AN EXPERIMENTAL STUDY ON BACTERIOLOGICAL ANALYSIS OF WATER TANKS IN HALLS OF RESIDENCE

ABSTRACT

The availability of bioresources is a precondition for life science research, medical applications, and diagnostics, but requires a dedicated quality management to guarantee reliable and safe storage. Anecdotal reports of bacterial isolates and sample contamination indicate that organisms may persist in liquid nitrogen (LN) storage tanks. To evaluate the safety status of cryocollections, we systematically screened organisms in the LN phase and in ice layers covering inner surfaces of storage tanks maintained in different biobanking facilities. We applied a culture-independent approach combining cell detection by epifluorescence microscopy with the amplification of group-specific marker genes and high-throughput sequencing of bacterial ribosomal genes. In the LN phase, neither cells nor bacterial 16S rRNA gene copy numbers were detectable (detection limit, 10² cells per ml, 10³ gene copies per ml). In several cases, small numbers of bacteria of up to 10⁴ cells per ml and up to 10⁶ gene copies per ml, as well as Mycoplasma, or fungi were detected in the ice phase formed underneath the lids or accumulated at the bottom. The bacteria most likely originated from the stored materials themselves (Elizabethingia, Janthibacterium), the technical environment (Pseudomonas, Acinetobacter, Methylobacterium), or the human microbiome (Bacteroides, Streptococcus, Staphylococcus). In single cases, bacteria, Mycoplasma, fungi, and human cells were detected in the debris at the bottom of the storage tanks. In conclusion, the limited microbial load of the ice phase and in the debris of storage tanks can be effectively avoided by minimizing ice formation and by employing hermetically sealed sample containers.

CHAPTER ONE

Introduction

The long-term storage of biomaterials (biobanking) is a precondition for modern life sciences, enabling follow-up scientific investigations, medical diagnostics, biotechnological applications, and the conservation of genetic resources and diversity (Overmann 2015; Overmann and Smith 2017; Schüngel et al. 2014; Stock et al. 2018). To guarantee the safe storage of biological material, dedicated quality management procedures and controls need to be improved continuously (Chatterjee et al. 2017; Lauterboeck et al. 2016; Rittinghaus and Glasmacher 2018).

Cryopreservation constitutes a key component of contemporary biobanking. Specific cryopreservation protocols have been established for different organisms and cell types. Living biological material may be prepared for cryopreservation under both, sterile or unsterile conditions. As a result, the biological materials themselves as well as the storage facilities may contain additional, accompanying organisms. For instance, plant material and human or animal cell material may be colonized by viral or bacterial pathogens (Bielanski et al. 2003; Knierim et al. 2017; Uphoff et al. 2015). Some cryopreservation techniques also require the direct contact of biomaterials with the liquid nitrogen (Rall and Fahy 1985) and are therefore particularly prone to contamination (Bielanski and Vajta 2009). However, LN and other liquefied gases are commonly manufactured in so-called air separation units which separate the atmospheric gases at very low temperatures. During this process, the air is filtered and dried. Tests conducted by one manufacturer of liquefied gases using validated methods could not detect any pathogens (personal communication Dr. Carsten Pilger, AIR LIQUIDE Medical GmbH).

So far, only anecdotal reports exist on the types of organisms occurring in LN storage tanks outside of the stored sample material. Some bacteria and fungi were determined in the debris at the bottom of LN storage tanks (Bielanski et al. 2000), but only Stenotrophomonas maltophilia was found also as contaminant in the cryopreserved material (Bielanski et al. 2003). An exchange of biological materials between individual samples may occur if stored in non-hermetically sealed containers in the same LN storage tank as indicated by reports of the transmission of human hepatitis B virus during cryopreservation of bone marrow transplants (Tedder et al. 1995), and by the infection of bovine embryos with bovine viral diarrhea virus and bovine herpes virus-1 after contact with contaminated LN (Bielanski et al. 2000). In a few studies, a few single microbial species were isolated directly from the LN storage tanks using culture-dependent approaches (Fountain et al. 1997; Ramin et al. 2014). However, these culture-based methods provide only very limited insights into the presence of microorganisms in complex samples since the majority of microorganisms still escapes cultivation (Overmann 2013; Overmann et al. 2017).

In the present study, we assessed the occurrence of microorganisms in LN storage tanks by state-of-the-art microscopic and culture-independent molecular approaches. In order to elucidate the types of organisms occurring in LN storage tanks, to infer possible routes of entry, and to deduce suitable strategies for quality management, we systematically screened bacteria, fungi, plant, and human cells in different phases of LN storage tanks maintained in ten different biobank facilities.

