Celiac disease (CD) is a multifactorial autoimmune disease induced by ingestion of gluten in genetically predisposed individuals. Despite technological progress, the diagnosis of CD is still based on duodenal biopsy as it was 50 years ago. In this study we analysed the expression of CD-associated genes in small bowel biopsies of patients and controls in order to explore the multivariate pathway of the expression profile of CD patients. Then, using multivariant discriminant analysis, we evaluated whether the expression profiles of these genes in peripheral blood monocytes (PBMs) differed between patients and controls.
Thirty-seven patients with active and 11 with treated CD, 40 healthy controls and 9 disease controls (Crohn’s disease patients) were enrolled.
Several genes were differentially expressed in CD patients versus controls, but the analysis of each single gene did not provided a comprehensive picture. A multivariate discriminant analysis showed that the expression of 5 genes in intestinal mucosa accounted for 93% of the difference between CD patients and controls. We then applied the same approach to PBMs, on a training set of 20 samples. The discriminant equation obtained was validated on a testing cohort of 10 additional cases and controls, and we obtained a correct classification of all CD cases and of 91% of the control samples. We applied this equation to treated CD patients and to disease controls and obtained a discrimination of 100%.
The combined expression of 4 genes allows one to discriminate between CD patients and controls, and between CD patients on a gluten-free diet and disease controls. Our results contribute to the understanding of the complex interactions among CD-associated genes, and they may represent a starting point for the development of a molecular diagnosis of celiac disease.
Citation: Galatola M, Izzo V, Cielo D, Morelli M, Gambino G, et al. (2013) Gene Expression Profile of Peripheral Blood Monocytes: A Step towards the Molecular Diagnosis of Celiac Disease? PLoS ONE 8(9): e74747. doi:10.1371/journal.pone.0074747
Editor: Valli De Re, Centro di Riferimento Oncologico, IRCCS National Cancer Institute, Italy
Received: April 9, 2013; Accepted: August 7, 2013; Published: September 17, 2013
Copyright: © 2013 Galatola et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was funded by the PREVENT-CD project: EU-FP6-2005-FOOD4B-contract no. 036383. The authors thank Associazione Italiana Celiachia (AIC), European Laboratory for food induced disease (ELFID) and the Italian Ministry of Health, International Affairs, (EUROMED action - MEDICEL project) for a supporting grant. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Celiac disease (CD) is gluten-induced autoimmune disease closely linked to a specific genetic profile. More than 95% of celiac patients are HLA-DQ2/8 carriers, although HLA genes account for only around 35% of the genetic variation [1,2]. A recent genome-wide association study identified 13 known, 13 new and 13 “suggested” genomic variants . Although these 39 risk variants account for less than 15% of the genetic variance, they help to shed light on the immunological factors involved in the gluten-induced abnormal response (i.e., T-cell development, innate immune detection of viral RNA, T- and B-cell co-stimulation/inhibition and cytokines, chemokines and their receptors) [3,4]. But much more work is needed to explain the missing heritability and to understand the functional consequences of associated alleles in this complex disease.
Based on expression quantitative trait meta-analysis, Dubois et al. identified celiac risk variants correlated with cis gene expression in 20 out of 38 (52.6%) tested loci . A study of gene expression in duodenal mucosa of a Spanish celiac sample  confirmed most of the results reported by Dubois et al.
In a previous study , we evaluated the expression of genes clustered on chromosome 4q27 (KIAA1109, IL-2 and IL-21) and of the c-REL gene in intestinal mucosa of controls and CD patients (active and treated with a gluten-free diet [GFD]). KIAA1109 and c-REL mRNA expression in intestinal mucosa was significantly higher in CD patients on GFD than in CD patients on gluten. Another study found that IL-21 expression was higher in CD patients than in controls . This increase seems to be gluten-dependent because IL-21 expression returned to the levels of controls after at least one year of GFD . In the same study, IL-2 mRNA did not differ among CD patients, CD-GFD patients and controls, although there was a trend towards over-expression in the CD group.
We recently analysed these genes in a family cohort association study . Similarly we explored their contribution to the progression of the potential CD phenotype . These mono-dimensional observations of the expression of single genes do not provide a valid picture of the complex inter-relationship of these molecules at cellular level: a multivariate approach is needed because the expression of each gene cannot be independent from the expression of the other genes in a functional pathway.
A discriminant analysis of gene expression was recently proposed as a promising diagnostic tool to distinguish celiac atrophic mucosa from normal mucosa . As expected, genes involved in the alteration of the crypt-villi architecture in the small intestinal mucosa were identified, and they matched the histological alterations.
