Metabolomics of Human Intestinal Transplant Rejection



Surveillance endoscopy with biopsy is the standard method to monitor intestinal transplant recipients but it is invasive, costly and prone to sampling error. Early noninvasive biomarkers of intestinal rejection are needed. In this pilot study we applied metabolomics to characterize the metabolomic profile of intestinal allograft rejection. Fifty-six samples of ileostomy fluid or stool from 11 rejection and 45 nonrejection episodes were analyzed by ultraperformance liquid chromatography in conjunction with Quadrupole time-of-flight mass spectrometry (UPLC-QTOFMS). The data were acquired in duplicate for each sample in positive ionization mode and preprocessed using XCMS (Scripps) followed by multivariate data analysis. We detected a total of 2541 metabolites in the positive ionization mode (mass 50–850 Daltons). A significant interclass separation was found between rejection and nonrejection. The proinflammatory mediator leukotriene E4 was the metabolite with the highest fold change in the rejection group compared to nonrejection. Water-soluble vitamins B2, B5, B6, and taurocholate were also detected with high fold change in rejection. The metabolomic profile of rejection was more heterogeneous than nonrejection. Although larger studies are needed, metabolomics appears to be a promising tool to characterize the pathophysiologic mechanisms involved in intestinal allograft rejection and potentially to identify noninvasive biomarkers.




Human Metabolome Database


Madison Metabolomic Consortium Database


mass spectrometry


nuclear magnetic resonance


Nucleotide-Binding Oligomerization Domain 2


total ion chromatogram


ultraperformance liquid chromatography in conjunction with Quadrupole time of flight mass spectrometry


Intestinal transplantation has evolved from being an experimental procedure to become an effective treatment option for patients with irreversible intestinal failure suffering life-threatening complications from parental nutrition (1). The results of intestinal transplantation have continued to improve over the last decade. The availability of more potent immunosuppression compared to the early era, especially tacrolimus, and antibody induction, resulted in decreased incidence of acute rejection from 70–90% to 30–50% within 3 months posttransplant. As a result, current patient survival of 90% or higher at 1 year posttransplant, comparable to the survival of patients on home parenteral nutrition, is achieved in specialized centers. It is therefore likely that the indications for intestinal transplantation will expand in the future.

Intestinal transplantation poses a greater immunologic challenge compared to other solid organs due to its high immunogenicity, rich composition of immune cells and heavy colonization by microbes. Severe acute rejection remains the most common cause of graft loss and death. Early diagnosis and treatment of acute rejection are critical for the successful reversal of the rejection process.

Currently intestinal transplant recipients are monitored with frequent surveillance endoscopies and biopsies. Limits of endoscopic biopsy include invasiveness, sampling error, and cost. In addition, biopsy findings are at times insufficient to discriminate between allograft rejection and infectious enteritis thus making it difficult to obtain a differential diagnosis between these two processes based purely on histopathologic criteria. Frequently the manifestations of rejection and infectious enteritis overlap and it is essential to differentiate the two causes of allograft dysfunction because the augmented immunosuppression required to treat rejection would have disastrous consequences in presence of viral infection.

