The Immunosuppressant Mycophenolic Acid Alters Nucleotide and Lipid Metabolism in an Intestinal Cell Model

The study objective was to elucidate the molecular mechanisms underlying the negative effects of mycophenolic acid (MPA) on human intestinal cells. Effects of MPA exposure and guanosine supplementation on nucleotide concentrations in LS180 cells were assessed using liquid chromatography-mass spectrometry. Proteomics analysis was carried out using stable isotope labeling by amino acids in cell culture combined with gel-based liquid chromatography-mass spectrometry and lipidome analysis using 1H nuclear magnetic resonance spectroscopy. Despite supplementation, depletion of guanosine nucleotides (p < 0.001 at 24 and 72 h; 5, 100, and 250 μM MPA) and upregulation of uridine and cytidine nucleotides (p < 0.001 at 24 h; 5 μM MPA) occurred after exposure to MPA. MPA significantly altered 35 proteins mainly related to nucleotide-dependent processes and lipid metabolism. Cross-reference with previous studies of MPA-associated protein changes widely corroborated these results, but showed differences that may be model- and/or method-dependent. MPA exposure increased intracellular concentrations of fatty acids, cholesterol, and phosphatidylcholine (p < 0.01 at 72 h; 100 μM MPA) which corresponded to the changes in lipid-metabolizing proteins. MPA affected intracellular nucleotide levels, nucleotide-dependent processes, expression of structural proteins, fatty acid and lipid metabolism in LS180 cells. These changes may compromise intestinal membrane integrity and contribute to gastrointestinal toxicity.

Scientific RepoRts | 7:45088 | DOI: 10.1038/srep45088 The etiology of MPA-related GI adverse effects is not yet fully understood and the underlying molecular mechanisms have never comprehensively been studied. However, several hypotheses exist regarding the origin of MPA's adverse effects on the GI tract. It has been suggested that the main mediators of toxicity are MPA's acyl glucuronide metabolite (AcMPAG) and eventually the morpholino ester moiety N-(2-hydroxyethyl) morpholine, which is cleaved from the prodrug mycophenolate mofetil (MMF) to result in the active MPA 13 . AcMPAG can form protein adducts 13,14 , N-(2-hydroxyethyl) morpholine may cause local irritation of the epithelium 13 . Moreover, it has been hypothesized that MPA promotes inflammation by proliferation inhibition of the GI tract's rapidly dividing epithelial cells 6 . This may lead to disruption of the GI barrier 13 and Crohn's disease-like symptoms. This hypothesis has been challenged as purines, released during the ingestion of cells from dietary sources, are highly abundant in the GI lumen 13 . Purines can enter the cell lumen via passive diffusion, by utilization of a transporter for nucleotides, or via a carrier-mediated process 13 .
Toxicodynamic mechanisms are not necessarily linked to a drug's mechanism of action. Therefore, we investigated mechanisms of MPA toxicity at the cellular level in the presence of extracellular guanosine as provided under physiologic conditions in the intestinal human colon adenocarcinoma cell line LS180 by untargeted analysis of protein and metabolite changes. Compared to an in vivo system, a cellular system allows control of local drug and guanosine concentrations while avoiding secondary pathological processes, such as inflammation, which may interfere with the assessment of the underlying toxicity mechanisms in an in vivo setting. Human epithelial-like colon cancer LS180 cells were chosen based on a comprehensive literature review of available models applicable to elucidate adverse effects of drug treatment on the GI tract. Most importantly, LS180 cells express the pregnane X receptor, and are therefore able to upregulate/induce the expression of drug metabolizing enzymes as seen in vivo 15 .
Cross reference with proteome alterations in rat models and other human cell lines of non-cancerous origin. For comparison of data collected using the LS180 cell model (human colon cancer) with those collected using common rat models (Wistar rats) proteins/genes from the three available publications on this topic (kidney tissue of MMF-treated rats 14 , liver and colon tissue of MMF-treated rats 21 , analysis of gene expression (complementary deoxyribonucleic acid microarray analysis) in liver and gut of MMF-treated rats 22 ) were pooled (Supplementary Table S1). For comparison with data collected through models using other human cell lines (HEK-293, CCRF-CEM; differential proteome analysis) protein/gene data from two available publications were pooled 23,24 (Supplementary Table S2). DAVID software was used for comparison.
