Sample characteristics
Particulate materials were collected from the ISS HEPA filter element (ISS HEPA) and vacuum cleaner bag components (ISS Debris) used aboard the ISS as well as at two cleanrooms at JPL (one each originating from the SAF and building 103, Pasadena, CA), during this study. Table 1 lists sample characteristics, usage time of the material collection devices or system(s), model, make, and cleanroom conditions where the devices were used. The materials collected using the HEPA filter system represented the air and the vacuum cleaner bag represented the surface locations.
ISS HEPA
The environmental control system (ECS) aboard the ISS includes a distributed ventilation system that contains HEPA filter elements to remove suspended particulate matter from the cabin atmosphere and protect humidity control and air purification equipment from debris accumulation and biofouling [53]. Particle counts are not routinely measured aboard crewed spacecraft. A study conducted during space shuttle mission STS-32 (January 1990) found that the particle size distribution approximated a class 100,000 cleanroom for particles <2.5 μm, a class 400,000 cleanroom for particles 2.5 μm to <10 μm, and a class 3000 cleanroom for particles >10 μm [53, 54]. The particle size distributions are defined using cleanroom standard definitions (FED-STD-209 1992). The HEPA filters used aboard the ISS retain 99.97 % particles of 0.3 μm via a pleated borosilicate media. A 20-mesh (841-μm sieve openings) inlet screen located at the filter element’s face removes larger debris and fibers. The pleated, non-woven borosilicate HEPA filter media was installed by Flanders Filters, Inc. (Washington, NC) in the filter element housing. Flanders Filters, Inc.’s commercial “NaturalAire” cut-to-fit filtration media has the greatest similarity to the ISS HEPA filter media (http://flandersfilters.com/products/naturalaire/). Twenty-one filter elements are distributed throughout the ISS in several modules and astronauts replace them on a scheduled maintenance cycle ranging from 2.5 to 5 years, depending on the location. The part number of the HEPA filter system analyzed during this study was SV810010-1 and the serial number was 0049. The HEPA filter element analyzed during this study was manufactured in September 1998, installed in ISS on January 2008, and returned aboard space shuttle flight STS-134/ULF6 in late May 2011. This filter was installed in ISS Node 2 and was in service for 40 months, although the typical service time is only 30 months for this location. Typically, ground testing is performed to characterize a filter element’s pressure drop at various process air flow rates [53] after retrieval from the ISS, but in this case, the HEPA filter remained untouched in its shipment packaging from the time it was removed from the ISS until particulates were recovered at JPL for microbial characterization. The particulate materials collected from this sample were designated as ISS HEPA during this study.
ISS debris
The vacuum cleaner bag contained debris that had collected on the HEPA filter inlet screens. The vacuum cleaner bag components are representative of the collected debris (lint, food particles, skin/hair, and miscellaneous debris) that had been airborne in the cabin but had collected over time on the filter element inlet screen. The molecular microbial community analysis for the ISS vacuum bag debris (ISS Debris) was previously documented [14]. These results were utilized here for comparative purposes. The previous molecular characterizations of the ISS Debris sample were extended here with Illumina-based deep sequencing using archived DNA aliquots [14]. Sample characteristics of the ISS Debris are described in Table 1. Briefly, during the period of ISS Expeditions 30 and 31, reports by some crewmembers cited differences in the cabin environment compared to earlier experience as well as allergic responses to the cabin environment. One of the noted observations was a high level of visible dust in the ISS Node 3 cabin, to the extent it was sticking to the walls. Flight surgeons indicated that this had been reported not just in Node 3 but also throughout the US on-orbit segment and expressed a concern for crew health. Dust on the ISS is expected, with humans being major contributors (via skin shedding, eating, exercising, etc.). Other sources such as on-orbit maintenance activities can release dust from sources such as payloads and systems, clothing, and visiting vehicles. As a precautionary measure, in the middle of 2012, an investigation was launched to define and mitigate dust sources and to determine if exposure to dust might elicit an adverse effect on crew health. As a result of these crewmember reports, particulate and fiber debris samples were collected during ISS Expedition 31 using a handheld portable vacuum cleaner and returned to Earth aboard Soyuz flight 29S in early July 2012 for analyses. Details of the novel “Prime” vacuum cleaner developed for NASA habitats were reported elsewhere [55]. Portions of the samples were aseptically collected for microbiological testing at NASA’s Johnson Space Center (JSC), with the remaining bag and its contents repacked in a biological hood, sealed, and shipped to NASA’s Marshall Space Flight Center (MSFC) for particle size testing, as well as to JPL to determine their molecular microbiological composition. The vacuum cleaner bag was shipped via FedEx at room temperature with instructions that the samples not be irradiated in transport.
