TY - JOUR
AU - Lin, Xiaoxia, N
AB - Abstract While the ‘unculturable’ majority of the bacterial world is accessible with culture-independent tools, the inability to study these bacteria using culture-dependent approaches has severely limited our understanding of their ecological roles and interactions. To circumvent cultivation barriers, we utilize microfluidic droplets as localized, nanoliter-size bioreactors to co-cultivate subsets of microbial communities. This co-localization can support ecological interactions between a reduced number of encapsulated cells. We demonstrated the utility of this approach in the encapsulation and co-cultivation of droplet sub-communities from a fecal sample collected from a healthy human subject. With the whole genome amplification and metagenomic shotgun sequencing of co-cultivated sub-communities from 22 droplets, we observed that this approach provides accessibility to uncharacterized gut commensals for study. The recovery of metagenome-assembled genomes from one droplet sub-community demonstrated the capability to dissect the sub-communities with high-genomic resolution. In particular, genomic characterization of one novel member of the family Neisseriaceae revealed implications regarding its participation in fatty acid degradation and production of atherogenic intermediates in the human gut. The demonstrated genomic resolution and accessibility to the microbial ‘dark matter’ with this methodology can be applied to study the interactions of rare or previously uncultivated members of microbial communities. microfluidics, microbial communities, microdroplet, human gut, microbiome, co-cultivation, metagenomics INSIGHT, INNOVATION, INTEGRATION Crucial for human and environmental health, natural microbial communities are highly complex structures composed of uncharacterized and ‘unculturable’ bacteria, which have eluded cultivation by traditional methods. The number of these ‘dark matter’ microbes greatly outweighs cultivated isolates, severely limiting the capability to study ecological interactions in natural systems. Studying the interspecies interaction of ‘unculturables’ within their communities requires the integration of high-resolution experimental and genomic methods. The work demonstrated in this study combines the low-volume and highly parallel nature of microfluidic droplets with metagenomics to demonstrate the high-resolution genomic characterization of previously uncharacterized members of complex natural communities. INTRODUCTION Culture-independent tools, particularly 16S rRNA gene sequencing and shotgun metagenomics, have greatly enhanced our understanding of the microbial world in the past 15 years. By avoiding the challenges of cultivation, researchers have profiled the diversity and functional potential of environmental and host-associated communities through the recovery of genomes from uncultivated taxa. Such methods have revealed the prevalence and importance of uncultivated phyla of bacteria critical to biogeochemical processes [1–3], host health [4, 5], and biotechnology [6]. Nevertheless, culture-independent methods are subject to fundamental limitations and cannot replace culture-dependent methods. While culture-independent approaches provide ready access to the uncultivated majority, well-controlled experimentation and replication is inherently difficult, especially for highly complex communities or environments with unexplored taxonomic diversity. In addition, inferences derived from culture-independent methods cannot fully emulate the direct phenotypic and ecological observations that culture-based methods can provide [7]. Traditional culture-dependent approaches involving experimental manipulation and cultivation using laboratory media have been staple to advancing microbial ecology. Phenotypic observations—such as growth kinetics, metabolism under different conditions, and the physiological response to other organisms in co-cultures—are critical to our understanding of microbial ecology. Synthetic and defined microbial communities have been used as model systems for studying physiology in a community context [8, 9] to yield valuable insights beyond those attained through culture-independent studies of complex communities. However, the reliance on cultivation limits how applicable these approaches are to the study of complex microbial communities, which are largely comprised of uncultured or uncharacterized members. A combination of top-down culture-independent approaches and bottom-up culture-dependent approaches holds promise for addressing the challenges in microbial ecology [3, 10–13]. However, there are numerous barriers to traditional cultivation [14], including microbial lifestyles such as obligate syntrophy [15] and slow-growth [16], competition, and extremely low abundances in diverse natural communities. Microfluidic droplet technology presents a unique solution to circumvent many of these barriers by manipulating microbial communities at the nanoliter scale. While used extensively in the past for the high-throughput encapsulation and manipulation of single cells [17, 18], microfluidic droplets can also be utilized as micron-scale bioreactors to encapsulate and propagate subsets of natural microbial communities, decomposing a bulk community into a large number of much smaller sub-communities. This compartmentalization limits complexity while allowing the study of microbe–microbe interactions. While microfluidic droplets have been employed for single-cell or ‘mini-metagenomic’ investigation of the microbial dark matter [19–23] or for synthetic systems [24, 25], these studies did not incorporate co-cultivation for the investigation of microbial communities. Microfluidic co-cultivation of microorganisms provides a bridge between culture-dependent and -independent tools, allowing for the exploration of ecological interactions between cells through co-growth at microscopic scales. This circumvents the technical difficulties of culture-dependent work while enabling the use of high-resolution ‘omics tools in studies of interacting microbial consortia comprised largely of uncultured species. To demonstrate the potential of utilizing microfluidic droplets to study the ‘dark matter’ of natural communities, this study uses microfluidic droplets to dissect a complex human fecal sample into sub-communities for highly parallel co-cultivation. Afterwards, 22 individual droplets with strong bacterial co-growth were selected with microfluidic techniques. Multiple displacement amplification (MDA), a whole genome amplification method, amplified nucleic acid from individual droplet sub-communities for 16S rRNA amplicon sequencing, and one sub-community was metagenomically shotgun sequenced. Of particular interest relating