Metabolomics of Acute vs. Chronic Spinach Intake in an Apc–Mutant Genetic Background: Linoleate and Butanoate Metabolites Targeting HDAC Activity and IFN–γ Signaling

There is growing interest in the crosstalk between the gut microbiome, host metabolomic features, and disease pathogenesis. The current investigation compared long–term (26 week) and acute (3 day) dietary spinach intake in a genetic model of colorectal cancer. Metabolomic analyses in the polyposis in rat colon (Pirc) model and in wild–type animals corroborated key contributions to anticancer outcomes by spinach–derived linoleate bioactives and a butanoate metabolite linked to increased α–diversity of the gut microbiome. Combining linoleate and butanoate metabolites in human colon cancer cells revealed enhanced apoptosis and reduced cell viability, paralleling the apoptosis induction in colon tumors from rats given long–term spinach treatment. Mechanistic studies in cell–based assays and in vivo implicated the linoleate and butanoate metabolites in targeting histone deacetylase (HDAC) activity and the interferon–γ (IFN–γ) signaling axis. Clinical translation of these findings to at–risk patients might provide valuable quality–of–life benefits by delaying surgical interventions and drug therapies with adverse side effects.


Materials and Methods
Animals-Studies in Pirc (F344/NTac-Apc am1137 ) and wild-type (WT) F344 male rats were approved by the Institutional Animal Care and Use Committee. For complete details on preclinical methodologies refer to Chen et al. [6]. In brief, Pirc and WT rats were assigned randomly to basal AIN93 control diet or AIN93 diet containing 10% w/w freeze-dried baby SPI. Rats were fed SPI from 4 to 30 weeks of age (26-wk SPI intake), or for 3 days only (SPI3d), starting in the final week of the 30 week study. At necropsy, tissue sampling for metabolomic analyses included Pirc colon tumors, adjacent normal-looking colonic mucosa, colonic mucosa scrapings, colon 'punch' biopsies, and normal colon from WT rats, with biological replicates as indicated in Figure 1A (see Section 3.1).
Metabolomics-Pre-weighed samples of rat colon tumor and normal colonic mucosa, collected at the time of necropsy, were homogenized in 0.5 mL cold methanol and 0.2 mL chloroform in pre-cooled Garnet bead tubes using a Precellys ® 24 beadbeater (Zymo Research, Irvine, CA, USA). Samples were centrifuged at 3000 rpm for 10 min at 4 • C and 0.7 mL cold water was added to the supernatant. The aqueous phase was collected by centrifugation at 3000 rpm for 1 min and passed through a sterile nylon cell strainer and lyophilized. Samples were reconstituted in 50 µL methanol/water (1:1, v/v) and stored at −80 • C. Liquid chromatography high-resolution accurate-mass spectrometry was conducted as reported [6]. A Synergi Fusion-RP C-18 column (Phenomenex, Torrance, CA, USA) was used with a methanol/acetonitrile solvent gradient, and mass scanning in the positive mode was in the range 50 to 750. MS1 and MS1-dependent MS2 spectra were collected at an m/z resolution of 70,000 and 17,500, respectively, with the autosampler maintained at 4 • C. Raw metabolomic data were imported into Progenesis QI (Waters, Milford, MA, USA) for alignment, peak picking, and metabolite identification, with reference to the Human Metabolome Database (HMDB). Raw abundance data were normalized to initial sample weights, incorporating Partial Least Squares Discriminant Analysis (PLSDA). Features were filtered by their appearance in three independent metabolomic databases, with at least three biological replicates and a significant ANOVA test. Significant features were subjected to clustering and correlation by MetaboAnalyst 4.0. The p-values (twotailed t-test) and t-scores (standardized test statistic) were generated for multiple group comparisons of metabolic networks and functional metabolite prediction via Mummichog version 2 in R. Compound names were mapped to Kyoto Encyclopedia of Genes and Genomes (KEGG) COMPOUND, and pathway analyses by Mummichog were ranked according to p-value, using p = 0.05 as the cutoff. For further information on the untargeted metabolomics, see Chen et al. [6].
Microbiome-Detailed methodologies were reported by Chen et al. [6]. In brief, rat fecal samples were submitted for bacterial genomic DNA extraction at the Center for Metagenomics & Microbiome Research (CMMR), Baylor College of Medicine, Houston, TX. The 16S rDNA V4 region was amplified and barcoded via PCR and sequenced using the MiSeq platform (Illumina, San Diego, CA, USA) with a 2 × 250 bp paired-end protocol. OTUs at a similarity cutoff value of 97% were generated by the UPARSE algorithm and mapped to SILVA database. OTU tables and Agile Toolkit for Incisive Microbial Analyses (ATIMA) were provided by CMMR for primary data visualization. ATIMA microbiome data were subjected to the Kruskal-Wallis test, as before [6].
