A case-control study was conducted. Fecal samples were obtained from 10 healthy individuals, between 61 and 86 years old, who were not undergoing antimicrobial therapy and not hospitalized for at least the previous two months, and from 10 patients, between 64 and 98 years old, who had been diagnosed with CDI at the moment of stool collection (Table 1). Stool samples were stored at −20˚C until used. All samples were taken under patient consent (Research Ethics Committee protocol number 927/07).

DNA was extracted from stool samples using the QIAamp DNA Stool Mini-Kit (Qiagen) according to the manufacturer’s instructions.

Fecal samples from healthy subjects were inoculated directly or after an alcohol shock procedure on the selective medium cycloserine-cefoxitin-fructose agar (CCFA) and incubated in anaerobic jars for 48 hours at 37˚C (80% N₂, 10% CO₂ and 10% H₂) [12] . Characteristic yellow colonies grown on this medium are identified as C. difficile. Polymerase chain reaction (PCR) targeting the species-specific triose phosphate isomerase gene, tpi, was also employed to verify the presence or absence of C. difficile in stool samples [13] . All fecal samples were also tested for the presence of toxins A and B by the Ridascreen^® C. difficile Toxin A/B (r-biopharm) immunoenzymatic assay (ELISA), according to the manufacturer’s instructions. The presence of C. difficile in stool samples from patients diagnosed with CDI (Table 1) was confirmed.

Two universal (bacteria-specific) primers, L1401R (CGG TGT GTA CAA GGC CCG GGA ACG) and U968F-CG (CGC CCG GGG CGC GCC CCG GGC GGG GCG GGG GCA CGG GGG GAA CGC GAA GAA CCT TAC), were used to amplify the 16S rRNA gene coding region [14] . Amplification reactions were performed in a total volume of 50 µL containing 50 ng of DNA that was extracted from each fecal sample, 10 μM of each primer, 2.5 U of Taq DNA polymerase (Promega), 0.2 mM of each dNTP, 50 mM Tris-HCl (pH 8.3), 40 mM KCl, 5 µg Bovine Serum Albumin (1 mg/mL), 0.5 µL

^aSH: systemic hypertension; DM: diabetes mellitus; EVA: hemorrhagic encephalic vascular accident; ARDS: acute respiratory distress syndrome; CAD: coronary artery disease; UGIB: upper gastrointestinal bleeding; DVT: deep vein thrombosis; *ND: not determined; ^bVAN: vancomycin; MTZ: metronidazole; MOX: moxifloxacin; LVX: levofloxacin; CIP: ciprofloxacin; MEM: meropenem; TZP: piperacillin-tazobactam; FEP: cefepime; TEC: teicoplanin; TBZ: thiabendazole; IPM: imipenem; LZD: linezolid; ETP: Ertapenem; RIF: rifampin; CLR: clarithromycin; FLC: fluconazole; AMK: Amikacin; CXM: cefuroxime; AMC: amoxicillin-clavulanic acid; CFG: caspofungin.

formamide and 2.5 mM MgCl₂. The PCR amplifications were performed in a Veriti^® Thermal Cycler (Applied Biosystems) using the following procedure: 94˚C for 4 minutes, followed by 35 cycles of 94˚C for 1 minute, 58˚C for 1 minute, 72˚C for 2 minutes, and a final extension period of 72˚C for 10 minutes. PCR products were visualized on 1.5% (w/v) agarose gels stained with Blue Green Loading Dye I (LGC Biotecnologia) under UV light (MiniBis Pro^®, Dnr Bio-Imaging System). PCR amplification products were stored at −20˚C.

DGGEs were carried out using the D-Code system from Bio-Rad Laboratories. PCR products (10 - 15 μl) were loaded onto 6% (w/v), 1 mm thick, polyacrylamide gels, in 1X TAE buffer (20 mM Tris-acetate, pH 7.4, 10 mM acetate, 0.5 mM disodium EDTA), containing a denaturing gradient of urea and formamide varying from 45% to 65%. Electrophoresis was performed at a constant voltage of 75 V for 16 hours. After electrophoresis, the gels were stained for 40 min with SYBR Green I nucleic acid gel stain (Molecular Probes). The gel bands were visualized and digitized using a STORM apparatus (Amersham Pharmacia Biotech). Some bands were retrieved from the gels, reamplified and sequenced. A dendrogram was constructed based on the presence and absence of bands with the unweighted pair group method (UPGMA) with mathematical averages and similarity coefficient of Jaccard (tolerance and optimization of 1%) using the BioNumerics software (Applied Maths).

Bands excised from the DGGE gels were eluted in 30 μL of water (4˚C overnight). Each supernatant (1 μL) was used as the PCR template with primers U968F (without the GC clamp) and L1401R, as described above but without BSA and formamide. PCR amplicons were purified with the GFX PCR DNA & Gel Band (GE Healthcare) kit. The purified products were ligated into a pGEM^® T-Easy plasmid vector (Invitrogen) and then transformed into competent Escherichia coli DH10B cells. After bacterial transformation, DNA from selected clones was used as the template in an amplification reaction using M13 primers. The amplicons obtained were purified (Illustra GFX^TM PCR DNA and Gel Band Purification Kit, GE Healthcare) and sequenced with an ABI Prism 3100 automatic sequencer (Applied Biosystems Inc.). Sequence quality was analyzed using Codon Code, and nucleotides with a Phred score less than 10 were discarded. Qualified sequences were manually edited using the BioEdit software to remove vector and primer sequences. The resulting sequences were identified by BLASTN analyses using the GenBank database (http://blast.ncbi.nlm.nih.gov/Blast.cgi).