The aim of this study was to explore the expression of genes associated to CD in the target tissue in order to estimate the contribution of each single gene to the development of the gluten-induced immune response. Then, using a multivariate model, we planned to evaluate the same set of genes in peripheral blood monocytes (PBM). The rationale for using PBMs is that they are more readily available than mucosal tissue: monocyte-derived cells (MDCs) were found to accumulate in the inflamed intestine of CD patients [10,11]. In a recent study, it was shown that the density of CD14+ CD11c+ MDCs was increased in the inactive form of the disease whereas the density of CD14- cells and macrophages was decreased in the active celiac lesion .
Expression of the candidate genes in duodenal mucosa
Duodenal intestinal mucosa samples of controls, CD and CD-GFD patients were examined for the expression of KIAA1109, IL-2, IL-21, LPP, RGS1, cREL, SH2B3, TAGAP, TNFAIP3, TNFSF14, and TNFRSF14 genes, according to our previous results . The expression of the LPP gene did not differ among the three groups (Figure S1 A). TNFAIP3 and RGS1 mRNA levels were up-regulated in CD patients versus controls but the difference was not statistically significant (Figure S1 B and C). Interestingly, TNFRSF14 mRNA levels in duodenal mucosa were similar in controls and CD-GFD patients, and higher, albeit not significantly so, in CD patients (Figure S1 D).
SH2B3 expression was higher in CD mucosa than in control mucosa, and significantly lower in CD-GFD subjects than in either controls or CD patients (Figure 1 A). Mucosal TNFSF14 levels were higher in CD patients and in CD-GFD patients than in controls (Figure 1 B). TAGAP expression was significantly higher in CD mucosa than in control mucosa (Figure 1 C). As we expected, the expression levels of IL-21 were significantly higher in CD than in controls (p<0.01), which confirms our previous data (Figure 2 A) . On the contrary, IL-2, KIAA1109 and cREL mRNA did not differ between the two groups, only IL-2 shows a very small trend (although not significant) of increase in CD compared to controls (Figure 2 B, C and D).
A) SH2B3 expression was slightly higher in CD patients than in controls, and significantly lower in CD-GFD patients than in controls (p<0.01) and CD patients (p<0.01); B) TNFSF14 expression was higher in CD versus controls, and remained higher also after one year of GFD versus controls (p=0.04) and CD patients (p<0.01); C) TAGAP expression was higher in CD patients versus controls (p=0.04);.
RQ: relative quantification; Ctr: controls; CD: celiac disease; CD-GFD: celiac patients on a gluten-free diet; * p<0.01, **p<0.05.
A) IL-21 is over-expressed in CD compared to controls (p<0.01), B) IL-2 shows a very small trend of increase in CD compared to controls; C) and D) Expression of the KIAA1109 and cREL genes: the patterns are very similar and did not show any variations among the three groups.
RQ: relative quantification; Ctr: controls; CD: celiac disease; CD-GFD: celiac patients on a gluten-free diet; * p<0.01, **p<0.05.
Expression of the candidate genes in peripheral blood monocytes
The expression of the panel of candidate genes was evaluated in monocytes extracted from peripheral blood samples of 18 controls, 17 CD and 5 CD-GFD patients. Monocytes extracted from peripheral blood of 9 Crohn patients served as positive controls. We did not evaluate the expression of IL-2 and IL-21 because they are not produced by monocytes but by CD4+ T-cell lines TCLs after antigen activation.
TAGAP, TNFSF14 and TNFRSF14 were expressed at similar levels in CD and CD-GFD monocytes (Figure S2). The KIAA1109 gene was over-expressed in CD, Crohn and CD-GFD patients versus controls (p<0.05) (Figure 3 A). Differently, c-REL, and SH2B3 expression was lower in CD monocytes than in controls, but significantly higher than in CD-GFD and Crohn monocytes (Figure 3 B and C). LPP expression was down-regulated in CD patients, whereas it was similar to controls in CD-GFD and Crohn patients (Figure 3 D). TNFAIP3 mRNA expression was moderately lower in CD patients than in controls and CD-GFD patients, whereas it was over-expressed in inflamed positive controls (Figure 3 E).
A) KIAA1109expression was higher in CD (p=0.05) and in CD-GFD patients (p=0.05) than in controls, but also in Crohn peripheral monocytes (p=0.02); B) c-REL expression was lower in CD peripheral monocytes than in controls (p<0.01), but become higher than controls (p<0.01) and CD (p<0.01) after one year of GFD; the same profile was observed in Crohn peripheral monocytes; C) SH2B3 expression was lower in CD versus controls (p=0.04) whereas it was significantly higher in Crohn and CD-GFD patients versus controls (p<0.01) and CD (p<0.01). D) LPP expression was lower in CD peripheral monocytes than in controls (p=0.04); E-F) TNFAIP3 and RGS1 genes expression were lower in CD peripheral monocytes versus controls (p<0.01) and higher in Crohn patients versus controls (p<0.01) and CD patients (p<0.01). Both genes expression levels normalized after one year of GFD;.