to the human gut microbiome, a partial-genome of a representative of a novel genus within the Neisseriaceae was found in this droplet, highlighting the capability of microfluidic co-cultivation to access and study uncharacterized microbial diversity. MATERIALS AND METHODS Microfluidic devices and fabrication The work here utilizes two microfluidic devices: a droplet generation device and a droplet spacing device. The droplet generation device is a modified cross-flow droplet generation device made from polydimethylsiloxane (PDMS) as described previously by Carruthers et al. [26]. The droplet spacing device used to separate individual droplets is a modified droplet generation device with three layers to construct a pressure-controlled membrane valve to precisely manipulate droplet flow through a cross-flow junction (Supplementary Fig. 1) [27, 28]. The valve is controlled manually by an external pressure pump to switch on and off the flow of individual droplets (Supplementary Video 1). When pressure is on, the membrane valve closes and stops the flow of droplets while allowing the oil phase flow to continue. To fabricate molds with the microfluidic device features, photo-mask creation and SU-8 mold etching were performed in the Lurie Nanofabrication Facility at the University of Michigan. SU-8 molds were made by negative etching on a silicon wafer, which was spin-coated with SU-8 2035 at a thickness of 50 μm. The wafer was pre-baked at 65°C and at 95°C, exposed, and post-exposure baked at 95°C. After baking, the wafer was silanized with tridecafluoro-1,1,2,2,-tetrahydrooctyl-1-trichlorosilane in a desiccator. For the PDMS layers with device features, PDMS with Sylgard® 184 curing agent (10:1 mass ratio of PDMS to curing agent) was poured on top of respective SU-8 molds, vacuumed to remove air bubbles, and heated at 70°C to solidify the polymer overnight. The devices were then cut to size. The membrane between the valve layer and the channel layers for the spacing device was made from PDMS (15:1 mass ratio of PDMS monomer to curing agent) spun on a glass wafer at 1000 rpm to create a 50-μm membrane. The membrane was heated to 80°C for 15 min and plasma-activated bonded to the other layers of the spacing device using a corona discharge wand. To complete the fabrication, the devices were punched with holes by a biopsy punch (1.25-mm ID) to create openings for channel flow, cleaned with acetone and rinsed with water, and bonded on cleaned glass microscope slides via plasma-activated bonding. Droplet generation, cultivation, and processing All droplet work was performed in an anaerobic chamber (Coy© vinyl anaerobic chamber) with an atmosphere of 5% hydrogen, 10% carbon dioxide, and balance nitrogen to protect and provide favorable conditions for obligate anaerobes in our sample. To remove any trace oxygen, a fan box to recirculate air through a Coy© palladium catalyst Stak-Pak is employed in the chamber. The cellular suspension for flow through the droplet generation microfluidic device was derived from a fecal sample obtained from a healthy human subject at the University of Michigan Hospital in October 2011. The fecal sample was stored at −80°C in a 2-ml cryovial and thawed anaerobically. A total of 1.5 ml of phosphate buffer saline (PBS) was added to the fecal sample, and the mixture was centrifuged at 1000 rpm to separate fecal debris from the bacterial suspension for microfluidic droplet generation. The cell density of the bacterial suspension was determined with a hemocytometer (C-Chip™ disposable hemacytometer, Fisher Scientific 22–600-100) under a Nikon inverted contrast phase microscope (Nikon Eclipse Ti-S) and was determined to be 1.9 × 108 cells/ml. The suspension was diluted to the appropriate concentration determined by the desired average cell number per droplet (λ) divided by the volume of droplet. To provide different droplet sub-community complexities, two λ values were used for droplet generation: 2 and 10. Control droplets without cells from the fecal sample were also generated to demonstrate no bacterial contamination from reagents was present (Supplementary Fig. 2). To dilute the volume and provide the nutrients for growth, two different media for the cultivation of anaerobic intestinal bacteria were used: Brain Heart Infusion (BHI) [29] (Becton-Dickinson) and the Schaedler medium (SM) [30, 31] (Oxoid). Droplet generation was performed in the anaerobic chamber. The droplet generation device was placed on a lab compound microscope (Amscope, M150C) with a USB connected camera. Syringes were connected to tubing (Cole-Parmer, EW-0641721) leading to the droplet generation device and placed on a syringe pump (CMA 102 Syringe Pump). Oil phase—fluorocarbon oil (HFE-7500, 3M) containing 2% perfluoropolyether-polyethyleneglycol surfactant (RAN Biotechnologies)—was placed into one syringe and the bacterial suspension in the other. The syringe pumps adjusted oil and aqueous flows to achieve generation on-chip. Droplets generated flowed outwards from the droplet generation device into Eppendorf tubes attached with tubing. Approximately 500 μl of droplets were collected for each sample and covered with sterile mineral oil to prevent evaporation. The droplets were incubated at 37°C in the anaerobic chamber for 1 week, which provided sufficient time for co-growth to occur. For imaging of droplets, 10 μl of droplets were inserted into a C-Chip hemocytometer before and after incubation and visualized under the inverted phase contrast microscope. Spacing of high cell density droplets Incubated droplets were stored and transferred in 1.5-ml micro-centrifuge tubes using two ports installed in the tube lid. In one port, Teflon tubing was inserted to the bottom of the tube, and the other port had a syringe needle. Droplets were transferred from the bottom of the tube to the spacing device by applying pressure through the syringe tip. Droplets were introduced into the middle junction of the device and held at the junction by flowing 1.5% surfactant in HFE oil while the membrane valve was closed. After the device reached steady state, Teflon tubing was inserted into the outlet channel. Spacing of individual droplets was performed through the manual control of a pressure pump connected by tubing to the valve channel. After the spacing of a droplet, the membrane valve was shut for 1 min to ensure that the isolated droplet had traveled through the entirety of the tubing, and was collected into a 1.5-ml microcentrifuge tube. While in the spacing device, droplets were individually visualized with microscopy for identification of droplets with clear bacterial co-growth (Supplementary Fig. 3). Through usage of the spacing device, chosen droplets were isolated into separate 1.5-ml microcentrifuge