Metabolomic analysis (metabolomics) consists of a comprehensive assessment of small molecule metabolites (< 800 Da) in a biologic sample utilizing high resolution analytics (Nuclear Magnetic Resonance [NMR] or Mass Spectrometry [MS] together with chemometric statistical tools (2–4). Using nontargeted and targeted analysis, metabolomics allows the detection and quantification of thousands of metabolites due to the high sensitivity and resolution of MS. Metabolites include amino acids, carbohydrates, nucleic acids, organic acids, polyphenols, vitamins and others. The characterization of small molecule profiles can be utilized to assess the functional phenotype of a cell, tissue or organism. Like genomics and proteomics, this ‘system-biology’ approach aims to characterize comprehensively the response of a biological system to disease (5). Metabolomics potentially offers a highly informative approach to monitor intestinal transplant recipients and to diagnose noninvasively allograft rejection before the appearance of pathologic changes in biopsy specimens. The rationale for its use is that inflammation of the intestinal allograft during a rejection episode perturbs the equilibrium of the mucosa and of the intraluminal flora, producing metabolites that can be detected in the effluent fluid from ileostomy or in fecal extracts. These metabolites derive from exfoliation, secretion, or leakage by the intestinal epithelium and/or from intraluminal production by microbes or from an exogenous source. We previously used metabolomics to characterize the microbial composition of the transplanted intestine and found that a shift toward facultative anaerobes occurs in the microflora of the intestinal lumen after transplant and that the flora returns to normal composition after closure of the ileostomy (6). In this pilot study we investigated the feasibility of metabolomics to identify noninvasive biomarkers of intestinal allograft rejection.

Materials and Methods

We collected 5 mL of effluent fluid from ileostomy (n = 21) or colostomy (n = 1) or stool (n = 11) or fluid aspirate during colonoscopy (n = 3) from 36 patients (16 male 20 female, 16 children with median age 5 years [range: 6 months–15 years] and 20 adults with median age 46.5 years [range 21–65]) after a median interval of 66 months (range 0–118) from intestinal transplantation. Samples were collected during routine scheduled ileoscopy/colonoscopy or follow-up clinic visits. The etiologies of intestinal failure are listed in Table 1. The grafts were 27 isolated intestine, 6 liver-intestine, 2 multivisceral (i.e. liver-intestine including stomach and duodenum +/− colon) and 1 modified multivisceral (i.e. multivisceral without the liver). Routinely a loop ileostomy was created at the time of transplant for posttransplant monitoring and taken down after 3–6 months. Immunosuppression consisted of induction with nondepleting antibodies (basiliximab) or depleting antibodies (thymoglobulin) in highly sensitized patients followed by maintenance with tacrolimus, sirolimus and prednisone. Posttransplant monitoring included endoscopy and biopsy twice weekly for the first 6 weeks, weekly until month 3, biweekly until month 6 and monthly thereafter. Rejection was diagnosed by standard histological criteria (7,8) in the presence of increased and/or bloody ileostomy output and graded as mild, moderate, or severe. Treatment of rejection consisted of steroid pulse therapy, increased tacrolimus levels and/or thymoglobulin depending on the severity of the rejection episode. All patients provided written informed consent to participate in this IRB approved study (IRB/GUACUC #: 2004–008).

Table 1. Causes of intestinal failure
  n =
Short gut syndromeMesenteric thrombosis7
 (n = 27)Necrotizing enterocolitis4
 Volvolus after gastric bypass4
 Crohn's disease1
 Gardner's syndrome1
 Radiation enteritis1
Dysmotility disordersPseudo-obstruction5
 (n = 9)Congenital secretory diarrhea1
 Hirschsprung's disease1
 Microvillous inclusion disease1
 Tufting enteropathy1

We collected and analyzed a total of 56 samples of fluid from ileostomy or stool from 36 patients: 24 patients provided 1 sample, 6 patients had 2 samples, 4 patients had 3 samples and 2 patients had 4 samples. Samples collected from the same patient at different time points posttransplant and under different clinical circumstances (before and during rejection) were included and analyzed individually as representative of a distinct clinical episode with regard to rejection/nonrejection. Each sample was labeled and refrigerated at 4° C for 1–12 h immediately after collection, aliquoted in 1 mL samples and stored at –80°C. A total of 150 μL of the sample was processed using the sequential metabolite extraction as described previously (9). LC/MS-grade acetonitrile (ACN), water and methanol were purchased from Fisher Scientific (Sommerville, NJ, USA). High-purity formic acid (99%) was purchased from Thermo-Scientific (Rockford, IL, USA). Debrisoquine was purchased from Sigma (St. Louis, MO, USA).