Material and methods

Biobanking facilities and sampling methodology

A total number of 121 samples were obtained across ten different biobank facilities in 2015 (Table S1). The LN storage tanks were located in buildings with or without air conditioning for supply and exhaust; five institutes (A, D, F, G, I) used a filtered air supply. Individual LN storage tanks varied with respect to manufacturer and type. The longest time of continuous usage without intermittent cleaning of LN storage tanks amounted to 30 years; the shortest usage interval was less than one year. Most of the tanks had not been cleaned on a regular basis in order to avoid potential damage of the stored biological materials during the transfer to another LN storage tank. Furthermore, most of the tanks had been opened regularly at least twice a week. The biological samples stored were of human (blood, stem cells), animal (rodents, fish, mussel, dove, monkey, pig, cat), or plant origin, or were microorganisms (bacteria, fungi, archaea, bacteriophages). Biomaterials were stored in cryotubes, cryobags, or straws and either in the gaseous or the LN phase of the LN storage tanks or in both (Table S1).

Wherever accessible, the LN phase, ice layers underneath LN storage tank lids, and debris accumulated at the bottom of LN storage tanks were sampled (Figs. 1a, b). For each LN sample, 15 individual subsamples, each amounting to 50 ml LN, were collected in Falcon tubes (Fig. 1a). The LN subsamples were incubated until all LN had evaporated. Ice samples were scraped off the inner rim or from the bottom face of the lid into a Falcon tube (Fig. 1b). Each ice sample amounted to 10–100 ml of thawed ice depending on accessibility. LN and ice samples were collected in three consecutive months (Table S1). All samples were stored frozen and shipped on (dry) ice to the Leibniz Institute DSMZ for subsequent analyses.

Liquid nitrogen (LN) storage tanks and sampling procedure of LN and ice. (a) Sampling of the LN phase using a reaction tube and grip tongs. (b) Sampling of the ice phase formed underneath the lids (and rim)

For further processing, of each LN sample, residuals from eleven pooled 50 ml-subsamples (total 550 ml) were used for DNA extraction. A total volume of 10 ml of 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES; Serva, Heidelberg, Germany) buffer (25 mM, pH 7.3) was added and the tubes were incubated at room temperature on a horizontal shaker to resuspend the residuals for 15 min. Samples were filtered through a 0.1 μm pore-size polycarbonate filter and stored at −20 °C for DNA extraction. For microscopic counting, the residuals from four pooled 50 ml (total 200 ml) LN subsamples were resuspended in 5 ml HEPES by shaking for 15 min and then fixed with glutaraldehyde (final concentration, 2% w/v; Serva, Heidelberg, Germany). The ice samples were thawed and HEPES buffer was added up to 10 ml in samples containing less than 10 ml thawed ice. An aliquot of 1.8 ml of each sample was fixed with glutaraldehyde for microscopic counting and the remaining suspension filtered through a 0.1 μm pore-size polycarbonate filter.

Three types of negative controls were included. Firstly, empty Falcon tubes provided along with the samples by each participant served as negative controls for the contamination of laboratory equipment, they were filled with 25 ml HEPES (Negative Control = NC eq). Secondly, reference samples processed at DSMZ consisting of 550 ml of the LN were filled into a sterilized Dewar, 0.1 μm Isopore™ polycarbonate filters (Merck Millipore Ltd., Tullagreen, Carrigtwohill, Irland) added, and the LN evaporated (NC ref). Thirdly, two of the 0.1 μm polycarbonate filter were treated by filtering 25 ml HEPES (NC HEPES). All controls were processed in parallel and in the same way as the samples.

Microscopy

For fluorescence microscopy, fixed cells were collected onto polycarbonate filters (25 mm diameter; 0.1 μm pore size), the filters were stained with 50 μl SYBR Green I (LifeTechnologies, 1:10000 in DMSO; Sigma-Aldrich, Darmstadt, Germany) and mounted in a drop of immersion oil on a glass slide. The samples were analyzed using a Zeiss (Oberkochen, Germany) Axio Imager.M2 microscope at excitation/emission wavelengths of 470/525 nm, and the Axio vision software Rel. 4.8.2. Twenty microscopic fields were counted in triplicate for each sample. Conspicuous structures were analyzed further for the presence of chlorophyll a autofluorescence as indicator of the presence of algae or plant cells using a Nikon (Düsseldorf, Germany) Ti microscope at an excitation wavelength of 425 nm and an emission wavelength of 607 nm and Nikon software NIS-Elements AR 4.13.01.