tubes and kept for downstream processing. Droplets not selected were discarded. Cell lysis and whole genome amplification of isolated droplets For each isolated droplet, 5 μl of surfactant destabilizer (RainDance Technologies, RDT 1000 droplet destabilizer), an additional 4 μl of PBS, and 3 μl of cell lysis solution (3 μl) was added to each tube. After a 10-min incubation period at 65°C and the addition of a lysis stopping solution (3 μl), the MDA reaction mixture (Single Cell Repli-G kit, Qiagen 150343) for whole genome amplification was added to the aqueous phase, and the reaction was allowed to proceed for 8 h at 30°C. The polymerase was deactivated by heating the solution to 65°C, and amplified DNA was collected from the mixture by pipetting out the aqueous phase and then diluted 100-fold. 16S and metagenomic shotgun sequencing The V4 region of the 16S rRNA gene was amplified and sequenced from MDA-amplified samples for all 22 droplet samples at the University of Michigan Center for Microbial Systems Sequencing Core using a dual-index PCR library preparation and sequencing strategy as specified by Kozich et al. [32]. Information on the library preparation procedure is found at https://github.com/SchlossLab/MiSeq_WetLab_SOP/blob/master/MiSeq_WetLab_SOP_v4.md. Metagenomic sequencing was conducted on a selected MDA-amplified sample at the University of Michigan Advanced Genomics Core. Sequencing core staff prepared barcoded Illumina libraries of a target fragment size of 450 bp using the IntegenX Apollo 324 PrepX ILM DNA Library kit and custom Illumina compatible BioScientific barcoded adapters. Libraries were amplified with the KAPA library amplification kit, quality controlled on the Advanced Analytical Fragment Analyzer, quantified using the KAPA Illumina library quantification kit, and then sequenced for paired end 150-cycle Illumina HiSeq 4000 sequencing. 16S OTU clustering and analysis 16S V4 reads were processed for quality control, chimera removal, and analysis with mothur [33] according to the MiSeq standard operating procedure (https://www.mothur.org/wiki/MiSeq_SOP). Classification was performed with the Ribosomal Database Project [34] v. 16 training set from February 2016 (https://www.mothur.org/wiki/RDP_reference_files). Operational taxonomic units (OTUs) were clustered at the 97% similarity with OTUs less than 78 reads, 1% of the lowest sample’s sequencing depth, removed. To determine if these droplet OTUs were abundant in the bulk fecal sample, the partial V4 region overlap between previously performed 16S V45 454-pyrosequencing reads (processed through mothur according to https://www.mothur.org/wiki/454_SOP, resulting in 8017 reads from the bulk) and the droplet amplicons was compared. Distances less than 0.05 signified the matching of a droplet OTU to an OTU from bulk sequencing. Analysis of metagenomic shotgun sequencing data Metagenomic shotgun reads were processed in the University of Michigan’s high-performance computing cluster Flux. To improve the quality of the sequence reads, reads for both microdroplet samples underwent dereplication (script available at https://github.com/Geo-omics/scripts/blob/master/DerepTools/dereplicate.pl), Illumina adapter residual removal by Scythe (v. 0.993) (https://github.com/vsbuffalo/scythe), low-quality region removal by Sickle (v. 1.33.6) (https://github.com/ucdavis-bioinformatics/sickle), and interleaving (script available at https://github.com/Geo-omics/scripts/blob/master/AssemblyTools/interleave.pl) to associate forward and reverse reads. FASTQC (v. 0.10.1) (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/) was used to assess the quality of the reads before quality control and after. To normalize the highly non-uniform coverage of reads generated from stochastic amplification of MDA across the sample and subsequently improve assembly, bbnorm.sh from BBTools (http://jgi. doe.gov/data-and-tools/bbtools/bb-tools-userguide/bbnorm-guide) was used to remove high coverage reads to a target k-mer coverage of 50 and remove reads with a k-mer coverage less than two. Before metagenomic assembly, bbsplit.sh from BBTools was used to remove sequence reads from the human genome by mapping to a reference genome provided by the 1000 Genomes Project [35]. This resulted in a removal of 36 Gb of reads associated with the human genome and a remainder of around 6 Gb presumably prokaryotic-associated. Metagenomic assembly of these cleaned reads was done with metaSPAdes (v. 3.9.0) [36], using the flags ‘sc’ for MDA amplified DNA and ‘meta’ for metagenomic reads. As suggested for a multi-cell data set with longer Illumina paired reads lengths, the iterative k-mer lengths used for assembly were: 21, 33, 55, 77, 99, and 127. QUAST (v. 4.3) [37] was used to check assembly quality. Visualization and manual binning of scaffolds was done with anvi’o (v. 2.3.0), following the metagenomic workflow [38]. Due to uneven MDA amplification, coverage information was not relevant for binning; therefore, read mapping information and the anvi-merge command were not used. As a result, anvi-profile was run with the ‘blank-profile’ parameter as specified (http://merenlab.org/2016/06/06/working-with-contigs-only/). Binning was performed using the automatically generated tree based on tetranucleotide frequency profiles and assisted with taxonomic information provided by Centrifuge [39] for contig splits in anvi-interactive visualization. Bin assessment was performed initially with anvi’o, utilizing the marker gene set from Campbell et al. [40], and finally with CheckM [41] to determine degree of genome completeness and contamination. Annotation of genes for genomes of high completeness and low contamination was performed using RAST [42]. Pathway inference and subsequent visualization and analysis was performed with the PathoLogic [43] component of Pathway Tools [44] utilizing the MetaCyc database (v. 22.0) [45]. A custom python script was used to convert the Genbank annotation files for each individual metagenomic scaffold for each genome from RAST into files compatible for PathoLogic genome inference. Incorporation of genomes into the microbial tree of life was done with PhyloPhlAn [46], which uses percent identity of proteins from reconstructed microbial genomes with conserved protein sequences from microbial genomes from the Integrated Microbial Genomes (IMG). Data availability All draft genomes recovered from metagenomic assembly are available in GenBank under Bioproject ID PRJNA597463. RESULTS General workflow: microfluidic encapsulation, co-cultivation, and metagenomic processing of microbial sub-communities We present a microfluidic droplet-based technological pipeline that extends previous development in droplet co-cultivation and