Scientific RepoRts | 7:45088 | DOI: 10.1038/srep45088 nucleotide concentrations remained significantly lower than in controls not treated with guanosine (p < 0.001) for MPA concentrations ≥ 5 μ M despite guanosine supplementation for 24 and 72 h ( Supplementary Fig. S1). Even for LS180 cells treated with 1 mM exogenous guanosine; GTP, GDP, and GMP levels were only approximately 30% of the controls. Addition of guanosine to culture media reversed changes in UMP and UDP levels after 24 and 72 h that were observed without guanosine supplementation ( Supplementary Fig. S1). UTP levels, on the other hand, were significantly higher (p < 0.01 and 0.001, respectively) after exposure to MPA concentrations ≥ 5 μ M and in the presence of 200 μ M and 1 mM guanosine after 72 h. In most cases, changes in cytidine nucleotide levels were reversed by supplementation with guanosine ( Supplementary Fig. S2). Slightly elevated values suggest imbalances in cytidine nucleotide levels in LS180 cells despite of, or possibly due to, guanosine supplementation of the culture media. Disturbances in NAD + , NADP + , and FAD concentrations, which mainly occurred after 72 h of exposure to higher MPA concentrations in the absence of guanosine supplementation, were similar to control levels when incubation media were supplemented with guanosine ( Supplementary Fig. 2). While nucleotide energy charges in MPA-treated cells did not show marked differences compared to control cells with guanosine supplementation <1 mM, uridylate energy charges after 72 h of treatment with 250 μ M MPA and 1 mM guanosine increased remarkably (125.3 ± 5.0% of controls, p < 0.05, Supplementary Fig. S2). In summary, some negative effects of MPA on nucleotide levels were partially or fully reversed by supplementation with guanosine. Despite supplementation with 1 mM guanosine, levels of all guanosine nucleotides were significantly decreased (p < 0.001) and levels of uridine and cytidine nucleotides were upregulated with consequent effects on the uridylate energy charge. Levels of the second messenger cyclic adenosine monophosphate and cyclic guanosine monophosphate were mostly unaffected by MPA, regardless of exogenous guanosine concentrations ( Supplementary Fig. 3), when examined by enzyme-linked immunosorbent assays.
Investigation of proteome alterations in LS180 cells after exposure to MPA and cross reference of results with proteome alterations in rat models and other human cell lines of non-cancerous origin. Expression levels of 35 proteins were significantly changed (p < 0.05, >1.2-fold) in MPA-treated LS180 cells despite supplementation with 1 mM guanosine as identified by SILAC in combination with GeLC-MS ( Table 2). As a representative example, HPLC-MS data for the VIILMDPFDDDLK peptide, one of the unique peptides that served to identify the protein long-chain acyl-coenzyme A synthetase 5 (ACSL5) is shown in Supplementary Fig. S4). Functional annotation clustering of the 35 affected proteins using DAVID software yielded the four annotation clusters "guanyl nucleotide-binding", "lipid catabolic process", "protein polymerization", and "mitochondrial membrane" (enrichment scores >2.7, Supplementary Table S3). Analysis using additional gene ontology tools repeatedly indicated links to purine/lipid metabolism as well as the subcellular   Figs S5 and S6). DAVID analysis of data from rat models pooled form the literature 14,21,22 showed effects of MPA on "carbohydrate catabolic processes" and "nucleotide/purine nucleotide metabolic processes" as well as on pathways of carbohydrate and amino acid metabolism (functional annotation clustering and pathway enrichment analysis, please see Supplementary Tables S7 and S8, respectively). Analysis of proteins/genes that were affected by MPA treatment in other human cell types (HEK-293 and CCRF-CEM 26,27 ) did not yield any significantly enriched annotation terms or pathways.
