Secondary Logo

Journal Logo


Open:Mucosal-luminal interface proteomics reveals biomarkers of pediatric inflammatory bowel disease-associated colitis

Deeke, Shelley A. MSc1; Starr, Amanda E. PhD1; Ning, Zhibin PhD1; Ahmadi, Sara MSc1; Zhang, Xu PhD1; Mayne, Janice PhD1; Chiang, Cheng-Kang PhD1; Singleton, Ruth RN, CCRP2; Benchimol, Eric I. MD, PhD2,3,4; Mack, David R. MD2,3; Stintzi, Alain PhD1; Figeys, Daniel PhD1,5,6

Author Information
American Journal of Gastroenterology: May 2018 - Volume 113 - Issue 5 - p 713-724
doi: 10.1038/s41395-018-0024-9



Inflammatory bowel diseases (IBD) are characterized by chronic inflammation of the gastrointestinal tract, having two main subtypes (Crohn's disease (CD) and ulcerative colitis (UC)), with no curative medical therapy. The highest incidence and prevalence of IBD is observed in Europe and North America, while the incidence for pediatric onset IBD continues to rise worldwide [1].

Due to nonspecific initial symptoms and disease presentation similar to other disorders (e.g., infectious or allergic colitis, and irritable bowel syndrome) diagnosis of IBD is challenging, requiring multiple and invasive tests associated with blood analysis, imaging, and endoscopy. The most frequently assayed biomarker used to distinguish IBD from non-inflammatory disorders is fecal calprotectin (FC), outperforming blood inflammatory markers (erythrocyte sedimentation rate and C-reactive protein) to indicate intestinal inflammation [2]. While limited utility has been shown within the adult population, the diagnostic accuracy of FC is inferior for pediatric patients [3] where the specificity reaches only 0.682 for suspected pediatric IBD [4]. The low specificity is due to similarly elevated levels of FC in children suffering from disorders including celiac disease, cystic fibrosis, infection, neoplasia and polyps [5], allergic diseases [6, 7], and even in apparently healthy children [8]. A positive FC result necessitates further testing for suspected IBD, including endoscopy. Biomarkers with the ability to lower the false-positive rate associated with FC would reduce the number of unnecessary colonoscopies, and thus avoid the risk, discomfort, and economic burden associated with invasive procedures.

After IBD diagnosis establishment, additional aspects of the disease require assessment to select an appropriate therapeutic strategy, including disease subtype (CD or UC), disease severity, and extent of disease (UC) [9, 10]. Recently, we evaluated colon biopsy proteomes to identify a panel of 12 proteins which differentiated CD from UC with an accuracy of 80% [11]. Additionally, two of the panel proteins correlated with disease severity.

Disease extent may partially dictate the method (oral or rectal), type of treatment administered, and the recommended time to begin monitoring for colorectal cancer [10]. Biomarkers able to determine extent of disease have not been implemented in the clinic, rather assessment is achieved by endoscopy and imaging. The extent of disease in UC is defined as the macroscopic degree of continuous mucosal inflammation in the colon, extending proximally from the rectum.

Herein, we report a quantitative proteomic analysis of mucosal-luminal interface (MLI) in treatment-naive pediatric IBD patients at two distinct colon sub-regions. Bioinformatic analysis identified candidate biomarkers for classification of IBD patients with colonic involvement compared with controls with normal appearing colons. Furthermore, we identified biomarkers that indicate extent of disease in UC (pancolitis vs. nonpancolitis). Finally, we showed that the biomarker candidates leukotriene A-4 hydrolase (LTA4H) and catalase (CAT) identified from the MLI proteome exhibited consistent results when assessed in stool samples by enzyme-linked immunosorbent assay (ELISA).


Patient cohort

This study was approved by the Research Ethic Board of the Children Hospital of Eastern Ontario (CHEO). We conducted a cross-sectional study of patients <18 years old undergoing diagnostic colonoscopy for IBD from November 2013 to November of 2015 at CHEO. To assess host proteomic landscape alterations in IBD, the following exclusion criteria were implemented as they are known to alter the intestinal microbiota and thus influence the host response: (1) body mass index >95th percentile; (2) diabetes mellitus; (3) infectious gastroenteritis within the preceding 2 months; and (4) use of any antibiotics, probiotics, or immunosuppressives within 1 month of colonoscopy. Patients with inconclusive IBD diagnosis at the time of sample collection were excluded from the analysis.

Reference standard test methods

IBD was diagnosed by clinical examination, endoscopy, imaging, and laboratory testing [2]. The Pediatric Crohn's Disease Activity Index (PCDAI) was utilized for CD [12] and the Pediatric Ulcerative Colitis Activity Index (PUCAI) was utilized for UC [13]. Inflammation of the mucosa of the ascending colon (AC) or descending colon (DC) was assessed by visual appearance at colonoscopy. Extent of macroscopically inflamed mucosa was classified using the Paris modification of the Montreal Classification for IBD [14].