Statistics-Unless stated otherwise, findings are representative outcomes from three or more biological and technical replicates, using Student's t-test for paired comparisons and ANOVA for group comparisons, as reported [6,9,14,15].

Metabolomics Segregated Pirc and WT Rats According to Acute vs. Chronic SPI Intake
Pirc and WT rats fed basal AIN93 control (Ctrl) diet or AIN93 diet containing 10% w/w freeze-dried baby SPI for 26 weeks or 3 days ( Figure 1A) were designated as AIN, SPI, and SPI3d groups, respectively. Among 17,243 metabolomic features identified in colon tissues at the end of the study, PLSDA segregated groups according to treatment and genotype ( Figure 1B). For example, colon scrapings from WT rats (top right) and colon tumors from Pirc rats (lower left) had AIN and SPI3d groups clustered together, separate from SPI 'chronic' treatment. In other cases, namely, WT and Pirc normal 'punch' biopsies and Pirc colon scrapings, the SPI3d group segregated between AIN and SPI groups.   Five groups were distinguishable among the AIN controls, with Pirc tumor being segregated furthest from WT normal colon ( Figure 1B, lower right). 'Heatmaps' averaged across the replicates in each group produced a distinct tumor metabolomic signatureespecially in rats given SPI for 26 weeks ( Figure 1C, top right). In correlation analyses, Pirc tumor was distinct from other groups, most notably for metabolomic features with increased relative abundance ( Figure 1D, lower right). Thus, the metabolomic signature from chronic SPI treatment was distinct from AIN and SPI3d groups, especially for Pirc colon tumors.

Fatty Acids and Other Compound Categories Were Altered by SPI Intake
Mummichog coupled to KEGG prioritized eight categories of small molecules. In Pirc colon tumors, SPI increased four compound categories significantly, namely, Lipids and Fatty acids, Phytochemicals, Carbohydrates, and Organic acids ( Figure 2A, lower right). This was not observed for SPI3d in Pirc tumors, and no statistically significant changes were detected in normal tissues from Pirc or WT rats. In pairwise comparisons, Omega-3 fatty acid metabolism, Butanoate metabolism, and Prostaglandins from linoleate were implicated in Pirc T SPI vs. Pirc T AIN groups ( Figure 2B, bottom). Linoleate metabolism also featured significantly for WT N AIN vs. WT N SPI and Pirc N AIN vs. Pirc N SPI3d pairwise comparisons. Thus, among other changes, linoleate and butanoate metabolism were altered markedly by SPI intake in the rat.
In metabolomic heatmaps, AIN group comparisons revealed a Pirc tumor signature that was distinct from Pirc normal and WT normal ( Figure 3A). Noteworthy in tumors was the lower Linoleate metabolism in six out of seven rats ( Figure 3A, blue square). Reduced Arachidonate, Purine, and Eicosapentaenoate metabolism and increased Carnitine shuttle and ω-3 fatty acid, β-Alanine, and Glutathione metabolism also was detected in tumors. In SPI and SPI3d groups, tumors had a lower relative abundance of Butanoate metabolism ( Figure 3B, blue square). In the Pirc scrape SPI dataset there was increased Butanoate, Purine, Pyrimidine, and Selenoamino acid metabolism ( Figure 3B, red square). We inferred that butanoate and other metabolites were increased by chronic SPI but not SPI3d in the zone closest to the colonic crypts, captured by colonic scraping.
Low linoleate metabolism in colon tumors of Pirc AIN controls ( Figure 3A, blue square) was taken for further analyses ( Figure 4, top left, blue square). Groups from our prior work [6] are reproduced here as the first four bars in each dataset, with additional comparisons not reported previously. From the y-axis ranges in Figure 4, linoleate was present at 3-4-fold higher levels than its 15-lipoxygenase-1/15-LOX-1 metabolites. Linoleate and its 15-LOX-1 metabolites had a lower relative abundance in Pirc T AIN controls (red bars), which was reversed or 'normalized' in the SPI and SPI3d groups. Exceptions were noted, however, including higher relative metabolite abundances in Pirc normal colon (third bar in each dataset). In Pirc tissues, the highest 15-LOX-1 metabolite levels were detected in SPI3d normal colon scrapings, as exemplified by 13(S)-HODE ( Figure 4, *** p < 0.001). In general, the highest 15-LOX-1 metabolite abundances were for WT rats in SPI and SPI3d groups ( Figure 4, datasets at right side). For example, 13(S)-HODE in WT N SPI3d scrape was significantly higher than in WT N AIN scrape (*** p < 0.001), as well as Pirc T AIN. We concluded that SPI treatment for 3 days or 26 weeks markedly increased linoleate and its 15-LOX-1 pathway intermediates in Pirc and WT colon tissues, including 13(S)-HODE. These observations are noteworthy given the proapoptotic antitumor activity reported for 13(S)-HODE in colorectal cancer [18][19][20][21][22][23][24].