A protein-protein interaction network was generated from the 13 proteins of significantly enriched KEGG functional categories (EASE score < 0.5), i.e. ACSL5, acetyl-coenzyme A acetyltransferase (ACAT2), very long-chain specific acyl-coenzyme A dehydrogenase (VLCAD), peroxisomal acyl-coenzyme A oxidase 1 (AOX), trifunctional enzyme subunit α (TFP), acetyl-coenzyme A acyltransferase (ACAA2), dihydrolipoamide dehydrogenase (DLD), tubulin α -1C chain (TUBA1C), tubulin α -4A chain (TUBA4A), tubulin β -4A chain (TUBB4A), tubulin β chain (TUBB), succinyl-coenzyme A ligase [ADP/GDP-forming] subunit α (SCS-α ), and fatty acid-binding protein 1 (FABP1) with the purpose of identifying additional potentially affected proteins. The protein-protein interaction network generated based on the BioGrid database (Fig. 1, Panel a) and Human Protein Reference Database (Fig. 1, Panel b) revealed 18 and 55 first-order direct neighbors with more than two interactions, respectively. First-order direct neighbors were sorted based on number of interactions with differentially expressed proteins (Supplementary Table S9). Candidate proteins to be affected by MPA (chosen based on biological/metabolic proximity to differentially expressed proteins and/or interactions with multiple differentially expressed proteins) were interrogated by Western blot analysis, but no differences in protein expression for representative proteins were found.
Western blot analysis. Western blot analysis was carried out for five representative proteins to verify SILAC GeLC-MS results, namely ACSL5 (increased), annexin A1 (ANXA1; increased), solute carrier family 12 member 2 (SLC12A2, decreased), polymeric immunoglobulin receptor (PIgR, decreased), and regenerating islet-derived protein 4 (REG-4; decreased; Fig. 2, Panel a and b), but SILAC results could only partially be confirmed. Two western blots for ACSL5, conducted with different antibodies raised against different epitopes, showed differential ACSL5 expression. While analysis using the first antibody (Sigma, WH0051703M1) showed no significant change for ACSL5 expression for MPA-treated LS180 cells, analysis using the second antibody (Abcam, ab104892) showed significant decreases of ACSL5 for treatment with 100 μ M (p < 0.05) and 250 μ M (p < 0.001) MPA. To illustrate these findings and visualize the discrepancy between MS and western blot results as well as western blot results using the two different antibodies, the ACSL5 amino acid sequence is listed in Fig. 2, Panel c; unique peptides and epitopes are indicated.
In addition, levels of protein expression were investigated by western blot analysis for three representative proteins identified by Pathway Palette analysis as first-order direct neighbors and therefore potentially affected due to metabolic proximity to altered proteins as identified by SILAC and GeLC-MS, i.e. tight junction protein ZO-1 (ZO-1), 14-3-3 protein θ (14-3-3 θ ; simultaneous identification by BioGRID and Human Protein Reference Database (HPRD) databases), and polyubiquitin-C (UBC; 9 interactions assigned by BioGRID). No differences in expression levels of these proteins were found (Fig. 2, Panel d and e).

Discussion
The most important molecular mechanisms of mycophenolate intestinal toxicity described in the literature, as of today, are limited to the covalent binding of the acyl glucuronide metabolite of MPA (AcMPAG) to various intestinal proteins 21 and the disruption of tight junctions 27 . Two follow up studies showed that the disruption of intestinal tight junctions was associated with chromatin histone modifications, including an increase in numbers of unique peptides identified, SILAC heavy/light (H/L) ratios for four MPA concentrations, Pearson product-moment correlation coefficients r for the correlation/linear dependence of H/L ratios and MPA concentrations, and the increase in protein levels after exposure to 250 μ M MPA compared to a H/L ratio = 1 as expected for controls (= calculation as % control). H/L ratios are given as means ± standard deviations (n = 3). Significance was determined for effects of increasing MPA concentrations using one-way ANOVA combined with Scheffe's post-hoc test with */#/○ p < 0.05, **/##/○○ p < 0.005, ***/###/○○○ p < 0.001 vs. treatment with 0.1, 5, and 100 μ M MPA. *Significance vs. treatment with 0.1 μ M MPA, # significance vs. treatment with 5 μ M MPA, ○ significance vs. treatment with 100 μ M MPA. Due to a lack of H/L-labeled controls, only treatments of different MPA concentrations were compared among each other. midkine concentrations 28,29 . In contrast to these targeted studies, as aforementioned, in the present study, we took a comprehensive, non-targeted, unique combined proteo-metabolomic approach that resulted in a much more complete picture of the time-and dose-dependent effects of mycophenolic acid on intestinal cell protein concentrations and cell metabolism at various levels of guanosine supplementation.