Sample collection and processing

Detailed methodologies are available in the Supplementary Material and methods online. Briefly, colonic MLI aspirates were obtained at the time of diagnostic colonoscopy following a standard 1 day clean-out preparation [15]. Samples were immediately put on ice for debris and bacteria removal prior to storage at -80 °C. After thawing, proteins were precipitated and combined with an equal amount of reference protein. Trypsin digest was performed by filter-aided sample preparation, fractionated, desalted, and analyzed on an Orbitrap Elite mass spectrometer (MS).

Frozen stool samples were obtained from treatment-naive participants, thawed, and diluted 5:1 in extraction buffer (extraction buffer volume:stool weight) for protein extraction and ELISA evaluation.

Bioinformatic analysis

Peptides were assigned and quantified using MaxQuant version [16] in a single run against the human Uniprot database (downloaded 2012/07/11). Patients with inconclusive IBD diagnosis at the time of sample collection were excluded from downstream bioinformatics analysis.

The MaxQuant output (ratio heavy/light normalized) was uploaded into Perseus v1.5.2.6 for heavy/light inversion, log2 transformation, Pearson's correlation coefficient determination, and hierarchical clustering of correlation values; data filtering was performed in Excel. Partial least-squares discriminant analysis (PLS-DA) and receiver operating characteristic (ROC) curve analyses were performed on k-nearest neighbor imputed data in the Biomarker Analysis module of MetaboAnalyst 3.0 [17]. Using MetaboAnalyst generated values, predictive class probability and ROC curves were plotted in Prism (version 7, GraphPad). Calculation of area under the curve (AUC), sensitivity, and specificity confidence intervals were performed in Prism. Principal component analysis (PCA) was performed using the prcomp argument in R studio. Protein interaction networks were generated using STRING version 10.0 and Cytoscape version 3.4.0. For suspected pediatric IBD biomarker panel generation, proteins identified by PLS-DA in both the AC and DC with the highest combined AUC upon ROC curve analysis were considered, and biomarker panel assembled based on iterative analysis. If a protein leads to a decrease in sensitivity or specificity without an increase in its counterpart in either colon sub-region it was skipped. Biomarker panel generation for UC extent of disease utilized features with the highest AUC from comparison of AC with and without macroscopic evidence of inflammation. Relative protein expression graphs with statistical analyses were generated in Prism. Biomarker panels and calprotectin (S100-A8 and S100-A9) were evaluated using PLS-DA in the ROC curve-based model evaluation module of MetaboAnalyst 3.0. Gene ontology was performed using DAVID Bioinformatics Resources 6.8 [18] and plotted in Prism.

Study data were collected and managed using REDCap (Research Electronic Data Capture) electronic data capture tools hosted at the CHEO Research Institute. REDCap is a secure, webbased application designed to support data capture for research studies [19].

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [20] partner repository with the dataset identifier PXD006383.

Enzyme-linked immunosorbent assay

LTA4H (abx572445, Abbexa Ltd, Cambridge, UK) and CAT (ab171572, Abcam, Cambridge, UK) from stool samples was measured from 50 and 0.5 μg of protein, respectively, by ELISA, which was then performed according to the manufacturer's protocol. A reference standard was added to all plates for inter-plate normalization.


Patient cohort

The entire cohort consisted of 60 patients, 18 controls without the presence of macroscopic inflammation, and 42 IBD patients. Notably, two of the control patients were ultimately diagnosed with irritable bowel syndrome. The colonic MLI aspirates of 93 colon sub-regions (57 AC (18 control, 39 IBD), 36 DC (10 control, 26 IBD)) were collected during diagnostic endoscopy from 60 patients prior to the administration of any treatment. For 33 patients (control = 10, CD = 14, UC = 9) both colon sub-regions were analyzed, whereas the AC or DC proteomes were exclusively analyzed for 24 patients (control = 8, CD = 7, UC = 9) and 3 patients (CD = 1, UC = 2), respectively (Fig. 1). No adverse events occurred during sampling or endoscopy. The patient characteristics are summarized in Table 1, with full details listed in Supplemental Table S1. No significant difference in age was observed between groups in either colon sub-region (Supplementary Figures 1A, B) or between sexes in the IBD group of the AC and DC and the control group of the DC, although we observed a significantly higher number of females in the AC control group (binomial one-tailed; p = 0.0481). In accordance with previous reports, the majority (70%) of pediatric UC patients in our cohort presented with extensive disease (Paris E3/E4 classification) [21]. Similarly, the majority (77%) of CD patients had disease involvement that was either isolated to the colon (L2) or had ileocolonic (L3) involvement [22].