Earth analogs
For comparison, particles collected from two JPL cleanroom floors using a Nilfisk GM80CR vacuum cleaner (the disposable bag part number is 81620000; Morgantown, PA) were analyzed (Table 1). Cleanroom certification is based on the maximum number of particles greater than 0.5 μm per cubic foot of air. The air within Class 1 K cleanrooms is maintained at fewer than 1000 particulates per cubic foot, Class 10 K cleanrooms are allowed to harbor a density of 10,000 particles per cubic foot, and so on. The cleanroom samples were from (a) the JPL-SAF cleanroom floor where various Mars mission spacecraft were built (Class 10 K; JPL-SAF Debris) and (b) JPL building 103 where non-mission critical activities were conducted (Class 1 K; JPL-103 Debris). The vacuum cleaner bag debris from the JPL-SAF and JPL-103 buildings was transported to the microbiology laboratories for storage. Sample processing was carried out immediately after aseptic collection of the materials, usually within 7 days from the retrieval of the vacuum bags from the cleanrooms. Both Earth-analog samples were processed at the same time as the ISS Debris samples. However, the ISS HEPA sample was received ~3 months later and hence was analyzed separately.
Sample processing
All samples were subjected to a variety of microbiological and molecular techniques to elucidate composition of cultivable, viable, and total microorganisms. Independent of the samples taken to cultivate bacterial and fungal analyses, subsamples of the same samples were taken for DNA extraction. Weighing vacuum debris samples was possible, whereas the HEPA filter elements were divided into small pieces and particulates associated with the pieces were aseptically collected using sterile scalpels before being quantitatively measured. Approximately 1 g of each vacuum debris and HEPA filter associated particle was weighed and placed into a sterile tube containing 25 mL of sterile phosphate-buffered saline (PBS) and vortexed for 1 min. After vigorous mixing, large particles were allowed to settle, and aliquots of samples were carefully siphoned and allocated for culture-based (1 mL) and culture-independent analyses (15 mL).
Traditional culture-based microbial examination
For estimating bacterial populations, after suitable serial tenfold dilution in sterile PBS, 100 μL of the sample suspension was spread onto two plates of R2A media (BD Difco, Franklin Lakes, NJ) and incubated at 25 °C for 2–7 days. For the fungal population enumeration, 100 μL was spread onto two plates of potato dextrose agar (PDA, BD Difco, Franklin Lakes, NJ) and incubated at 25 °C for 2–7 days. Bacterial and fungal colony-forming units (CFUs) were counted and reported as CFU/g of material. Identifications and phylogenetic affiliations were carried out via sequencing for both bacteria and fungi by targeting the 16S rRNA gene [42] and the ITS region [56], respectively. When identifications were ambiguous, sequencing of an additional gene (gyrB) was performed to confirm the phylogenetic identity of the purified bacterial isolates [57, 58]. The nucleotide sequences of bacteria (KT763339–KT763368) and fungi (KT832780-KT832790) were deposited in GenBank.
Sample processing for molecular analysis
The biological materials associated with each sample (15 mL) were further concentrated using Amicon Ultra-50 Ultracel centrifugal filter tubes (Millipore, Billerica, MA). Each filter unit has a molecular mass cutoff of 50 kDa, which facilitates the concentration of microbial cells, spores, and exogenous nucleic acid fragments greater than 100 bp in a final volume of 2.5 mL. All filtered samples were then divided into three separate aliquots: the first aliquot (1000 μL) was subjected to PMA pretreatment (viability assessment), the second (1000 μL) was an untreated environmental sample (viable + nonviable; total DNA), and the third (500 μL) was used for adenosine triphosphate (ATP) analysis (see below).