DNA extraction and PCR

DNA was extracted from the filters using the DNA Micro Kit (Qiagen, Hilden Germany) according to the protocol of the manufacturer. Filters were cut into strips and incubated with lysozyme (final concentration, 20 mg per ml; Serva, Heidelberg, Germany) at 37 °C on a shaker (800 rpm) for 1 h. In the second lysis step, 20 μl proteinase K (final concentration 50 μg per μl; Applichem, Darmstadt Germany) was added and the samples were incubated at 56 °C over night. In the final step of the protocol, DNA was eluted in 20 μl PCR-clean water (Promega, Mannheim, Germany).

Bacterial 16S rRNA genes, eukaryotic (human) transposable elements Line1, and fungal ITS region were PCR amplified using the respective primer sets 8F-1492R, Line1 and ITS1F-ITS4 at a final concentration of 0.2 pmol per μl (Table S2). The PCR was performed in an Applied Biosystems cycler (Foster City, USA) using Thermo Scientific DreamTaq Green (0.02 U per μl; Waltham, USA) and buffer (Table S2). Bacterial 16S rRNA gene copy numbers (V3 region; specific primers at 0.2 pmol per μl final concentration, Table S2) were determined in a quantitative real-time PCR using LightCycler® (Roche, Basel, Schweiz) 480 and SYBR Green I. Mycoplasma was detected using a previously established PCR-based detection method (Uphoff and Drexler 2002). This endpoint PCR was performed in an Applied Biosystems cycler using Invitrogen Platinum Taq (0.02 U per μl; Carlsbad, USA) and buffer (Table S2).

Library preparation and sequencing

The V3-region of the bacterial 16S rRNA gene was sequenced using amplicons generated with specific primers 341F wobble and 515R (0.2 pmol per μl each), Qiagen Phusion polymerase (0.04 U per μl; Hilden Germany) and GC-buffer with the addition of dNTPs (0.2 mM), BSA (0.8 mg per ml), MgCl₂ (0.5 mM), DMSO (3.0%), and PCR-clean water. Between 1 and 20 ng, DNA template was used. The PCR product (60 μl) was cleaned up using DNA Clean & Concentrator™-5 (ZymoResearch, Irvine, USA) eluting the product in 30 μl water. After adding 0.1X TE (1 mM Tris-HCl, pH 8.0, 0.1 mM EDTA) to a final volume of 50 μl, the amplicon was processed using the NEBNext® Ultra™ II DNA Library Prep Kit for Illumina® (New England Biolabs, Frankfurt a. Main, Germany) according to the protocol of the manufacturer. Amplicons were prepared for adapter ligation using the NEBNext End Prep enzyme mix, and the 25-fold diluted adapter was ligated in a subsequent step. Adapter-ligated fragments were cleaned up without size selection using Agencourt AMPure XP Beads (Beckman Coulter GmbH, Krefeld, Germany). Then, the adapter-ligated DNA was enriched by 13 PCR cycles of using NEBNext® Multiplex Oligos for Illumina® (Index Primers Set 1, New England Biolabs, Frankfurt a. Main, Germany). The size distribution of the purified PCR product (AMPure XP Beads) was checked on an Agilent Bioanalyzer (high sensitivity chip; Santa Clara, USA). Adapter dimers of the combined library pool (~ 10 ng PCR product per sample) were removed by gel purification (MetaPhor® agarose; Lonza, Basel, Switzerland) using the NucleoSpin® Gel and PCR Clean-up Kit (Macherey-Nagel, Merck, Darmstadt, Germany) and amplification products were sequenced on a HiSeq 2500 Ultra-High-Throughput Sequencing System (Illumina, San Diego, CA, USA) as described recently (Gossner et al. 2016).

Raw sequence reads were organized based on unique barcodes and denoised into amplicon sequence variants (“sequence variants” in the following) using plugins implemented in Quantitative Insights into Microbial Ecology (Qiime2, ver. 2017.12.0; Caporaso et al. 2010; team 2016-2018 https://qiime2.org/) creating a Feature Table. Default settings were used unless otherwise noted. The forward and reverse reads were joined, chimera-filtered and clustered (vsearch, Rognes et al. 2016), quality filtered (Bokulich et al. 2012) and trimmed to a length of 150 bp (minimum size = 2, minimum reads = 5; deblur, Amir et al. 2017). A phylogenetic tree was constructed with FastTree (Price et al. 2010) after performing multiple sequence alignment using MAFFT (Katoh and Standley 2013) and Mask (Bailey and Gribskov 1998). Samples were rarefied to 99.0% sequence coverage (Chao and Jost 2012). Taxonomy was determined using a pre-trained Naive Bayes classifier based on the SILVA Database (v.128, Quast et al. 2013; Yilmaz et al. 2013) with the Qiime 2 plugin feature-classifier (https://github.com/qiime2/q2-feature-classifier). The reads were then compared against SILVA 132 SSURef Nr99 with an initial identity cutoff of 97% with vsearch 2.7 (-strand both, Rognes et al. 2016). Each read was then taxonomically assigned to the hit with the best bit score. When multiple best hits were present, the first one listed was chosen. The origin of the sequence variants was analyzed using the microbial isolation sources search implemented in bacterial metadatabase BacDive (Reimer et al. 2018). All Illumina datasets were submitted to the SRA database under accession number PRJNA558333.