RQ: relative quantification; Ctr: controls; CD: celiac disease; CD-GFD: celiac patients on a gluten-free diet; * p<0.01, **p<0.05.
The trend of RGS1 expression was similar to that of c-REL, but did not differ between CD-GFD and control monocytes (Figure 3 F).
Discriminant analysis in duodenal mucosa
To identify genes whose expression best characterizes celiac tissue versus controls, a linear discriminant equation was fitted to the standardized values of expression (RQ). By a stepwise multivariate approach 5 genes (stepwise: TNFAIP3, IL-21, c-REL, RGS1 and LPP) were selected for discriminating capacity (Table 1). As previously described in the methods section, by multiplying the canonical unstandardized coefficients produced by the analysis to the actual values of the RQ of the candidate genes a D-score was obtained for each individual as follows:
|Step||Candidate genes||Wilks’ Lambda||Exact F|
This multivariate equation discriminated efficiently CD patients from controls: 92.9% of individuals were correctly classified (95% of controls and 90.9% of CD patients). One control/20 (5%) was misclassified as a celiac and two celiacs/22 (9.1%) were misplaced as controls.
Discriminant analysis in peripheral blood monocytes
Encouraged by the results obtained with duodenal mucosa, we developed a similar linear discriminant analysis for the gene expression observed in PBMs. We randomly divided our celiac and control cases into two balanced groups: a training set to develop the equation and a validation set to verify its efficiency. By a stepwise multivariate approach the expression of 4 candidate genes were selected with a pattern quite similar to that observes in the duodenal tissue. LPP, c-REL, KIAA1109 and TNFAIP3 genes help to discriminate cases from controls, reaching a very low Wilks’ lambda (0.048) (Table 2). The low Wilk’s lambda obtained by this set of genes, close to 0 = complete discrimination, supports the confidence into the classification capacity of the equation in the clinical setting.
|Step||Candidate genes||Wilks’ Lambda||Exact F|
Indeed 91% of controls and all CD patients were correctly classified (Table 3). To verify the efficiency of the discriminant model obtained in the training set we applied the equation to the gene expression of a new cohort of patients (validation set) made up by 7 controls, 8 CD, 5 patients on GFD and 9 disease controls (Crohn patients). We obtained four clustered D-scores, one for each group (Controls, CD, Crohn and CD on GFD) (Figure 4) with no overlap with the active celiacs. Figure 4 shows the distribution of the D-Scores of CD, CD patients on GFD, controls and Cohn’s disease patients. The D-Score of active celiac patients was negative in all cases, while it was positive for all the other groups on differentiated clusters.
|Predicted Group Membership|
|Real Group Membership||Control||10 (91%)||1 (9%)||11|
|Celiac||0 (0%)||9 (100%)||9|
The D-score clearly separated the four groups of subjects evaluated. Only CD patients had a negative D-score. The D-score of CD patients on gluten free diet was intermediate between the scores of controls and Crohn patients.
This score produces a group membership probability for each individual, allowing us to correctly classify all controls and CD patients; none of the controls, neither CD on GFD nor Crohn patients were misclassified as CD patients (Table S1).
As indicated in the new ESPGHAN diagnostic algorithm for CD , small bowel biopsy may now be avoided in a sizeable proportion of patients who have clinical symptoms: anti-tTG antibodies levels 10 times the normal values, predisposing HLA genotype. Unfortunately, not all patients have such a high production of anti-tTG antibodies, and many are asymptomatic. Moreover, this new protocol is rarely applicable to at-risk relatives . Therefore, the aim of this study was to determine whether the gene expression profile of CD-associated genes in PBMs could help to differentiate patients affected by CD from controls as a step towards the molecular diagnosis of the disease. We studied the genes that have most often been associated with CD and those that show interesting functional profiles [3,4,14]. As expected, the analysis of the expression of each single gene did not unequivocally differentiate between celiac and non-celiacs, but a multivariate combination of genes that are implicated in the pathogenesis of CD was selected in the target tissue.
We previously showed that the NF-kB complex is specifically and precociously implicated in the gluten-induced inflammatory response in celiac mucosa: in fact, the NF-kB complex is fully activated just 6 hours after gluten exposure . The cREL and TNFAIP3 genes are both involved in the regulation of this nuclear activating complex. The former is one subunit of the complex, while the latter is a negative regulator of its activation. TNFAIP3 (also known as A20) expression is up-regulated by NF-kB activation, as confirmed in our study, and it acts in a negative feedback loop to control NF-kB-dependent gene expression. Furthermore, the TNFAIP3 gene is a susceptibility locus for several human inflammatory and autoimmune diseases, including inflammatory bowel disease, rheumatoid arthritis, psoriasis, lupus and type 1 diabetes. TNFAIP3 is also frequently inactivated in subsets of B-lineage lymphomas that are characterized by NF-kB hyper activation and was therefore suggested to be a novel tumour suppressor .