Butyrate can attain millimolar concentrations in the gut [25][26][27][28], but information often is lacking for its metabolites. Test compounds were screened using a reported HDAC activity assay [8][9][10], with Trichostatin A (TSA) as a positive control in some experiments. In a cell-free assay with whole cell lysates from HCT116 human colon cancer cells, concentration-dependent inhibition of HDAC activity was observed by (S)-2HB and sodium butyrate (NaB) at 62.5, 125, 250, 500 and 1000 µM ( Figure 6A), and by 0.625, 1.25, 2.5 and 5 µM 13(S)-HODE ( Figure 6B). Cytoplasmic and nuclear lysates from HCT116 cells also were treated with selected inhibitor doses. Compared to vehicle control, 2.5 µM 13(S)-HODE alone or in combination with 100 µM (S)-2HB inhibited HDAC activity significantly, similar to 1 mM NaB ( Figure 6C). Thus, deacetylase activities in nuclear and non-nuclear compartments were susceptible to inhibitor treatments. Analogous results were obtained when HCT116 cells were incubated with test agents for 48 h and the whole cell lysates were added to HDAC activity assays ( Figure 6D).

Discussion
Recent human clinical intervention trials have assessed diverse aspects of SPI intake, including anthropometric measures, muscle fitness, metabolic profiles, arterial stiffness, and urinary biomarkers [5,[34][35][36], extending prior research on the health benefits of green leafy vegetables and the functional properties of spinach-derived phytochemicals and bioactives [37][38][39][40][41]. We reported that long-term feeding of freeze-dried SPI at 10% w/w in the diet for 26 weeks exhibited significant antitumor efficacy in the Pirc model, resulting in >60% reduced tumor multiplicity in the colon and small intestine [6]. In Apc-mutant and WT rats, increased gut microbiome diversity after SPI consumption coincided with reversal of taxonomic composition. Metagenomic prediction implicated linoleate and butanoate metabolism, tricarboxylic acid cycle, and pathways in cancer, which was supported by transcriptomics and metabolomics. Thus, tumor suppression by SPI involved marked reshaping of the gut microbiome and changes in host RNA-miRNA networks. When colon polyps were compared with matched normal-looking tissues via metabolomics, anticancer outcomes were linked to SPI-derived linoleate bioactives with known antiinflammatory/proapoptotic mechanisms in colorectal cancer [18][19][20][21][22][23][24].
The current investigation confirmed and extended these observations, and sought to compare long-term vs. acute (26-week vs. 3d) SPI consumption in the rat, incorporating colonic mucosa scrapings and tissues from WT animals. Partial least squares discriminant analyses of 17,243 metabolomic features aligned SPI3d with AIN controls, or distributed SPI3d midway between AIN and SPI groups. These findings hinted at SPI3d starting to reshape metabolomic features towards the more marked changes observed after 26 weeks of SPI intake. Heatmaps revealed a distinct Pirc colon tumor metabolomic signature, with SPI (but not SPI3d) increasing lipids and fatty acids, organic acids, carbohydrates, and phytochemicals. Based on our prior report [6], we focused initially on linoleate and its downstream metabolites. Interestingly, 13-HPODE, 13(S)-HODE and 13-oxoODE had some of the highest relative abundance levels in WT tissues, especially from colon scrape samples in SPI and SPI3d groups (Figure 4). This implicated SPI-derived (rather than tumor-specific) linoleate metabolites, consistent with their detection in the freeze-dried baby SPI incorporated into AIN basal diet, using unbiased metabolomic analyses [6].
Interestingly, among the 700+ metabolomic features in baby SPI, no (S)-2A2HB was detected [6]. The significant increase in (S)-2A2HB in Pirc colon tumors by SPI and its absence following SPI3d treatment ( Figure 5A, green square) suggested the necessity for reshaping of the gut microbiome over several weeks, to increase α-diversity and enhance butyrate-producing bacteria. This was confirmed in a time-course investigation, in which α-diversity increased in the Pirc model after approximately 7-14 days of dietary SPI intake, and plateaued thereafter for up to 60 days ( Figure 5B).