Statistics

Statistical tests were performed using R (version 3.3.4, R-Core-Team 2017). Two-sample t-test and variance F-test were calculated for the gene copy numbers and relative abundances of single taxa comparing reference and single LN storage tank samples. The variance between different groups was determined by one-way-ANOVA with multiple comparisons of means using Tukey Contrasts (package multcomp, Herberich et al. 2010) shown as compact letter display (cld, Piepho 2004). Correlations for the association between paired samples were tested (R, corr.test) using two-sided Spearman’s rank correlation rho. A multivariate analysis of variance of the distance matrices was performed with permutation tests (n = 999) using the adonis2 function of the vegan package in R (McArdle and Anderson 2001). Different linear models (generalized with mixed effects, Kuznetsova et al. 2015) were applied to evaluate the effect of predictor variables (institute, storage phase, surrounding condition, stored material, storage device, number of openings and usage time) on a response variable (gene copies or cell numbers). The Akaike information criterion (AIC) and residual plots (DHARMa, Hartig 2017) were taken into account. All response variables were log-transformed.

Employing the phyloseq package (McMurdie and Holmes 2013) of R (version 3.3.4, R-Core-Team 2017), a principal coordinates analysis (PCoA) based on weighted UniFrac distances was calculated on species level for sequence variants defined at 3% sequence dissimilarity. Student’s t-Test was performed to compare the weighted UniFrac distances between samples above (ice, debris) and below the threshold of the negative controls (NC, LN). For further analysis, the values determined for negative controls were taken as a threshold. Therefore, all samples containing cell counts of < 10² cells per ml were excluded from the analysis. Alpha-diversity (Chao1, Shannon diversity index; for institute comparison all samples and NCs were included), Constrained Analysis of Principal Coordinates (CAP, based on weighted UniFrac distances), and relative abundances of the bacterial communities were calculated on genus level removing sequence variants with abundances of less than 5 reads per sample phyloseq package (McMurdie and Holmes 2013) of R (version 3.3.4, R-Core-Team 2017). The parameter cells per ml, surrounding condition (building), storage phase, time of usage, frequency of openings, storage device, and stored material were used as constraining variables.

Results

Microbial cell counts and PCR detection

Bacterial cell counts in both, negative controls as well as in the LN samples were low and at ≤ 10² cells per ml LN (Figure 2a, Table S3). Correspondingly, all LN samples and negative controls showed 16S rRNA gene copy numbers of ≤ 10³ per ml LN (Table S3, Fig. 3). Variance analysis of gene copy numbers between the different sample types showed that samples taken directly from the LN phase could not be distinguished from the negative controls (p = 0.158, ANOVA, Table S4), indicating that the microbial entry in the LN samples is below the detection limit determined by the negative controls and reference samples. The negative control (NC_B) with the highest cell number determined the threshold for the detection limit (Fig. 3). Therefore, a threshold of 277 cells per ml LN was applied to choose samples to be included in all subsequent analyses. All samples which had cell numbers below the threshold were only considered for selected comparative analyses, in particular to access the potential origin of the cells.

References

1. Amir A, McDonald D, Navas-Molina JA, Kopylova E, Morton JT, Zech Xu Z, Kightley EP, Thompson LR, Hyde ER, Gonzalez A, Knight R (2017) Deblur rapidly resolves single-nucleotide community sequence patterns. MSYSTEMS 2(2). https://doi.org/10.1128/mSystems.00191-16
2. Bach EM, Williams RJ, Hargreaves SK, Yang F, Hofmockel KS (2018) Greatest soil microbial diversity found in micro-habitats. Soil Biol Biochem 118:217–226. https://doi.org/10.1016/j.soilbio.2017.12.018