The gastrointestinal epithelium is locally exposed to high MPA concentrations, especially at the locations where MMF/MPA is released from the formulation, which are in the range of the concentrations tested in the present study. Other studies used MPA at lower concentrations (e.g. 0.1-5 μ M 30 , 10 μ M 31 ) in vitro than the present study and guanosine concentrations necessary to reverse MPA-induced effects were around 50-100 μ M 30,31 . Such studies evaluated inhibition of cell proliferation, which depends on availability of guanosine nucleotides, but does not necessarily reflect levels of intracellular guanosine pools and the concentrations present at the intestinal epithelium. Due to the differences in MPA and guanosine concentrations, our results cannot directly be compared to such previous studies. As in the present study supplementation even with concentrations as high as 1 mM guanosine could not restore guanosine nucleotide levels, it is likely that a similar dysregulation of nucleotides occurs intracellularly in vivo. In particular GDP-, UDP-, and CDP-linked intermediates play a key role in lipid metabolism, membrane synthesis, and protein glycosylation 32,33 , suggesting negative effects of MPA on epithelial barriers. Interactions between interconnecting proteins are not shown. 18 (BioGRID) and 57 (HPRD) protein pairs linked through first-order shared neighbors and 0 (BioGRID) and 3 (HPRD) direct interactions between the proteins of affected pathways were found. Star-shaped nodes represent proteins found in significantly affected pathways annotated through Pathway Enrichment Analysis using DAVID and KEGG with a color scheme corresponding to pathways as classified in the legend and Supplementary Table S1. Proteins are indicated by their gene names. Proteins with ≥ 3 interactions are listed in Supplementary Table S2 including abbreviation, gene name accession number, and physiological function.
Binding of guanosine nucleotides was associated with the function of eight of the 35 differentially expressed proteins in LS180 cells (Supplementary Table S4, #1: "guanyl nucleotide-binding"). Effects of MPA have been attributed to intracellular guanosine nucleotide depletion 34-37 , e.g. changes in protein glycosylation especially studied with respect to modifications of adhesion molecules 36,37 . Nevertheless, the notion that intracellular guanosine nucleotides in GI epithelial cells are abundant due to import of guanosine from the intestinal lumen so that MPA-mediated inhibition of de novo synthesis of GTP has little or no effect 13 , is not confirmed by the present study.
Effects of MPA and its metabolite AcMPAG on tubulin polymerization (even in the presence of exogenous GTP) have been described by Feichtinger et al. 37 . Effects on tubulins were also seen in LS180 cells; tubulin subunits constitute 50% of proteins of group #1 ("guanyl nucleotide-binding") and 100% of proteins of group #3 ("protein polymerization"). Interestingly, potentially affected proteins identified by Pathway Palette are also linked to enriched functions/annotation terms (i.e. regulation of cell cycle, signaling processes, transcription, tight junctions, cytoskeletal properties, and transport; Supplementary Table S3).
Feichtinger et al. 37 reported induction of tubulin polymerization, while in the present study the cellular amount of four tubulin subunits was decreased to 74.0-79.2% of control values in LS180 cells treated with 250 μ M MPA (Table 2). Morath et al., on the other hand, found cytoskeletal proteins (vinculin, tubulin) to be downregulated in human fibroblasts, suggesting that MPA exposure results in cytoskeletal dysfunction 15 . Literature on MPA's effect on lipid and fatty acid metabolism is scarce 38 . However, clinical studies show association of MMF and hyperlipidemia 38 , which is in accordance with our findings that expression of four proteins attributed to the term "Lipid catabolic process" (Supplementary Table S3) were significantly increased. Conversely, no significant changes occurred in serum cholesterol or triglyceride levels of rabbits on high-cholesterol diets with or without MMF in a study by Subramanian et al. 38 .
