoverlapping clustering patterns specific to LCC and RCC groups, indicating metabolic separation between left- and right-sided CRC samples within this cohort (Figure G). Significant differences in the levels of key differential metabolites, including linoleic acid, amino acids, purines, sugars, and others, were observed between LCC and RCC (Figure H). These findings provide complementary cohort-level support for the metabolic distinctions initially observed by MALDI-MSI analysis and suggest pathway-level divergence between LCC and RCC.

Collectively, our integrated spatial MALDI-MSI and bulk UHPLC-HRMS/MS analyses provide preliminary evidence that anatomical tumor location and histological subtype could play a role in CRC metabolic heterogeneity. They may do so by interacting with distinct metabolite landscapes and pathway activities, offering a potential explanation for metabolic discrepancies between left- and right-sided tumor subtypes.

3.2. Spatial Metabolic Features of Bilateral Primary Colon Tumors

We performed spatially weighted nearest shrunken centroid (NWSC) clustering on the intensity profiles of all identified metabolites; this approach partitioned LCC and RCC tissues into 10 spatially distinct regions, respectively (Figure A,B). Euclidean distances between the centroids of these spatially demarcated clusters are presented in Figure C,D, where the magnitude of distance potentially reflects the extent of metabolic divergence between paired regions within this single individual.

2.

Spatial Metabolic Features of Bilateral Primary Colon Tumors. (A,B) 10 regions of LCC (A) or RCC (B) divided by spatially aware nearest shrunken centroids clustering analysis based on MALDI-MSI data from this patient; (C,D) the Euclidean distance between the centroids of different regions in spatial partitioning of LCC (C) and RCC (D) based on MALDI-MSI data; and (E,F) spatial distribution of regionally characterized metabolites and t-statistics in the individual LCC (E) and RCC (F) tissues based on MALDI-MSI data.

The ten spatially demarcated regions of the LCC tissue were arranged radially from the tumor center to the invasive margin (Figure A), a distribution suggestive of spatial heterogeneity in metabolite expression profiles. Specifically, segments 7, 4, 9, and 5 were identified as the LCC-characteristic segments. These regions showed spatially varying metabolic features from the tumor core to the tumor periphery, with representative metabolites for each segment summarized in Figure E. Phosphatidylethanolamine (PE)-related metabolites were more abundant in the tumor core, indicating accelerated cancer cell proliferation and adaptive responses to oxidative stress within the core tumor microenvironment (Figure E). In contrast, the elevated expression of phosphatidic acid metabolites at the tumor periphery could potentially be linked to the invasive capacity of cancer cells and the formation of an inflammatory microenvironment around the tumor edge. This observed metabolic heterogeneity underscores the complexity of the microenvironmental landscape in this sample, providing an exploratory reference framework for future investigation of intratumor metabolic variation in CRC. The other regions, including segments 1, 2, 3, 6, 8, and 10, were localized closer to the tissue periphery and may have been confounded by signal interference derived from the sample embedding procedures.

In contrast to LCC, the ten segments of the studied RCC tissue were primarily distinguished by the histological subtype. Virginiamycin S1 exhibited marked enrichment in segment 7, corresponding to moderately differentiated tubular adenocarcinoma (Figure F), which may reflect localized metabolic features potentially associated with microenvironmental variation. For mucinous adenocarcinoma, PE (20:4­(5Z,8Z,11Z,14Z)/P-18:1­(11Z)) was identified as the signature metabolite, which tentatively points to the possible involvement of arachidonic acid and plasmalogen metabolism in shaping a pro-inflammatory microenvironment and mediating oxidative stress resistance. In addition, segment 5 showed accumulated levels of 5-bromo-3-fluoropyridine-2-carboxylic acid, which may represent a localized metabolite accumulation potentially associated with tumor-associated biochemical processes.

By referring to the H&E staining results (Figure B), we identified segment 1 of RCC as the connective tissue, where metabolites 1,3,5-triazine-2,4-diamine and hemine were highly expressed (Figure S1A). Furthermore, we conducted colocalization network analysis of the metabolites and found that the metabolites in segment 1 existed independently, with no apparent interaction with other regions (Figure S1B). Within segment 1, however, an interaction was observed between m/z hemine and (E)-1-[4-[(2S,3R,4S,5S,6R)-4,5-dihydroxy-6-(hydroxymethyl)-3-[(2S,3R,4R,5R,6S)-3,4,5-trihydroxy-6-methyloxan-2-yl]­oxyoxan-2-yl]­oxy-2,6-dihydroxyphenyl]-3-(4-methoxyphenyl)­prop-2-en-1-one within the region (Figure S1B). The metabolism and spatial distribution of hemine might be associated with oxidative stress, inflammatory responses, and cell proliferation. Histopathological alignment further defined segment 2 of RCC as normal intestinal epithelium, with 1-hexadecanoyl-2-(9Z,12Z-octadecadienoyl)-sn-glycero-3-phosphocholine and dioleoylphosphatidic acid identified as its characteristic metabolites (Figure S1C). Segment 8 of RCC was classified as mixed cancer/noncancer areas, where the characteristic metabolites include dioleoylphosphatidylcholine, 1-oleoyl-2-palmitoyl-sn-glycero-3-phosphocholine, and PC­(18:0/18:1­(11Z)) (Figure S2A). The interaction network showed that metabolites from segments 2 and 8 exhibit interactive relationships (Figure S2B). This result implies a potential metabolic crosstalk between the normal intestinal epithelium and intermixed regions, which may reflect localized metabolic interactions within the tumor microenvironment.