IL-21 plays a crucial role in the activation of the gluten-induced mucosal activation through the NK system. It is most abundantly produced by CD4+ T cells and natural killer T (NKT) cells and is important for the development of the pro-inflammatory Th17 lineage. It has a protean function principally oriented to the activation of intraepithelial T-cell, which play a pivotal role in the development of the mucosal damage .
The RGS1 gene controls the homing of intraepithelial lymphocytes (IELs), which are essential for the production of the gluten induced epithelial damage, and is less active in celiacs than in controls. It is the activation of IELs that drives the destruction of the intestinal epithelium. The main role of IELs is to promote immune protection by preventing the entry and spread of pathogens while avoiding unwanted and excessive inflammatory reactions capable of damaging the intestinal epithelium. To that end, IELs exert a cytolytic function to eliminate infected and damaged cells, and regulatory functions that contribute to epithelium healing and repair. Deregulated activation of IELs is a hallmark of CD and is critically involved in epithelial cell destruction and the subsequent development of villous atrophy. In addition, lymphocytic infiltration of the small intestinal epithelium in the absence of villous atrophy has been observed in patients with dermatitis herpetiformis , an autoimmune skin manifestation of CD; a finding that supports the concept that intraepithelial lymphocytosis is a marker of CD even in the absence of intestinal damage . It was recently demonstrated that elevated RGS1 levels profoundly reduce T cell migration to lymphoid-homing chemokines, whereas RGS1 depletion selectively enhances such hemotaxis in gut T cells . Its capacity to limit the egress of inflammatory and/or autoimmune cells could clearly promote immunopathology.
Finally, the LPP gene appears to have a relevant activity in modulating cell adhesion since it is an integral component of cell migration. It is not surprising that up- or down-regulation of LPP expression results in an increase or decrease, respectively, in cell migration . Recent data showed that the over-expression of LPP increased epidermal growth factor-stimulated migration of vascular SMCs induced by TGF-β1, suggesting the participation of LPP in cell motility [22,23].
Since the expression of these genes is not independent, it is not possible to give a priority of function to any of them: indeed genes that are not selected by the stringent criteria of the multivariate analysis may well have their own relevant function, but their contribution is no longer significant when other more “discriminating” genes are included in the equation. Nevertheless, it is very interesting to note that this model is built by genes that are more likely to exert a relevant function in the abnormal gluten-induced response in individuals with a specific genomic profile.
When we moved from the target tissue to PBMs, to our surprise, 3 of the 4 genes selected for their discriminating capacity are the same as those included in the multivariate model of the small intestinal mucosa tissue. Again the NF-kB complex appears implicated in development of CD, as well as the fascinating LPP gene. The KIAA1109 gene, located in the region encompassing KIAA1109/Tenr/IL2/IL21 in chromosome 4q27, has often been replicated in association studies and provides a significant contribution to the discrimination between celiacs and non celiacs. This cluster region is involved in the differentiation of naïve human CD4+ T cells into Th17 cells . It is important to note that Th17 cells produce a variety of cytokines, among which, IL-17A, IL-17F, IL-21 and IL-22. Genetic alterations in the 4q27 locus could result in non-functional IL-21 and hence lack of IL-17A or vice versa. The regulation of this process may be an important factor in determining the risk for CD, as shown in such other autoimmune diseases as rheumatoid arthritis and uveitis [25,26].
In conclusion, we report an intriguing picture of the possible relationships among the expressions of candidate CD genes in the target mucosa, but also, with a minor difference, in PBMs. The analysis of the expression of each single gene is non-informative and does not reveal the specific isolated function of any of these candidates. Indeed, only the search for functional pathways may shed light on the complex gluten-induced abnormal response in genetically predisposed individuals.
We suggest that the expression of a small set of candidate genes in PBMs can be used to distinguish CD patients from healthy controls and from disease controls (patients affected by Crohn’s disease), without considering clinical data, HLA or anti-tTG antibodies. In fact, the procedure we used resulted in a distance between groups (very low Wilk’s lambda) that is unusual with ordinary diagnostic tools. We did not add anti-tTG antibodies or HLA data to the multivariate equation because it is well recognized that the former has a very high sensitivity and specificity, and the latter has a very high negative predictive value. Since we reached a correct diagnostic classification in the validation set (above 95%) by gene expression data only, we would have shown an overoptimistic estimate by adding these strong discriminators. However, this should be done in clinical practice in order to reinforce the sensitivity and the specificity of the diagnosis when duodenal biopsy is not available or desirable.