For measuring viable microbial population, one aliquot of filter-concentrated sample suspension (1000 μL) was treated with 12.5 μL of PMA (2 mM; Biotium, Inc., Hayward, CA) to a final concentration of 25 μM [26, 32], followed by thorough mixing and incubation in the dark for 5 min at room temperature [15]. The sample was exposed to PhAST blue-Photo activation system for tubes (GenIUL, S.L., Terrassa, Spain) for 15 min (in parallel with the non-PMA-treated sample). This step enabled blocking DNA from dead cells [15]. The samples were then split in half and one half was subjected to bead beating with the Fastprep-24 bead-beating instrument (MP Biomedicals, Santa Ana, CA) with parameters set at 5 m/s for 60 s. The second half of the unprocessed sample was then combined with the mechanically disrupted counterpart before DNA was extracted via the Maxwell 16 automated system (Promega, Madison, WI), in accordance with manufacture instructions [59]. Resulting DNA suspensions (100 μL each) were stored at −20 °C.
Quantitation of total and viable microorganisms using molecular methods
ATP assay
A bioluminescence assay was performed to determine the total ATP and intracellular ATP from all samples using the CheckLite HS kit (Kikkoman Corp., Noda, Japan), as described previously [60]. Briefly, to determine the total ATP (dead and viable microbes), 100-μL sample aliquots (four replicates) were each combined with 100 μL of a cell lysing detergent (benzalkonium chloride) then incubated at room temperature for 1 min prior to the addition of 100 μL of a luciferin-luciferase reagent. The sample was mixed, and the resulting bioluminescence was measured with a luminometer, the Lumitester K-210 (Kikkoman Corp.). To determine intracellular ATP (viable microbes), 50 μL of an ATP-eliminating reagent (apyrase, adenosine deaminase) was added to a 500-μL portion of the sample and allowed to incubate for 30 min to remove any extracellular ATP, after which the assay for ATP was carried out, as described above, in four replicates, including sterile PBS as negative controls. As previously established, one RLU, the unit of measurement of ATP, was assumed to be approximately equal to one CFU [42].
qPCR assay
Real-time quantitative polymerase chain reaction (qPCR) assay, targeting the 16S rRNA gene, was performed in triplicate with a CFX-96 thermal cycling Instrument (Bio-Rad, Hercules, CA) to quantify the bacterial burden. Bacteria-directed primers targeting the 16S rRNA gene 1369 F (5′-CGG TGA ATACGT TCY CGG-3′) and modified 1492R (5′-GGW TAC CTTGTT ACG ACT T-3′) were used for this analysis [61]. Each 25-μL reaction consisted of 12.5 μL of 2X iQ SYBR Green Supermix (BioRad, Hercules, CA), 1 μL each of forward and reverse oligonucleotide primers (10 μM each), and 1 μL of template DNA. Purified DNA from B. pumilus cells served as the positive control and DNase/RNase-free molecular-grade distilled water (UltraPure, Gibco) was used as the negative control. These controls were included in all qPCR runs. Reaction conditions were as follows: a 3-min denaturation at 95 °C, followed by 35 cycles of denaturation at 95 °C for 15 s, and a combined annealing and extension at 55 °C for 35 s.
Molecular microbial diversity analysis using next-generation sequencing methods
Pyrosequencing conditions
Bacterial primers 28 F (5′-GAG TTT GAT CNT GGC TCA G-3′) and 519R (5′-GTN TTA CNG CGG CKG CTG-3′) were used to amplify ~500-bp fragments spanning the V1–V3 hypervariable regions of the bacterial 16S rRNA gene. Archaeal primers 341 F (5′-GYG CAS CAG KCG MGA AW-3′) and 958R (5′-GGA CTA CVS GGG TAT CTA AT-3′) were used to amplify ~600-bp fragments spanning the V3–V5 hypervariable regions of the archaeal 16S rRNA gene. A fungal primer set ITS1F (5′-CTT GGT CAT TTA GAG GAA GTA A-3′) and ITS4R (5′-TCC TCC GCT TAT TGA TAT GC-3′) was employed to amplify ~600-bp fragments of the fungal ITS region. These primer pairs were tailored for pyrosequencing by adding a fusion linker and a proprietary 8-nt barcode sequence at the 5′ end of the forward primer and a biotin and fusion linker sequence at the 5′ end of the reverse primer [62]. A HotStarTaq Plus master mix kit (Qiagen, Valencia, CA) was used to catalyze the PCR under the following thermal cycling conditions: initial denaturing at 95 °C for 5 min, followed by 35 cycles of denaturing at 95 °C for 30 s, annealing at 54 °C for 40 s, and extension at 72 °C for 1 min, finalized by a 10-min elongation at 72 °C. Resulting PCR products were purified via Diffinity Rapid Tip (Diffinity Genomics, Inc., West Henrietta, NY) chemistry and were then pooled accordingly. Small fragments were removed with Agencourt Ampure Beads (Beckman Coulter, Brea, CA).