single-droplet processing [27, 28] with metagenomic sequencing (Fig. 1). A complex environmental sample, in this case a human fecal sample, has been processed to form a microbial suspension that flows through a microfluidic droplet generation device [47] (Fig. 1a). The microfluidic device utilizes the biphasic flows of the aqueous microbial suspension and an oil phase through a T-junction to generate monodispersed, nanoliter-sized microdroplets at a rate of 1–500 droplets/second (Fig. 1b). The distribution of cells encapsulated in individual droplets follows Poisson statistics and can be manipulated through droplet size and initial cell concentration of the microbial suspension. Afterwards, the droplets are incubated under appropriate conditions (Fig. 1c) and visualized with microscopy. Droplets meeting selection criterion are isolated and transferred to a separate receptacle using a microfluidic droplet spacing device (Fig. 1d). Lysis reagents and droplet destabilizer are added to the droplets to release and lyse cells, and MDA is used to amplify minute amounts of DNA to a quantity required for sequencing (Fig. 1e). Based on results from 16S amplicon sequencing, droplets containing sub-communities with interesting taxa can be selected for metagenomic shotgun sequencing (Fig. 1f) for further study, followed by bioinformatic analysis. Figure 1 Open in new tabDownload slide Overview of the microfluidic droplet cultivation and processing pipeline. (a) A microbial suspension derived from a human fecal sample is prepared. (b) Using biphasic flows of aqueous and oil phases, random combinations of bacteria are encapsulated in microdroplets at frequencies according to a Poisson distribution. (c) These droplets are incubated anaerobically for a week to allow for co-cultivation of the subcommunities. (d) With a droplet spacing device, microdroplets are isolated and processed individually. (e) Upon droplet destabilization, cells released from individual droplets are lysed and their genomes are amplified with MDA to generate sufficient nucleic acid material for downstream sequencing. (f) 16S amplicon and metagenomic libraries are prepared with amplified DNA and sequenced. 16S profiling of individual droplets is used to determine which droplet to submit for metagenomic shotgun sequencing. Figure 1 Open in new tabDownload slide Overview of the microfluidic droplet cultivation and processing pipeline. (a) A microbial suspension derived from a human fecal sample is prepared. (b) Using biphasic flows of aqueous and oil phases, random combinations of bacteria are encapsulated in microdroplets at frequencies according to a Poisson distribution. (c) These droplets are incubated anaerobically for a week to allow for co-cultivation of the subcommunities. (d) With a droplet spacing device, microdroplets are isolated and processed individually. (e) Upon droplet destabilization, cells released from individual droplets are lysed and their genomes are amplified with MDA to generate sufficient nucleic acid material for downstream sequencing. (f) 16S amplicon and metagenomic libraries are prepared with amplified DNA and sequenced. 16S profiling of individual droplets is used to determine which droplet to submit for metagenomic shotgun sequencing. Microfluidic droplets allow for highly parallel co-cultivation of microbial sub-communities Two rich media for anaerobe cultivation, BHI and the SM, were used to cultivate microbial sub-communities in individual droplets (Fig. 2). Initially, most sampled droplets contained only a few cells (Fig. 2a and c), but a subset of droplets exhibited substantial co-growth after incubation (Fig. 2b and d). A range of morphologies were observed across these cultivated sub-communities. For instance, some droplets showed communities with distinctly long-rod morphologies (Fig. 2b-3, d-8), while others consist of communities with distinct cocci morphologies (Fig. 2b-4, d-7). Interestingly, only a subset of droplets supported growth in rich media and with a high λ of 10 to ensure all droplets contained viable cells, demonstrating that not all sub-communities encapsulated were capable of co-growth in these conditions. Droplets that were individually selected from droplet spacing based on dense co-growth are shown in Supplementary Fig. 4. Figure 2 Open in new tabDownload slide Microbial sub-communities in generated microfluidic droplets before and after co-cultivation. A sample pool of droplets with encapsulated microbial sub-communities before (a, c) and after anaerobic cultivation for a week (b, d). Droplets were cultivated in two rich media: BHI (a, b) and Schadler media (SM) (c, d). Droplets are not tracked over time, so each droplet viewed is distinct. Dashed boxes on the left correspond to the magnified droplets in each subpanel on the right, identified by the numerical marker. Arrows distinguish single cells in the pre-incubation microfluidic droplet. Scale bar is 100 microns. Figure 2 Open in new tabDownload slide Microbial sub-communities in generated microfluidic droplets before and after co-cultivation. A sample pool of droplets with encapsulated microbial sub-communities before (a, c) and after anaerobic cultivation for a week (b, d). Droplets were cultivated in two rich media: BHI (a, b) and Schadler media (SM) (c, d). Droplets are not tracked over time, so each droplet viewed is distinct. Dashed boxes on the left correspond to the magnified droplets in each subpanel on the right, identified by the numerical marker. Arrows distinguish single cells in the pre-incubation microfluidic droplet. Scale bar is 100 microns. Sub-community composition is a result of specific growth medium enrichment and stochasticity in cell encapsulation A total of 22 droplets with high cell densities at the end of the incubation period were isolated and underwent 16S amplicon sequencing. Because amplification bias introduced with MDA heavily obscures any quantitative signal in sequencing [48], OTUs were analyzed by presence or absence to compare community composition across droplet sub-communities. The community membership of these 22 droplets was highly variable between droplets (Fig. 3). Due to the stochastic nature of cell encapsulation during droplet generation and the high diversity of the original human gut microbiome sample, a relatively large number of OTUs appear in a relatively small number of co-cultivated sub-communities. However, OTUs of certain genera, such as Staphylococcus (OTU1), Propionibacterium (OTU27), and Corynebacterium (OTU4) (Supplementary Table 1), are pervasive across these sub-communities. Hierarchal clustering of droplet sub-communities demonstrated that two major clusters emerged, partitioning generally by growth medium. While this clustering is not significant (unbiased