Expression may be regulated by the specific polymorphism associated to the disease but, in the case of CD, none of the identified polymorphisms contributed greatly to the pathogenesis of the gluten-induced immune response. Indeed, expression data should be examined within the framework of a reasonable pathogenic pathway. The same gene may be over- or under-expressed and produce significant downstream stimuli. Expression is one of the several regulations that control the production of functional molecules from a specific protein-coding gene. Epigenetic mechanisms have recently emerged as important partners in this domain, and our group reported a specific microRNA that is over-expressed in celiac patients versus controls .
Genome-wide studies have identified many encoding variants associated to CD. Several of these are implicated or are in linkage with regulatory DNA marked by deoxyribonuclease hypersensitive sites. Most of these sites are active during fetal development and associated with gestational exposure phenotype. Indeed, disease-associated variants often perturb transcription factor recognition sequences, altering the normal regulatory networks. In common human disease, and certainly in CD, regulatory DNA variations are likely to play a pivotal role, since there are no ‘missing’ or ‘failed’ genes .
The missing variance of heredity in CD is probably due to the thousands of expression quantitative loci, expression ‘hot spots’, where a polymorphism at a locus is responsible for changes in gene expression of many other genes, and finally by gene-by-environment interactions . Gene expression is currently the best tool with which to explore the final results of genetic variance; it is quite robust and reproducible and may be tailored to specific target and non-target tissues. We cannot, therefore, predict a precise functional model of the gluten-induced immune response by studying a small, albeit important, set of genes, but we can try to obtain clues about this complexity. Interaction among genes, which is not considered in genome-wide association studies, is estimated by ordinary multivariate analysis, which is likely to provide an independent model of a possible function, or, at least, point to the genes whose expression is important in the differentiation between the affected and the unaffected.
Our discriminant function is proposed in the attempt to improve the diagnosis of CD and as a support to limit invasive techniques. Molecular analysis to discriminate a pathogenic from a healthy phenotype has become increasingly popular with the advent of innovative applications in many types of cancer and complex diseases . Esophago-gastro-duodenoscopy is still the gold standard for the diagnosis of CD, but it can decrease the patient’s compliance and is indeed a major bottleneck in developing countries: a simple blood sample, which can also be easily dispatched, may help to disseminate the diagnostic coverage to the majority of patients that cannot reach a specialized reference centre .
In the near future, because of the new ESPGHAN protocol , we may have no information about the status of the traditional target tissue in many patients: gene expression on a blood sample may well add safety and sensitivity to a biopsy-free diagnostic protocol, thereby providing a good proxy of the mucosal status.
Patients and Methods
The project was discussed in detail with parents to obtain their approval for the use of specimens collected for diagnostic proposal (small bowel biopsies and blood samples) for research purposes. Patients did not undergo specimen sampling over and above those required for routine diagnostic procedures as indicated in the ESPGHAN guidelines . Written consent is included on the clinical form that the patient signs prior to collection of specimen: confidentiality between parents and doctor was an accepted proxy of a second written consent that might have been considered invasive by parents. Parents were informed of each result of the expression study that was contained in a written report included in the clinical file. The study protocol was approved by the Ethics Committee of the University of Naples “Federico II”.
For gene expression analysis, duodenal biopsies were obtained during esophago-gastro-duodenoscopy (EGD) procedures and fresh-frozen in liquid nitrogen. Celiac disease was diagnosed according to ESPGHAN criteria, and their clinical characterization was based on the Marsh Stage classification . Controls consisted of patients with a normal duodenal mucosa with no atrophy (Marsh lesion stage M0) .
We analyzed 48 biopsies: 20 active CD patients, 6 CD patients on a GDFD, 22 healthy controls. Controls underwent EGD because of gastritis, gastroesophageal reflux disease or suspected Helicobacter pylori infection. The clinical features of Crohn’s disease patients were evaluated according to the Crohn Pediatric Disease Activity Index (PDCAI) . The clinical parameters of our patients are listed in Table 4 and Table 5.
|Code||Sex||Age||Sample type||Clinical status||Histology*|
|Code||Sex||Age||Sample type||Clinical status||Histology/disease status*†||Cohort|
We used the Dynabeads® My Pure™ Monocyte kit (Life Technologies, Foster City, CA) to isolate monocytes from other peripheral blood cell types (B- and T-lymphocytes, NK cells, erythrocytes, dendritic cells etc.). Monocytes were extracted from 10 ml of peripheral blood of 18 healthy controls, 17 CD patients, 9 Crohn’s disease patients and 5 CD patients on a GFD.
