2.1. Group 1a and 1b.

At the end of the elective surgical procedure and during termination of general anaesthesia endobronchial secretion (EBS) was obtained through the endobronchial tube (to avoid contamination of samples from oral or pharyngeal bacterial and fungal microbiota during sampling) by the anaesthesiologist and the study physician, captured in a sterile cup and immediately brought to the inhouse microbiology laboratory. An oral swab was obtained prior to sampling EBS. In accordance to recent literature showing similar community compositions by 16s rRNA sequencing between EBS and BAL from single individuals and according to the local ethical approval (not allowing lavage in group 1) BAL was not performed in group 1 patients [16].

2.2. Group 2a and 2b.

Samples from the LRT were obtained by bronchoscopy through endotracheal tubes (to avoid contamination of samples from oral or pharyngeal bacterial and fungal microbiota during sampling) and BAL (20 ml normal saline) of the right lung. Oral swabs were obtained prior to bronchoscopy.

2.3. Group 3b.

Samples from the LRT were obtained within 24 hours of clinically (including x-ray) established diagnosis of CAP, ASP, NAP or VAP by bronchoscopy (to avoid contamination of samples from oral or pharyngeal bacterial and fungal microbiota during sampling) and BAL (20 ml normal saline) through endotracheal tubes and directed to pulmonary infitrates suggestive of CAP, ASP, NAP or VAP. In case of infiltrates in both lungs samples from both sites were obtained and processed separately. BAL was captured in a sterile cup and used for routine and study purpose. An oral swab was performed prior to bronchoscopy. A second bronchoscopy and BAL was performed 4–7 days after the first sampling in patients still intubated and mechanically ventilated. Samples for calculation of Candida colonization index were obtained concomitantly to bronchoscopy.

2.4. Group 4.

Routine blood cultures obtained from patients in our Department of Internal Medicine and routinely processed in our inhouse laboratory were observed for Candida positivity. Candida spp. were subsequently identified by routine microbiological methods including MALDI-TOF-mass-spectrometry, stored at -70°C and processed as described below.

Respiratory samples from all patients were immediately brought to the inhouse microbiology laboratory, aliquoted, processed and remaining samples stored at -70°C until further analysis (e.g. microbiome analysis). Serum samples were obtained from all patients simultaneously with microbiological sampling, were centrifugated and stored at -70°C until further analysis. Three pairs of blood cultures were obtained concomitantly to blood samples mentioned above in patients with clinical and laboratory signs of systemic inflammatory response syndrome (i.e. tachycardia, leucocytosis, tachypnea, and temperature >38°C or <36°C), those were all patients from groups 1b, 2b, 3b and 4. Candida cultures for calculation of the Candida colonization index (CI) were performed in patient groups 2 and 3b. Swabs from wounds, catheter insertion sites, perineal region, oropharynx, as well as BAL and urine were cultured for calculation of CI as described previously [10].

2.5. Comparison of sampling techniques (EBS vs. BAL).

Sampling techniques (collection of endobronchial secretion, EBS, versus BAL) and sterile saline used for BAL were compared in additional five intubated and mechanically ventilated ICU patients. Prior to routinely scheduled bronchoscopy (and BAL) endobronchial secretion was sampled and collected in a sterile cup. The bronchoscope was then flushed with 10ml of a given bottle of sterile saline used for subsequent BAL and the fluid was collected in a sterile cup. Then BAL was obtained by bronchoscopy and the sample collected in a sterile cup. Consequently, we obtained three sterile cups with fluid samples for bacterial and fungal microbiota comparison of EBS, BAL and sterile saline used for BAL corresponding to one patient. BAL was additionally processed and cultured in the routine laboratory if necessary for routine purpose.

3.1. Quantitative Candida cultures and bacterial cultures.

Quantitative Candida cultures and bacterial cultures from respiratory samples were performed by routine microbiological procedures. Briefly, samples from LRT were vortexed and 100 μL transferred onto a Candida CHROMagar (Becton Dickinson, Heidelberg, Germany), chocolate agar, blood agar, McConkey agar and was plated evenly with a sterile swab. After incubation at 37°C for 48 h in ambient air, the colonies were counted and classified as C. albicans, C. glabrata, C. krusei, C. tropicalis, or other Candida spp., according to the color of the colonies and were further identified by MALDI-TOF-mass-spectrometry. Quantitative bacterial cultures were performed by routine microbiological procedures.