In preparation for FLX-Titanium sequencing (Roche, Nutley, NJ), resulting PCR amplicon fragment size and concentration were accurately measured with DNA 1000 chips using a Bioanalyzer 2100 automated electrophoresis station (Agilent, Santa Clara, CA) and a TBS-380 Fluorometer (Turner Biosystems, Sunnyvale, CA). The total volume of initial PCR product used for subsequent emulsion PCR was 2 μL for strong positives (>10 ng/μL), 5 μL for weak positives (5 to 10 ng/μL), and 20 μL for samples that failed to yield PCR products (<5 ng/μL). This normalization step helped to ensure minimal bias favoring downstream amplification from initially strong PCR products. Approximately 9.6 × 106 molecules of ~600-bp double-stranded DNA were combined with 9.6 × 106 DNA capture beads and then subjected to emulsion PCR conditions. Following recovery and enrichment, bead-attached DNA molecules were denatured with NaOH and sequencing primers were annealed. A four-region 454 pyrosequencing run was performed on a GS PicoTiterPlate (PTP) using the Genome Sequencer FLX System, in accordance with manufacturer instructions (Roche, Nutley, NJ). Twenty-four to 30 tagged samples were applied to each quarter region of the PTP. All pyrosequencing procedures were performed at the Research and Testing Laboratory (Lubbock, TX) in accordance with well-established protocols [62].
Illumina sequencing conditions
The library preparations for next-generation sequencing and Illumina MiSeq sequencing were conducted by GENEWIZ, Inc. (South Plainfield, NJ). The DNA samples were quantified using a Qubit 2.0 fluorometer (Invitrogen, Carlsbad, CA) and DNA quality was confirmed by electrophoresis (0.8 % agarose gel). The sequencing library was constructed using a MetaVx™ Library Preparation kit (GENEWIZ, Inc., South Plainfield, NJ). In brief, 5–50 ng of DNA was used for amplicon generation to cover the V3 and V4 hypervariable regions of 16S rDNA. Indexed and universal adapters were added to the ends of the 16S rDNA amplicons by limited-cycle PCR. The DNA libraries were validated with an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA) and quantified using Qubit and real-time PCR (Applied Biosystems, Carlsbad, CA). The DNA libraries were multiplexed and loaded on an Illumina MiSeq following the instructions from the manufacturer (Illumina, San Diego, CA). A 2 × 150 paired-end (PE) configuration was used for sequencing. The image analysis and base calling were processed using MiSeq Control Software. The initial taxonomy was performed on Illumina’s BaseSpace cloud computing platform.
Bioinformatic analyses of bacterial pyrosequences
High-throughput 16S rRNA sequencing data were processed. Bacterial and archaeal sequences were processed and analyzed using the mothur software [16, 63]. Sequences were quality filtered by removing sequences that (a) did not contain the primer sequence, (b) contained an uncorrectable barcode, (c) were <250 nt in length, (d) had homopolymers longer than 8 nt, or (e) had a quality score of <25; and then demultiplexed using the respective sample nucleotide barcodes. These sequences were searched against the Greengenes reference database [64, 65] and clustered into OTUs based on their sequence similarity (97 %) with UCLUST [66]. A representative sequence was picked from each OTU and taxonomic classification was assigned using mothur’s Bayesian classifier [67] and Greengenes training sequences and taxonomy [64, 65].