Gene expression studies
Total RNA was extracted from duodenal biopsies and blood monocytes with the Ambion® RiboPure™ kit. The quantity of RNA was measured using the Nanodrop® spettrophotometer, and then RNA quality was analyzed by Agarose gel electrophoresis in Tris/Borate/EDTA buffer (TBE). Two µg by each biopsy, and 100 ng by monocytes of total RNA were reverse-transcribed into cDNA with the High Capacity cDNA Reverse Transcription kit, as per the manufacturer’s protocol. After retro-transcription, we carried out a linear pre-amplification step to enhance the low amount of RNA recovered from monocytes. Pre-amplification was performed with the TaqMan® PreAmp Master Mix. Experiments were performed on the 7900HT Fast Real Time PCR system using the TaqMan® Gene Expression Assay, and about 40 ng of cDNA according to the manufacturer’s protocol. The gene expression assay used for candidate genes is reported in the supplementary materials (Table S2). The relative expression was calculated with the comparative Ct method. The expression of each gene was normalized to an endogenous housekeeping gene (GUSb), GUSb was chosen as reference gene after it had been determined as the most stable reference gene out of 5 candidates (β-actin, B2M, GAPDH, GUSb, and HPRT1). The SDS software (ABI, version 1.4 or 2.4) was used to analyze the raw data and then an additional statistic analysis was performed on GraphPad Prism 5.01®. Relative quantification was performed using the ΔΔCt method. All gene expression experiments (original raw data available as supporting materials Table S3 and Table S4) were conducted according to MIQE guidelines (http://www.gene-quantification.de/miqe-bustin-et-al-clin-chem-2009.pdf).
The non-parametric Mann-Whitney U test for independent variables was used to assess the difference between data sets; first degree error was set at ≤0.05. ANOVA was used to estimate differences among mean expression levels, when appropriate. A discriminant analysis was performed to estimate the contribution of the expression of each gene to distinguish CD patients from healthy individuals and disease controls. The aim of this analysis is to weigh the discriminating capacity of each single gene to obtain a new composite variable, the discriminant score (D-score) which provides a group-specific score for each individual. Wilks’ lambda is an estimate of the discriminant capacity ranging from 1 (complete overlap) to 0 (maximum distance). The variable that minimizes the overall Wilks’ lambda is entered at each step. According to this analysis, only a few specific genes were selected for discriminating capacity, giving a significant contribution to the Variance ratio F, with a first degree error always less than 0.001.
By multiplying the canonical unstandardized coefficients produced by the analysis to the actual values of the RQ of the candidate genes a D-score was obtained for each individual. The discriminant score provides a probability of membership to the cases or to the controls groups for each individual. The highest membership probability for each case allows the classification into the diagnostic groups.
Statistical analyses were performed using the SPSS 17.0 (SPSS Inc., Chicago, IL, USA) and GraphPad Prism 5.0 (GraphPad software, San Diego, CA, USA) software packages.
Gene expression in duodenal tissue. The expression of A) LPP, B) TNFAIP3 C) RGS1 and D) TNFRSF14 genes did not differ significantly among the three groups.
Gene expression in peripheral blood monocytes. TAGAP,TNFSF14 and TNFRSF14 genes were expressed at similar levels in CD and CD-GFD monocytes.
Validation of the discriminant analysis. The analysis conducted to classify controls, CD, Crohn and CD-GFD patients in 2 groups: Controls and Celiacs. The Highest Group corresponds to the first prediction choice, and the Second Highest Group to the second one. Effective control and celiac patients were all correctly predicted; Crohn and CD-GFD patients were predicted as controls.
List of TaqMan Gene Expression assays used in the expression experiments (Life Technologies, Foster City, CA).
Raw data of gene expression analysis in biopsy. *For the diagnosis of CD has been applied Marsh classification, all controls have a normal duodenal mucosa with no atrophy (Marsh lesion stage M0).
Raw data of gene expression analysis in monocytes. *For the diagnosis of CD has been applied Marsh classification, all controls have a normal duodenal mucosa with no atrophy (Marsh lesion stage M0). † For Crohn’s patients is indicated disease activity index (CDAI) are accepted outcome, all patients are treated and in remission with a score of < 150.
The authors thank Jean Ann Gilder (Scientific Communication srl., Naples, Italy) for writing assistance.
Conceived and designed the experiments: MG VI MPS LG RA. Performed the experiments: DC MM GG. Analyzed the data: MG VI LG. Contributed reagents/materials/analysis tools: CS DZ. Wrote the manuscript: MG VI LG.
- 1. Schuppan D, Junker Y, Barisani D (2009) Celiac disease: from pathogenesis to novel therapies. Gastroenterology. 137: 1912-1933. doi:10.1053/j.gastro.2009.09.008. PubMed: 19766641.