Data from studies involving LS180 cells differ from corresponding rat model data (Supplementary Table S5 vs. S8). These differences are likely due to differences in model systems (in vitro vs. in vivo), species (human vs. rat), doses, platforms, and data analysis methods. Comparing proteomics data in the present study with  Supplementary Table S5 vs. #6 in Supplementary Table S8), four proteins involved in fatty acid metabolism were affected in rats, but Functional Annotation Clustering of rat model data (Supplementary Table S7) yielded only clusters related to carbohydrate or nucleotide processes. Sufficient data on MPA's effects on glucose metabolism is lacking 38 . No effects of MMF on glucose metabolism were observed in clinical trials and effects of MMF on insulin secretion and insulin gene expression seem to differ between species in vitro (MMF inhibits insulin secretion in rat islets 39 , but does not show effects on insulin secretion or insulin gene expression in human islets 40 ). Although multiple other factors such as the cancerous nature of LS180 cells, methodological differences, and sample size may have an effect, this pattern can be traced when comparing the present results from the LS180 cell model with in vivo data from rat models. As shown in Supplementary Table S7 (rat), the term "carbohydrate catabolic process" is listed as the most enriched term, but is not listed among Functional Annotation Clusters from analysis of LS180 cell data (Supplementary Table S3). Pathway Enrichment Analyses of 35 proteins in LS180 cells affected by MPA exposure and 76 proteins/genes affected by MPA exposure in rats (Supplementary Table S3) show the same model-dependent pattern using the KEGG database and DAVID's Functional Annotation Chart tool (Supplementary Table S5 vs. S8). For LS180 cells, several of the listed terms were linked to lipid metabolism (e.g. #1: fatty acid metabolism, #5: propanoate metabolism, #6: fatty acid elongation in mitochondria, #7: peroxisome proliferator-activated receptor (PPAR) signaling pathway; Supplementary Figs S5 and S6). Analysis of rat proteome data revealed several enriched terms linked to lipid metabolism (#4: propanoate metabolism, #6: fatty acid metabolism, #11: butanoate metabolism). On the other hand, alterations in the expression of proteins involved in glucose and protein metabolism/amino acid degradation were primarily observed in rats undergoing MMF treatment (i.e. #1: glycolysis/gluconeogenesis, #2: pyruvate metabolism, #3: arginine and proline metabolism, #7: tryptophan metabolism, #8: phenylalanine metabolism, #10: fructose and mannose metabolism), while only two pathways involved in amino acid degradation were significantly affected in LS180 cells (i.e. #2: valine, leucine, and isoleucine degradation, #5: propanoate metabolism). To corroborate an increase in ACSL5 observed in the SILAC proteomics experiment after exposure of LS180 cells to MPA, samples were also analyzed using western blot. At first only a small, non-significant change in expression levels was found. Nevertheless, repeat analysis using a different antibody showed a marked decrease in ACSL5 protein levels after exposure to 100 μ M and 250 μ M MPA (66.4 ± 8.3% and 8.7 ± 1.2% of controls). Only unmodified peptides and standard variable modifications were included for protein quantification in MaxQuant. The approximately 2-fold increase in ACSL5 assessed in SILAC GeLC-MS experiments related to these unmodified peptides. The decrease seen in western blots using the second anti-ACSL5 antibody was likely due to a modification of the antibody binding site (marked in blue in the ACSL5 amino acid sequence in Fig. 2, Panel c), as this site is not part of a unique peptide that served for protein identification (marked in bold and brackets in Fig. 2, Panel c).
Moreover, compared to controls, UTP and GTP levels were dysregulated in MPA-treated LS180 cells despite supplementation with 1 mM guanosine. Nucleotides are intermediates in the glycosylation of proteins and lipids 41,42 : glucose, galactose, and various amines are transferred to proteins via UDP intermediates, fucose and mannose are transferred via GDP. Although MPA induces inhibition of glycosylation of proteins through depletion of guanosine nucleotides 12,43 , augmentation of glycosylation in LS180 cells may occur due to increased UTP ( Supplementary Fig. S1).
While increases in ANXA1 expression during MPA exposure as observed in the SILAC proteomics analysis could be verified by western blot (Fig. 2, Panel a and b), no changes in SLC12A2, PIgR, and REG-4 expression were detected by western blot. Nevertheless, for the following reasons, this does not preclude the SILAC GeLC-MS results from being correct. As discussed for the ACSL5 western blot results, the generally semi-quantitative nature of western blot analyses and MPA's known ability to strongly influence protein glycosylation may have compromised the western blot analysis. Moreover, the results from nucleotide HPLC-MS, DAVID, and NMR-based metabolomics analyses and the literature 22 support that MPA indeed affects SLC12A2 and PIgR expression.