CAS Article Google Scholar 

• Bailey TL, Gribskov M (1998) Combining evidence using p-values: application to sequence homology searches. Bioinformatics(Oxford, England) 14(1):48–54

CAS Article Google Scholar 

• Bajerski F, Stock J, Hanf B, Darienko T, Heine-Dobbernack E, Lorenz M, Naujox L, Keller ERJ, Schumacher HM, Friedl T, Eberth S, Mock H-P, Kniemeyer O, Overmann J (2018) ATP content and cell viability as indicators for cryostress across the diversity of life. Front Psychol 9:921. https://doi.org/10.3389/fphys.2018.00921

Article Google Scholar 

• Bielanski A, Bergeron H, Lau PCK, Devenish J (2003) Microbial contamination of embryos and semen during long term banking in liquid nitrogen. Cryobiology 46(2):146–152. https://doi.org/10.1016/S0011-2240(03)00020-8

CAS Article PubMed Google Scholar 

• Bielanski A, Nadin-Davis S, Sapp T, Lutze-Wallace C (2000) Viral contamination of embryos cryopreserved in liquid nitrogen. Cryobiology 40(2):110–116

CAS Article Google Scholar 

• Bielanski A, Vajta G (2009) Risk of contamination of germplasm during cryopreservation and cryobanking in IVF units. Hum Reprod 24(10):2457–2467. https://doi.org/10.1093/humrep/dep117

CAS Article PubMed Google Scholar 

• Bokulich NA, Subramanian S, Faith JJ, Gevers D, Gordon JI, Knight R, Mills DA, Caporaso JG (2012) Quality-filtering vastly improves diversity estimates from Illumina amplicon sequencing. Nat Methods 10:57. https://doi.org/10.1038/nmeth.2276 https://www.nature.com/articles/nmeth.2276#supplementary-information. Accessed 28 Aug 2018

Article Google Scholar 

• Buell CR, Joardar V, Lindeberg M, Selengut J, Paulsen IT, Gwinn ML, Dodson RJ, Deboy RT, Durkin AS, Kolonay JF (2003) The complete genome sequence of the Arabidopsis and tomato pathogen Pseudomonas syringae pv. tomato DC3000. Proc Natl Acad Sci USA. 100(18):10181–10186

CAS Article Google Scholar 

1. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, Costello EK, Fierer N, Peña AG, Goodrich JK, Gordon JI (2010) QIIME allows analysis of high-throughput community sequencing data. Nat Methods 7(5):335–336

CAS Article Google Scholar 

1. Chao A, Jost L (2012) Coverage-based rarefaction and extrapolation: standardizing samples by completeness rather than size. Ecology 93(12):2533–2547

Article Google Scholar 

1. Chatterjee A, Saha D, Niemann H, Gryshkov O, Glasmacher B, Hofmann N (2017) Effects of cryopreservation on the epigenetic profile of cells. Cryobiology 74:1–7

CAS Article Google Scholar 

1. Drexler HG, Uphoff CC (2002) Mycoplasma contamination of cell cultures: incidence, sources, effects, detection, elimination, prevention. Cytotechnology 39(2):75–90. https://doi.org/10.1023/A:1022913015916

CAS Article PubMed PubMed Central Google Scholar 

1. Fountain D, Ralston M, Higgins N, Gorlin JB, Uhl L, Wheeler C, Antin JH, Churchill WH, Benjamin RJ (1997) Liquid nitrogen freezers: a potential source of microbial contamination of hematopoietic stem cell components. Transfusion 37(6):585–591. https://doi.org/10.1046/j.1537-2995.1997.37697335152.x

CAS Article PubMed Google Scholar 

1. Gossner MM, Lewinsohn TM, Kahl T, Grassein F, Boch S, Prati D, Birkhofer K, Renner SC, Sikorski J, Wubet T (2016) Land-use intensification causes multitrophic homogenization of grassland communities. Nature 540(7632):266–269

CAS Article Google Scholar 

1. Hartig F (2017) DHARMa: residual diagnostics for hierarchical (multi-level/mixed) regression models. R package version 0.1. 5.
2. Herberich E, Sikorski J, Hothorn T (2010) A robust procedure for comparing multiple means under heteroscedasticity in unbalanced designs. PLoS One 5(3):e9788 https://github.com/qiime2/q2-feature-classifier. Accessed 29 Aug 2018

Article Google Scholar 

1. Kabiri L, Alum A, Rock C, McLain JE, Abbaszadegan M (2013) Isolation of Bacteroides from fish and human fecal samples for identification of unique molecular markers. Can J Microbiol 59(12):771–777. https://doi.org/10.1139/cjm-2013-0518