3.2. Blood cultures.

Three pairs of blood culture bottles were drawn as recommended by local blood culture instructions and incubated at 37°C in an automatic blood culture detection system (BACTEC^® 9240 or BACTEC FX^®, Becton Dickinson, Heidelberg, Germany) for 5 days that allows continuous monitoring of blood cultures for microbial growth. Broth from positive blood culture bottles was Gram-stained and subcultured on specific agars depending on the Gram stain results. Fungi and/or bacteria were identified by routine microbiological procedures including API (bioMérieux, Marcy-l’Etoile, France), and peptid nucleic acid (PNA) FISH tests allowing quick identification of certain microorganisms within 20–90 minutes. Candida isolates were also identified by MALDI-TOF-mass-spectrometry. Isolated strains were stored at a temperature of -70°C until further analysis (e.g. susceptibility testing, multilocus sequence typing).

3.3. Urine and swab cultures.

Cultures from urine were performed by routine microbiological procedures. Candida isolates were counted and identified by MALDI-TOF-mass-spectrometry. Cultures from swabs were performed on Candida CHROMagar (Becton Dickinson, Heidelberg, Germany) at 37°C for 48 h in ambient air and Candida isolates were counted and identified as described above.

3.4. Candida colonization index.

In patient groups 2 and 3b swabs from wounds, catheter insertion sites, perineal region, oropharynx, as well as BAL and urine were cultured for calculation of Candida colonization index as described previously [10].

5.1. DNA extraction for MLST.

DNA extraction was carried out by using the PrepMan^® Ultra Sample Preparation Reagent (Applied Biosystems, Foster City, CA). 200 μL of PrepMan Buffer and 1 loop from a pure culture of each sample were mixed thoroughly in 1.5-mL tubes and heated in boiling water for 12 min. After a 3 min centrifugation step at 16,100g, the DNA-containing supernatant was decanted and transferred into a new tube.

5.2. Multilocus sequence typing (MLST).

5.2.1. Primer design and PCR: Seven primer pairs specific for DNA sites encoding housekeeping genes were chosen as recommended previously [19–21]: AAT1a, (Aspartateaminotrasferase), fragment size 373, chromosome 2, Fwd 5’-ACTCAAGCTAGATTTTTGGC- 3’, Rev 5’-CAGCAACATGATTAGCCC- 3’; AAC1 (Acetyl-coenzyme A carboxylase), fragment size 407, chromose R, Fwd 5’-GCAAGAGAAATTTTAATTCAATG- 3’, Rev 5’-TTCATCAACATCATCCAAGTG- 3’; ADP1 (ATP-dependent permease), fragment size 443, chromose 1, Fwd 5’-GAGCCAAGTATGAATGATTTG- 3’, Rev 5’-TTGATCAACAAACCCGATAAT- 3’, PMI1b^a, (formerly MPI1b) (Mannose phosphate isomerase), fragment size 375, chromosome 2, Fwd 5’-ACCAGAAATGGCCATTGC- 3’, Rev 5’-GCAGCCATGCATTCAATTAT- 3’; SYA1 (Alanyl-RNA synthetase), fragment size 391, chromosome 6, Fwd 5’-AGAAGAATTGTTGCTGTTACTG- 3’, Rev 5’-GTTACCTTTACCACCAGCTTT- 3’, VPS13 (Vacuolar protein sorting protein), fragment size 403, chromosome 4, Fwd 5’-TCGTTGAGAGATATTCGACTT- 3’, Rev 5’-ACGGATGGATCTCCAGTCC- 3’; ZWF1b, (Glucose-6-phosphate dehydrogenase), fragment size 491, chromosome 1, Fwd 5’-GTTTCATTTGATCCTGAAGC- 3’, Rev 5’-GCCATTGATAAGTACCTGGAT- 3’. According to previous literature these genes enable high discriminatory power in C. albicans MLST investigations [20,21]. Primers were commercially produced by Invitrogen^™ Custom DNA Oligos (life technologies, Carlsbad, CA). The primer stock solution was diluted with PCR-grade water to obtain a concentration of 10 μM/L. PCR was carried out with total reaction volumes of 50 μL, containing 37 μL water, 5 μL 10x Taq Buffer, 1 μL dNTP Mix, 1 μL Taq DNA Polymerase (5prime, Hilden, Germany), 2 μL of the forward and reverse primer (10 μM) and 2 μL DNA template. PCRs were performed on an Applied Biotechnologies 2720 Thermal Cycler (life technologies, Carlsbad, CA). The duration and temperature of the cycles were chosen as previously described [22].

