Two groups of children referred to the Pediatric Gastroenterology and Liver Unit of the "Sapienza" University of Rome were included in this study: 20 CD (mean age 8.3 years, range 1.2-16.1 years) in active and in remission state (at diagnosis and after at least 9 months of gluten-free diet, respectively) and 10 controls undergoing upper gastrointestinal endoscopy for functional dyspepsia (mean age 11.7 years, range 7.8-20.8 years). The latter tested negative for antitransglutaminase and antiendomysial antibodies with normal IgA levels, while histology of duodenum did not reveal features of CD. Diagnosis of CD had been performed according to ESPGHAN criteria . Table 2 summarizes clinical features of the studied population.
Size appropriate and well oriented endoscopic biopsy specimens were obtained from the second part of the duodenum. The histopathological diagnosis was based on typical mucosal lesions with crypt cell hyperplasia, villous atrophy, and increased number of intra-epithelial lymphocytes (IELs) . All untreated CD patients were positive for antiendomysial and antitransglutaminase antibodies at the time of diagnosis. In all patients there was an endoscopic improvement of duodenal mucosa following gluten withdrawal, but only in 2 of them (patients number 12 and 19) there was also a full histological improvement. None of the children included in the study was treated with antibiotics for at least 3 months before the sampling time. The study protocol was approved by the Committee on Ethical Practice of the 'Policlinico Umberto I' hospital. Children were enrolled in the study after written informed consent from their parents. The biopsy samples were placed in liquid nitrogen immediately after their emission and stored at -80°C until analysis.
The strains listed below were obtained from the American Type Culture Collection (ATCC) and used as marker on TTGE gel electrophoresis: Bacteroides fragilis ATCC 23745, Bacteroides thetaiotaomicron ATCC 29148, Bacteroides vulgatus ATCC 8482, Parabacteroides distasonis ATCC 8503, Escherichia coli MG1655. Bacterial DNA was extracted with UltraClean kit (MO BIO Laboratories, Solana Beach, California, USA) according to the manufacturer's instructions.
Duodenal biopsy specimens from CD and control patients were first quickly washed in 500 μL of physiologic saline with 0.016% dithiothreitol to remove luminal bacteria from the mucus, and then utilized for DNA extraction procedure by DNeasy tissue kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. In order to obtain maximum yield of both Gram-positive and Gram-negative bacteria, a special step in DNA purification protocol was added, following DNeasy tissue kit manual. Briefly, 180 μL of ATL buffer were added to sample followed by 180 μL volume of enzymatic lysis buffer (20 mM Tris·Cl, pH 8.0, 2 mM sodium EDTA, 1.2% Triton® X-100, lysozyme to 20 mg/ml), and incubated for 30 minutes at 37°C. Next, 25 μL of proteinase K solution and 200 μL of buffer AL were added, followed by an incubation step at 56°C for 30 minutes.
DNA concentration was determined using an Eppendorf biophotometer at 260 nm.
We obtained similar DNA concentrations after kit extraction both from celiac patients and controls biopsies. A Mann-Whitney U test was performed on total DNA concentration (P = 0.11), indicating a similar amount of extracted DNA in both celiac and controls.
Polymerase chain reaction (PCR) was performed, as previously described  using 400 ng of metagenomic DNA, with minor modification. Briefly, to rule out unspecific PCR products we performed touchdown PCR with a starting annealing temperature of 58°C and decreasing it by 0.5°C each cycle to reach 53°C, then 30 cycles at 53°C were achieved. Same amounts of amplified DNA were also obtained. A Mann-Whitney U test was performed on PCR amplicons (P = 0.23), indicating a similar amount of PCR products in both celiac and controls. To minimize heteroduplex formation and single-stranded DNA (ssDNA) contamination during PCR amplification that might cause sequence heterogeneity in a single TTGE band, an additional 5 cycles of reconditioning PCR was performed, taking 1/10 of the previous PCR volume as template in a new reaction. Moreover, we used 16S rDNA V6-V8 region instead of V3-V4 region that showed coamplification with human DNA. To avoid the problem due to the low bacterial load we performed six individual PCR reactions for each sample. The individual PCR reactions were unified, analyzed by electrophoresis on 2% agarose gels containing ethidium bromide to determine their size (498 bp), and concentrated with SpeedVack (Savant, Holbrook, NY, USA). The unified PCR reactions, before and after the concentration step, were titrated using two different methods: first, densitometry analysis of agarose gel by GelQuest software (Sequentix, Klein Raden, Germany); second, measure of DNA density by biophotometer at 260 nm. The results obtained by such measures were in agreement each other.