The majority of affected proteins (Table 2) is involved in lipid metabolism. Proteins of "Fatty acid metabolism", "Fatty acid elongation in mitochondria" (Supplementary Table S5), and "Lipid catabolic processes" (Supplementary Table S3 and S6) were significantly upregulated in LS180 cells after MPA treatment (with the exemption of ACAT2, which was downregulated). The term "mitochondrial membrane" (Supplementary Table S3) lists almost exclusively proteins that were found to be increased (with the exemption of SCS-α , which was downregulated). These results also suggested imbalances in membrane composition.
To further examine the impact of changes in the expression of proteins involved in lipid metabolism, intracellular lipid patterns were assessed using an 1 H-NMR-based metabolomics approach. Cholesterol is a precursor for signaling molecules and is a fundamental constituent of cell membranes 44 and changes of its intracellular levels may mediate drug toxicity. With the observed increases in phosphatidylcholines and cholesterol levels, homeostasis of two major membrane constituents is affected in LS180 cells exposed to MPA. In fact, phosphatidylcholines are the most prominent membrane phospholipids and crucial for maintenance of GI barrier function 45 . Intracellular phosphatidylcholines are secreted by epithelial cells and passaged across tight junctions into the apical mucus layer 46 , where they contribute to the establishment of a hydrophobic surface of the colonic mucus layer 45 . In the plasma membrane of enterocytes, phosphatidylcholine modulates the mucosal signaling state as lipid composition of membranes is a regulatory parameter of inflammatory responses 45 . Decreased levels of luminal phosphatidylcholines in colonic mucus have been linked repeatedly to ulcerative colitis, an inflammatory bowel disease similar to Crohn's disease 45,46 , but this finding does not apply to Crohn's disease patients 47,48 . Upregulation of intracellular phosphatidylcholines in MPA-treated LS180 cells suggests disturbances of lipid levels linked to mucosal defense. Phosphatidylcholine is synthesized from choline via the Kennedy-pathway involving cytidine nucleotides 49 . Phosphocholine and CTP are formed from CDP-choline and pyrophosphate, consecutively CDP-choline and diacylglycerol (or alkyl-acylglycerol) are converted to phosphatidylcholine (with CMP as byproduct). Slightly elevated levels of cytidine nucleotides observed at 72 h with 100 μ M MPA and 1 mM guanosine ( Supplementary Fig. S2, Table 1) may be associated with increased phosphatidylcholine biosynthesis. Furthermore, increases in fatty acid, diacylglycerol, and triacylglycerol may also be related to impaired composition of membrane lipids, e.g. other phospholipids than phosphatidylcholines, which constitute membranes 50 .
The major limitation of this study is the use of a cancer cell line and the associated potential differences in metabolism, protein expression and their regulation compared to normal human intestinal cells. Nevertheless, as aforementioned, the results of the present study were cross-referenced to currently available proteomics data on MPA toxicity in rat models and other human cell culture models using cells of non-cancerous origin using DAVID (pathway enrichment analysis using KEGG). Overlap of the results with those of the present study support the validity of the LS 180 cell model. Although it was not our objective to validate the LS180 cells culture model, the results of the present study are further proof that LS180 cells are an attractive model as a fast and simple screen for potential intestinal toxicity of drugs and drug candidates beyond MPA and are an alternative to more cumbersome but widely used Caco-2 cell models. An interesting question is if, and how, the observed proteomics and metabolomics changes are associated with impairment of the intestinal barrier. The present study was not designed to assess this question and this will be evaluated in a follow up.
In conclusion, the present study based on the LS180 cell model and a metabolomics-proteomics profiling strategy suggests that MPA-induced GI disturbances involve the dysregulation of nucleotide-dependent processes and lipid metabolism. Our data support that MPA's GI toxicity is linked to the drug's mechanism of action resulting in disturbance of intracellular guanosine levels, that are important not only for cell proliferation but also other vital intracellular processes. Importantly, the results of the present study suggested that, other than hypothesized, guanosine supplementation does not fully reverse the negative effects of MPA on nucleotide metabolism. The negative effects of MPA in the present study were concentration-dependent suggesting that the avoidance of high local concentrations in the intestine, for example by the development of sustained release formulations, may improve its GI tolerability.