As noted above, (S)-2A2HB was not available commercially, but the presumed deacetylated metabolite (S)-2HB was viewed as a possible mechanism-based HDAC inhibitor, analogous to sulforaphane metabolites [8,34]. Using molecular docking in silico, favorable interactions were predicted for (S)-2A2HB, (S)-2HB and the enantiomeric metabolite (R)-2HB with allosteric sites in HDAC1-and HDAC3-containing corepressor complexes ( Figure 5C). Notably, 13(S)-HODE also had favorable interactions with HDAC3 allosteric sites and with the catalytic pocket of HDAC1. Docking scores were in the range -4.2 to -5.4 kcal/mol ( Figure 5D), synonymous with the degree of HDAC inhibition detected in cell-free and cell-based HDAC activity assays, relative to 0.1 µM TSA. Tissues in the Pirc model were prioritized for metabolomics, but future work should examine changes in selected HDACs and histone acetylation or methylation marks following SPI and SPI3d intake.
In human colon cancer cells, a threshold was observed for apoptosis induction at 48 h using 100 µM (S)-2HB and 2.5 µM 13(S)-HODE. When combined, these metabolites caused marked induction of cleaved PARP and Caspase-3, comparable to 1 mM NaB. Decreased viability in cell-based assays coincided with loss of β-catenin, but this was not recapitulated in vivo. Thus, expression of β-catenin, Cyclin D1, and Mmp7 remained high in adenomatous polyps compared to normal colon, and was unaffected by SPI3d or SPI treatment, although SPI increased cleaved PARP and Caspase-3 in Pirc colon tumors, indicative of apoptosis induction. Quantifying nuclear 'active' β-catenin might yield greater insights, as described in recent studies that provided genetic and molecular corroboration for specific LOX enzymes and linoleic metabolites in suppressing LPR5 recycling, Wnt/βcatenin signaling, and colon carcinogenesis [24].
The IFN-γ signaling axis was defined as a top priority in Pirc SPI colon tumors at 30 weeks [6]. In the current investigation, Pirc tumor SPI samples had increased expression of β2m, interferon-γ and Canx as compared to Pirc tumor SPI3d and Pirc tumor AIN groups, and decreased levels of Foxp3, Iκbα and Survivin, implicating NFκB signaling and apoptosis induction. Several of the immune biomarkers were similarly altered in cell-based assays involving human and murine colon carcinoma cells incubated with (S)-2HB±13(S)-HODE. Notably, increased cell surface occupancy of β2m was confirmed in FACS-based analyses, and the functionality of MHC-I complexes was corroborated in MC38-OVA+B3Z co-culture experiments. Downregulation of MHC-I components is a potential oncogenic driver [29][30][31][32][33], and the targeting of epigenetic 'readers', 'writers' and 'erasers' might facilitate re-expression of cell surface MHC complexes to reengage host immune pathways in cancer cells. These mechanisms also might be pertinent at earlier stages, as in the case of adenomatous colon tumors from the Pirc model and in FAP or Lynch Syndrome patients, which harbor predicted MHC neoantigens [42]. Llosa et al. [43] noted that 'an altered amino acid due to a coding mutation is only relevant as a tumor neoantigen for T cells if it can be processed and presented on self-MHC . . . individual tumors with lower mutational load can nonetheless generate good T-cell neoepitopes if the mutations are appropriately positioned'. A roadblock to appropriately positioned neoepitopes involves epigenetic silencing of MHC components, and the ability of linoleate and butanoate metabolites to inhibit HDAC activity and to re-express MHC functional complexes at the surface of colon cancer cells is worthy of further investigation.

Conclusions
This investigation compared long-term vs. acute SPI intake in a preclinical model of hereditary colon cancer, and corroborated our prior findings vis-à-vis SPI-derived linoleate bioactives and butanoate metabolites linked to increased α-diversity of the gut microbiome. This is the first report to demonstrate HDAC inhibition, apoptosis induction, and altered IFN-γ signaling in colon cancer cells treated with specific butanoate and linoleate metabolites in combination. Future work should seek to corroborate the concentrations of these metabolites in vivo in the context of apoptosis induction in colon tumors after long-term dietary SPI intake, and phenotypic outcomes following treatment of FAP patient organoids. Clinical translation of freeze-dried whole foods, such as SPI, to at-risk patients might provide valuable quality-of-life benefits via inflammasome/immune mechanisms, delaying colectomy and drug intervention [44][45][46][47][48][49][50].