DGGE fingerprinting analyses of all bacteria, Clostridium cluster IV and Clostridium cluster XIVa indicate a highly diverse dataset between individuals and uniqueness of fecal microbiota. Table 1 shows the average number of bands in cancer patients at the three time points and for controls over all time points. It becomes apparent that the average number of bands within Clostridium cluster IV declined immediately after chemotherapy (T1), followed by a recovery at T2. The average number of Clostridium cluster XIVa bands decreased after onset of chemotherapy and remained low also at T2. The datasets were subjected to principal component analysis (PCA). PCA extracts underlying components of samples according to their variance. Figure 1A illustrates the bacterial fingerprints of sample P01 over time. Figure 1B displays the PCA analysis of all bacteria. Most samples taken after chemotherapy are grouped together with all other samples. Patients who receive antibiotics are indicated as black symbols. They cluster together with the samples taken after chemotherapy and also with the majority of samples before chemotherapy and healthy controls. There are two exceptions though: Two samples from P07 after chemotherapy under antibiotic treatment are outliers in the lower right part of the PCA plot. P07 received blood stem cell transplantation resulting in a sharp decline in bacterial abundances as measured with quantitative PCR. The first two principal components explain 17.4% of variance.

Figure 1C shows the principal components analysis of Clostridium cluster IV. DGGE fingerprints of individuals after chemotherapy are found to be less variable than healthy controls and patients before onset of treatment. Although overlapping, PCA resulted in grouping of band patterns before and after chemotherapy. Additional effects by antibiotic treatment became evident: Antibiotic treatment significantly reduced the diversity within the Clostridium cluster IV (p = 0.00003) with Shannon diversity index being 1.4±0.7 compared to patients under chemotherapy alone 2.1±0.6. In the PCA plot, samples affected by antibiotics are found in the lower right corner of the plot. This means that they are grouped according to their variance along principal component (PC) 1 and 2. These two PCs explain 17.9% and 9.15% of the variance in the dataset, underlining the validity of this interpretation. Principal components analysis of Clostridium cluster XIVa is not shown.

To study whether chemotherapy with or without antibiotics changes the human GI microbiota composition in contrast to healthy individuals and over time, we investigated absolute numbers and relative percentages of bacterial subgroups. Absolute numbers give an indication about the direct antimicrobial effects of the treatments. Relative quantification is able to identify which bacterial subgroups are particularly affected and helps to describe the community disruption induced by chemotherapy with or without antibiotics. In absolute numbers, oncology patients harbored significantly less bacteria (p<0.05) than healthy control (figure 2). From already low bacterial counts before chemotherapy, bacterial abundance significantly declined further (p = 0.037) immediately after chemotherapy (T1) and recovered 5–9 days later (T2) in comparison to time points before treatment (T0). Absolute numbers of bacteria in different time points of healthy controls are following a lognormal distribution in contrast to microbiota abundances in oncology patients. The decrease in total bacteria following chemotherapy (p = 0.037) was significantly greater than any variation in copy numbers observed in healthy controls (p = 0.027). The observed decrease after chemotherapy affected the Bacteroides (p = 0.044), the bifidobacteria (p = 0.034) and Clostridium cluster IV (p = 0.049) as shown in figure 3. There were also fewer absolute numbers of Clostridium cluster XIVa, but this difference was not significant. All patients with fever showed an increase in total fecal microbiota (see figure 3). In sample P07 a sharp decline affecting all bacteria and bacterial subgroups was observed at T1 (figure 3), following blood stem cell transplantation and medical intervention.

Patients who received antibiotics had highest abundances of all bacteria (p = 0.000003) amongst all patients (data not shown). This bacterial overgrowth affected the Bacteroides, the bifidobacteria and Clostridium clusters IV and XIVa, since relative abundances of those subgroups did not stand out significantly. Thus, patients were grouped according to their chemotherapeutic cycle regardless whether or not they received antibiotics. The influence of antibiotics on the species composition as assessed with PCR-DGGE fingerprinting is discussed in the previous section.

Relative quantification of Clostridium cluster XIVa as percentage of total bacterial DNA showed that oncology patients harbored significantly less Clostridium cluster XIVa (p = 0.047) than healthy controls. The mean proportion of Bacteroides in stool samples was 26±12% in chemotherapy patients and 22±14% in healthy individuals. The mean percentage of bifidobacteria in patients was 0.8±1.4% and 0.3±0.6 in controls. Patients harbored on average 16±9% of Clostridium cluster IV and 18±12% of Clostridium cluster XIVa, while controls harbored 20±12% and 24±15% of clostridial clusters IV and XIVa.