- 2. Sollid LM, Jabri B (2013) Triggers and drivers of autoimmunity: lessons from coeliac disease. Nat Rev Immunol 13: 294-302. doi:10.1038/nri3407. PubMed: 23493116.
- 3. Dubois PC, Trynka G, Franke L, Hunt KA, Romanos J et al. (2010) Multiple common variants for celiac disease influencing immune gene expression. Nat Genet 42: 295-302. doi:10.1038/ng.543. PubMed: 20190752.
- 4. Trynka G, Hunt KA, Bockett NA, Romanos J, Mistry V et al. (2011) Dense genotyping identifies and localizes multiple common and rare variant association signals in celiac disease. Nat Genet 43: 1193-1201. doi:10.1038/ng.998. PubMed: 22057235.
- 5. Plaza-Izurieta L, Castellanos-Rubio A, Irastorza I, Fernández-Jimenez N, Gutierrez G et al. (2011) Revisiting genome wide association studies (GWAS) in coeliac disease: replication study in Spanish population and expression analysis of candidate genes. J Med Genet 48: 493-496. doi:10.1136/jmg.2011.089714. PubMed: 21490378.
- 6. Sperandeo MP, Tosco A, Izzo V, Tucci F, Troncone R et al. (2011) Potential celiac patients: a model of celiac disease pathogenesis. PLOS ONE. 6: e21281. doi:10.1371/journal.pone.0021281. PubMed: 21760890.
- 7. Fina D, Sarra M, Caruso R, Del Vecchio Blanco G, Pallone F et al. (2008) Interleukin 21 contributes to the mucosal T helper cell type 1 response in coeliac disease. Gut. 57: 887–892. doi:10.1136/gut.2007.129882. PubMed: 17965065.
- 8. Izzo V, Pinelli M, Tinto N, Esposito MV, Cola A et al. (2011) Improving the estimation of celiac disease sibling risk by non-HLA genes. PLOS ONE. 6: e26920. doi:10.1371/journal.pone.0026920. PubMed: 22087237.
- 9. Bragde H, Jansson U, Jarlsfelt I, Söderman J (2011) Gene expression profiling of duodenal biopsies discriminates celiac disease mucosa from normal mucosa. Pediatr Res 69: 530-537. doi:10.1203/PDR.0b013e318217ecec. PubMed: 21378598.
- 10. Ráki M, Tollefsen S, Molberg Ø, Lundin KE, Sollid LM et al. (2006) A unique dendritic cell subset accumulates in the celiac lesion and efficiently activates gluten-reactive T cells. Gastroenterology. 131: 428–438. doi:10.1053/j.gastro.2006.06.002. PubMed: 16890596.
- 11. Harris KM, Fasano A, Mann DL (2010) Monocytes differentiated with IL-15 support Th17 and Th1 responses to wheat gliadin: implications for celiac disease. Clin Immunol 135: 430–439. doi:10.1016/j.clim.2010.01.003. PubMed: 20153259.
- 12. Beitnes AC, Ráki M, Brottveit M, Lundin KE, Jahnsen FL et al. (2012) Rapid accumulation of CD14+CD11c+ dendritic cells in gut mucosa of celiac disease after in vivo gluten challenge. PLOS ONE. 7: e33556. doi:10.1371/journal.pone.0033556. PubMed: 22438948.
- 13. ESPGHAN Working Group on Coeliac Disease Diagnosis, ESPGHAN Gastroenterology Committee, European Society for Pediatric Gastroenterology, Hepatology and Nutrition (2012) Guidelines for the diagnosis of coeliac disease. J PediatrGastroenterolNutr. 54:136-160. doi: 10.1097/01.mpg.0000413582.61110.4a
- 14. Hunt KA, Zhernakova A, Turner G, Heap GA, Franke L et al. (2008) Novel celiac disease genetic determinants related to the immune response. Nat Genet 40: 395–402. doi:10.1038/ng.102. PubMed: 18311140.
- 15. Maiuri MC, De Stefano D, Mele G, Fecarotta S, Greco L et al. (2003) Nuclear factor kappa B is activated in small intestinal mucosa of celiac patients. J Mol Med 81: 373-379. PubMed: 12743709.
- 16. Verhelst K, Carpentier I, Kreike M, Meloni L, Verstrepen L et al. (2012) A20 inhibits LUBAC-mediated NF-κB activation by binding linear polyubiquitin chains via its zinc finger 7. EMBO J 31: 3845-3855. doi:10.1038/emboj.2012.240. PubMed: 23032186.
- 17. Korn T, Bettelli E, Gao W, Awasthi A, Jäger A et al. (2007) IL-21 initiates an alternative pathway to induce proinflammatory T(H)17 cells. Nature. 448: 484–487. doi:10.1038/nature05970. PubMed: 17581588.