In summary, we present an exploratory characterization of spatial metabolic heterogeneity in bilateral primary colon tumors. In these specimens, the LCC lesion exhibited spatial variation in metabolic features from the tumor center to the periphery, whereas metabolic partitioning in RCC appeared to be more closely associated with distinct histological subtypes. These observations provide a preliminary spatial metabolomics view of CRC heterogeneity and support future validation in larger cohorts.

3.3. Metabolic Pathway Features of Location-Specific Phenotypes in Tubular Adenocarcinomas

After characterizing the spatial metabolic heterogeneity in this bilateral colon cancer case, we further focused on the paired moderately differentiated tubular adenocarcinomas in order to reduce the confounding effect of histological heterogeneity and to perform a more controlled subtype-specific comparison.

We first performed KEGG analysis on the spatial metabolomics (MALDI-MSI) data from this single patient, which revealed significant differences in ether lipid metabolism (p = 0.0256) and nucleotide metabolism (p = 0.0261) between Ltu (Left-sided colon tubular adenocarcinomas) and Rtu (Right-sided colon tubular adenocarcinomas) regions (Figure A). We identified two ether lipid metabolites enriched in the Ltu region, choline Alfoscerate and 2-(((R)-2,3-Dihydroxypropyl)­phosphoryloxy)-N,N,N-trimethylethanaminium (Figure B). Given that ether lipid metabolites are known to be β-oxidation substrates and play a role in membrane structure, these preliminary results suggest that this Ltu region may undergo membrane remodeling through ether lipid metabolic reprogramming, potentially acquiring an antioxidant and highly proliferative malignant phenotype.

3.

Tubular adenocarcinomas of LCC and RCC show different metabolic characteristics. (A) Differential abundance score plot showing overall KEGG pathway changes in segment 7 of LCC (Ltu) vs segment 7 in RCC (Rtu) from the single case. Ether lipid metabolism: p = 0.0256; nucleotide metabolism: p = 0.0261; (B) spatial distribution of choline alfoscerate and 2-(((R)-2,3-dihydroxypropyl)­phosphory-loxy)-N,N,N-trimethylethanaminium based on MALDI-MSI data; (C) OPLS-DA of Ltu and Rtu based on independent UHPLC-HRMS/MS data (Ltu, n = 10; Rtu, n = 10); (D) heatmap of differential metabolites analyzed by UHPLC-HRMS/MS (p < 0.05, Student’s t-test); and (E) pathway enrichment analysis of differential metabolites in Ltu versus Rtu identified by UHPLC-HRMS/MS, with Rtu used as the reference group for fold-change calculation. In the tree map, blue labels indicate pathways enriched in Ltu and red labels indicate pathways enriched in Rtu. Phenylalanine, tyrosine, and tryptophan biosynthesis: p = 0.049; fructose and mannose metabolism: p = 0.010; arachidonic acid metabolism: p = 0.045; glutathione metabolism: p = 0.034; amino sugar and nucleotide sugar metabolism: p = 0.010; starch and sucrose metabolism: p = 0.014; tyrosine metabolism: p = 0.049; pentose phosphate pathway: p = 0.013; glycerophospholipid metabolism: p = 0.044; arginine biosynthesis: p = 0.048; arginine and proline metabolism: p = 0.029; alanine, aspartate and glutamate metabolism: p = 0.047; and purine metabolism: p = 0.008.