Bioinformatic analyses of fungal pyrosequences
The sequences were run through ITSx 1.0.9 [68] to remove non-ITS sequences, assembly chimeras, and sequences for which none of the 3′ end of SSU rDNA or the 5′ end of 5.8S rDNA was detected. The ITS1 sub-region was extracted from the remaining sequences for further analyses. Chimera control was exercised through UCHIME 7 [69] using the UNITE chimera reference dataset [70] as reference corpus. The sequences were subjected to clustering and taxonomic assignment in the Sequence Clustering and Analysis of Tagged Amplicons (SCATA) next-generation sequencing analysis pipeline (http://scata.mykopat.slu.se) using the February 2014 release of UNITE as taxonomic reference corpus. The default sequence quality control settings of SCATA were used; however, the clustering threshold was set to 98.5 % [71]. All taxonomic affiliations proposed by SCATA were examined manually using Basic Local Alignment Search Tool (BLAST) 2.2.30 [72] in GenBank [73] and UNITE [74] and occasionally refined using the principles outlined in Koljalg et al. [75] and Nilsson et al. [76].
Preprocessing of bacterial Illumina sequences
The 12 sets of 16S rDNA V3 amplicon data were sequenced at GENEWIZ, Inc. (South Plainfield, NJ,), where the obtained reads were trimmed to remove primer sequences. The raw sequence dataset was composed of 10,241,173 paired-end reads, 150 bp in length, with exceptionally high quality (<0.1 % error rate) (Resphera DiscoveryTM protocol, Baltimore, MD). After trimming noisy reads and removal of low-quality and chimeric sequences, we identified a moderate level of expected contaminants, including chloroplast and mammalian mitochondrial sequence (range 0.0001–9.3 %), as well as a low level of unknown contaminants (range 0.01–1.8 %). Exploratory characterization of unknown contaminant representatives indicated nonspecific amplification mitochondrial sequences from various eukaryotic organisms. After completion of preprocessing, the resulting high-quality R1 and R2 read sets contained 9.1 M and 9.3 M sequences, respectively, with an average length of 128 bp. Due to the primer set and sequencing technology selected by the vendor, we were unable to merge paired-end sequences into longer consensus fragments as they did not overlap. Therefore, in this analysis, we compared our findings between the R1 (forward)- and R2 (reverse)-associated datasets to evaluate consistency. Across R1 and R2 datasets, each sample had on average 917,060 clean sequences. After clustering sequences into OTUs, OTU tables were rarefied to an even coverage of 525,000 sequences per sample. On average, Good’s coverage statistic for rarefied samples was 99.78 %, indicating we have observed the vast majority of OTU diversity in each community.
Bioinformatic analysis of bacterial Illumina sequences
Sequences were de-multiplexed using 5′ barcode identifiers and analyzed using the Resphera Discovery™ protocol (Baltimore, MD). Briefly, 16S rRNA sequence fragments were first screened in Quantitative Insights into Microbial Ecology (QIIME) [77] for quality and length requiring: (a) trimming at the first 5-bp run of Phred quality scores below 20, (b) one ambiguous base call or less, and (c) a minimum final length of 125 bp after trimming of forward and reverse primer sequences. Passing sequences were screened for PhiX-174 spiked fragments and PCR-associated chimeras by UCHIME [69] (de novo mode). Non-chimeric reads were then filtered for contaminant chloroplast and mitochondrial sequences using the RDP classifier [17], followed by a broad nucleotide BLAST (BLASTN) [78] search of the GreenGenes 16S rRNA reference database [65] (v1.1) to identify potential unknown contaminants. The resulting high-quality dataset was clustered into de novo OTUs using UCLUST [66] with a 95 % identity threshold. OTU representatives were assigned to a taxonomic lineage using the RDP classifier trained on the Resphera 16S rRNA database (v1.1) requiring a minimum confidence score of 0.50.
Statistical analyses
To perform pairwise statistical comparisons of sample groups of interest, a negative binomial test implemented in DESeq [79] or Fisher’s exact test was utilized and controlled for false positives using the false discovery rate (FDR) [80]. For the comparison of bacterial pyrosequence and Illumina data and fungal pyrosequence analysis, multivariate statistical analyses of community profiles were performed using an “in-house R-script” employing the libraries vegan, ape, gplots, mgcv, and GUniFrac [81]. First, each dataset composed of OTU abundances per sample was rarefied to the lowest number of reads and a Bray-Curtis index was calculated. This procedure was repeated 10,000 times and the average Bray-Curtis distance was calculated for each dataset in order to avoid biases arising from rarefication. The Bray-Curtis distance was then utilized to calculate principal coordinate (PCoA) analyses, PERMANOVA (1000 permutations), and MRPP (1000 permutations).