Fig. 1
Fig. 1:
Overall workflow: Flow chart illustrating the active IBD biomarker panel and UC extent of disease biomarker panel generation. The proteomes of 93 (AC, n = 57; DC, n = 36) colonic mucosal-luminal interface aspirates were analyzed by LC-MS/MS. Bioinformatic analysis was performed on proteins quantified in ≥75% of samples (AC = 972; DC = 995). A panel of four proteins yielded the minimum number of proteins to achieve the highest sensitivity and specificity for both the IBD biomarker panel and UC pancolitis vs. non-pancolitis panel (UC extent of disease). The expression of the biomarker candidates CAT and LTA4H was evaluated in stool samples by ELISA
Table 1
Table 1:
Patient characteristics

Proteomic dataset assessment

Samples were processed and analyzed by MS over a period of 22 months. A super-SILAC approach was implemented for accurate quantification of proteins obtained from colonic MLI aspirate samples, and for use as a consistent reference over time. The majority (87.7% (AC), 88.5% (DC)) of proteins were quantified within a 10-fold median ratio (normalized MLI proteins/super-SILAC reference proteome) (Supplementary Figure 2A, B). There was an average of 49 (SD = 13) proteins per sample which did not have a heavy intensity counterpart, yielding an average of 2.28% (SD = 0.81) of proteins with light intensities without heavy intensity counterparts per sample. Furthermore, there were no significant differences observed in the total number of quantified proteins between groups in either colon sub-region (Supplementary Figure 2C, D). A total of 3,537 and 3,132 proteins were quantified from the AC and DC MLI, respectively, of which 972 (AC) and 995 (DC) proteins were quantified in ≥75% of samples (Q75) with 80% protein ID overlap between regions (Supplementary Table 2). Pearson's correlation analysis of the Q75 was performed on the log 2 light/heavy ratios (Supplementary Figure 3A, B), yielding 89% and 84% of values with a correlation >0.75 in the AC and DC, respectively, indicating consistent MS performance and sample preparation. Hierarchical clustering of the Pearson correlation values and PCA tended to segregate samples according to diagnosis and inflammatory status rather than by MS batch analysis (Supplementary Figures 3C, D and 4A, B).

Proteomic landscape alterations in treatment-naive pediatric IBD

To assess the proteomic alterations in an unbiased manner, PCA was performed on the Q75. The control samples generally clustered separately from those IBD samples that were deemed by endoscopy to have macroscopic evidence of inflammation (IBD CoA) (Fig. 2a, b). In the AC, the IBD samples with no evidence of macroscopic inflammation (IBD CoN) distributed between control samples and IBD CoA samples, whereas in the DC the IBD CoN samples tended to cluster with the IBD CoA. Segregation according to sex was not observed in either colon sub-regions (Supplementary Figure 4C, D). In order to identify discriminate features between control patients and IBD patients with active disease (IBD CoA), a multivariate analysis approach (PLS-DA) was applied, identifying 130 and 208 proteins in the AC and DC, respectively (Supplementary Table 2). There were 78 proteins common to both the AC and DC (Supplementary Figure 5A), which included proteins known to be altered in IBD such as the S100-A8 subunit of calprotectin (Supplementary Table 2). Gene ontology enrichment of the PLS-DA results yielded several expected biological processes related to IBD including defense response to bacterium, innate immune response, inflammatory response, phagocytosis, and neutrophil chemotaxis (Supplementary Figure 5B, C). Upon PLS-DA analysis of controls and IBD samples without macroscopic evidence of inflammation (CoN), pathways related to inflammation were observed (Supplementary Figure 6), possibly indicating the presence of microscopic inflammation. Notably, of the 130 proteins in the current AC IBD CoA vs. control dataset, 26 were also identified in our previous study [11] (Fig. 2c). Our previous study was performed on AC biopsies, wherein 225 proteins were identified as discriminant features of IBD CoA vs. control by PLS-DA analysis of the discovery cohort. Among these 26 common proteins, 11 (42%) and 9 (35%) were increased and decreased, respectively, in both datasets (Supplementary Table 2). The remaining six (23%) common proteins exhibited opposite relative expression between the biopsy and MLI datasets, with the majority (4/6) being elevated in the IBD samples relative to control in the MLI proteome vs. a relative reduction in the biopsy proteome. Within these 26 proteins, 22 localize to the extracellular region and an enrichment of proteins related to the immune response was identified (11/26).