To further examine whether these pathway-level alterations were recurrent at the bulk tissue level, we subsequently conducted untargeted metabolomic analysis (UHPLC-HRMS/MS) on 20-paired samples from patients with Ltu (n = 10) or Rtu (n = 10). OPLS-DA analysis revealed a tight clustering within Ltu and Rtu samples (Figure C). Using R (version 4.4.2) to perform detailed statistical analysis and hierarchical clustering of the UHPLC-HRMS/MS data, we identified 72 differential metabolites that were significantly altered between Ltu and Rtu (Figure D). The enrichment analysis for differential metabolites highlighted alterations in amino acid metabolism, nucleotide metabolism, and lipid metabolism between the two tumors (Figure E), corroborating the pathway-level divergence observed in our spatial analysis. Specifically, glycerophospholipid metabolism demonstrated heightened activity in Ltu, marked by elevated levels of lysophosphatidylcholine (LysoPC) metabolites, including LysoPC­(18:1/0:0) and LysoPC­(0:0/18:2­(9Z,12Z)) among others (Figure A). During Ltu progression, this pathway may regulate membrane fluidity and signal transduction, mirroring the role of the ether lipid pathway observed via MALDI-MSI (Figure A). Complementing these bulk-level alterations, single-case spatial mapping captured distinct intratumor topographies of these glycerophospholipid metabolites within Ltu regions: choline Alfoscerate, lysophosphatidylcholine, LPC (0:0/20:4), and 2-(((R)-2,3-Dihydroxypropyl)­phosphoryloxy)-N,N,N-trimethylethanaminium were predominantly concentrated in tumor core regions, while 1-O-Hexadecyl-2-deoxy-2-thio-R-(heptanoyl)-sn-glyceryl-3-phosphocholine was primarily localized to the relatively peripheral area of the tumor (Figures B and B). While these single-sample spatial distributions require wider validation, their potential clinical relevance was further explored in public prognostic data sets. Glycerol-3-phosphate acyltransferase (GPAT) is the rate-limiting enzyme that initiates triacylglycerol synthesis. It catalyzes the conversion of glycerol-3-phosphate to lysophosphatidic acid, which acts as a G protein-coupled receptor agonist to promote cell migration and proliferation. Our analysis demonstrated that elevated GPAT expression in LCC was significantly associated with an unfavorable prognosis (Figure C).

4.

Increased phosphocholine and amino acid metabolism are associated with the Ltu phenotype. (A) Relative abundance of Glycerophospholipid pathway metabolites in bulk Ltu and Rtu, analyzed by UHPLC-HRMS/MS (means ± SD; Student’s t-test); (B) exploratory spatial distribution of LPC (0:0/20:4) and 1-O-hexadecyl-2-deoxy-2-thio-R-(heptanoyl)-sn-glyceryl-3-phosphocholine based on single-case MALDI-MSI data; (C) Kaplan–Meier survival analysis of the correlation between glycerol-3-phosphate acyltransferase (GPAT) expression and overall survival (OS) of LCC patients. n = 233; cutoff value: 431. (D) Relative abundance of amino acid metabolites in Ltu and Rtu, based on UHPLC-HRMS/MS data (means ± SD; Student’s t-test); (E) exploratory spatial distribution of gamma-glutamyl-beta-cyanoalanine, Met–Phe, and gamma-glutamyl-gamma-glutamyl-S-methylcysteine based on single-case MALDI-MSI data; and (F) Kaplan–Meier survival analysis of the correlation between γ-glutamyl transpeptidase (GGT1) expression and Recurrence-Free Survival of LCC patients. n = 233; cutoff value: 325.

The rapid proliferation of tumor cells necessitates a substantial supply of energy and biosynthetic precursors. In this context, amino acids and glycolytic metabolism play pivotal roles in supporting energy demands, biosynthesis, and signal transduction pathways of tumor cells. The increased alanine in the Ltu group detected by UHPLC-HRMS/MS suggests an active pyruvate–lactate–alanine cycle and a dependence on glycolysis for energy production, potentially supporting proliferative metabolic demands (Figure D). Parallel to these bulk tissue findings, our exploratory MALDI-MSI data also identified several metabolites elevated in the Ltu region (the LCC segment 7) implicated in amino acid metabolism, including gamma-glutamyl-beta-cyanoalanine, Met–Phe, and gamma-glutamyl-gamma-glutamyl-S-methylcysteine (Figure E). Additionally, elevated expression of γ-glutamyl transpeptidase (GGT1), an enzyme involved in amino acid metabolism, was associated with an unfavorable prognosis of LCC (Figure F).

Unlike Ltu, the Rtu tissues exhibited enrichment in nucleotide, amino sugar, and fatty acid metabolism for energy production at the bulk level. Specifically, Guanosine, guanine, fructose 6-phosphate, mannose 6-phosphate, sedoheptulose 7-phosphate, and glucose 6-phosphate were enriched in bulk Rtu tissues (Figure A), which are associated with nucleotide biosynthesis pathways. Consistent with these findings, spatial metabolomics also detected the enrichment of a nucleotide metabolite, adenosine-5-diphosphate (ADP), in the Rtu region of this individual, where its concentration exceeds that in Ltu (Figure B). ADK regulates extracellular adenosine and intracellular purine nucleotide concentrations. Notably, Kaplan–Meier analysis indicated that higher expression of ADK exhibited a trend toward poorer overall survival in RCC (Figure C). In fatty acid metabolism, stearic acid, myristic acid, tetradecanoic acid, 13-hydroxydocosapentaenoic acid, and 9-hydroxyoctadecadienoic acid were significantly increased in the Rtu group compared to the Ltu group (Figure D). Spatially, single-case MALDI-MSI data showed distinct intratumoral distributions of two lipid metabolites within the Rtu tissue, namely, 11,12,15-Trihydroxyeicosatrienoic acid and PE (18:2­(9Z,12Z)/18:2­(9Z,12Z)) (Figure E). Similarly, elevated transcript levels of diacylglycerol lipase β (DAGLβ, encoded by DAGLB), a key enzyme involved in lipid metabolism, were linked to a potential stratification trend toward unfavorable patient prognosis in RCC (Figure F).