P-value <95%) most likely due to the relatively small number of droplets analyzed, certain OTUs are highly associated with SM, such as Parabacteroides (OTU5) (indicator value = 90.9, P-value = 0.021) and Butyricimonas (OTU3) (indicator value = 75.3, P-value = 0.046) (Supplementary Table 2). This suggests that the medium composition is one factor determining that microbes are co-cultivated in the droplets. Figure 3 Open in new tabDownload slide 16S OTU profiles of 22 isolated droplets sub-communities. OTU classification and bootstrap values are provided. Droplet identity nomenclature is based on Schaedler media (S) or BHI media (B) and with the initial λ value [2 or 10] and a numerical identifier. Because MDA introduced significant bias, quantitative information is not shown and OTUs are presented as either present (black) or absent (light gray) in a sample. Hierarchical clustering of the droplet taxonomic profiles provides two distinct clusters. P-values provided by pvclust in R do not signify statistical significance (AU < 95%). Figure 3 Open in new tabDownload slide 16S OTU profiles of 22 isolated droplets sub-communities. OTU classification and bootstrap values are provided. Droplet identity nomenclature is based on Schaedler media (S) or BHI media (B) and with the initial λ value [2 or 10] and a numerical identifier. Because MDA introduced significant bias, quantitative information is not shown and OTUs are presented as either present (black) or absent (light gray) in a sample. Hierarchical clustering of the droplet taxonomic profiles provides two distinct clusters. P-values provided by pvclust in R do not signify statistical significance (AU < 95%). Interestingly, while the droplets were generated such that the average initial cell number per droplet (λ) was either 2 or 10, most of the selected and analyzed sub-communities were found to each contain more than 10 OTUs. In fact, the difference in the species richness between droplets with λ of 2 and those of 10 was not statistically significant (P-value = 0.711, student’s two-tailed T-test). In terms of community composition, a subset of the λ = 10 droplets appeared to group together in the cluster dominated by droplets with SM, but no obvious patterns emerged in the other cluster dominated by droplets with BHI (Fig. 3). Additionally, statistical testing showed that there were no strong associations between individual OTUs and either group of droplets with a specific λ (Supplementary Table 2). There are several possible reasons for this somewhat unexpected result. First, it is important to note that even when the λ value is 2 in the Poisson distribution, there are still droplets that each contain a relatively large number of cells (e.g. droplets containing 6 or more cells per droplet exist at a frequency of over 1.6% theoretically). Second, the droplets generated in this study were non-uniform, resulting in higher variation in the distribution of cell numbers than we would expect theoretically; in particular, larger droplets had more cells per droplet than expected and smaller droplets had less. Finally, our criterion for selecting droplets heavily favored those with strong bacterial co-growth. All these factors combined likely have led to the observed high diversity of the analyzed droplets, even those from the experiment with a low λ value of 2. It is also important to note that the incubation time was 1 week, biasing our analysis towards sub-communities that demonstrated co-growth by the end of the incubation time. Given a longer period, slower growing bacteria could have been enriched and included in the analysis as well. Microfluidic droplet cell encapsulation provides accessibility to microbial ‘dark matter’ Many genera observed in the droplet sub-communities are representative gut commensals. For example, representatives of the glycan-degrading Bacteroides and Prevotella [49–51], Akkermansia [52, 53], and lactic acid bacteria [54, 55] have established functional roles in host health that are relatively well understood as a result of culturing-dependent experimentation. However, commensals of the Neisseriaceae [56] and Clostridiales [57] are present in droplet sub-communities and have low sequence identities to NCBI 16S representatives (Supplementary Table 1). In addition to many of these members being largely uncharacterized, many of them are low-abundance members. To demonstrate this, we compared the amplicons in the droplet sub-communities with those in the bulk fecal sample from which our microfluidic droplet sub-communities were derived (Supplementary Table 3). A small fraction (15/44) of the OTUs in droplet sub-communities were detected in the bulk community, and 9 OTUs of those were represented at an abundance lower than 0.1% in the bulk sample (Supplementary Table 4). This result demonstrated the ability of microfluidic droplets to encapsulate and co-cultivate representatives at very low abundances in microbial communities. Metagenomic reconstruction of a single droplet sub-community We performed metagenomic shotgun sequencing, assembly, and binning on an individual droplet sub-community of interest (B2–2) (Table 1). B2–2 was chosen for further analysis due to the presence of lactic acid bacteria and several phyla that have been characterized to a very limited extent. From 16S V4 amplicon sequencing, 13 OTUs were detected in B2–2. Out of 13 metagenomically assembled genomes (MAGs), 6 were of high quality (completeness above 75% and contamination below 10%) and were recovered, 6 were of lower quality, and 1 (Alistipes) was missing. Additionally, it is interesting to note that in the metagenomic library, a partial Lactobacillus genome was recovered despite Lactobacillus not being detected in the 16S library of this droplet sub-community. We also report a high degree of contamination by host DNA: 85.5% (347 million out of 406 million pair-end reads) of metagenomic reads mapped onto the human genome, resulting in a small fraction of sequencing effort contributing toward recovery of the bacterial community. Nevertheless, even from this relatively small fraction of sequencing reads, a substantial portion of the droplet sub-community, including several nearly complete genomes, was recovered. Table 1 Assembly statistics of genome bins recovered from the bacterial sub-community in droplet B2–2 and their inferred taxonomies from CheckM and NCBI BLASTN. 