PCR protocol was optimized to obtain maximum yield from starting total DNA. The band intensity was quantified at every step (touchdown PCR, reconditioning PCR, concentrated PCR) to ensure an equal DNA concentration. A first-step assessment of DNA suitability for subsequent PCR was achieved through a β-globin gene amplification for each starting sample. Briefly, aliquots of each DNA sample (50 ng) were amplified with specific primers: forward primer, 5'-CAACTTCATCCACGTTCACC-3; reverse primer, 5'-GAAGAGCCAAGGACAGGTAC-3'. Amplification reactions were carried out in a 50-μl volume containing 1× PCR buffer II (Applied Biosystems, Roche, California, USA), 3 mM magnesium chloride, 200 μM each deoxynucleoside triphosphate, 50 pmol each primer and 5 Uμ/l AmpliTaq Gold polymerase (Applied Biosystems). The PCR was carried out under the following conditions: 1 cycle of 95°C for 7 min, 35 cycles of 94°C for 1 min, 55°C for 1 min and 72°C for 1 min and 1 cycle of 72°C for 7 min. 500 ng of DNA of PCR product from each sample were used to perform the subsequent TTGE experiments.
TTGE analysis of PCR amplicons
We used the DCode Universal mutation detection system (Bio-Rad, Paris, France) for the sequence-specific separation of PCR products. Electrophoresis was performed as previously described . TTGE runs were conducted in triplicate and gel photographed with DigiDoc-It system (UVP, Cambridge, UK).
We choose to detect those particular species whose presence seems to be involved in celiac disease [7, 9]. 16S rDNA gene-targeted primers were utilized to detect them. The primers used were ECO-1 5'-gacctcggtttagttcacaga-3', ECO-2 5'-cacacgctgacgctgacca-3' for Escherichia coli (585 bp); BV-1 5'-gcatcatgagtccgcatgttc-3', BV-2 5'-tccatacccgactttattcctt-3' for Bacteroides vulgatus (287 bp); g-Ccoc-F 5'-aaatgacggtacctgactaa-3', g-Ccoc-R 5'-ctttgagtttcattcttgcgaa-3' for Clostridium coccoides group (438-441 bp), g-Bifid-F 5'-ctcctggaaacgggtgg-3', g-bifid-R 5'-ggtgttcttcccgatatctaca-3' for Bifidobacterium spp (549-563 bp). The PCR were performed as previously described .
Agglomerative Hierarchical Classification (AHC.) Dendrogram generated with XLStat 7.5 (Addinsoft, NY, USA) on binary matrix of TTGE variables was evaluated by one-tailed chi-squared test. Data were automatically mean centred and unit variance (UV) scaled. A P value equal or less 0.05 was considered statistically significant. Dice similarity index (S
, mean % ± SD) was calculated within the respective HC and CD groups to assess inter-individual similarity by the formula S
)/(nA + nB), where n
is the total number of bands in pattern A, n
is the total number of bands in pattern B and n
is the number of bands common to pattern A and B. Ecological features. Doc-It LS software (UVP, Cambridge, UK) was used for TTGE bands densitometry peak height quantification, and the correspondent data were analyzed for the microbial biodiversity by Shannon-Wiener index with SigmaPlot 9.0 software. Intra-group variance value (V value) was also calculated. V value defines the variance of data points in each cohort, representing the data dispersion, and indicating the homogeneity/heterogeneity between individuals within a population. In addition, the range-weighted richness (Rr), reflecting the carrying capacity of the duodenal system, was calculated by the formula Rr = N2 XTg, where N is the total number of bands in the TTGE profile and Tg the temperature gradient comprised between the first and the last band of the same pattern . Principal Component Analysis (PCA). Linearly-dependent TTGE variables were ortogonalized in new factorial axes (F1,F2...Fn) through PCA by XLStat 7.5 (Addinsoft). The coordinates of the observations on the factorial axes were considered new variables for subsequent factorial discriminant analysis (FDA). Factor Discriminant Analysis (FDA). FDA included in XLStat 7.5 software was performed to create a predictive model useful to classify the patients into one of the three groups according to their TTGE profile. Wilk's Lambda test was used and a P value less than or equal to 0.05 was considered statistically significant. Partial Least Square Discriminant Analysis (PLS-DA). PLS-DA included in SIMCA+ software (UMETRICS, Umea, Sweden) was performed to depict score plot of TTGE profiles by means of principal components PC1 and PC2, and to assess TTGE band importance. Data were automatically mean centred and unit variance (UV) scaled by the statistical software. Each TTGE band was hierarchically classified based on a software-assigned variable importance (VIP) value. The variables with VIP value > 1 were chosen as discriminatory. Non-parametric statistical methods. For Shannon-Weaver index, species-specific PCR, FDA and PLS-DA, a bilateral Wilcoxon signed rank test was utilized to compare active and inactive CD patients' groups, whilst a bilateral Mann-Whitney U-test was utilized to compare active/inactive CD patients with control group. A P value less than or equal to 0.05 was considered statistically significant.