5.

Purine nucleotide, glycolysis, and fatty acid metabolism are the metabolic signatures that characterize Rtu. (A) Relative abundance of purine nucleotide and amino sugar pathway metabolites in Ltu and Rtu based on UHPLC-HRMS/MS data (means ± SD; Student’s t-test); (B) spatial distribution of adenosine-5-diphosphate based on MALDI-MSI data; (C) Kaplan–Meier survival analysis of the correlation between the ADK expression and OS of RCC patients. n = 233; cutoff value: 1586; (D) relative abundance of fatty acid metabolites in Ltu and Rtu based on UHPLC-HRMS/MS data (Means ± SD; Student’s t-test); (E) spatial distribution of 1,12,15-trihydroxyeicosatrienoic acid and PE (18:2­(9Z,12Z)/18:2­(9Z,12Z)) based on MALDI-MSI data; and (F) Kaplan–Meier survival analysis of the correlation between diacylglycerol lipase β (DAGLB) expression and OS of RCC patients. n = 233; cutoff value: 116.

Together, these analyses suggest distinct metabolic reprogramming tendencies distinguishing between Ltu and Rtu. At the bulk-tissue level, consistent findings were observed in independent cohorts, the Ltu tissue exhibits a preference for membrane lipid remodeling and glycolytic support to fuel rapid proliferation, whereas Rtu exhibited relatively stronger nucleotide- and fatty-acid-associated metabolic signatures. While the localized intratumor distributions remain based on an exploratory single-case observation, the consistent alignment between bulk metabolic profiling and broader clinical prognostic validation (via GPAT, GGT1, ADK, and DAGLB cohorts) provides preliminary evidence from spatial, bulk, and prognostic analyses in CRC.

3.4. Functional Shift of Sphingolipid Metabolism in Rmu from Apoptosis-Promoting Signaling to Structural Functions

Given that mucinous adenocarcinoma and tubular adenocarcinoma exhibit significant differences in terms of pathology, biological behavior, response to treatment, and prognosis, − we next investigated the subtype-specific metabolic characteristics of these two tumor types in RCC. MALDI-MSI data indicated significant differences in the sphingolipid metabolic pathways between right tubular adenocarcinoma (Rtu) and mucinous adenocarcinoma (Rmu) of RCC (p = 0.029; Figure A). Among these metabolites, N-tetracosanoylsphingosine was downregulated, while trihexosylceramide (d18:1/22:0) was enriched within the Rmu region (segment 9) of RCC tissue (Figure B).

6.

Rmu exhibits increased metabolites in the sphingolipid metabolism. (A) Clustering heatmap of metabolites of mucinous adenocarcinoma (R9) and tubular adenocarcinoma (R7) in RCC analyzed by MALDI-MSI; (B) spatial distribution of N-tetracosanoylsphingosine and trihexosylceramide (d18:1/22:0) based on MALDI-MSI data; (C) OPLS-DA of Rmu and Rtu based on independent UHPLC-HRMS/MS data (Rmu, n = 10; Rtu, n = 10); (D) heatmap of differential metabolites analyzed by UHPLC-HRMS/MS (P < 0.05, Student’s t-test); and (E) enrichment analysis of differential metabolites identified by UHPLC-HRMS/MS, with Rmu as the reference group for fold-change calculation. In the tree map, red labels indicate pathways enriched in Rtu, and blue labels indicate pathways enriched in Rmu. Linoleic acid metabolism: p = 0.005; alpha-linolenic acid metabolism: p = 0.005; histidine metabolism: p = 0.002; glycerophospholipid metabolism: p = 0.013; sphingolipid metabolism: p = 0.044; (F) relative abundance of sphingolipid pathway metabolites in Rmu and Rtu, based on UHPLC-HRMS/MS data (means ± SD; Student’s t-test); (G) Kaplan–Meier survival analysis of the correlation between SPHK1 expression and OS of RCC patients. n = 237; cutoff value: 416; and (H) schematic diagram of the related sphingolipid metabolism features in mucinous adenocarcinoma and normal tissues.

