The whole cell microbial reference (WCMR) standard used to validate the μTitan system consists of nine bacteria, five of which are Gram positive and four of which are Gram negative. The strains used, their amounts, and other metadata are listed in Table 1. The WCMR was produced as part of the Association of Biomolecular Resource Facilities (ABRF) metagenomics research group (Tighe et al., 2017). Briefly, individual bacteria were cultured on either tryptic soy agar or marine 2216 agar until early log phase growth was achieved, followed by harvesting and resuspension in 1x phosphate buffered saline (PBS) without calcium or magnesium. The cell suspension was washed twice with 1x PBS followed by vortexing, to ensure that cells were in a unicellular suspension and not aggregated (confirmed by microscopy) before the fixation step with ethanol. Fixation was performed by adding 100% ethanol in a drop wise fashion while vortexing until the final concentration was between 90–95%. Cells were allowed to settle overnight, and the supernatant was discarded to remove any free-floating DNA. The final sample was resuspended in fresh 90% ethanol and stored at 4°C.

Enumeration of each individual ethanol fixed bacterial suspension was performed by diluting 1 μl of the suspension in 48 μl of 1X PBS and 1 μl of 0.1% sodium dodecyl sulfate. Staining was performed by adding 1 μl of Sytox staining reagent, 1 μl of SYBR green (2.5X), 1 μl of enhancer, and 7 μl of buffer (all reagents from Logos BioSystems, Seoul, S. Korea) to this dilution and incubating in the dark at room temperature for 5 min. Enumeration was performed by adding 4 μl of the stained cell preparation to a INCYTO counting slide (INCYTO C-Chip Hemocytometers, DHC-S025) and counted at 400x magnification using 485 nm/525 nm epifluorescent microscopy according to the manufacturer’s protocol. Combining the individual bacterial preparation into one pool was based on a staggered design. Cell concentrations ranged from 10⁴ to 10⁶ cells/μl, depending on the organism (Table 1), resulting in a mixed microbial community containing 7.3 × 10⁶ total cells/μl.

Before using the WCMR to validate the μTitan system, a controlled laboratory extraction of DNA was performed using a modified Qiagen procedure by first washing the cells in 1× PBS followed by enzymatic digestion of the cell walls with 20 μl of Metapolyzyme (MPZ) [stock concentration = 10μg/μl] (Millipore Sigma, St. Louis, MO, United States) at 35°C for 4 h, after which DNA extraction was completed using a QIAamp DNA Micro Kit (Qiagen 56304, Germantown, MD) with a pretreatment bead beating step (Matrix A beads MP biomedical 116910050CF) to increase lysis efficiency. Measurements from a Qubit spectrofluorometer indicated that 3.16 ng of DNA was obtained per 10⁶ cells.

The μTitan system is based on a fused deposition modeling 3D printer in which the motion and temperature controls have been repurposed for automated NA extraction, with the capability to perform both heated incubation and NA amplification (Figures 1A,B). The system is portable and robust and was able to perform extractions while being transported in the back of a sports utility vehicle (SUV) at highway speed (Figure 1C). The extraction efficiency of Chlamydia trachomatis DNA from urine samples using an early prototype was demonstrated to be comparable to the spin column method from Qiagen (Chan et al., 2016a). Extractions of RNA from the Zika virus in urine and saliva samples also proved to be successful (Chan et al., 2016b).

The current version of μTitan is 34×34×36 cm when encased with a cover. It weighs about 8 kg including the power adaptor. A laptop is currently used to operate the device, but the protocol program can be saved on a secure digital (SD) card if needed. A comparison between the Maxwell^TM and the μTitan specifications are presented in Table 2.