Metabolite extraction from ileostomy fluid or stool sample was performed using the protocol described by Sheikh et al. (9). Briefly, a thawed suspension of 25 μL of ileostomy fluid or stool sample was diluted in water to a final volume of 150 μL and sonicated. Subsequently, 600 μL of methanol containing internal standard [debrisoquine(1 mg/mL)] was added, the samples were vortexed and incubated on ice for 15 min, followed by addition of equal volume of chloroform. The tubes were vortexed and centrifuged at 13,000 rpm for 10 min at 15°C. The two solvent phases were recovered separately, dried, combined for MS analysis and transferred to a fresh tube carefully avoiding the interface. Protein was precipitated by adding chilled ACN (600 μL) to each tube, vortexed and incubated at −80°C overnight. The tubes were centrifuged at 13,000 rpm for 10 min at 4°C. The supernatant was transferred to fresh tubes and dried under vacuum. The contents of two tubes for respective sample were combined by resuspending in 100 μL of solvent a (98% water 2% ACN), followed by UPLC-TOFMS analysis. Each sample (5 μL) was injected onto a reverse-phase 50 × 2.1 mm ACQUITY 1.7 μm C18 column (Waters Corp, Milford, MA) using an ACQUITY UPLC system (Waters) with a gradient mobile phase consisting of 2% ACN in water containing 0.1% formic acid (A) and 2% water in ACN containing 0.1% formic acid (B). Each sample was resolved for 10 min at a flow rate of 0.5 mL/min. The gradient consisted of 100% A for 0.5 min then a ramp of curve 6 to 100% B from 0.5 min to 10 min. The column eluent was introduced directly into the mass spectrometer by electrospray. Mass spectrometry was performed on a QTOF Premier (Waters) operating in positive-ion (ESI+) electrospray ionization mode with a capillary voltage of 3200 V and a sampling cone voltage of 35 V in positive mode. The desolvation gas flow was set to 800 liters/h and the temperature was set to 350°C. The cone gas flow was 25 L/h, and the source temperature was 120°C. Accurate mass was maintained by introduction of LockSpray interface of sulfadimethoxine (311.0814 [M+H]+) at a concentration of 250 pg/μL in 50% aqueous ACN and a rate of 150 μL/min. The data were acquired in duplicate using an ACQUITY UPLC system attached with an electrospray Quadrupole time-of-flight tandem (ESI-Q-TOF) as described in Ref. 10. The variability in the starting amounts of the material greatly contributes to downstream interpretation of results. Hence, we performed total protein estimation in each sample using Bradford's method. This was used to normalized the raw ion intensity before proceeding for statistical and multivariate data analysis. The centroided data from UPLC-TOFMS acquired in the positive mode were preprocessed using XCMS (Scripps) to generate a data matrix containing feature intensities, mass to charge (m/z) and retention time values followed by multivariate data analysis using the SIMCA-P software (Umetrics, Inc.) and the Random Forests (11). The online Madison Metabolomic Consortium Database (MMCD) (12) and the Human Metabolome Database (HMDB) (13) were interrogated for accurate mass based identification of metabolites.

Statistics: we calculated sample size to identify metabolites that are differentially abundant in rejectors and nonrejectors, controlling the false discovery rate (FDR). Conservatively, we focused on 600 metabolites in positive or negative mode. At worst, we projected that 60 (10%) of the 600 metabolites investigated will be differentially expressed. We anticipated the observed standard deviation of the log-expression difference to be approximately 0.5 on a log to base 2 scale. A sample of size 10 per group achieves a 90% power for each metabolite to detect a true difference in expression of at least twofold with an estimated standard deviation of 1.0 unit with a false discovery rate of 1% using a two-sided paired t-test, with anticipated true mean difference of expression over twofold. Of these, at least 30% are expected to be identified. Thus our sample size will provide ample power to be used as a pilot set for discerning informative metabolites.