5.2.2. Purification: The PCR products were purified using the MSB^® Spin PCRapace purification system (STRATEC molecular, Berlin, Germany). Each PCR sample was mixed with 250 μL binding buffer and transferred onto the enclosed spin filter. After 3 min centrifugation at 13,400 g, the filter was removed and transferred into a new tube. 50 μL elution buffer were added prior to a further centrifugation step for 1 min at 9,200g. The filter was then discarded, the supernatant contained the purified PCR product.

5.2.3. Cycle sequencing: Cycle sequencing was arranged using the BigDye^® Terminator v3.1 Cycle Sequencing Kit (life technologies). The primer stock solution was diluted with PCR-grade water to obtain a concentration of 1.6 μMol/L. Some amplicons, most notably SYA1 products, had low chromatogram signals or a high noise signal. In this case, agarose gel electrophoresis was performed. Whether gel electrophoresis revealed narrow bands, the amount of template was raised. Cycle sequencing was carried out on an Applied Biosystems 2720 Thermal Cycler. An initial step of 94°C for 18 sec was followed by 25 cycles of 94°C for 15 sec, 50°C for 10 sec and 60°C for 2 min with a final extension step at 60°C for 4 min. Further purification steps followed the cycle sequencing. The cycle sequencing products were vortexed in a 1.5-mL tube with 25 μL 98% ethanol (v/v) and 2.5 μL 125 μM EDTA and then centrifuged for 20 minutes at a temperature of 4°C at 16,100 g. After discarding the supernatant, 200 μL of 70% ethanol (v/v) were added and another centrifugation step for 15 min at a temperature of 4°C at 16,100g was done. Thereafter the supernatant was removed and the pellet was dried in an incubator at 37°C for at least 1 h. Then the pellet was resuspended in 15 μL Applied Biosystems^® Hi-DiTM Formamide (life technologies) and pipetted onto a 96-well plate. Analysis was performed on the Applied Biosystems^® 3130 Genetic Analyzer (life technologies).

5.2.4. MLST Data analysis: Analysis of the raw sequencing data succeeded with Applied Biosystems DNA Sequencing Analysis Software (life technologies, Carsbad, CA). After analysis and post-processing, each forward and reverse sequence pair was assembled with the help of Lasergene^® SeqMan Pro software (DNAStar, Madison, WI). Any assembly was searched automatically and manually for single nucleotide polymorphisms (SNPs). The SNPs were badged using the IUPAC nucleotide code according to previous literature [23]. Each consensus sequence was compared with the database on the MLST website to obtain the allele number (ST—sequence type) and the MLST profile number (DST—diploid sequence type). New alleles and DSTs were sent to the curator of the MLST website to obtain the allelic and DST numbers. The 7 matching sequences of each isolate were concatenated and afterwards aligned by means of muscle (http://drive5.com/muscle/) according to previous literature [24,25]. Our strains were added to the C. albicans MLST database (calbicans.mlst.net) and they are available using the following headings: id = 2275–2315 or Sequencer = Buzina Walter. In order to assign the strains to their clades, an eBURST analysis was performed acording to previous literature [26].

5.2.5. Phylogenetic analysis of MLST data: Evolutionary analyses, using neighbor-joining and Unweighted Pair Group Method with Arithmetic mean (UPGMA method), were conducted in MEGA6 (www.megasoftware.net) [27].