Individually enclosed environments for sample extraction are typically used to prevent cross-contamination. This is essential in the ISS environment to avoid the unintentional release of extraction reagents and potentially harmful microbes. Currently, μTitan uses aluminum-foil-sealed, pre-filled reagent strips placed inside a strip holder that can be closed after sample input (Figure 1D). Using this approach, magnetic coupling can be used to perform extraction with multiple mounted cartridges. The operation of the coupling mechanism is similar to that of a magnetic fish tank cleaner, where the motion of NA extraction can be controlled inside the cartridge without direct contact from outside. Figure 2A shows the basic set-up where an 8-well reagent strip is placed inside an enclosure to become a cartridge. The mechanical movements inside each individual cartridge are enabled by a unique magnetic coupling approach (Figures 2B,C). These figures show how nucleic acid extraction is performed by the stepper motors that are programmed to automatically control the movement of the external magnets, which in turn drive the movement of the components inside the cartridge. The “driver magnets” outside of the cartridge precisely control the movement of “follower magnets” within the cartridge to process nucleic extraction and elution. The two sets of magnetic particles (MPs) are held together by the magnetic field that penetrates through the cartridge’s plastic wall. Synchronized motions of the magnetic tip-combs enable reproducible and controllable action inside each cartridge, including MP mixing. Figure 2D shows the process of effective mixing and washing of MPs. These external “driver magnets” allow for precise positioning of the “MP manipulating tip” located inside the cartridge. The bottom pair of magnets below the wells leads to release and capture of the MPs by this tip, allowing the MPs to move throughout the wash buffer, allowing for effective mixing and washing of these MPs. Under normal gravity, the magnet inside the tip falls back down when the bottom repelling magnet is removed (Figure 3A), but this is not the case in reduced gravity as the magnet inside the tip will just remain “floating” inside (Figure 3B). For the microgravity adapted machine, a set of opposing field magnets was installed to pull the extraction tip magnet down (Figure 2D). These innovative steps allow the elimination of pipetting in space under microgravity, which is difficult, laborious, and often results in mishandling.

A commercially available magnetic particle–based NA isolation kit (NucliSENS Magnetic Particle Extraction Kit, bioMerieux, Durham, NC) was purchased, and the reagents from this kit were used to pre-fill the μTitan cartridges. These reagents included lysis buffer, magnetic particle solution, wash buffer #1, wash buffer #2, wash buffer #3, and elution buffer. The μTitan extraction protocol (e.g., volume, time of incubation, and number of repeated washing steps) is as follows: 100 μL of the pre-processed sample was added to the second well of a μTitan cartridge containing 400 μL of lysis buffer and incubated for 10 min. Eight microliters of NucliSENS magnetic particles, for capturing NAs, were then added to the lysed sample solution in the second well. The magnetic particles were intermittently mixed for 10 min, allowing for NA to bind to the MPs. The cartridge was then secured on the μTitan system. An extraction tip mounted on the follower magnet set was placed on the first well of the cartridge. Wells # 3–7 contained wash buffer # 1 (600 μL and 250 μL), wash buffer # 2 (600 μL and 250 μL), and wash buffer # 3 (250 μL). Well # 8 contained an elution buffer (60 μL). Each pre-processed sample was run in triplicate. To ensure the quality of the work and to check for contamination, molecular-grade water was used for extraction during each run of the machine, instead of sample, and is considered the machine control (“Tx_μT_CTL” or “YNP_μT_CTL”).

The μTitan system was operated using a laptop computer (connected with a USB cable), loaded with an open-source software called Repetier-Host (Hot-World GmbH & Co., Willich, Germany) that uses G-code based programming commonly used for 3D printing and CNC milling. The G-code written by AI Biosciences for μTitan, was based on that written for a standard 3D printer, which provides instructions on the movement, temperature, and amount of material to be deposited. Since this is pre-programmed, a user just needs to click on the “print” button icon to start the automated extraction process.

Once the program started, the extraction tip comb (driver) magnet picked up the extraction tip magnet inside the cartridge box using a magnetic coupling mechanism and moved it to the subsequent wells of the cartridge. The NA bound magnetic particles attached to the base of the extraction tip were then washed sequentially in wash buffer # 1 and # 2, twice for 2 min each, and wash buffer # 3 for 30 s. Next, the magnetic particles were air-dried for 4 min. When the air-drying process began, the heater strip at the bottom of the elution well of the cartridge was heated to bring the elution buffer temperature to ∼70°C by a cartridge heater commonly used as the extruder of a 3D printer. This extruder heater’s function was also controlled by the program, and the temperature was precisely monitored with a thermistor. The NAs captured on the magnetic particles were eluted for 5 min in 60 μL of elution buffer. After elution, the magnetic particles were captured again by the extraction tip and moved back to well # 1, leaving only the eluate in the elution well (well # 8). The eluate was collected and stored in a 500 μL tube for future analysis. The μTitan system has the capability of processing 6 samples in parallel.