A total of 2541 metabolites were detected with a mass range between 80 and 850 Daltons in 56 samples of effluent fluid from ileostomy or stool, including 11 samples collected during an episode of acute rejection and 45 samples collected during nonrejection. Metabolite rankings demonstrated different metabolomic profiles between rejection and nonrejection (Figure 1). We used the SIMCA-P+ software to generate PLS-DA two-component models. PLS-DA is a supervised method that uses multiple covariance between a data set and the class labels. It sharpens the separation between groups of observations by rotating principal component analysis (PCA) components so as to achieve maximum separation among classes, and to delineate the variables that carry the class separating information.

Figure 1.

OPLS loadings score plot comparing metabolite changes in rejection and nonrejection.

Additionally, we used the Random Forests which is an ensemble of unpruned classification or regression trees that uses random feature selection in the tree induction process. It is a group separation method based on “growing” an ensemble of decision tree classifiers. A measure of the importance of classification variables is also calculated by considering the difference between the results from original and randomly permuted versions of the data set. Random Forests avoids over fitting of data and is independent of ion intensity for the selection of features. The display of metabolites on a heatmap (Figure 2) revealed a number of upregulated (panel B) and downregulated (panel C) metabolites in the rejection group compared to nonrejection. Random Forests analysis confirmed the presence of a significant interclass separation between rejection and nonrejection when all metabolites and the top 50 metabolites were considered with accuracy >86% and >97%, respectively (Figure 3 and 4). The two computational methods use different algorithms to build classifiers for group segregation and hence a slight difference in the pattern of classification of samples was observed. However, the overlap between the top 50 metabolites determined by each method imparts significance and “in silico” validation of the putative biomarkers. The group of metabolites from rejection episodes appeared more dis-homogeneously clustered compared to nonrejection episodes. We repeated this study twice for reproducibility and confirmed that the metabolic profiles of nonrejectors exhibited significant homogeneity while the profiles from rejectors showed large intraclass spread. This could well represent different phenotypes of rejection. Future studies with larger cohorts would allow subgroup classification-based analysis of this heterogeneity.

Figure 2.

Multivariate analysis of the UPLC-TOFMS data. A. Heatmap visualization of the metabolite rankings comparing relative levels in ileostomy fluid or stool from episodes of rejection versus nonrejection. Each column represents a patient's sample and each row represents a unique metabolite with a characteristic mass to charge and retention time value. B. Box plots of selected metabolites upregulated in rejection from panel 1 of heatmap. C. Box plots of selected metabolites downregulated in rejection from panel 2 of heatmap.

Figure 3.

Random Forests analysis. Interclass separation of all 2541 metabolites between nonrejection (red) and rejection episodes (blue).

Figure 4.

Random Forests analysis. Interclass separation of top 50 metabolites between nonrejection (red) and rejection episodes (blue).

A total of 477 (19%) of the 2541 detected metabolites demonstrated a significant fold change between rejection and nonrejection: 249 metabolites had fold change ≥2 (range 2–20) and 228 showed fold change ≤ 0.49 (range 0.49–0.006) in the rejection group compared to nonrejection.

Metabolites with significant fold change between rejection and nonrejection were mass-based searched by consulting the online MMCD and HMDB. This search yielded the identification of 104/477 (22%) single match metabolites, including 49 with fold change ≥2 and 55 with fold change ≤ 0.49, respectively. The fold change was calculated as the ratio of normalized intensity in rejectors and nonrejectors for the given metabolite. The remaining 373/477 (78%) detected metabolites with significant fold change between rejection and nonrejection remain to be identified due to absence of matching with known metabolites of online databases (n = 187) or to matching with multiple (range 2–26) known metabolites (n = 186).