Subsequently, we performed untargeted metabolomics analysis using UHPLC-HRMS/MS on paired Rtu (n = 10) and Rmu (n = 10) tissues from 20 cases of RCC. Distinct clustering of Rtu and Rmu in the OPLS-DA plot revealed markedly divergent metabolic profiles, suggesting that metabolic reprogramming is closely associated with the pathological features of these two subtypes (Figure C). Integration of UHPLC-HRMS/MS data from positive and negative ion modes identified a total of 92 differentially abundant metabolites (Figure D), of which 83 were significantly upregulated in Rmu versus Rtu. Consistent with the MALDI-MSI results, Rmu demonstrated significantly higher enrichment of the sphingolipid metabolism pathway than Rtu (p = 0.044; Figure E). Within this pathway, sphinganine (p = 0.044), N-Acetylneuraminic acid (p = 0.005), dehydrophytosphingosine (p = 0.004), and Cer­(d18:0/18:0) (p = 0.045) were enriched in Rmu (Figure F). Notably, no differences in ceramide levels were detected in our study. SPHK1 plays a crucial role in sphingolipid metabolism, primarily by phosphorylating sphingosine to generate Sphingosine-1-Phosphate (S1P), thereby regulating intracellular lipid homeostasis. As shown in Figure G, higher SPHK1 levels were linked to a potential trend toward unfavorable prognosis in the RCC cohort. These results suggest that mucinous adenocarcinoma may exhibit altered sphingolipid metabolic flux, characterized by accumulation of upstream (sphinganine/dehydrophytosphingosine) and intermediate products (Cer­(d18:0/18:0)) within this pathway. Ceramides may serve as precursors for sphingomyelin synthesis, potentially contributing to membrane structure maintenance and cellular growth demands (Figure H). Additionally, ceramides are conjugated with significantly increased sialic acid to form glycosphingolipids, which may be associated with the mucin-rich phenotype characteristic of this subtype.

Furthermore, carnitine metabolism was a characteristic feature of metabolic reprogramming in Rmu, which is involved in fatty acid transport and oxidation (Figure A–D). Eicosapentaenoic acid (EPA), an omega-3 polyunsaturated fatty acid, was notably abundant (Figure A). EPA relies on the carnitine shuttle system, involving carnitine palmitoyltransferase 1 and CPT2, to translocate into the mitochondria for beta-oxidation, ultimately contributing to acetyl-CoA production and cellular energy metabolism. 5-Hydroxyhexadecanoic acid (5-PAHSA) and 12-hydroxystearic acid were also elevated in the Rmu group, and they may indirectly depend on the carnitine transport system for omega-oxidation and beta-oxidation pathways (Figure A). UHPLC-HRMS/MS data also revealed significant upregulation of multiple carnitine metabolites in Rmu tissues compared with paired adjacent normal tissues (N-Rmu, Figure S3), suggesting a potential role of carnitine metabolism in Rmu initiation and progression.

7.

Rmu exhibits increased metabolites in the fatty acid-carnitine and glycerophospholipid pathways. (A) Relative abundance of fatty acid carnitine metabolites in Rtu and Rmu based on UHPLC-HRMS/MS data (means ± SD; Student’s t-test); (B) spatial distribution of oleoyl-l-carnitine based on MALDI-MSI data; (C) Kaplan–Meier survival analysis of the correlation between solute carrier family 22 (organic cation transporter), member 5 (SLC22A5) expression and OS of RCC patients. n = 237; cutoff value: 502; (D) relative abundance of glycerophospholipid pathway metabolites in Rtu and Rmu based on UHPLC-HRMS/MS data (means ± SD; Student’s t-test); and (E) Kaplan–Meier survival analysis of the correlation between PLCG1 expression and OS of RCC patients. n = 237; cutoff value: 464.

Consistently, MALDI-MSI data showed a characteristic spatial distribution of carnitine-related metabolites, carnitine-2-methyl-C4 and oleoyl-l-carnitine, within the Rmu region (segment 9; Figure B). Organic Cation Transporter 2 (OCTN2), a sodium-dependent carnitine transporter located on the cell membrane, is responsible for transporting carnitine from the extracellular space into the cell. As illustrated in Figure C, Kaplan–Meier analysis indicated that higher expression of Solute carrier family 22 member 5 (SLC22A5), the gene encoding OCTN2, exhibited a trend toward poorer overall survival in patients with RCC, further implicating carnitine metabolism in mucinous adenocarcinoma biology. In addition, UHPLC-HRMS/MS data demonstrated that glycerophospholipid metabolites were more abundant in Rmu than in Rtu (Figure D). PLCG1, a key enzyme in phospholipid metabolism, similarly correlated with a potential stratification trend toward unfavorable survival in the RCC cohort (Figure E).

Collectively, our integrated spatial and global metabolomics data suggest that Rmu is characterized by enhanced sphingolipid synthesis, activated carnitine shuttle systems, and upregulated glycerophospholipid metabolism. Moreover, the association of key regulatory molecules in these pathways with a poor prognosis further validates the functional relevance of the identified metabolic signatures.

4. Discussion

CRC is a highly heterogeneous disease in which the anatomical tumor location plays a decisive role in shaping molecular features and metabolic programs. To interrogate location-specific metabolic reprogramming while minimizing interindividual variability, we conducted an exploratory, spatially resolved metabolomic analysis of a synchronous bilateral primary CRC case, comprising left- and right-sided tumors from a single patient. To support and complement these case-derived spatial findings, we performed untargeted metabolomic profiling in an independent cohort of 30 matched CRC pairs, including Ltu, Rtu, Rmu, and their corresponding adjacent normal tissues.