6.1. DNA isolation and PCR amplification.

Fifty-four randomly selected LRT samples were submitted to NGS-based fungal and bacterial microbiota analyses comprising 8 specimens from group 1a, 7 from group 1b, 7 from 2a, 6 from 2b and 26 from group 3b. For 10 samples, the ITS PCR did not produce any amplicons and samples were therefore excluded from analyses for fungal microbiota (4 in group 1a, 2 in group 1b, 4 in group 3b). In one sample the 16S rRNA gene PCR did not produce any amplicons and this sample was therefore excluded from analysis of bacterial diversity. For total DNA isolation from 5 mL BAL/EBS samples, Dithiothreitol (DTT) at a final concentration of 0.15% (w/v) was added, samples were incubated until dissolved and centrifuged at 4,000g for 30 min. Cells pellets were homogenized in 250 μL PBS and mixed with 250 μL bacteria lysis buffer in MagnaLyser bead tubes by using the MagnaLyser Instrument (Roche Diagnostics, Mannheim, Germany) according to manufacturer’s instructions. DNA was extracted using the MagNA Pure LC DNA Isolation Kit III (Bacteria, Fungi) with 25 μL lysozyme (100 mg/mL) (Roche Diagnostics, Mannheim, Germany), 20 μL Proteinase K (20 mg/mL) and three consecutive freeze-thaw cycles in a MagNA Pure LC 2.0 Instrument according to manufacturer’s instructions. DNA was quantified by quantitative real time PCR as described previously [28]. 16S rRNA gene hypervariable regions V1-2 (F27-5’-AGAGTTTGATCCTGGCTCAG-3’, R357-5’CTGCTGCCTYCCGTA-3’) [29,30] were amplified from total DNA by PCR using FLX 454 one way read (Lib-L kit, Primer A, Primer B, Roche 454 Life Science, Branford, CT, USA) fusion primers bearing the titanium FLX adaptor (5’-CCATCTCATCCCTGCGTGTCTCCGAC-3’), a key sequence (5’-TCAG-3’), a 10-mer Multiplex Identifier (MID) and the target specific forward sequence (5’-AGAGTTTGATCCTGGCTCAG-3’). The reverse fusion primer sequence was 5’-CCTATCCCCTGTGTGCCTTGGCAGTC-TCAG-CTGCTGCCTYCCGTA-3’. For each sample a mix of 25 μL containing 1 x Fast Start High Fidelity Buffer (Roche Diagnostics, Mannheim, Germany), 2.5 U High Fidelity Enzyme (Roche Diagnostics, Mannheim, Germany), 200μM dNTPs (Roche Diagnostics, Mannheim, Germany), 0.4μM barcoded primers (Eurofins MWG, Ebersberg, Germany), PCR-grade water (Roche Diagnostics, Mannheim, Germany) and 6.5ng genomic DNA. PCR was performed in triplicates. Thermal cycling was of initial denaturation at 95°C for 3 min followed by 35 cycles of denaturation at 95°C for 45 sec, annealing at 56°C for 45 sec and extension at 72°C for 1 min with a final extension of 7 min at 72°C. Amplicons were purified according to standard procedures. Amplicon DNA quantification, pooling and sequencing was performed as recently described [31]. For internal transcribed spacer (ITS) amplification FLX Titanium Fusion primers with the target specific forward ITS1F 5’-CTTGGTCATTTAGAGGAAGTAA-3’ and reverse ITS2R 5’-GCTGCGTTCTTCATCGATGC-3’ sequences were used [32]. For ITS PCR the same PCR mix was run with 50ng total genomic DNA and the cycling conditions as previously described [32]. Subsequent procedure was performed according to the 16S rRNA gene protocol.