16S V4 OTU identity . Genome identity . Length (Mb) . # of contigs . GC% . Completeness (%) . Contamination (%) . Propionibacterium (100) Propionibacterium acnes 2.8 70 60 99.35 2.01 Corynebacterium (100) Corynebacterium sp. 2.4 54 58.7 98.46 0.37 Actinomycetales unclassified (100) Lawsonella sp. 1.6 25 52.4 95 0 Staphylococcus (100) Staphylococcus sp.a 2.6 95 32.7 94.76 14.93 Lactococcus (100) Lactococcus lactisa 2.1 198 35.8 85.72 13.26 Neisseriaceae unclassified (99) Neisseriaceae nov. gen. nov. sp. 1.7 127 42.3 76.05 8.56 Streptococcus (100) Streptococcus sp. 3 115 39.7 97.72 81.63 Enterobacteriaceae unclassified (100) Escherichia sp. 3.9 249 50.9 69.17 2.96 N/A Lactobacillus sp. 1.2 124 46.5 36.21 0 Enterobacteriaceae unclassified (99) Klebsiella sp. 2.3 171 55.8 30.7 0 Chryseobacterium (100) Chryseobacterium sp. (100) 1.5 143 38.4 25.16 5.17 Bacteroides (100) Bacteroides sp. (100) 2.4 283 43.6 22.41 3.45 Bacteroides (100) Bacteroides sp. (100) 1 119 38.3 8.88 1.75 Alistipes sp. Not detected in metagenomic dataset 16S V4 OTU identity . Genome identity . Length (Mb) . # of contigs . GC% . Completeness (%) . Contamination (%) . Propionibacterium (100) Propionibacterium acnes 2.8 70 60 99.35 2.01 Corynebacterium (100) Corynebacterium sp. 2.4 54 58.7 98.46 0.37 Actinomycetales unclassified (100) Lawsonella sp. 1.6 25 52.4 95 0 Staphylococcus (100) Staphylococcus sp.a 2.6 95 32.7 94.76 14.93 Lactococcus (100) Lactococcus lactisa 2.1 198 35.8 85.72 13.26 Neisseriaceae unclassified (99) Neisseriaceae nov. gen. nov. sp. 1.7 127 42.3 76.05 8.56 Streptococcus (100) Streptococcus sp. 3 115 39.7 97.72 81.63 Enterobacteriaceae unclassified (100) Escherichia sp. 3.9 249 50.9 69.17 2.96 N/A Lactobacillus sp. 1.2 124 46.5 36.21 0 Enterobacteriaceae unclassified (99) Klebsiella sp. 2.3 171 55.8 30.7 0 Chryseobacterium (100) Chryseobacterium sp. (100) 1.5 143 38.4 25.16 5.17 Bacteroides (100) Bacteroides sp. (100) 2.4 283 43.6 22.41 3.45 Bacteroides (100) Bacteroides sp. (100) 1 119 38.3 8.88 1.75 Alistipes sp. Not detected in metagenomic dataset aContamination values are below 10% for alternative marker gene set [40] Open in new tab Table 1 Assembly statistics of genome bins recovered from the bacterial sub-community in droplet B2–2 and their inferred taxonomies from CheckM and NCBI BLASTN. 16S V4 OTU identity . Genome identity . Length (Mb) . # of contigs . GC% . Completeness (%) . Contamination (%) . Propionibacterium (100) Propionibacterium acnes 2.8 70 60 99.35 2.01 Corynebacterium (100) Corynebacterium sp. 2.4 54 58.7 98.46 0.37 Actinomycetales unclassified (100) Lawsonella sp. 1.6 25 52.4 95 0 Staphylococcus (100) Staphylococcus sp.a 2.6 95 32.7 94.76 14.93 Lactococcus (100) Lactococcus lactisa 2.1 198 35.8 85.72 13.26 Neisseriaceae unclassified (99) Neisseriaceae nov. gen. nov. sp. 1.7 127 42.3 76.05 8.56 Streptococcus (100) Streptococcus sp. 3 115 39.7 97.72 81.63 Enterobacteriaceae unclassified (100) Escherichia sp. 3.9 249 50.9 69.17 2.96 N/A Lactobacillus sp. 1.2 124 46.5 36.21 0 Enterobacteriaceae unclassified (99) Klebsiella sp. 2.3 171 55.8 30.7 0 Chryseobacterium (100) Chryseobacterium sp. (100) 1.5 143 38.4 25.16 5.17 Bacteroides (100) Bacteroides sp. (100) 2.4 283 43.6 22.41 3.45 Bacteroides (100) Bacteroides sp. (100) 1 119 38.3 8.88 1.75 Alistipes sp. Not detected in metagenomic dataset 16S V4 OTU identity . Genome identity . Length (Mb) . # of contigs . GC% . Completeness (%) . Contamination (%) . Propionibacterium (100) Propionibacterium acnes 2.8 70 60 99.35 2.01 Corynebacterium (100) Corynebacterium sp. 2.4 54 58.7 98.46 0.37 Actinomycetales unclassified (100) Lawsonella sp. 1.6 25 52.4 95 0 Staphylococcus (100) Staphylococcus sp.a 2.6 95 32.7 94.76 14.93 Lactococcus (100) Lactococcus lactisa 2.1 198 35.8 85.72 13.26 Neisseriaceae unclassified (99) Neisseriaceae nov. gen. nov. sp. 1.7 127 42.3 76.05 8.56 Streptococcus (100) Streptococcus sp. 3 115 39.7 97.72 81.63 Enterobacteriaceae unclassified (100) Escherichia sp. 3.9 249 50.9 69.17 2.96 N/A Lactobacillus sp. 1.2 124 46.5 36.21 0 Enterobacteriaceae unclassified (99) Klebsiella sp. 2.3 171 55.8 30.7 0 Chryseobacterium (100) Chryseobacterium sp. (100) 1.5 143 38.4 25.16 5.17 Bacteroides (100) Bacteroides sp. (100) 2.4 283 43.6 22.41 3.45 Bacteroides (100) Bacteroides sp. (100) 1 119 38.3 8.88 1.75 Alistipes sp. Not detected in metagenomic dataset aContamination values are below 10% for alternative marker gene set [40] Open in new tab Droplet co-cultivation allows for cultivation and study of novel functional and phylogenetic diversity within the Neisseriaceae From droplet B2–2, we recovered the genome sequence of an uncultivated member of the Neisseriaceae. Several lines of evidence indicate that this genome represents a novel genus within the family Neisseriaceae. BLASTN results against the NCBI 16S rRNA database signify the closest phylogenetic match as Snodgrassella alvi with 93% identity (Supplementary Table 1, OTU21). By phylogenetic analysis of a comprehensive set of conserved bacterial proteins [58], this member clearly falls within the Neisseriaceae but is quite distinct from other representatives (Fig. 4a). Figure 4 Open in new tabDownload slide Phylogenetic and metabolic description of a novel member Neisseriaceae observed in droplet B2–2. (a) Phylogenetic tree comparing conserved protein sequences between the recovered genome with other Betaproteobacteria genomes from IMG. (b) Metabolic reconstruction of the most distinctive pathways. Figure 4 Open in new tabDownload slide Phylogenetic and metabolic description of a novel member Neisseriaceae observed in droplet B2–2. (a) Phylogenetic tree comparing conserved protein sequences between the recovered genome with other Betaproteobacteria genomes from IMG. (b) Metabolic reconstruction of the most distinctive pathways. As its phylogeny suggests, the functional diversity of this novel representative of the Neisseriaceae is also distinct from other members of this family and the gut microbiome as a whole. Key metabolic characteristics from pathway reconstruction include a complete fatty acid oxidation pathway, acetate fermentation from pyruvate via acetyl-CoA, complete glyoxylate cycle, and carnitine degradation (Fig. 4b). Fatty acids are primarily absorbed in the small intestine; however, under high fat diets, a significant fraction of fatty acids reach the large intestine and affect the microbiome and host health [59–61]. Noting that the genome of this member was not completely recovered, it is still worth pointing out that the TCA cycle, among the best characterized and documented pathways, was not detected. This along with the presence of the acetate fermentation pathway suggests an anaerobic lifestyle, like many other members of the distal gut microbiota. In addition, the pathway for the conversion of carnitine, a nutrient in red meat, to γ-butyrobetaine (γBB) was present. γBB is an important intermediate for the microbe-dependent conversion to trimethylamine (TMA) [62, 63], which is eventually converted to trimethylamine-N-oxide (TMAO). TMAO has been demonstrated to increase the risk to cardiovascular diseases [64, 65], and the utilization of carnitine by gut microbiota has been demonstrated to play a crucial role in accelerating atherosclerosis in mouse models and potentially in human hosts as well [63]. DISCUSSION Recognizing the fact that bacterial species rarely live in isolation, we utilized microfluidic techniques and metagenomic approaches to dissect, cultivate, screen, and analyze complex communities comprised of uncharacterized bacteria. Rare and uncharacterized members of the microbiome were analyzed with this methodology, including a novel Neisseriaceae with metabolic characteristics potentially important for human host health. The detection of this novel representative, which has complete fatty acid oxidation and carnitine degradation pathways, suggests possible niches that members of the Neisseriaceae occupy in the gut. This study is also the first to apply whole genome amplification, sequencing, and state-of-the-art bioinformatic tools to reconstruct and resolve multiple genomes from a co-cultivated sub-community in a microfluidic droplet. This approach can be used to address questions in a variety of natural systems where bacteria cannot be studied with traditional cultivation methods, such as those that have slow-growing or fastidious life styles. We would like to note that the improved genome recovery of certain MAGs (up to 99% completeness) in this study compared with previous studies using MDA to amplify single cell genomes, which recovered an estimated 70–78% of the total genome [66, 67]. This improvement is enabled by droplet co-cultivation, which allows growth-mediated ‘amplification’ from a small number of single cells to populations, thereby increasing the DNA material and aiding the recovery of nearly complete genomes. While other microfluidic studies to recover genomes from complex microbiomes were able to achieve a completeness of 90% or higher, they required multiple droplets for sequencing and an amalgamation of contigs from the sequencing of multiple droplets [19, 20]. In contrast, all the draft genomes reconstructed in this study were from sequencing a single droplet. As much potential as this approach has, there are technical limitations. Because whole genome amplification, demonstrated here with MDA, introduces significant stochastic amplification bias, quantitative insights regarding community composition are lost. Alternative 16S amplification methods, which better preserve quantitative community composition, could be employed in parallel with MDA-enabled metagenomic sequencing. For example, touchdown polymerase chain reaction (TD-PCR) has been used to study low-biomass microbial systems [68] and could be applied to 16S amplicon sequencing of individual microfluidic droplets. In addition, although bacterial contamination was not present in appropriate co-cultivation controls (Supplementary Fig. 2), contamination during metagenomic sequencing and genomic recovery of droplet sub-communities was present, which is common for amplification from low-biomass samples. In addition to MDA indiscriminately amplifying host DNA from the human fecal sample, highly pervasive OTUs, such as Staphylococcus, Propionibacterium, Corynebacterium, and other members of Actinomycetales, are more typically skin commensals [69] and are suspicious of being molecular contamination as well. In the future, more thorough washing steps to reduce nucleic acid contamination from microbial suspensions and extensive sequencing controls of reagents should be performed as well. One crucial aspect of the workflow demonstrated in this work is how co-cultivated sub-communities in individual droplets are selected for further sequencing and analysis. We carried out manual selection using a previously developed microfluidic device for droplet spacing [27, 28]. Future enhancement could leverage automated droplet sorting [70, 71] to increase the throughput and robustness of the technology pipeline. It is also important to note that we selected droplets with high cell densities hypothesizing that a high biomass yield may be the result of positive interactions between bacteria, which are prevalent in natural systems [72] and have been shown to occur effectively, for instance in the form of metabolic cross-feeding, in the microfluidic droplet environment [47, 73]. However, due to the complexity of the growth media employed in this work, inferring what ecological mechanisms led to enhanced co-growth is difficult. In addition, most of the analyzed droplet sub-communities comprised of a relatively large number of species, in part due to our selection criteria for high total biomass after co-cultivation. Communities with a higher diversity corresponding to a higher biomass productivity are a typical ecological phenomenon [74]. Possible mechanisms for higher productivity increasing with species richness in these droplets include higher overall usage of droplet resources through niche complementarity, facilitation through cross-feeding, and higher chances of including faster-growing members that grow abundantly regardless of surrounding microbial partners. While the methodology demonstrates potential for microbial co-cultivation in droplets and the resolution of analysis, the specific experimental design is limited in capacity to determine mechanisms for co-growth. With more specific questions, metagenomic analyses of co-cultivated sub-communities in microfluidic droplets have the capability to shed light on the ecology of many uncharacterized microbial systems. In particular, cross-feeding is believed to be widespread throughout complex bacterial communities [72, 75], but most demonstrated examples of this are in highly simplified synthetic communities [8, 76, 77]. Such ecological interactions can be studied not just for simple synthetic systems, but sub-communities of naturally occurring complex microbiomes, including those which have eluded cultivation-based efforts. Possible applications of this methodology can especially be applied to less complex, yet undefined, natural communities. Dissecting and co-cultivating these communities in droplets in defined media can elucidate complementarity of genomic pathways across species which leads to co-growth. Further development of single droplet resolution applied to bacterial transcriptomics and metabolomics would provide the means to elucidate mechanisms for even more complex systems. ACKNOWLEDGEMENTS We thank Dr Patrick Schloss for providing 16S V45 amplicon sequences derived from the bulk fecal sample as well as colleagues at the University of Michigan Host Microbiome Initiative for their work in sample collection. We thank Clarisse Betancourt and other staff at the University of Michigan Microbial Systems Sequencing Core for assistance with 16S library