Although this was a nontargeted approach to characterize metabolomic profiles rather than to identify specific metabolites or metabolic pathways, we were able to detect leukotriene E4, a proinflammatory cysteinyl leukotriene product of metabolism of arachidonic acid, as a metabolite with the highest intensity and fold change (>20) in the rejection group compared to nonrejection. In addition, a preliminary identification of other significant known metabolites revealed that D-pantethine, the dimeric form of pantothenic acid (vitamin B5), pyridoxal-5-phosphate (vitamin B6), taurocholate (a salt of taurocholic acid, one of the bile acids) and riboflavin (vitamin B2) were detected with fold change 9.3, 3.9, 3.7 and 3.2 in the rejection group compared to nonrejection, respectively. The characterization of metabolic pathways involving several other metabolites detected in this preliminary study is ongoing.

The composition of metabolites was analyzed in each sample and represented graphically by total ion chromatogram (TIC). The chromatogram displays the metabolites detected at different time points between 0 and 10 min (abscissa). The metabolomic profile is composed of multiple peaks representing different metabolites whose height reflects their intensity (ordinate). The analysis of chromatograms of samples collected longitudinally from the same patient at different time point after transplant revealed the changes in the metabolomic profile occurring at the time of rejection. An example of change in metabolomic profile before and during rejection in the same patient is reported in Figure 5. In this patient the profile of a sample of effluent fluid from the ileostomy collected during rejection demonstrated upregulation of two metabolites detected at 5.81 and 7.06 min. These metabolites were detected also in the prerejection sample, although at a much lower intensity, and their identification is ongoing.

Figure 5.

Total ion chromatogram of two samples from the same patient before and during rejection episode: two metabolites detected at 5.81 and 7.06 min are upregulated during rejection compared to baseline.

Likewise, we analyzed sequentially four chromatograms of samples collected from another patient at different time points before rejection and during a rejection episode (Figure 6). The analysis of ileostomy effluent collected at 1 week, 4 weeks and 8 weeks posttransplant in the absence of rejection shows a comparable metabolomic profile for metabolites detected between 4.7 and 8.13 min. However, the sample collected at 10 weeks posttransplant at the time of biopsy proven acute rejection shows a downregulation of metabolites detected in the previous nonrejection samples. Furthermore, metabolites detected at a low intensity in the prerejection samples at 2.84 and 4.2 min of the mass spectrometry analysis demonstrated upregulation during the rejection episode compared to previous nonrejection samples.

Figure 6.

Total ion chromatogram of four samples of ileostomy fluid from the same patient. Three metabolites detected at 2.86, 3.12 and 4.2 min showed upregulation during rejection compared to before. Several metabolites detected between 4.8 and 8.17 min were downregulated during rejection compared to before.


So far in intestinal transplantation there is no molecular profile or individual biomarker comparable to serum creatinine in kidney transplant or transaminases and bilirubin in liver transplant with useful clinical application to diagnose allograft rejection. In this pilot study we tested the feasibility of metabolomics to monitor intestinal transplant recipients with the goal of characterizing the metabolomic profile of allograft rejection and identifying noninvasive biomarkers from the effluent fluid from ileostomy or stool sample.

Metabolomics consists of a comprehensive analysis of small molecules (mass range 50–850 Da) in a biologic sample. The composition of low-molecular weight metabolites in cells or tissues can be viewed as the functional phenotype of biologic processes. The analysis of the metabolome is an attractive approach to understand and characterize disease processes since small molecules including amino acids, carbohydrates, fatty acids, nucleic acids, vitamins, inorganic species and others represent the composite output of cellular processes from the genomic level downstream. Metabolomics has only recently been applied to organ transplantation (14,15) and few groups have published the results of metabolomics for the evaluation of acute rejection (16), graft dysfunction (17), ischemia–reperfusion injury (18) and for drug monitoring (19).

In kidney transplantation a number of urinary and serum biomarkes have been identified to correlate with a rejection episode including trimethylamine-N-oxide (TMAO) (20) and others (21,22). In liver transplantation a change in the bile acids profile between pre- and postreperfusion bile samples has been recently reported by Legido-Quigley et al. (23).