Fig. 2
Fig. 2:
Proteomic landscape alterations in treatment-naïve pediatric IBD: PCA of Q75 proteins from a ascending colon and b descending colon. CoN without macroscopic inflammation, CoA with macroscopic inflammation. c Protein interaction network of features identified by PLS-DA of control vs. IBD CoA, common to both the current colonic mucosal-luminal interface (MLI) dataset (Supplementary Table 2) and biopsy dataset [11]. Grouping is based on relative IBD CoA/control expression levels between the two datasets. Border color indicates relative expression in the biopsy data, whereas the internal color represents the relative expression level from the MLI. High expression (blue) indicates elevated protein expression in IBD CoA compared to control, whereas low expression (orange) indicates decreased protein expression in IBD CoA compared to control. Squared boxes represent proteins involved in immune response. Small shape (22/26) indicates proteins that localize to the extracellular region. Arrows indicate protein-protein interaction

Biomarker panel for suspected pediatric active IBD-associated colitis

To identify biomarkers of active IBD, discriminant features identified by PLS-DA (control vs. IBD CoA) were further considered for biomarker panel generation. Proteins with the highest combined AUC (Supplementary Table 2) values from both colon sub-regions were considered for biomarker panel assembly. A maximum sensitivity and specificity utilizing the minimum number of features was reached in the AC using a panel of four features, whereas two features were sufficient in the DC (Supplementary Figure 7); the four panel proteins ultimately utilized for IBD CoA vs. control are listed in Table 2. The relative expression levels are depicted in Fig. 3a and compared to that of IBD CoN in Supplementary Figure 8. While not included in the development of the panel, IBD CoN levels of all panel proteins were significantly different from control samples, with the exception of transketolase in the AC.

Table 2
Table 2:
Proteins in panel for pediatric active IBD diagnosis and for extent of disease in UC (pancolitis vs. non-pancolitis)
Fig. 3
Fig. 3:
Biomarker panel for suspected pediatric active IBD diagnosis: a Relative expression levels of proteins included in active IBD diagnosis biomarker panel. P values were generate by t test; ****p < 0.0001. b Receiver operating characteristics curve utilizing panel of features listed in Table 2 for both the AC and DC, as indicated. c Predictive class probabilities in each colon sub-region (as indicated) wherein samples predicted to be control are to the left of 0.5 and those predicted to be IBD are on the right of 0.5. The number of True and Test samples is displayed in the table. Samples from control patients are shown in blue, and those from IBD patients are shown in red

Applying a ROC curve utilizing this panel of four proteins to the discovery cohort achieves an AUC value of 0.989 (95% confidence interval (CI): 0.967–1.0) and 0.999 (95% CI: 0.999–1.0) for the AC and DC, respectively (Fig. 3b). Predictive class probabilities yielded a sensitivity of 0.954 (95% CI: 0.7716–0.9988) and 1.0 (95% CI: 0.8235–1.0) for the AC and DC, respectively, and a specificity of >0.999 (AC 95% CI: 0.8147–1.0; DC 95% CI: 0.6915–1.0) for both the AC and DC (Fig. 3c), providing a classification accuracy of 97.5% and 100% in the AC and DC, respectively. Individual patient predictive class probability values are listed in Supplementary Table 1. Disease attribute shuffling using 500 permutations indicated the high performance of our panel, yielding significant values of p < 0.000118 and p < 0.000125 for the AC and DC, respectively. When IBD patient sample obtained from areas without macroscopic inflammation (CoN) were also included in the analysis, the biomarker panel ROC values were 0.9245 (95% CI: 0.8492–0.9998) and 0.9846 (95% CI: 0.9538–1.015) for the AC and DC, respectively (Supplementary Figure 9). This panel of biomarkers for suspected pediatric IBD diagnosis outperformed the classification accuracy of calprotectin (S100-A8 and S100-A9) using PLS-DA, by a direct comparison of MS data obtained from the MLI in both colon sub-regions; calprotectin yielded a sensitivity of 0.682 (95% CI: 0.4513–0.8614) and 0.684 (95% CI: 0.4345–0.8742) for the AC and DC, respectively, and a specificity of 0.833 (95% CI: 0.5858–0.9642) and 0.700 (95% CI: 0.3475–0.9333) in the AC and DC, respectively, which yielded classification accuracies of 75.0% and 69.0% in the AC and DC, respectively (Supplementary Figure 10 and Supplementary Table 1). Notably, the majority of patient calprotectin values were outside the accurate limit of detection which may influence its classification performance.

Biomarker panel for extent of disease in UC (pancolitis vs. non-pancolitis)

In accordance with previous studies [21], the majority (70%) of new-onset pediatric UC patients recruited in our discovery cohort displayed extensive disease (E3 or E4). To generate a biomarker panel to evaluate disease extent, features identified by PLS-DA upon comparison of colonic aspirate proteins arising from the AC of patients with (UC CoA) and without (UC CoN) evidence of macroscopic inflammation were further considered (Supplementary Table 2). As listed in Table 2, a panel of four proteins achieved a sensitivity of >0.9999 (95% CI: 0.5904–1.0) and specificity of >0.9999 (95% CI: 0.7151–1.0) and yielded a classification accuracy of 100% (Fig. 4 and Supplementary Figure 11). Individual patient predictive class probability values are listed in Supplementary Table 1.

