This work presents a framework for the in silico metabolic modeling of Crohn's disease (CD) to identify metabolic pathway alterations using archival transcriptomic data and in vitro enteroid models. Cell-to-cell metabolic variability occurs within cell populations, as these populations execute a diverse range of functions, which change depending upon environmental impacts on transcriptomic and epigenetic regulation. Metabolic modeling provides a functional snapshot of the ongoing biological processes in a given cell or organism based on complex genotype-phenotype-biochemical relationships. These findings can be validated externally through in vitro models such as enteroids, untargeted metabolomics, and lipidomics to clarify actionable biomarkers or therapeutic targets. Identifying specific transcriptional signatures of disease through bulk RNA sequencing poses a significant challenge given the heterogeneity in gene expression that exists between patients with the same disease. To distill large magnitudes of transcriptomic data into a contextualized disease model capable of elucidating alterations in metabolic pathways, which may serve as either disease biomarkers or targets for therapeutic intervention, we have employed metabolic modeling to distinguish alterations in metabolism in diseased versus control tissue across two different datasets. To develop an in silico model, we applied our metabolic modeling approach to the archival RISK dataset, which showed that mevalonate metabolism, fatty acid oxidation, and uridine metabolism have variable flux, or flow when comparing patients with CD to controls (Fig. 2), suggesting the relevance of these three classes of biochemical pathways in CD pathogenesis. Our in vitro model comprised organoids generated from patients with CD and controls ileal tissue samples. These organoids were previously found to have a similar transcriptomic signature as in vivo epithelium and retain disease-specific gene expression patterns [27]. In enteroids, we found glycerophospholipid, linoleic acid, and sphingolipid metabolism to be altered (Fig. 3). Lastly, we piloted an additional validation step by using untargeted metabolomics to reveal individual metabolic pathways associated with CD, in which we found vitamin B3, pyrimidine, and glutamate metabolism to be significantly enriched in CD.
The metabolic modeling pipeline detailed in this study showed that reactions involved in mevalonate metabolism were among the top reactions that differentiated CD from controls in the RISK dataset (Fig. 2B). Alterations in mevalonate metabolism in T-lymphocytes have been implicated in decreased inflammatory suppressive activity [30]. In addition, derivatives of mevalonate are known to be directly involved in limiting the cytotoxic effector response of T-lymphocytes in inflammation and the production of immunomodulatory precursors [31]. Alterations of this pathway, such as mevalonate kinase deficiency (MKD), lead to inflammatory bowel disease, or IBD-like intestinal inflammation, possibly due to decreased immunosuppressive isoprenoid intermediates formed through the mevalonate kinase pathway [32]. Therefore, the changes in mevalonate metabolism between CD and controls tissue identified here may contribute to establishing the inflammatory environment associated with CD.
Fatty acid metabolism is a highly-regulated process where dysregulation can lead to an imbalance of pro- and anti-inflammatory mediators. Of note, the dysregulation of fatty acid metabolism contributes to both the type and degree of inflammatory responses in the intestine during IBD [33–35]. Recently, in-depth gene array analysis has also revealed that various genes involved in fatty acid uptake and synthesis are differentially expressed in the ileum of IBD patients [36]. One such pathway differently expressed in IBD patients involves uridine transport and exchange. Uridine is a critical regulator of lipid metabolism and a pathway identified in our current study (Fig. 2D) [37]. As such, pharmaceuticals targeting uridine metabolism are a promising candidate for future treatment of inflammatory bowel disease (IBD). Remarkably, in studies utilizing uridine to treat mice with Dextran Sulfate Sodium (DSS)-Induced Colitis, levels of pro-inflammatory cytokines IL-6, IL-1β, and TNF in the serum and mRNA expression in the colon were significantly decreased in the uridine-treated groups [38], further demonstrating the critical role for fatty acid metabolism in regulating gastrointestinal inflammation.
In previous studies using cultured ileal organoids (enteroids), no morphological characteristics were observed that could be used to differentiate between CD and controls. Further, when gene expression profiles from enteroids were compared to profiles from freshly isolated intestinal crypts, a strong positive correlation was observed in the mean expression levels of many genes [27, 39]. This result suggests that most genes expressed in vivo in the epithelium are also expressed in organoid cultures, demonstrating the immense value of enteroids as an important tool for IBD research. As with the RISK dataset, we grouped altered metabolic reactions identified in organoids based on broad biochemical pathways. We noted that four of the top ten reactions were involved in the dysregulation of fatty acid and phospholipid metabolism (Fig. 3B-C). These findings agree with previous lipid profiling studies of CD patients compared to controls. In these studies, fatty acid and phospholipid metabolism was significantly altered, with most alterations found in glycerophospholipid, linoleic acid, and sphingolipid metabolism pathways [40]. Various metabolites involved in these pathways play a role in inflammation, intracellular signaling, pain, immune function, reproduction, and appetite and perpetuate colitis when dysregulated [41]. Future work will test whether drugs targeting these pathways can alleviate the characteristic inflammation of IBD in patient-derived enteroids, laying the groundwork for further studies into potential therapeutic use [39]. Additionally, the observed alterations of different metabolites of fatty acid oxidation between tissue and organoids suggest mitochondrial dysfunction [42]. Fatty acid biosynthesis has previously been shown to be part of a mitochondrial-to-cytosolic stress response that results when mitochondrial protein synthesis is disrupted [43]. Staining our patient-derived organoids with a mitochondrial membrane potential dye may further help clarify this relationship.