The mean percentage of Clostridium cluster XIVa before chemotherapy was 22±13% compared to after chemotherapeutic cycles with 19±12%. The average amount of Bacteroides, bifidobacteria and Clostridium cluster IV were 26±11%, 1.4±2% and 16±9% at time points before chemotherapy and 28±14%, 0.5±1.2% and 18±12% after chemotherapy. Figure 3 illustrates the development of the microbiota in the course of antibiotic treatment. Data were normalized for clarity, so that changes in abundances from time point T0 (before onset of treatment) to T1 (1–4 days after chemotherapy) and T2 (5–9 days after chemotherapy) rather than relative abundances are shown. It can be seen that chemotherapy causes a dramatic reduction of microbiota abundance immediately after chemotherapy, affecting all subgroups. As mentioned above, the significant decrease in all bacteria following chemotherapy was significantly greater than any variation in copy numbers observed in healthy controls (p = 0.027).

To find out whether the chemotherapeutic and antibiotic disruption favors the growth of pathogens, we investigated the abundance of C. difficile. Three out of seventeen patients receiving chemotherapy harbored C. difficile (data not shown). Patient P09 harbored C. difficile at all time points investigated. Mean proportion over all four samples of P09 was recorded as 0.4±0.7%, yet the highest level (1.22% of total bacteria) occurred at sampling point T1 immediately after chemotherapeutic and antibiotic treatment. C. difficile was detected in P11 (3.90% of all analyzed bacteria) after chemotherapeutic intervention at time point T1. P14 carried C. difficile in low abundance directly after onset of chemotherapy (0.003% of all analyzed bacteria). Samples of patients P09 and P11 at T0 and T1 were further analyzed in 454 sequencing.

High throughput sequencing showed a dramatic increase in sequences within the Peptostreptococcaceae towards sequences 98.9–100% similar to C.difficile^T (figure 4): Clostridium bartletti related sequences (98.1–100% similarity) were only detected before chemotherapy (T0). After chemotherapy (T1), 63 sequences 98.9–100% similar to C.difficile appeared in samples P11 and P09. In accordance with the phylogenetic classification by the ribosomal database project, they are shown as ‘unclassified Peptostreptococcaceae’ in figure 5.

Furthermore pronounced reductions of Faecalibacterium spp. as well as lactobacilli, Veillonella spp., bifidobacteria (in P09) and E.coli/Shigella became apparent in response to chemotherapy (figure 5). The abundance of lactobacilli decreased in both patients after chemotherapy, in P09 from already low levels. Individual P11 did not receive concomitant antibiotics, whereas P09 did. In P09 and P11 Faecalibacterium spp. decreased dramatically from 9.5% and 8.3% to 0.07% and 0.00%, respectively. In both individuals Enterococcus faecium increased following chemotherapy. Furthermore, less abundant sequences appeared that were attributable to bacterial genera not detected before chemotherapy. These genera are: Eggerthella, Megasphaera, Parvimonas (only in P11), Anaerostipes, Eubacterium, Anaerococcus, Methylobacterium, Holdemania, Turicibacter, Akkermansia, Sutterella (only in P09), Sphingomonas, Anaerotruncus, Coprococcus, Streptococcus and Dorea. Species with abundance <0.01% of all sequences are not shown in figure 5. The number of Blautia species from Clostridium cluster XIVa remained constant in the 454 sequencing datasets before and after chemotherapy.

The Viennese Human Ethics committee (3., Thomas-Klestil-Platz 8/2) under the chair of Dr. Karin Spacek, approved the proposal of the project “Analysis of microbiota in feces of patients with immunosuppression”. Votum: EK 07-153-VK, 2008. From all participants involved in the study written consent was obtained.

Seventeen subjects receiving ambulant chemotherapy with or without antimicrobial therapy (aged 59±13 y, BMI 27±6) from the Sozialmedizinisches Zentrum Ost (SMZ Ost) in Vienna and seventeen healthy individuals (aged 65±18 y, BMI 24±5) joined this study. Four fecal samples within two weeks were collected of each ambulant oncology patient in order to collect samples before and after a single immune-suppressive chemotherapy cycle. The four samples obtained from every patient were grouped into three groups: samples taken before the day of chemotherapy (T0), samples taken 1–4 days after chemotherapy (T1) and samples taken 5–9 days after chemotherapy (T2). Healthy individuals also donated four samples during two weeks. Gender ratio among healthy controls was 56% female, 44% male. Oncology patients were 47% female and 53% male. Three out of seventeen patients (P04, P08, P13) had never received any chemotherapy before, while the others had a history of chemotherapy. Anonymous medical records reported types of malignancies as well as chemotherapeutic and antimicrobial treatment as shown in table 2.