Fig. 4
Fig. 4:
Biomarker panel for extent of disease in UC (pancolitis vs. non-pancolitis): a Relative expression levels of proteins included in the UC extent of disease biomarker panel. P values were generated by t test; ***p < 0.001. b Receiver operating characteristics curve for differentiation of pancolitis from non-pancoloits utilizing the expression of proteins in UC extent of disease biomarker panel. c Predictive class probabilities of inflammatory status in the AC (pancolitis vs. non-pancolitis) wherein samples predicted to be inflamed are to the left of 0.5 and those predicted to be non-inflamed are on the right of 0.5. The number of True and Test samples is displayed in the table

Validation of differential protein expression in stool samples on an independent cohort

In order to implement easily accessible biomarkers that would be more readily translatable to the clinic, we assessed the expression of two of the IBD biomarker proteins identified herein on an independent cohort, namely CAT and LTA4H, in stool samples. ELISA was performed on stool samples from 24 children for CAT and LTA4H (Supplementary Table 3). The increased expression in IBD compared to control observed by proteomics at the MLI was reflected by ELISA in stool samples, in which the expression of CAT (P < 0.0001) and LTA4H (P = 0.0002) were significantly different in IBD patients compared to controls (Fig. 5a, b). Furthermore, when considering both independent participants and participants for which their MLI samples were utilized for biomarker discovery, the expression of LTA4H in stool correlated with the PUCAI (Spearman's r = 0.567 (95% CI: 0.1378–0.817); P = 0.0114) (Supplementary Figure 12C). LTA4H also correlated with albumin (r = -0.3696; P = 0.0265) but did not correlate with erythrocyte sedimentation rates (ESR) or hematocrit (HCT), which are used in the calculation of the PCDAI, nor with PCDAI (Supplementary Table 4). The expression of CAT in stool samples correlated with the level of albumin measured in the serum (r = -0.4293; P = 0.008) but not with PCDAI, PUCAI, HCT, or ESR (Supplementary Table 4).

Fig. 5
Fig. 5:
Validating differential protein expression in stool on an independent cohort: a Catalase and b leukotriene A-4 hydrolase, biomarker candidates proposed for the diagnosis of pediatric IBD, were assessed by ELISA from a cohort of independent patients. P values were calculated using the Mann-Whitney test, ****p < 0.0001; ***p < 0.001


This is the first proteomic study to investigate alterations in newonset pediatric IBD patients at the MLI. We have identified a panel of four biomarkers that accurately classify > 95% of patients as active IBD or control, and propose the first biomarkers for disease extent in UC. Furthermore, we confirmed the differential expression of select biomarker panel proteins in a non-invasive biological specimen (stool samples).

To address the need for improved non-invasive biomarkers of IBD, and of UC extent of disease, we have built upon findings from our previous studies [11] to identify and validate biomarkers from alternative biosources, namely, the MLI and stool samples, respectively. Although the exact etiologies of IBD remain unknown, aberrant host-microbe interactions at the MLI are considered an important factor for the pathogenesis of disease [23]. We recently evaluated the bacterial component of the MLI in pediatric patients, finding differences in the bacteria present at this location in CD patients when compared with non-IBD controls [24]. Investigating biofluids in proximity to the disease site can alleviate biomarker discovery-associated challenges, including dynamic range and low relative abundance of a disease-specific biomarker [25]. However, there has been limited investigation of the host proteome alteration at the MLI in IBD [26], with results limited to adults and a therapeuticintervention patient cohort.

Comparison of our colonic MLI dataset with our reported biopsy dataset (Fig. 2c) shows that a portion of proteins demonstrate opposite relative expression for IBD vs. control. This could be partially explained by the difference in locale under assessment. For example, complement C3 is increased within the MLI but decreased in the biopsy of IBD patients; we postulate that tissue macrophages or epithelial cells are releasing C3 into the extracellular space, where it is then detected at higher levels within the MLI.

Proteins in the active IBD biomarker panel include both novel and previously reported IBD biomarkers. Annexin A3 and transketolase have not previously been investigated as biomarkers for suspected IBD, although the elevated gene expression of the former has been proposed as a biomarker for colorectal cancer [27], for which IBD patients are at an increased risk. Transketolase has been linked to inflammation; during the inflammatory response, the enzyme indirectly restores antioxidants to minimize ROSinduced damage [28]. Furthermore, the loss of transketolase-like 1, a protein closely related to TKT, has been shown to exacerbate colitis in a dextran sulfate sodium-induced mouse model of colitis [29]. The elevated expression of TKT that we observe in IBD patients compared to controls (Fig. 3) may be a response toward reducing the ROS-induced damage that is observed within the colon of IBD patients [30]. CAT, also in our IBD biomarker panel, is an enzyme with antioxidant activity and has been reported as one of the target antigens of the anti-neutrophil cytoplasmic autoantibodies (ANCA) [31]. This serological biomarker (ANCA) is occasionally used in the clinic to aide in distinguishing CD from UC, although its low sensitivity has prevented it from becoming routinely used [32]. We did not observe a significant difference in CAT expression between CD and UC at the MLI (Supplementary Figure 8). However, there is significant elevation of the enzyme in MLI from both macroscopic (CoA) and non-macroscopic (CoN) inflammatory aspirates from IBD patients when compared with controls (Supplementary Figure 8). Similar to TKT, the elevated levels of CAT may be to protect cells from the damage caused by hydrogen peroxide which is elevated in the inflamed mucosa of IBD patients [33]. LTA4H is the only protein included in both the active IBD and UC extent biomarker panels. We first reported it as a biomarker to differentiate CD from UC in intestinal biopsies [11]. In this study, we report the first indication of its use as a biomarker of IBD diagnosis and UC extent of disease. LTA4H catalyzes the biosynthesis of leukotriene B4, which in turn is a potent neutrophilic chemoattractant and has been implicated in chronic inflammation [34]. The reported significant decrease in LTA4H expression observed in UC compared to CD within AC biopsies [11] was not reflected at the colonic MLI in either colon location (Supplementary Figure 8).