Metabolomics has also been applied to other disease processes including, among others, pancreatic cancer (24), colon cancer (25) and inflammatory bowel disease (26–28).

Severe acute rejection is the most common cause of graft loss and mortality in intestinal transplantation. It is critical to diagnose rejection at an early stage in order to successfully reverse the process and save the graft. The advantage of adding molecular profiling to the current pool of diagnostic studies and monitoring protocol would be to potentially increase the accuracy of early diagnosis by detecting molecular changes not yet manifested histopathologically or clinically. Furthermore, if the phenotypic expression of severe intestinal allograft rejection usually manifests clinically and endoscopically with a rather uniform picture dominated by increased ileostomy output or diarrhea with bleeding and sloughing of the mucosa, nonetheless it results from a complex interactions of genetic, environmental (microflora), dietary and local tissue factors. To understand the dysregulations of the intestinal homeostasis occurring in the intestinal graft during a rejection episode at this level of complexity an integrated approach is required that includes analysis of perturbations in the network of genes, proteins and metabolites. The application of new analytical and bioinformatics techniques allows a comprehensive approach to complex biological samples and the detection and characterization of a large number of chemically diverse structures. Marchesi et al. first published the results of NMR analysis of fecal extracts from patients with inflammatory bowel disease and reported a reduced concentration of short chain fatty acids (acetate, butirate) and increased level of aminoacids (alanine, leucine, isoleucine, lysine, valine) in patients with Crohn's disease compared to controls (29). Jansson et al. performed a nontargeted analysis of fecals samples from patients with Crohn's disease and found a different composition in amino-acids (tyrosine, tryphtophane, phenylalanine), bile acids, arachidonic acid and other fatty acids between patients with intestinal involvement of Crohn's disease compared to patients with disease predominantly affecting the colon (30).

In our pilot study we applied metabolomic analysis to the effluent fluid from ileostomy or stool of intestinal transplant recipients using nontargeted mass spectrometry and found it to be a very sensitive technique to detect a large number of metabolites. We demonstrated an interclass separation of metabolites detected during rejection and nonrejection episodes. This approach revealed that different clusters of metabolites appeared upregulated and others downregulated in rejection samples compared to nonrejection (Figure 1). The analysis of all 2541 metabolites (Figure 3) and of the top 50 metabolites (Figure 4) showed homogeneous clustering of metabolites in nonrejection samples compared to rejection samples. Although this may in part reflect the smaller sample size of rejection episodes, nevertheless it raises the possibility that a different composition of metabolites in the effluent fluid from ileostomy during rejection compared to nonrejection results from the tissue modifications occurring in the graft as seen on biopsy.

Among the large number of detected metabolites, 477 (19%) compounds were identified with significantly different normalized intensity in rejection compared to nonrejection. Of those, the consultation of metabolomics databases allowed the identification of 104 single match metabolites. The characterization of these metabolites and their metabolic pathways is ongoing.

In this study we elected to adopt a nontargeted approach with the aim of describing the metabolomic profile of the effluent fluid from ileostomy of intestinal transplant recipients rather than identifying individual biomarkers of rejection. However, from a preliminary analysis of detected metabolites, we identified N-acetyl-leukotriene E4 as the metabolite with the highest fold change between rejection and nonrejection. Leukotrienes are potent biologically active molecules produced by leukocytes and other cells during an inflammatory response (31). Interestingly, Shibata et al. reported an increased leukotriene level in the exhaled breath and urine of children with active asthma (32). Based on these preliminary findings, we hypothezise that the detection of increased inflammatory biomarkers in the effluent fluid from ileostomy may potentially provide a noninvasive tool to diagnose rejection.