- 18. Vecchi M, Crosti L, Berti E, Agape D, Cerri A et al. (1992) Increased jejunal intraepithelial lymphocytes bearing gamma/delta T-cell receptor in dermatitis herpetiformis. Gastroenterology. 102: 1499–1505. PubMed: 1314748.
- 19. Abadie V, Discepolo V, Jabri B (2012) Intraepithelial lymphocytes in celiac disease immunopathology. Semin Immunopathol 34: 551-566. doi:10.1007/s00281-012-0316-x. PubMed: 22660791.
- 20. Gibbons DL, Abeler-Dörner L, Raine T, Hwang IY, Jandke A et al. (2011) Cutting Edge: Regulator of G protein signaling-1 selectively regulates gut T cell trafficking and colitic potential. J Immunol 187: 2067-2071. doi:10.4049/jimmunol.1100833. PubMed: 21795595.
- 21. Gorenne I, Jin L, Yoshida T, Sanders JM, Sarembock IJ et al. (2006) LPP expression during in vitro smooth muscle differentiation and stent-induced vascular injury. Circ Res 98: 378-385. doi:10.1161/01.RES.0000202802.34727.fd. PubMed: 16397143.
- 22. Grunewald TG, Pasedag SM, Butt E (2009) Defining the Role of the Novel Proto oncogene LPP m. Transl Oncol 2: 107–116. PubMed: 19701494.
- 23. Qu Z, Yu J, Ruan Q (2012) TGF-β1-induced LPP expression dependent on Rho kinase during differentiation and migration of bone marrow-derived smooth muscle progenitor cells. J Huazhong Univ Sci Technol Med Sci 32: 459-465. doi:10.1007/s11596-012-0080-0. PubMed: 22886954.
- 24. Yang L, Anderson DE, Baecher-Allan C, Hastings WD, Bettelli E et al. (2008) IL-21 and TGF-beta are required for differentiation of human T(H)17 cells. Nature. 454: 350-352. doi:10.1038/nature07021. PubMed: 18469800.
- 25. Yang MM, Lai TY, Tam PO, Chiang SW, Chan CK et al. (2012) Complement factor H and interleukin gene polymorphisms in patients with non-infectious intermediate and posterior uveitis. Mol Vis 18: 1865-1872. PubMed: 22876110.
- 26. Hollis-Moffatt JE, Chen-Xu M, Topless R, Dalbeth N, Gow PJ et al. (2010) Only one independent genetic association with rheumatoid arthritis within the KIAA1109-TENR-IL2-IL21 locus in Caucasian sample sets: confirmation of association of rs6822844 with rheumatoid arthritis at a genome-wide level of significance. Arthritis Res Ther 12: R116. doi:10.1186/ar3053. PubMed: 20553587.
- 27. Capuano M, Iaffaldano L, Tinto N, Montanaro D, Capobianco V et al. (2011) MicroRNA-449a overexpression, reduced NOTCH1 signals and scarce goblet cells characterize the small intestine of celiac patients. PLOS ONE;6: e29094. doi:10.1371/journal.pone.0029094. PubMed: 22194996.
- 28. Maurano MT, Humbert R, Rynes E, Thurman RE, Haugen E et al. (2012) Systematic localization of common disease-associated variation in regulatory DNA. Science. 337: 1190-1195. doi:10.1126/science.1222794. PubMed: 22955828.
- 29. Orozco LD, Bennett BJ, Farber CR, Ghazalpour A, Pan C et al. (2012) Unraveling inflammatory responses using systems genetics and gene-environment interactions in macrophages. Cell. 26;151: 658-670. PubMed: 23101632.
- 30. Greco L, Timpone L, Abkari A, Abu-Zekry M, Attard T et al. (2011) Burden of celiac disease in the Mediterranean area. World J Gastroenterol 17: 4971-4978. doi:10.3748/wjg.v17.i45.4971. PubMed: 22174546.
- 31. Working Group of European Society of Paediatric Gastroenterology and Nutrition. (1990) Revised criteria for diagnosis of coeliac disease. Report of Working Group of European Society of Paediatric Gastroenterology and Nutrition. Arch dis child 65: 909-911. doi: 10.1136/adc.65.8.909
- 32. Oberhuber G, Caspary WF, Kirchner T, Borchard F, Stolte M et al. (2001) Study Group of Gastroenterological Pathology of the German Society of Pathology. Recommendations for celiac disease/sprue diagnosis. Z gastroenterol 39: 157-166.
- 33. Irvine EJ (1995) Usual therapy improves perianal Crohn’s disease as measured by a new disease activity index. McMaster IBD Study group. J ClinGastroenterol 20: 27–32.