To our knowledge, this is the first report identifying the proteins TXNDC17, TMSB10, and VASP as biomarker for IBD diagnosis, herein part of the panel for disease extent in UC. Interestingly, TXNDC17 can regulate tumor necrosis factor-α (TNF-α) signaling [35]; TNF-α is an important signaling molecule in inflammation, and is the target of anti-TNF-α agents that are utilized for the induction and maintenance of remission in CD. Unlike the role of TXNDC17, TMSB10, and VASP are involved in cytoskeleton organization [36, 37]. VASP has been reported to impede the transmission between cells of the diarrhea-causing Gram-negative bacterium Shigella flexneri [38]; perhaps, elevated VASP observed in inflamed UC is associated with altered host-microbe interactions that are thought to contribute to IBD [26]. Detecting VASP within our MLI samples is in accordance with the literature as it can localize to the plasma membrane [39], whereas TMSB10 has only been reported to localize intracellularly. Despite both having roles in cytoskeletal organization, they demonstrate opposite expression trends in inflamed tissue compared to uninflamed tissue (Fig. 4a). The mechanism describing the decreased expression of TMSB10 observed in inflamed UC compared to non-inflamed UC requires further investigation.

The extent of disease in UC patients can indicate response to therapy, since healing progresses proximally to distally through the colon. Therefore, further investigations evaluating the UC extent of disease biomarker panel on patients undergoing treatment would be beneficial to measure the capacity of the markers to not only identify disease extent, but also to evaluate the response to therapy. Considering that the majority of pediatric UC patients present with pancolitis [21], a non-invasive biomarker indicating a status of pancolitis vs. non-pancolitis would be beneficial as it would permit assessment of response to therapy without the need for colonoscopy.

Notably, due to limitations of compatible commercial ELISAs and sufficient sample, we were able to validate only two of the four candidate biomarkers from the panel with stool samples; these two proteins resulted in an ROC of 0.8929 (95% CI: 0.7428–1.043). However, we cannot predict the comparability of the ROC curve with all four proteins from stool samples to that of aspirate samples. Preliminary findings, using orthogonal methods, indicate the differential expression of ANXA3 and TKT in stool samples, which we intend to expand upon in future studies with continued patient recruitment. Moreover, further studies are required to test our biomarker panel within the adult population, on a multicenter level, and inclusion of infectious colitis (i.e., non-IBD inflammatory) to evaluate the capacity of our panels to extend beyond the main pediatric and single center cohort. In addition, the influence of bowel preparation on colonic MLI proteomic expression is unknown and requires further investigation. Finally, although super-SILAC permits accurate quantification, this approach is limiting for proteins for which their heavy counterparts’ expression is beyond a 10-fold difference which would result in their poor performance as potential biomarkers. In summary, this study identified biomarkers that are better at predicting active IBD than calprotectin at the MLI, and elevated expression of CAT and LTA4H, identified as biomarker candidates at the MLI was reflected in stool samples. Further, a panel of UC extent of disease biomarkers was identified, classifying 100% of patients with pancolitis or non-pancolitis; this panel was developed by comparison of UC patients with and without evidence of macroscopic inflammation, and thus at a level similar in nature to what is observed by endoscopy. This study represents the initial work to develop measures for more accurate diagnosis of IBD, answering the increasing need for sensitive and accurate IBD diagnostic biomarkers, as worldwide incidence of IBD continues to rise.


Guarantor of the article: Daniel Figeys, PhD.