We interviewed all study participants assessing age, gender, body length, weight, health status (chronic and acute diseases), and life-style aspects such as alcohol consumption and physical activity. Dietary habits were assessed using a food frequency questionnaire. Exclusion criteria for healthy controls were (a) antimicrobial medication (b) chemotherapeutic treatment and (c) pre- and probiotics at least three months before sample collection. Approval for this study was obtained from the Viennese Human Ethics committee (3., Thomas-Klestil-Platz 8/2).

After collection, study participants immediately stored their samples at -18°C in their homes. Stool samples were still frozen when brought to the laboratory and then immediately stored at −70°C. A 200 mg aliquot of each sample was treated twice for 45 s in a bead-beater (Mini-Beadbeater-8). Thereafter DNA was extracted using the QIAamp® DNA Stool Mini Kit (QIAGEN) following the manufacturer's protocol. The DNA was stored at −20°C until analysis.

We used type strains, known to be part of the human gastrointestinal microbiota and cloned sequences to design a DGGE standard lane marker. Type strains Bacteroides thetaiotaomicron DSM 2079T, Enterococcus faecium DSM 20477T, Lactobacillus reuteri ATCC 55730T, Bifidobacterium longum ssp. longum DSM 20097T, Escherichia coli IMBH 252/07 and clones CL16 and CC34 (see below) were used for creating a comparable standard lane marker for DGGE gels analyzing all bacteria. E.coli IMBH 242/07 gave 4 bands due to its multiple operons for the 16S rRNA gene.

To create a standard lane marker for DGGE analysis and to identify members of the Clostridium cluster XIVa we constructed clone libraries from two stool samples of healthy volunteers. For this purpose PCR products amplified with primers 195-F [46] and Ccocc-R [47] were inserted into a p-GEM Easy Vector (Promega) following the instructions of the manufacturer. Nucleotide sequences were corrected for primer and vector sequences in CodonCodeAligner (www.codoncode.com) and taxonomically identified using the online tools of the ribosomal database project (http://rdp.cme.msu.edu/). The clone library used for creating a standard lane marker for DGGE analysis of Clostridium cluster IV has previously been described [48]. Clones CL16 (Clostridium leptum 16) and CC34 (Clostridium coccoides 34) were also used as positive controls in Taqman qPCR.

PCR was carried out amplifying 16S rRNA gene sequences from bacteria in fecal samples, type strains and cloned sequences for DGGE analysis as well as for creation of the clone library using group- and kingdom- specific primers (table 3). The PCR reaction mixture consisted of ready-to-use mastermix (Promega) with 1.5 mM MgCl2, 500 nM of primers and 2 µl of template DNA. When amplifying fecal samples, bovine serum albumin (Fermentas) was added to a final concentration of 400 µg/ml. We used a Robocycler (Stratagene) for all amplifications.

DGGE was performed as previously described [49]. Primer pairs and annealing temperatures to analyze the diversity of (a) all bacteria, (b) Clostridium cluster IV and (c) Clostridium cluster XIVa are described in table 3. PCR products were separated by polyacrylamide gels with a denaturing gradient of 30–60% for predominant bacteria, 30–50% for Clostridium cluster IV and 35–50% for Clostridium cluster XIVa. Electrophoresis was performed for 9 h at 129 V at 60°C (predominant bacteria), 5 h at 200 V at 60°C (Clostridium cluster IV) and 7 h at 200 V at 60°C (Clostridium cluster XIVa). Standard lane markers were created for each DGGE analysis assay to ensure reliable gel-to-gel comparison. These standard lane markers (described above) were loaded in triplicate on each gel to adjust for gradient-variations between gels. We analyzed PCR-DGGE fingerprints using GelComparII (www.applied-maths.com). When generating the band comparison, a 1% tolerance was selected. Principal components analysis (PCA) was applied on quantitative band comparison datasets in ‘R’ (www.r-project.org) using the default settings. Shannon diversity index was calculated on quantitative band information as well with the default settings implemented in the ‘vegan’ package in ‘R’ (www.r-project.org). Shannon index is defined as H = −∑ pi ln pi, where pi is the proportional abundance of species i. In short, the higher the Shannon index is, the higher is the diversity. For interpretation of results, samples were grouped into three groups: samples taken before the day of chemotherapy (T0), samples taken 1–4 days after chemotherapy (T1) and samples taken 5–9 days after chemotherapy (T2).