Parallel extractions of the pre-processed samples were carried out on the Maxwell^TM automated extraction system (Promega Corporation, Madison, WI, United States) using a DNA kit (AS1520) and the “Maxwell^TM RSC blood DNA” run program. Briefly, 100 μL of the preprocessed samples were added to the cartridge and placed on the cartridge holder tray, and the DNA was eluted in 60 μL of elution buffer. Each pre-processed sample was run in triplicate. To ensure the quality of the work and to check for contamination, molecular-grade water was used for extraction during each run of the machine, instead of sample, and is considered the machine control (“Tx_Max_CTL”). The purified DNA from both machines was stored at −20°C until further analysis.

Samples were quantified using two methods: the Q32854 Qubit^TM dsDNA HS Assay (ThermoFisher, Waltham, MA, United States) and qPCR using a standard curve SYBR green method using two control DNA standards. Qubit assay was performed using a 2 μl aliquot of DNA mixed with 198 μl of the dsDNA HS reagent and read on a Qubit 4 spectrofluorometer.

qPCR was performed on replicate 1 μl samples of 1:10 dilution of test DNA using the PerfeCTa SYBR Green SuperMix (95054 Quanta Bio, Beverly, MA, United States) using primers recommended by the Earth Microbiome Project -515 Forward GTGYCAGCMGCCGCGGTAA (Parada et al., 2016) and 806 Reverse GGACTACNVGGGTWTCTAAT (Walters et al., 2016). Standard curves were generated from two separate microbial reference control DNA, including (1) purified genomic DNA (gDNA) from the WCMR that had been extracted with Qiagen and (2) The ABRF/ATCC^® MSA-3001 gDNA (ATCC, Manassas, VA). Both standard curves were generated from 2, 0.2, 0.02, 0.002, and 0.0002 ng/μl of DNA. qPCR thermocycling was performed in 25 μl reactions on an ABI 7900HT (Applied Biosystems, Foster City, CA, United States) using the following three-step program: Denatured at 95°C -3 min, 40 cycles of PCR (95°C -15 s| 54°C-60 s| 72°C -60 s), followed by a standard dissociation curve (95°C-15s| 60°C-15s| 95°C-15s). Primers were included at a final concentration of 0.1 μM and low ROX as an internal reference.

DNA fragment size analysis (molecular weight) of the extracted DNA was evaluated using the Agilent Bioanalyzer 2100 with the High Sensitivity DNA chip (5067-4627) and the Advanced Analytical Technologies Fragment Analyzer 5200 with the HS Large Fragment 50kb Kit (DNF-493-0500) according to the manufacturer’s instructions.

Oxford Nanopore libraries were synthesized for all samples from input ranges of BDL (below detection limit of the Qubit) up to 5 ng for higher input samples using the Rapid PCR barcoding kit (RPB004 Oxford Nanopore Technologies, Oxford, United Kingdom) with the following modifications: high input samples were amplified using the standard method of 14 PCR cycles, while samples that were below the detection limit of Qubit were amplified using 22 PCR cycles. Barcoded libraries were pooled equally at 1 fmol each for the high samples, while the BDL samples were pooled at the maximum volume allowable for library input. Sequencing was performed using the MK1B MinIon with 9.4 flow cell yielding > 1 million reads of 2 Gbases with average read lengths of 2–8 kb. All quality control statistics were within specification as outlined by Oxford Nanopore.

Microbial reference controls were also sequenced on the Oxford Nanopore sequencer, including the purified gDNA from the Qiagen Benchmark sample. Rapid PCR barcoded libraries were synthesized from input concentration identical to the qPCR standard curve at 6, 0.6, 0.06, 0.006 ng. This titration was performed to determine the detection limits, linearity, and efficiency for ultra-low input samples.

Oxford Nanopore data was analyzed using the WIMP (What’s In My Pot) module of the EPI2ME software (Oxford Nanopore Technologies, Oxford, United Kingdom), and raw read count data was exported as CSV data for manual parsing and enumeration using Microsoft Excel.

Nanopore detection and sequencing of the low biomass samples revealed negligible concentrations above background. This is not surprising since the input concentrations were below the recommended input for the Rapid PCR barcoding kit. For these reasons, Nanopore data generated from low biomass samples were not included.