From the DGGE profiles, a qualitative numerical matrix was generated to non-metric multidimensional scaling (NMDS) using the Bray-Curtis similarity coefficient. To test the significance of the two groups, ANOSIM and PERMANOVA tests were performed using the Past 3.07 software (HAMMER, HARPER & RYAN, 2001). The total number of bands found for each sample was determined and the average was calculated and analyzed by Mann-Whitney test with the Statistical Package for Social Science (SPSS) software version 11.5 (IBM, Chicago, USA).

Samples obtained from healthy subjects were confirmed as negative for the presence of C. difficile and its toxins. On the other hand, samples from individuals diagnosed with CDI had the presence of C. difficile and toxins confirmed.

All samples used in this study were from subjects older than 61 years, with the same proportion of females and males in both the healthy (average of 72 years of age) and CDI groups (average of 82 years of age), to minimize differences regarding age or gender. All individuals from the CDI group presented one or more clinical conditions, with predominance of respiratory syndromes. These inpatients were submitted to a range of antibiotic therapies, but we observed that metronidazole, vancomycin, carbapenems and/or fluoroquinolones were administered to all individuals (Table 1).

DGGE analysis of the bacterial 16S rRNA gene was performed to obtain an overview of the community structure. DNA extracted from fecal samples was used as template for PCRs. The amplified 16S rDNA (expected size of 463 bp) were used for DGGE analysis. DGGE profiles containing the group of healthy individuals presented a similar pattern between samples, and this was also observed in the CDI group. Digitalized images obtained for each group were analyzed with BioNumerics 6.5 (Applied Maths NV) software, which allowed marking and counting of bands, as presented in Figure 1. After analysis with Bionumerics it was possible to observe that bacterial diversity was reduced in the CDI group, in comparison to the healthy group. The average of the number of bands was calculated for the healthy (13.5 ± 2.97) and CDI (7.4 ± 2.5) groups (Figure 2). Statistical non-parametric Mann-Whitney test was applied (p value < 0.001) and showed a significant reduction of the number of bands in the CDI samples. A dendrogram showed five main clusters, but an outstanding similarity among samples from the same group, with low similarity between both groups. In general, the similarity observed was higher in the healthy group (70%) when compared to the CDI group (50%). Comparing the two groups, the similarity was very low (30%) (Figure 1). In Figure 3, following the NMDS analysis the samples of the healthy group showed a different behavior from the CDI group, which was confirmed through the ANOSIM and PERMANOVA (p < 0.0001). All these data confirm the differences between the two groups, although a more expressive similarity is highlighted in samples from the healthy group.

One of the principles of the DGGE technique is that each sample with a different DNA sequence will denature at a different position. So, in theory every band in the gel will represent a distinct bacterium in the studied community. The intensity of the bands of the same sample also gives information about the amount of DNA present, where darker bands are composed of a higher amount of DNA than lighter bands. In this study, the DGGE profiles obtained demonstrated the presence of dominant bands,

represented by the letters A, B, C, D, E, F, G and H in Figure 1. Table 2 shows the occurrence of these bands. The total incidence was calculated by the number of bands that appeared at the same time in the gel divided by the total of 20 samples. The occurrence in the healthy group was calculated by the number of bands that appeared at the same time in the gel, divided by the 10 CDI negative samples. The same was done for the CDI group. The data in Table 2 shows that bands A and B are exclusive to the healthy group and the CDI positive group had no unique bands. There are bands present in both groups, but with a higher occurrence in the healthy group, such as bands C, E and H. Bands A and B represented bacteria present in most samples of the healthy group and absent in the CDI group. The D band occurred equally in both groups. The F band represents bacteria that appear in higher amounts in the CDI group when compared to the healthy group.

Taking advantage of these results, twenty-six bands (indicated with letters and numbers in Figure 1) from the DGGE profiles corresponding to the healthy and CDAD patients were extracted from DGGE gels, reamplified, cloned and sequenced. Table 2 presents summarized results generated by the NCBI BLAST tool for each of the sequenced bands. Most of the dominant bands A and E from the healthy group were identified as uncultured organisms, whereas band B was identified as a Eubacterium rectale. Furthermore, bands F, G and H were mainly identified as Akkermansia mucinophila. Nevertheless, although some bands were present in both clusters, the identified species was different for each group. For instance, Klebsiella variicola and Klebsiella pneumoniae (band D) were exclusively found in CDI samples.

As can be seen in Table 3, after editing the sequences, BLASTN analyses revealed that three bands (D6, D7 and D17) from the CDI group were closely related to a species

from the Proteobacteria phylum, particularly the member of the Enterobacteriaceae family, Klebsiella variicola/K. pneumonia. One band (F1) was identified as belonging to the Firmicutes phylum, an uncultured Ruminococcaceae, and the remaining sequences C4d and D4 were affiliated with Ruminococcus torques and uncultured bacteria, respectively. Concerning the healthy group, the dominating phyla were Firmicutes, represented by the Lachnospiracea family, Eubacterium rectale (B7) and Eubacterium eligens (A5) species, followed by Verrucomicrobia, with Akkermansia sp. being the most representative microorganism. Uncultured Lachnospiraceae (A4 and A8), Lachnospiraceae (B6 and C3), uncultured Clostridia (E2), Akkermansia mucinophila (C4, F4, F8, G6, H3 and H8), uncultured Roseburia (C2), and uncultured bacteria (B3, D5, E4, F6, F17 and H1) were also found.