Additionally, we found that three water-soluble vitamins (B2, B5 and B6) were detected with higher fold change in rejection compared to nonrejection. Since vitamins are not endogenously produced, the intestinal sources of vitamins are dietary and bacterial. Dietary vitamins are processed and absorbed in the small intestine. Diseases of the small intestine including inflammatory bowel disease have been associated with vitamin deficiency secondary to malabsorption (33). In addition, changes in the microflora can lead to vitamin deficiency. It is reasonable to speculate that during rejection the inflammation of the mucosa and/or changes in the microflora resemble a condition of vitamin malabsorption leading to increased vitamin content in the effluent fluid from ileostomy.

Taurocholate, a salt of the bile acid taurocholic acid was also detected with higher intensity in rejection samples compared to nonrejection. As previously reported by Jansson et al. (30) in patients with active Crohn's disease involving the small intestine, this may represent a form of malabsorption of bile acids at the level of the terminal ileum secondary to the inflammation associated with rejection.

The majority of metabolites detected in our study remain unidentified. It is not uncommon, given the high sensitivity of the technique, to detect a large number of metabolites in biologic samples, the majority of which remain unidentified for multiple reasons including the presence of isotopic variants of the same metabolite and the still incomplete library of human metabolites (34).

In this study we included samples of effluent from ileostomy collected from the same patient at different time-points after intestinal transplantation. For the purpose of this study, each sample was collected under different clinical circumstances (i.e. rejection/nonrejection) and considered expression of a biologically different phenomenon and therefore analyzed separately. The analysis of chromatograms of multiple samples obtained from the same patient at different time point posttransplant showed a change in the metabolomic profile between nonrejection and rejection. Although the nature of several detected metabolites remains to be determined, as in the case of nontargeted profile analysis, nevertheless a different metabolomic profile was observed in the sample of ileostomy fluid collected at the time of acute rejection compared to samples collected during nonrejection episodes from the same patient. A targeted analysis and quantification of selected metabolites with significant intensity difference between rejection and nonrejection is ongoing and will potentially allow the identification of a pool of metabolites whose aggregate intensity could represent a score of “rejection risk” in the individual patient. In fact, it would be challenging to find among so many metabolites a single compound that correlates with rejection due to the individual genomic variability, different composition of microbiota, metabolism and diet.

Our study highlights the potential role of a “system biology” approach to transplantation (35). In intestinal transplantation, other groups have recently reported on the application of proteomics and on the study of the microbiome. Kumar et al. detected the upregulation of 17 protein features during rejection, including human neutrophil peptide 1 and 2, highlighting the role of innate immune activation during rejection (36). Oh et al. recently described the changes in the ileal microbiota associated with rejection and found an increased proportion of Enterobacteriaceae and decreased proportion of Firmicutes and Lactobacilli compared to nonrejection (37).

Given that multiple genetic and environmental factors contribute to complex disease phenotypes including allograft rejection, integrated analysis of clinical, microbiological and molecular variables is required. The use of new high-throughput sequencing technologies in combination with bioinformatics and biostatistics potentially can increase the diagnostic accuracy of intestinal allograft rejection and allow the development of noninvasive biomarkers of allograft function.

To our knowledge this is the first study to apply metabolomics by mass spectometry to the effluent fluid from ileostomy to monitor intestinal transplant recipients. This approach yielded a large amount of information using a noninvasive and relatively nonexpensive technique. Although it is unlikely that a close relationship will be identified between a single metabolite in the ileostomy effluent and allograft rejection to allow the routine use of a biomarker comparable to serum creatinine in kidney transplant recipients, nevertheless metabolomics appears to be a potentially useful tool to investigate the patterns of molecular derangements associated with intestinal rejection and to uncover its pathophysiologic mechanisms. Future studies will use validation cohort data sets which will enable us to calculate the sensitivity and specificity of putative biomarkers. Larger and collaborative studies would further delineate the role of metabolomics in organ transplantation and bring it closer to the clinic.


The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.