The Illumina sequencing data was analyzed in R version 3.3.1 using the Hmisc, ggplot, compositions and cluster packages. For the PCA plot, the data was first center log ratio (clr) transformed and the Euclidian distances plotted. K-means clustering was used to determine the number of distinct groups and the samples that belonged to these groups.

DNA yield and qPCR graphs were generated in Prism version 7.0. Statistical analyses were executed using one-way analysis of variance (ANOVA) and then using the Benjamini- Hochberg false discovery rate (FDR) multiple test correction. Statistical significance was set at P < 0.05.

A whole cell microbial reference standard, that contained intact microbial cells (Table 1) was extracted using both μTitan and Maxwell^TM instruments in a laboratory setting. High biomass (10⁷ cells) and low biomass (10⁴) cellular concentrations were used in the validation tests. WCMR standard was extracted using both the μTitan and Maxwell^TM instruments in a laboratory setting. The composition of the WCMR (Table 1) contains fully intact microbial cells of Gram +/−, high and low GC, and a range of morphologies. Two cellular concentrations were used in the extraction tests of the two instruments representing high biomass (10⁷ cells) and low biomass (10⁴ cells) inputs and subject to pretreatment with a cell wall digesting enzyme mix (Metapolyenzyme) and mechanical bead beating. After pretreatment, replicate 100 μl aliquots were used as inputs to both instruments. Following extraction, the yield of recovered DNA was determined using a Qubit spectrofluorometer and qPCR (Figure 4). These results indicate that the μTitan system produced higher DNA yields, with both the high biomass (“Tx_x_High”) and low biomass (“Tx_x_Low”) samples; however, only the high cellular concentrations were statistically different.

Shotgun metagenomic sequencing was performed on all samples using the Illumina HiSeq 2500 platform to determine whether the isolated DNA could be used to prepare successful sequencing libraries and obtain high-quality sequencing results. Table 3 shows the WCMR members detected in each sample from both extraction instruments and their respective counts. The bar graph in Figure 5 shows the proportion of these nine bacteria in each sample and also includes the proportion of bacteria detected when a manual Qiagen QIAquick extraction (Qiagen Benchmarking Control) was performed in the lab with high efficiency pre-processing steps added. Both Table 3 and Figure 5 show that DNA from all nine bacteria could be isolated from the high biomass samples extracted with either the μTitan (“Tx_μT_High”) or the Maxwell^TM instrument (“Tx_Max_High”). For the low biomass samples, the μTitan system had increased performance over the Maxwell^TM and from the DNA extracted, six out of nine bacteria were identified (“Tx_μT_Low”), while from the Maxwell^TM extracted DNA, only three out of nine bacteria were identified (“Tx_Max_Low”). The proportions of identified bacteria in the high biomass samples were similar between the two instruments as well as for the Qiagen Benchmarking Control (Figure 5).

The μTitan system was further evaluated in a Phase II study at Yellowstone National Park (YNP) to test its performance in a remote setting with limited resources. The Maxwell^TM system was not included in this field study due to its size, weight, power, and calibration requirements. In this phase of testing, two concentrations of the WCMR were used; a high biomass sample (“YNP_μT_High”) similar to that used in the laboratory test (“Tx_μT_High”) and a medium biomass sample composed of a 10-fold dilution of the high biomass sample (“YNP_μT_Med”, 10⁶ cells).

The resulting DNA yield for the high biomass sample, as measured by qPCR was comparable to the yield obtained in the laboratory tests (“Tx”), indicating the μTitan system is able to produce consistent results even when used in a remote setting (Figure 4). Results of the Illumina-based shotgun metagenomic sequencing were also consistent with the laboratory tests in that all nine bacteria in the WCMR were detected in both the high and medium biomass samples (Table 3), in the same proportions as that observed in the “Tx_μT_High” and Qiagen benchmark sample (Figure 5).

As is expected from any shotgun metagenomics sequencing run, there will be some measure of contaminant reads that come from extraction and sequencing processes (the “kitome”), and this dataset was no exception. The heatmap in Supplementary Figure S1 shows the bacterial contaminants (i.e., any species that did not belong to one of the nine bacteria in the WCMR along with their counts) in all samples.