Specific author contributions: Study design performed by SAD, AES, DRM, AS, and DF. Patient enrollment, diagnosis, patient sample collection, and clinical data acquisition was performed by RS, EIB, and DRM Acquisition and analysis of data was performed by SAD, AES, and SA. Interpretation of data wa performed by SAD, AES, ZN, XZ, JM, CKC, EIB, DRM, AS, and DF. Drafting of the manuscript was completed by SAD, AES, and DF

Financial support: This work was supported by the Government of Canada through Genome Canada and the Ontario Genomics Institute (OGI-067), Canadian Institutes of Health Research (CIHR) grant number GPH-129340, CIHR grant numbers MOP-114872 and ECD-144627, and the Ontario Ministry of Economic Development and Innovation (REG1–4450). We acknowledge funding from the IBD Foundation, Crohn's and Colitis Canada (CCC), the CHEO Research Institute, and the Faculty of Medicine of the University of Ottawa. SAD was supported by the Ontario Graduate Scholarship. EIB was supported by a New Investigator Award from CIHR, CCC and the Canadian Association of Gastroenterology, and the Career Enhancement Program of the Canadian Child Health Clinician Scientist Program. DF was supported by a Canada Research Chair 1 in proteomics and systems biology.

Potential competing interests: AS, DRM, DF, SAD, and AES have a corresponding patent application “Markers for disease and disease extent in inflammatory bowel disease” filed in 2017 (62/520, 652). AS, DRM, and DF have co-founded Biotagenics, a clinical microbiomics company. The remaining authors declare that they have no competing interests.

Study Highlights


✓ Improved biomarkers are needed to decrease the number of unnecessary invasive procedures in IBD diagnosis.

✓ Need is greatest in children, where the most frequently used IBD biomarker (calprotectin) suffers from low specificity.

✓ Biomarkers are an unmet need for the assessment of the extent of disease in UC.


✓ First proteomic study to evaluate alterations occurring in new-onset pediatric IBD patients at the MLI.

✓ The differential expression of select biomarker panel proteins was reflected in stool samples.

✓ The first protein biomarker panel to delineate disease extent in pediatric ulcerative colitis is proposed.