Another biological process that was found to be significantly altered in enteroids derived from CD patients was sphinganine transport (Fig. 3D). Sphinganine is biosynthesized from serine and palmitoyl-CoA through decarboxylating condensation, resulting in a keto intermediate, which is reduced by NADPH. It is then further acylated, then dehydrogenated to form ceramide [44]. Ceramide decreases the release of tumor necrosis factor (TNF), most likely via post-translational regulatory mechanisms and the modulation of TNF-converting enzyme activity [45]. Furthermore, ceramides play an essential role in regulating autophagy, a process strongly linked to the pathogenesis of CD [46]. Sphinganine also blocks lysosomal cholesterol transport and has been linked to Niemann-Pick Type C disease, which predisposes to early-onset IBD with CD phenotype and granuloma formation [47]. However, the exact mechanism by which sphinganine contributes to CD remains unclear and will require further study.
In addition, oxidative stress, caused by increased ROS production, is present locally and systemically in patients with CD [48]. Here, we identified two reactions in this pathway: superoxide dismutase and catalase (Fig. 3A). These enzymes play a role in endogenous antioxidant mechanisms and counteract the effects of excess ROS [49, 50]. Notably, multiple enzymes catalyze the production of ROS. Among these enzymes, mucosal NOXs and dual oxidase 2 (DUOX2) have been reported as novel risk factors for IBD [51, 52], further demonstrating that an imbalance in redox homeostasis contributes to the pathogenesis of IBD. Interestingly, In the RISK study, the authors highlight the role of oxidative stress in gut inflammation. Their works detail the enrichment of pro-inflammatory genes, including antimicrobial dual oxidase (DUOX2), and the decreased expression of anti-inflammatory and antioxidant genes, such as those resulting in the production of apoprotein A1 (APOA1) in the ileal tissue of patients with CD [26].
Using mass spectrometry-based lipidomics, we aimed to pilot an additional level of validation complementary to our in silico metabolic modeling approach. The significant variations in metabolic alterations between diseased and not diseased individuals and within different diseased individuals provide a rationale for future metabolomic analyses of large cohorts examining phenotypic variations within patients with CD. Our metabolomics analysis showed overlaps with our in silico and in vitro modeling. For example, vitamin B3 was found to be significantly enriched, iterating the theme of the dysregulation of lipid metabolism present in CD. Interestingly, treatment with high dose vitamin B3 has been shown to ameliorate ulcerative colitis through increased prostaglandin D2 synthesis in mice [53]. Thus, niacin supplementation can be a potential therapeutic target to be investigated in CD as well. In addition, untargeted metabolomics showed the dysregulation of pathways related to pyrimidine, glutamate, and nitrogen metabolism. Similarly, in the RISK dataset, we found uridine transport and metabolism to be altered between patients with CD and controls, further warranting assessment of the potential efficacy of uridine in CD management.
Limitations in this study are shown in the discrepancy in the specific metabolic reactions that have altered flux when comparing the RISK dataset to the enteroid dataset presents a visible limitation to our study. Variations in the mean age and ethnicity of patients also have varying effects on the metabolome and lipidome of the enteroids and thus are confounders in our analysis. In addition, there is an incongruity between the sizes of the two cohorts, as enteroids were generated from a smaller population (Table 1). Furthermore, it is crucial to recognize that enteroids only consist of epithelial tissue. CD spans the entire thickness of the epithelium, involving a complicated interaction between genetic factors, growth factors, cytokines, environmental factors such as diet and smoking, and the constitution of the gut microbiome [26]. The absence of gut microbiome signaling in the enteroids may explain the dissimilarities between altered fatty acid pathways in the RISK dataset and enteroid models [54, 55]. In emerging work studying organoids' interaction with other elements of the gastrointestinal system, researchers added microbiome elements to characterize the carcinogenic effects of E. coli on intestinal organoids [56]. This work clearly demonstrated that microbial elements may significantly impact the genetic signature of organoids, revealing a technical limitation in using organoids. To address this limitation, other groups have started developing novel organoid culture systems having added resident innate immune cells and fibroblasts and observed their interactions with tissue to enable further use of enteroid as an in vitro model for changes to the gut epithelium [57]. Future studies with these complex organoid systems will allow us to better probe the relationship of the microbiome to the metabolic pathways underlying CD pathogenesis.
The notable strength of this study is the detailed framework for in silico prediction with in vitro validation, pruning genes and metabolic reactions from transcriptomic data, and lipidomics analysis to provide a high-order understanding of alterations in metabolism for specific disease states. These altered pathways may define measurable biomarkers specific to certain disease states or even represent targetable therapeutic options. In this study, we identified metabolic pathways with altered flow in both archival data and enteroid models of Crohn's disease and revealed overlap in the overall biochemical processes in which they occur. In the archival RISK dataset, we found mevalonate, fatty acid, and uridine metabolism to be altered, while in enteroids, we found glycerophospholipid, linoleic acid, and sphingolipid metabolism to be altered (Fig. 2, Fig. 3). Thus, fatty acid metabolic pathways offer promising therapeutic targets in managing CD. Due to the overlap between metabolic alterations observed in both enteroids and the RISK datasets, ileal enteroids can offer us insight into epithelial response to various pharmaceutical interventions. In addition, the other metabolic pathways identified by metabolic modeling offer a wide range of targetable metabolites and reactions that can be investigated further. Metabolomics analyses can also be used to further validate these findings by confirming whether the metabolites of the pathways with altered flow are present and measurable in ileal enteroids and biopsies. Future studies will determine whether the metabolites identified in this study can be used to monitor disease progression and treatment response and whether therapeutic targeting of these pathways may provide an effective approach to treating Crohn's disease.