By integrating spatially resolved discovery with this expanded cohort validation, we aimed to delineate the location-dependent metabolic pathways in CRC at both the microregional and tissue-wide levels. Our findings suggest that tumors arising from opposite sides of the colon exhibit marked metabolic divergence. Ltu tumors were preferentially enriched in ether lipid and glycerophospholipid metabolic pathways, whereas Rtu tumors showed an increased activity in nucleotide, amino sugar, and fatty acid metabolism. In contrast, Rmu tumors demonstrated distinct alterations in sphingolipid and carnitine metabolism compared to their tubular counterparts.

In this bilateral case, our spatially resolved analysis revealed distinct intratumoral metabolic architectures between LCCs and RCCs, which are typically masked by conventional tissue-wide bulk profiling. In this specific instance, the LCC lesion showed different metabolite distribution features from the tumor core to the invasive margin, whereas in RCC, metabolic differences were more closely associated with the histological subtype. Although derived from a single case, these observations suggest how the anatomical location and histology may synergistically shape space-dependent metabolic adaptation strategies. This interpretation is consistent with previous MALDI-MSI and spatial transcriptomic studies indicating that CRC is composed of spatially distinct functional compartments rather than a metabolically uniform tissue, and some specific metabolic activities are spatially enriched within the tumor core. − While these findings require validation in larger spatial cohorts, our data provide evidence that metabolic heterogeneity in CRC may vary across localized tumor regions.

These metabolic differences are informative not only at a descriptive level but also for understanding the broader biological and clinical divergence between LCC and RCC. Previous studies have already shown that LCCs and RCCs differ substantially in molecular characteristics, pathological features, and clinical behavior. − Our data add an exploratory spatial metabolic dimension to this divergence, suggesting that metabolic heterogeneity may be organized differently in LCC and RCC. In LCC, metabolite distributions differed between the tumor center and invasive margin, raising the possibility that local metabolic features are more influenced by the sampling position. In RCC, by contrast, metabolic variation was more closely associated with the histological subtype, suggesting a stronger role for a tissue architecture in shaping regional metabolic differences. Although the functional interpretation of the PE- and PA-related features remains tentative and requires validation in larger cohorts, prognostic associations of selected pathway-associated enzymes in public data sets support the potential clinical relevance of these localized metabolic programs.

RCC specimens comprised a heterogeneous mixture of Rtu, Rmu, and stromal regions, whereas left-sided CRC samples were composed of almost exclusively Ltu tumor cells. Consequently, bulk metabolomic analyses suggested a lower overall abundance of metabolites in right-sided CRC compared with left-sided CRC. However, when spatially matched Ltu and Rtu regions were directly compared, the metabolic activity was largely comparable. This contrast suggests that the lower metabolic activity observed in right-sided CRC at the bulk-tissue level may largely reflect differences in tissue composition rather than the intrinsic metabolic inferiority of the tumor cells themselves, further underscoring the exploratory value of spatially resolved metabolomic profiling in anatomically and histologically heterogeneous tumors. −

Having established that LCC and RCC differ in their spatial principles of metabolic organization within this case, we next examined whether these differences are reflected in their histological subtypes. In our independent untargeted metabolomic cohort, tubular adenocarcinoma, characterized by a well-defined glandular architecture, showed distinct left- and right-sided metabolic phenotypes: glycerophospholipid metabolism was more prominent in Ltu, whereas nucleotide-related alterations were more evident in Rtu. Rtu also showed marked arachidonic acid metabolism, which is of particular interest given our previous finding that the PLA2G4A/arachidonic acid axis promotes RCC progression through induction of CD39⁺ γδTreg polarization, highlighting the potential functional relevance of this pathway.

Mucinous adenocarcinoma is commonly associated with a hypoxic tumor microenvironment, which may in part relate to its abundant extracellular mucus and altered diffusion properties. Hypoxic conditions enhance carnitine-dependent transport of long-chain fatty acids into mitochondria to support β-oxidation-mediated energy production, which in turn promotes cancer cell proliferation and metastasis. , Stearyl-carnitine (CAR 18:0) can induce mitochondrial dysfunction in CD8⁺ T cells, impairing their secretion of IFN-γ and granzyme B and consequently blunting antitumor immunity. These findings suggest the hypothesis that dysregulated carnitine metabolism in Rmu tumors might exert both tumor-intrinsic and immunomodulatory effects, rendering this pathway a compelling subject for future multipatient spatial validation.