The abundance of bacteria and bacterial subgroups was measured by 16S rRNA gene-targeting TaqMan qPCR in a Rotorgene 3000 (Corbett Life Science). Primers, annealing temperatures and expected product sizes are shown in table 4. Each sample was analyzed in duplicate. Amplifications were carried out in a total volume of 10 µl consisting of 5 µl Taq-Man SensiMix DNA Kit (Quantance), 1 µl of each primer and Taq-Man probe (concentrations see table 4) and 10 ng of bacterial DNA. Amplification programs included an initial denaturation at 95°C for 10 min followed by 40 cycles consisting of denaturation at 95°C for 30 s, annealing at 55°C (all bacteria, Clostridium cluster IV), 56°C (Clostridium cluster XIVa), 58°C (C. difficile) or 60°C (Bacteroides, bifidobacteria) for 30 s and extension at 72°C for 50 s.

We used tenfold serial DNA dilutions of type strains Bacteroides thetaiotaomicronT, Bifidobacterium longum ssp. longum^T and C. difficile as well as cloned sequences and one fecal sample to construct standard curves for comparison of PCR reaction efficiencies among different experiments.

We quantified DNA of Bacteroides thetaiotaomicron^T, Bifidobacterium longum ssp. longum^T and C. difficile, using the nanodrop method and calculated DNA copies/µl through mean G+C content of each strain. Clones CL16 and CC34 were amplified with the SP6 Promoter Primer (Promega, Cat.# Q5011) and the T7 Promoter Primer (Promega, Cat.# Q5021) and the PCR product quantified using a nanodrop machine. Knowing the sequences of these two PCR products and their flanking vector sequences we could quantify the copy numbers and use it as standards. Relative percentages of bacterial subgroups were calculated in relation to total rRNA gene copies amplified with primer pair BAC-338-F and BAC-805-R [50].

We reviewed sensitivity of PCR reactions with stepwise dilutions of standard curve DNA until we achieved sensitive detection levels of PCR. The specificity was confirmed using non-target DNA.

In total, four samples (P09-T0, P09-T1, P11-T0, P11-T1) were amplified with primer 525F (5′- TCAGCAGCCGCGGTAATAC -3′) and 926R (5′-TCCGTCAATTCCTTTGAGTTT -3′) using a high-fidelity DNA polymerase (Phusion®, Finnzymes, Thermo Fisher Scientific) and submitted to 454 barcode sequencing (AGOWA, Berlin, Germany), resulting in a total of 113 000 reads. The sequences were trimmed and aligned using the pyro pipeline of the ribosomal database project (http://rdp.cme.msu.edu/). Only sequences longer than 150 bp were retained, resulting in 3886 to 6811 sequences per sample with the average lengths of 366 to 368 bp. All analyses were performed using the online tools of the ribosomal database project pyro pipeline (http://rdp.cme.msu.edu/). Results of the phylogenetic classification are shown as a heatmap [51]. The Peptostreptococcaceae, harbouring also C.difficile, from all four datasets were analyzed in more detail using the online tools of the ribosomal database project. 100% similar sequences were grouped and their abundances shown together with a phylogenetic tree.

Statistical evaluation of differences between groups (chemotherapy and control) and changes within the chemotherapy group (all time points before and after chemotherapy) was carried out using the OriginPro version 8 (OriginLab, Northampton, MA). For two group comparisons of independent ordinal and interval values we used the two-sample t-test and the nonparametric Mann-Whitney U-test. For the analysis of related data we used the paired sample t-test or the non-parametric Wilcoxon signed-rank test. P values <0.05 were considered statistically significant. To show the decline in abundance immediately after chemotherapy qPCR results were plotted in heatmaps [51]. Values were z-scored for presentation in this heatmap showing changes over time rather than absolute abundances.

We assessed the participants' dietary habits using a food frequency questionnaire. All study participants (patients and controls) were omnivores and showed similar consumption patterns of liquids, alcohol, fruits, vegetables, grains and milk products. Healthy controls stated more frequent consumption of fruits, whole grain products and alcohol several times a week compared to patients receiving chemotherapy.