6.2. Phylogenetic analysis.

To improve overall analysis quality as well as to avoid OTU inflation due to sequencing errors data was de-noised using previously published methods [33]. Additionally, contaminant sequences (such as sequences originating from the human host, or from other non 16S or ITS sources) were detected and removed by the Decontaminator with 80% identity and 15% query coverage. Briefly, amplicons were aligned using BLAT [34] against the greengenes 16S version (09.05.2011) and the UNITE ITS version (15.10.2013) database [35]. respectively. BLAT hits were sorted descending according their percentage of identity and quality coverage. Amplicons which fall below the specified criteria or with no hit at all were identified as contaminating sequence and removed. Furthermore, all sequences shorter than 150 bps containing any ambiguous characters, or not matching to the forward primer (distance>2) or the multiplexing barcodes were discarded. Subsequently, chimeric sequences were identified for both datasets with UCHIME [36] and removed. The remaining sequences after the pre-processing process were assigned to their respective samples by using the 10 bp barcode preceding the primer. In order to perform sample- and group-wide comparisons, operational taxonomic units (OTUs) were generated with two distinct phylotyping pipelines of SnoWMAn (http://SnoWMAn.genome.tugraz.at) [37]. For the 16S dataset OTUs were generated with an extended Ribosomal Database Project (RDP)-Pyrosequencing approach [38], by using the integrated RDP pipeline of SnoWMAn. Briefly, all sequences were pooled and aligned with Infernal (V1.0) using a 16S rRNA secondary structure based model for accurate position alignment of sequences [39]. The aligned sequences were clustered by complete linkage to form OTUs with sequence distances ranging from 0% to 5%. For each OTU a representative sequence was extracted and a taxonomic classification was assigned to it using the RDP Bayesian classifier 2.5 [37]. Finally, the pooled sequences were again separated according to their sample affiliation. Taxonomic classification and biostatistical analyses reported in this paper were performed on the clustering results for 3% distance at a classification confidence threshold of 80%.

Phylogenetic classification of the ITS dataset was done using the advanced BLAT approach of SnoWMAn. Briefly, unique sequences were aligned with BLAT against the UNITE ITS reference database version (15.10.2013). Subsequently, the best BLAT hit was used to assign taxonomic classification to each amplicon. Finally, samples were de-uniqued again and samples which result in the same taxonomic classification were assigned to the same OTU.

6.3. Statistical analysis and vizualization.

The analyses were conducted using the statistical and visualization capabilities of SnoWMAn. Species richness [40] was estimated according to Chao1 [41]. Additionally, the abundance-based coverage estimator (ACE) [42], evenness [40], Shannon and Simpson diversity index were calculated. Furthermore, OTUs significantly changing between the different patient groups were assessed by the R Bioconductor package edgeR (version 3.2.4). Briefly, the feature matrices (i.e. OTU counts per sample) were imported into R. All counts were increased by 1 to prevent taking the log from 0 and stored within a DGEList object. To scale the raw library sizes, the calcNormFactors [43] function was applied using the relative log expression (RLE) method [44]. Subsequently, a model matrix according to the experimental design was created. The common dispersion of all biological coefficients of variation (BCV) averaged over all OTUs, as well as the OTU-specific dispersion of the dataset, were calculated (estimateGLMTagwiseDisp, estimateGLMTrendedDisp, respectively). Prior to the likelihood ratio test (glmLRT) [45,46] the read counts for each feature were fit to a negative binomial generalized log-linear model (glmFit). To account for multiple testing and control of the Type I error (FDR), p-values were adjusted using a Benjamini-Hochberg method [47]. Only features with an adjusted p-value less than 0.05 were considered as differentially abundant.

6.4. Association and dissociation of bacterial (16S rRNA gene) and fungal (ITS) communities.

Relationships between bacteria and fungi in the samples of the different patient groups were calculated as odds ratio as described previously [48]. Briefly, for all bacteria and fungi pairs a two by two contingency table was created, presenting their presence and absence in the samples. Based on this information the odds ratio for each pair was calculated using the odds ratio function of R (package vcd_1.3–1. R version 3.0.3). The odds ratios above 2 were considered a positive association as this means that the detection of a particular organism with the other organism is twice as likely as with its absence. In contrast, an odds ratio below 0.5 was interpreted as negative association.

6.5. Sequence data availability.

Sequence data generated for this work can be accessed via the European Bioinformatics Institute Sequence Read Archive (EBI SRA) under the accession number ERP008535.