In sphingolipid metabolism, sphingosine and phytosphingosine combine with fatty acids to form dihydroceramide, which is then catalyzed by dihydroceramide desaturases (DES1/2) to generate ceramide. Under normal physiological conditions, ceramide primarily exerts a potent pro-apoptotic and tumor-suppressive function. , In Rmu, we observed a marked accumulation of sphingosine, dehydrophytosphingosine, and the intermediate dihydroceramide, Cer (d18:0/18:0), whereas ceramide itself did not show a significant increase. One possible explanation is reduced efficiency of the dihydroceramide-to-ceramide conversion step, potentially involving DES1/2, , although this remains to be tested. More broadly, these findings raise the possibility that sphingolipid metabolism in mucinous adenocarcinoma may be biased away from canonical signaling-related outputs and toward structural or biosynthetic functions. However, this interpretation requires further enzymatic and biochemical validations.

Additionally, UHPLC-HRMS/MS data, but not MALDI-MSI, indicated an upregulation of histidine metabolism in Rtu (Figures , Figure S4A, and Table S5). Histidine metabolites, such as histamine, are known to modulate immune cell function and can promote immune escape. The discrepancy between the two analytical platforms likely reflects both biological and technical factors. On the one hand, UHPLC-HRMS/MS captures averaged metabolic signals across the whole tissue and is therefore more sensitive to global metabolic alterations, whereas MALDI-MSI is optimized to resolve spatially localized metabolic differences; thus, histidine metabolism may be elevated at the whole-tissue level without showing a sufficiently distinct spatial enrichment for MSI detection. On the other hand, MALDI-MSI may show reduced sensitivity toward certain low-molecular-weight and highly polar metabolites involved in histidine metabolism because of matrix-related background interference, ion suppression, adduct variability, and the relatively narrow effective dynamic range of tissue imaging experiments. In addition, pathway enrichment in UHPLC-HRMS/MS is inferred from the combined behavior of multiple metabolites, whereas MALDI-MSI may not capture all relevant intermediates with comparable coverage. Importantly, this discrepancy highlights that pathway enrichment detected at the bulk-tissue level does not necessarily translate into a distinct spatially resolved enrichment pattern in situ, underscoring the complementary value of integrating bulk and spatial metabolomics approaches.

Importantly, paired tumor-versus-normal analyses partially recapitulated these observations, with lipid metabolic enrichment in Ltu and nucleotide metabolic activation in Rtu also observed relative to matched adjacent tissues, whereas the sphingolipid and carnitine signatures distinguishing Rmu from Rtu were less evident in Rmu versus matched normal tissues (Tables S6–S8 and Figure S4B–D), suggesting that these latter alterations may represent subtype-associated metabolic specialization during tumor progression rather than a generic tumor-associated shift.

Overall, our findings both align with and extend previous reports of CRC metabolic heterogeneity. For instance, dysregulated nucleotide metabolism has been reported in liver metastasis of RCC, consistent with our observation of activated nucleotide metabolism in Rtu. In contrast, enhanced FAO has been described in LCCs, whereas our data identify glycerophospholipid metabolism as the dominant pathway in Ltu. This discrepancy may reflect differences in histological composition as our study focused exclusively on tubular adenocarcinoma. Importantly, this work represents an exploratory spatial characterization of metabolic gradients in synchronous bilateral CRC, helping to reconcile inconsistencies arising from prior bulk metabolomics studies that lacked spatial and histological resolutions.

While this study provides novel insights into CRC metabolic heterogeneity, several limitations should be acknowledged. First, the spatial metabolomic profiling via MALDI-MSI was performed on tissue from a single, highly unique case of synchronous bilateral primary CRC with an equivalent TNM staging. Although this exceptional patient was selected from a large cohort of 1987 individuals to minimize genetic and host-specific confounding factors for intertumor comparison, this design may introduce individual-specific biases. We explicitly recognize that these spatial findings are exploratory and hypothesis-generating in nature and may not fully represent the heterogeneous spectrum of typical LCC and RCC across the general patient population. Second, although the independent cohort of 30 bulk tissue samples supported pathway-level metabolic differences, bulk metabolomics cannot directly confirm the spatial gradients identified in the index case. Third, although the 30-patient cohort used for untargeted metabolomic validation is sufficient for robust differential metabolite identification, expanding the sample size to include patients with more diverse clinical and demographic backgrounds, and further enhancing spatial resolution, would inherently improve the data depth and broader applicability. Finally, consensus molecular subtypes (CMS) classification was not incorporated in the current analysis.

To address these limitations and build on our findings, future studies should incorporate multicenter, multipatient spatial multiomics designs to validate the metabolic patterns identified here. Improved spatial resolution and more integrated analytical strategies will be essential for resolving metabolic heterogeneity with greater precision. Moreover, combining anatomical location, histological subtype, and CMS classification may help to establish a more comprehensive framework for metabolic stratification in CRC.