A PCA plot (Figure 6) comparing the diversity and relative abundances of the species present in the samples and controls (both WCMR bacteria and contaminants) showed that the high and medium biomass samples had similar microbiomes, regardless of whether they were extracted with the μTitan system (green points) or with the Maxwell^TM instrument (blue points), and were distinct from the kitomes and machine controls. With the low biomass samples, the μTitan samples (Tx_μT_Low, pink points) were distinct from their respective controls, but this was not the case with the Maxwell^TM samples (Tx_Max_Low, purple points), which were indistinguishable from their respective controls. Additionally, the PCA plot indicates that replicate extractions with the μTitan system were consistent and produced comparable results (replicates group together on the plot). More importantly, the plot shows that the field-operated μTitan system (YNP_μT_High) and lab-operated system (Tx_μT_High) produced similar results.

In addition to the Illumina HiSeq data, shotgun sequencing was also performed on the high and medium biomass samples using Oxford Nanopore Technologies’ MinIon MK1B. The importance of including Nanopore data in this study was to provide an alternative sequencing technique that would avoid any Illumina Nextera short-read sequencing biases due to GC content, to enable detection using long reads with an alternative library synthesis that may influence taxonomic classification due to genome complexity and long genomic repeats (e.g., Pseudoaltermonas), and to determine how DNA extracted from μTitan would sequence on the MinION MK1B as this is the sequencing platform that is currently being used on the ISS and in the field. Although read depth, read length, and Q-Score are different between Nanopore and Illumina sequencing technologies, the organisms belonging to the WCMR were just as easily detected and classified with the MinION using DNA isolated from both the Maxwell^TM and μTitan systems, in both the laboratory and field settings. Tabulation of organisms in each sample belonging to the WCMR was represented at the Genus level on raw data normalized to 50,000 reads to allow for direct comparison between samples (Table 4). These results indicate that, when compared to the Maxwell^TM instrument, the μTitan system generated a marginally greater number of reads of Staphylococcus and Bacillus, and an even greater number of Enterococcus, all of which are Gram-positive bacteria. On the other hand, more reads were observed by the Maxwell^TM instrument for the Gram-negative bacteria, Escherichia and Pseudomonas. These trends observed with the Nanopore data are consistent with that observed with the Illumina data.

The proportions of the WCMR were compared between Illumina and Nanopore data for the purpose of detecting any bias inherent to the sequencing method (Figure 7). These data indicates that Illumina short-read sequencing using Nextera XT libraries has increased ratios of detection for Chromobacterium, Enterococcus, and Micrococcus, while the Oxford Nanopore rapid PCR data has an increase detection ratio for Pseudomonas, Pseudoalteromonas, Bacillus, and Staphylococcus.

The extracted DNA from the high and medium biomass samples were also analyzed using an Agilent Bioanalyzer 2100 to determine the DNA fragment length. For both the Maxwell^TM and μTitan instruments, the fragment length amongst the replicates were very consistent, with the μTitan system producing slightly larger DNA fragments (average length: 6,625 bp) compared to the Maxwell^TM system (average length: 5,604 bp). These results were confirmed with a subsequent analysis using a high-sensitivity assay with the Advanced Analytical Fragment Analyzer 5200 (data not shown), which was closer to the fragment length obtained by using a Qiagen Benchmark control manual extraction (average length: 7,912 bp) (Table 5; Supplementary Figure S2).

While the μTitan cartridges used in these experiments were not designed for microgravity compatibility, the newest set of cartridges have been designed to exploit surface tension properties to prevent the unwanted release of fluids (Weislogel et al., 2009, 2011). These cartridges accommodate fluids and a magnetic probe and allow the fluids to preferentially remain at the base of the cartridge well, discouraging wetting of the upper portion of the well, thus preventing liquids from floating free when used in microgravity (Weislogel et al., 2009, 2011). The microgravity fluid physics behind this design comes from the same team who engineered the “space cup” in 2015, which allowed astronauts on the ISS to drink hot espresso from an open coffee cup for the first time (NASA, 2015). Several 2.1s drop tower tests have been performed on the μTitan system (used in the current study) and the newly designed cartridges, and during these low-gravity tests, the fluids remained at the bottom of the cartridge wells, as designed (Supplementary Figure S3), indicating its compatibility with use on the ISS.