1. Benchimol EI, Fortinsky KJ, Gozdyra P, et al. Epidemiology of pediatric inflammatory bowel disease: a systematic review of international trends. Inflamm Bowel Dis. 2011;17:423-39.
2. Levine A, Koletzko S, Turner D, et al. ESPGHAN revised Porto criteria for the diagnosis of inflammatory bowel disease in children and adolescents. J Pediatr Gastroenterol Nutr. 2014;58:795-806.
3. van Rheenen PF, Van de Vijver E, Fidler V. Faecal calprotectin for screening of patients with suspected inflammatory bowel disease: diagnostic meta-analysis. BMJ. 2010;341:c3369.
4. Henderson P, Anderson NH, Wilson DC. The diagnostic accuracy of fecal calprotectin during the investigation of suspected pediatric inflammatory bowel disease: a systematic review and meta-analysis. Am J Gastroenterol. 2014;109:637-45.
5. Mendoza JL, Abreu MT. Biological markers in inflammatory bowel disease: practical consideration for clinicians. Gastroenterol Clin Biol. 2009;33(Suppl 3):S158-73.
6. Merras-Salmio L, Kolho KL, Pelkonen AS, et al. Markers of gut mucosal inflammation and cow's milk specific immunoglobulins in non-IgE cow's milk allergy. Clin Transl Allergy. 2014;4:8.
7. Orivuori L, Mustonen K, de Goffau MC, et al. High level of fecal calprotectin at age 2 months as a marker of intestinal inflammation predicts atopic dermatitis and asthma by age 6. Clin Exp Allergy. 2015;45:928-39.
8. Zhu Q, Li F, Wang J, et al. Fecal calprotectin in healthy children aged 1-4 Years. PLoS ONE. 2016;11:e0150725.
9. Danese S, Fiocchi C. Ulcerative colitis. N Engl J Med. 2011;365:1713-25.
10. Stange EF, Travis SP, Vermeire S, et al. European Evidence-based Consensus on the diagnosis and management of ulcerative colitis: definitions and diagnosis. J Crohns Colitis. 2008;2:1-23.
11. Starr AE, Deeke SA, Ning Z, et al. Proteomic analysis of ascending colon biopsies from a paediatric inflammatory bowel disease inception cohort identifies protein biomarkers that differentiate Crohn's disease from UC. Gut. 2016;66:1573-83.
12. Hyams J, Markowitz J, Otley A, et al. Evaluation of the pediatric crohn disease activity index: a prospective multicenter experience. J Pediatr Gastroenterol Nutr. 2005;41:416-21.
13. Turner D, Otley AR, Mack D, et al. Development, validation, and evaluation of a pediatric ulcerative colitis activity index: a prospective multicenter study. Gastroenterology. 2007;133:423-32.
14. Levine A, Griffiths A, Markowitz J, et al. Pediatric modification of the Montreal classification for inflammatory bowel disease: the Paris classification. Inflamm Bowel Dis. 2011;17:1314-21.
15. Jimenez-Rivera C, Haas D, Boland M, et al. Comparison of two common outpatient preparations for colonoscopy in children and youth. Gastroenterol Res Pract. 2009;2009:518932.
16. Cox J, Matic I, Hilger M, et al. A practical guide to the MaxQuant computational platform for SILAC-based quantitative proteomics. Nat Protoc. 2009;4:698-705.
17. Xia J, Sinelnikov IV, Han B, et al. MetaboAnalyst 3.0—making metabolomics more meaningful. Nucleic Acids Res. 2015;43:W251-7.
18. Huang da W, Sherman BT, Lempicki RA. Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Res. 2009;37:1-13.
19. Harris PA, Taylor R, Thielke R, et al. Research electronic data capture (REDCap)—a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform. 2009;42:377-81.
20. Vizcaino JA, Csordas A, Del-Toro N, et al. 2016 update of the PRIDE database and its related tools. Nucleic Acids Res. 2016;44:11033.
21. Sauer CG, Kugathasan S. Pediatric inflammatory bowel disease: highlighting pediatric differences in IBD. Gastroenterol Clin N Am. 2009;38:611-28.
22. Vernier-Massouille G, Balde M, Salleron J, et al. Natural history of pediatric Crohn's disease: a population-based cohort study. Gastroenterology. 2008;135:1106-13.
23. Wlodarska M, Kostic AD, Xavier RJ. An integrative view of microbiomehost interactions in inflammatory bowel diseases. Cell Host Microbe. 2015;17:577-91.
24. Mottawea W, Chiang C-K, Mühlbauer M, et al. Altered intestinal microbiota -host mitochondria crosstalk in new onset Crohn's disease. Nat Commun. 2016;7:13419.
25. Rifai N, Gillette MA, Carr SA. Protein biomarker discovery and validation: the long and uncertain path to clinical utility. Nat Biotechnol. 2006;24:971-83.
26. Presley LL, Ye J, Li X, et al. Host-microbe relationships in inflammatory bowel disease detected by bacterial and metaproteomic analysis of the mucosal-luminal interface. Inflamm Bowel Dis. 2012;18:409-17.
27. Marshall KW, Mohr S, Khettabi FE, et al. A blood-based biomarker panel for stratifying current risk for colorectal cancer. Int J Cancer. 2010;126:1177-86.
28. Lassen N, Black WJ, Estey T, et al. The role of corneal crystallins in the cellular defense mechanisms against oxidative stress. Semin Cell Dev Biol. 2008;19:100-12.
29. Bentz S, Pesch T, Wolfram L, et al. Lack of transketolase-like (TKTL) 1 aggravates murine experimental colitis. Am J Physiol Gastrointest Liver Physiol. 2011;300:G598-607.
30. Cao W, Vrees MD, Kirber MT, et al. Hydrogen peroxide contributes to motor dysfunction in ulcerative colitis. Am J Physiol Gastrointest Liver Physiol. 2004;286:G833-43.
31. Roozendaal C, Zhao MH, Horst G, et al. Catalase and alpha-enolase: two novel granulocyte autoantigens in inflammatory bowel disease (IBD). Clin Exp Immunol. 1998;112:10-6.
32. Sabery N, Bass D. Use of serologic markers as a screening tool in inflammatory bowel disease compared with elevated erythrocyte sedimentation rate and anemia. Pediatrics. 2007;119:E193-E199.
33. Strus M, Gosiewski T, Fyderek K, et al. A role of hydrogen peroxide producing commensal bacteria present in colon of adolescents with inflammatory bowel disease in perpetuation of the inflammatory process. J Physiol Pharmacol. 2009;60(Suppl 6):49-54.
34. Palmer RM, Stepney RJ, Higgs GA, et al. Chemokinetic activity of arachidonic and lipoxygenase products on leuocyctes of different species. Prostaglandins. 1980;20:411-8.
35. Jung Y, Kim H, Min SH, et al. Dynein light chain LC8 negatively regulates NF-kappaB through the redox-dependent interaction with IkappaBalpha. J Biol Chem. 2008;283:23863-71.
36. Yu FX, Lin SC, Morrison-Bogorad M, et al. Effects of thymosin beta 4 and thymosin beta 10 on actin structures in living cells. Cell Motil Cytoskelet. 1994;27:13-25.
37. Ke Y, Tan Y, Wei N, et al. Yersinia protein kinase A phosphorylates vasodilator-stimulated phosphoprotein to modify the host cytoskeleton. Cell Microbiol. 2015;17:473-85.
38. Lee SY, Gertler FB, Goldberg MB. Vasodilator-stimulated phosphoprotein restricts cell-to-cell spread of Shigella flexneri at the cell periphery. Microbiology. 2015;161:2149-60.
39. Castellano F, Le Clainche C, Patin D, et al. A WASp-VASP complex regulates actin polymerization at the plasma membrane. EMBO J. 2001;20:5603-14.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, and provide a link to the Creative Commons license. You do not have permission under this license to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit

Supplemental Digital Content

© The American College of Gastroenterology 2018. All Rights Reserved.