5. Conclusions

This exploratory study combines spatial metabolomics from a rare synchronous bilateral CRC case with untargeted metabolomic profiling of an independent cohort of 30 paired tumor and adjacent normal tissues. The spatial analysis revealed localized metabolic differences across tumor regions and histological subtypes, while the bulk cohort provided complementary support for recurrent pathway-level alterations associated with tumor location, histological subtypes, and tumor versus normal status. These findings suggest that the tissue microenvironment and histological architecture may influence tumor metabolism and contribute to subtype- and region-related metabolic differences. Overall, our results highlight substantial metabolic heterogeneity in CRC and point to candidate metabolic features for a follow-up investigation. Larger studies incorporating spatial multiomics will be required to determine the robustness and broader relevance of these observations.

Glossary

Abbreviations

RCC

Right-sided colon cancer

LCC

Left-sided colon cancer

MALDI-MSI

Matrix-assisted laser desorption/ionization-mass spectrometry imaging

UHPLC-HRMS/MS

Ultrahigh-performance liquid chromatography tandem mass spectrometry

Ltu

Left-sided colon tubular adenocarcinomas

Rtu

Right-sided colon tubular adenocarcinomas

Rmu

Right-sided colon mucinous adenocarcinomas

BRAF

B-Raf proto-oncogene, serine/threonine kinase

KRAS

Kirsten rat sarcoma viral oncogene homologue

EGFR

Epidermal growth factor receptor

LysoPC

Lysophosphatidylcholine

GPAT

Glycerol-3-Phosphate Acyltransferase

GPCR

G protein-coupled receptor

LPA

Lysophosphatidic acid

DAGLB/DAGLβ

Diacylglycerol lipase β

S1P

Sphingosine-1-Phosphate

EPA

Eicosapentaenoic acid

CPT1

Carnitine palmitoyltransferase 1

5-PAHSA

5-hydroxyhexadecanoic acid

OCTN2

Organic Cation Transporter 2

SLC22A5

Solute carrier family 22 (organic cation transporter), member 5

The original contributions presented in this study are included in the article and its Supporting Information. The UHPLC-HRMS/MS data have been deposited in MetaboLights under accession number MTBLS13729 and are publicly available at https://www.ebi.ac.uk/metabolights/MTBLS13729. The MALDI-MSI data have been deposited in METASPACE and are publicly available at https://metaspace2020.org/dataset/2026-01-19_15h19m06s, under the project Spatial and Untargeted Metabolomics Define Location- and Histology-Specific Metabolic Reprogramming in Colorectal Cancer.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jproteome.5c01260.

• Metabolic characteristics of RCC Segment 1 and Segment 2; metabolic characteristics of RCC Segment 8; relative abundance of fatty acid carnitine metabolites in Rmu, relative to N-Rmu; and spatial distribution of histidine in Rtu and Rmu by MALDI-MSI and tree map analysis of differential metabolic pathways in Ltu, Rtu, and Rmu relative to matched normal tissues based on UHPLC-HRMS/MS­(PDF)

• Detailed pathological information and molecular data of the synchronous bilateral CRC case used for MALDI-MSI (XLSX)

• Detailed pathological information and molecular data for each sample in the UHPLC-HRMS/MS cohort (XLSX)

• Detailed metabolite identification results and annotation information for MALDI-MSI features (XLSX)

• Detailed metabolite identification results and annotation information for UHPLC-HRMS/MS (XLSX)

• Differential metabolites of Rmu relative to Rtu based on MALDI-MSI (XLSX)

• Differential metabolic pathway analysis of Ltu relative to matched normal tissues based on UHPLC-HRMS/MS (CSV)

• Differential metabolic pathway analysis of Rtu relative to matched normal tissues based on UHPLC-HRMS/MS (CSV)

• Differential metabolic pathway analysis of Rmu relative to matched normal tissues based on UHPLC-HRMS/MS (CSV)

†.

These authors contributed equally to this work. Conceptualization, Y.Z.; methodology, J.W., J.G., and Y.Z.; software, J.L. and J.G.; validation, J.W. and J.G.; formal analysis, J.W., J.G., and J.L; investigation, D.H.; resources, D.H.; data curation, J.G. and J.L.; writingoriginal draft preparation, Y.Z. and J.W.; writingreview and editing, Y.Z.; visualization, J.G.; supervision, Y.Z.; project administration, Y.Z.; and funding acquisition, J.W. and Y.Z. All authors have read and agreed to the published version of the manuscript.

This research was funded by the National Natural Science Foundation of China, grant number 82203542, the Science and Technology Development Fund of Tianjin Education Commission for Higher Education, grant number 2023ZD023, and the Tianjin Key Medical Discipline (Specialty) Construction Project, grant number TJYXZDXK-009A.

The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of Tianjin Medical University Cancer Hospital and Institute (Approval No. EK2022243, approved on 7 March 2022). Informed consent was obtained from all subjects involved in this study.

The authors declare no competing financial interest.

The statements, opinions, and data contained in this publication are solely those of the individual author(s) and contributor(s) and do not necessarily reflect those of the editor(s) or publisher. The editor(s) and publisher disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions, or products referred to in the content.