preparation and sequencing. We thank the staff at the University of Michigan Sequencing Core for assistance with metagenomic shotgun library preparation and sequencing. We would also like to thank members from the Geomicrobiology Laboratory under Dr Gregory Dick, particularly Sharon Grim and Robert Hein, for technical assistance with metagenomic analysis on the high-performance computing cluster and on the University of Michigan Department of Earth and Environmental Sciences Cayman server. FUNDING This work was funded in part by the NIH (3-R21-HG-005077-02-S1) and the University of Michigan MCubed Program. S.W. was partially supported by a predoctoral fellowship from the University of Michigan Cellular Biotechnology Training Program (T32GM008353). CONFLICT OF INTEREST The authors declare no conflict of interest. References 1. Beja O , Aravind L, Koonin EV et al. Bacterial rhodopsin: Evidence for a new type of phototrophy in the sea . Science 2000 ; 289 : 1902 – 6 . Google Scholar Crossref Search ADS PubMed WorldCat 2. Vanwonterghem I , Evans PN, Parks DH et al. Methylotrophic methanogenesis discovered in the archaeal phylum Verstraetearchaeota . Nat Microbiol 2016 ; 1 : 16170 . Google Scholar Crossref Search ADS PubMed WorldCat 3. Dick G . Genomic Approaches in Earth and Environmental Sciences , 1st edn. Hoboken, New Jersey: John Wiley & Sons, Inc. , 2019 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 4. Wang J , Qin J, Li Y et al. A metagenome-wide association study of gut microbiota in type 2 diabetes . Nature 2012 ; 490 : 55 – 60 . Google Scholar Crossref Search ADS PubMed WorldCat 5. Borrel G , McCann A, Deane J et al. Genomics and metagenomics of trimethylamine-utilizing archaea in the human gut microbiome . ISME J 2017 ; 11 : 2059 – 74 . Google Scholar Crossref Search ADS PubMed WorldCat 6. Crits-Christoph A , Diamond S, Butterfield CN et al. Novel soil bacteria possess diverse genes for secondary metabolite biosynthesis . Nature 2018 ; 558 : 440 – 4 . Google Scholar Crossref Search ADS PubMed WorldCat 7. Oremland RS , Capone DG, Stolz JF et al. Whither or wither geomicrobiology in the era of community metagenomics . Nat Rev Microbiol 2005 ; 3 : 572 – 8 . Google Scholar Crossref Search ADS PubMed WorldCat 8. Ponomarova O , Sevin D, Mülleder M et al. Yeast creates a stable niche for symbiotic lactic acid bacteria through nitrogen overflow . Cell Syst 2017 ; 5 : 345 – 57 . Google Scholar Crossref Search ADS PubMed WorldCat 9. Desai MS , Seekatz AM, Koropatkin NM et al. A dietary Fiber-deprived gut microbiota degrades the colonic mucus barrier and enhances pathogen susceptibility . Cell 2016 ; 167 : 1339 , e21– 53 . Google Scholar Crossref Search ADS PubMed WorldCat 10. Madsen EL . Identifying microorganisms responsible for ecologically significant biogeochemical processes . Nat Rev Microbiol 2005 ; 3 : 439 – 46 . Google Scholar Crossref Search ADS PubMed WorldCat 11. Lagier J-C , Khelaifia S, Alou MT et al. Culture of previously uncultured members of the human gut microbiota by culturomics . Nat Microbiol 2016 ; 1 : 16203 . Google Scholar Crossref Search ADS PubMed WorldCat 12. Cross KL , Campbell JH, Balachandran M et al. Targeted isolation and cultivation of uncultivated bacteria by reverse genomics . Nat Biotechnol 2019 ; 37 : 1314 – 21 . Google Scholar Crossref Search ADS PubMed WorldCat 13. Browne HP , Forster SC, Anonye BO et al. Culturing of ‘unculturable’ human microbiota reveals novel taxa and extensive sporulation . Nature 2016 ; 533 : 543 – 6 . Google Scholar Crossref Search ADS PubMed WorldCat 14. Stewart EJ . Growing unculturable bacteria . J Bacteriol 2012 ; 194 : 4151 – 60 . Google Scholar Crossref Search ADS PubMed WorldCat 15. McInerney MJ , Struchtemeyer CG, Sieber J et al. Physiology, ecology, phylogeny, and genomics of microorganisms capable of syntrophic metabolism . Ann N Y Acad Sci 2008 ; 1125 : 58 – 72 . Google Scholar Crossref Search ADS PubMed WorldCat 16. Pulschen AA , Bendia AG, Fricker AD et al. Isolation of uncultured bacteria from Antarctica using long incubation periods and low nutritional media . Front Microbiol 2017 ; 8 : 1346 . Google Scholar Crossref Search ADS PubMed WorldCat 17. Chen F , Zhan Y, Geng T et al. Chemical transfection of cells in picoliter aqueous droplets in fluorocarbon oil . Anal Chem 2011 ; 83 : 8816 – 20 . Google Scholar Crossref Search ADS PubMed WorldCat 18. Zhan Y , Wang J, Bao N et al. Electroporation of cells in microfluidic droplets . Anal Chem 2009 ; 81 : 2027 – 31 . Google Scholar Crossref Search ADS PubMed WorldCat 19. McLean JS , Lombardo M-J, Badger JH et al. Candidate phylum TM6 genome recovered from a hospital sink biofilm provides genomic insights into this uncultivated phylum . Proc Natl Acad Sci USA 2013 ; 110 : E2390 – 9 . Google Scholar Crossref Search ADS PubMed WorldCat 20. Dodsworth JA , Blainey PC, Murugapiran SK et al. Single-cell and metagenomic analyses indicate a fermentative and saccharolytic lifestyle for members of the OP9 lineage . Nat Commun 2013 ; 4 : 1854 . Google Scholar Crossref Search ADS PubMed WorldCat 21. Ma L , Kim J, Hatzenpichler R et al. Gene-targeted microfluidic cultivation validated by isolation of a gut bacterium listed in human microbiome Project’s most wanted taxa . Proc Natl Acad Sci USA 2014 ; 111 : 9768 – 73 . Google Scholar Crossref Search ADS PubMed WorldCat 22. Berghuis BA , Yu FB, Schulz F et al. Hydrogenotrophic methanogenesis in archaeal phylum Verstraetearchaeota reveals the shared ancestry of all methanogens . Proc Natl Acad Sci USA 2019 ; 116 : 5037 – 44 . Google Scholar Crossref Search ADS PubMed WorldCat 23. Yu FB , Blainey PC, Schulz F et al. Microfluidic-based mini-metagenomics enables discovery of novel microbial lineages from complex environmental samples . Elife 2017 ; 6 : e26580 . Google Scholar Crossref Search ADS PubMed WorldCat 24. Kehe J , Kulesa A, Ortiz A et al. Massively parallel screening of synthetic microbial communities . Proc Natl Acad Sci USA 2019 ; 116 : 12804 – 9 . Google Scholar Crossref Search ADS PubMed WorldCat 25. Hsu RH , Clark RL, Tan JW et al. Rapid microbial interaction network inference in microfluidic droplets . Cell Syst 2019 ; 9 : 229 , e4– 42 . Google Scholar Crossref Search ADS PubMed WorldCat 26. Carruthers DN , Byun CK, Cardinale BJ et al. Demonstration of transgressive overyielding of algal mixed cultures in microdroplets . Integr Biol 2017 ; 9 : 687 – 94 . Google Scholar Crossref Search ADS WorldCat 27. Wang S , Rossion M, Park J, Burns MA, Lin XN. Multiple displacement amplification of single droplet-cultivated microbes . (manuscript in preparation) , 2020 . 28. Wang S . Characterizing microbial communities using droplet microfluidic technology . [dissertation] . Ann Arbor, MI , University of Michigan , 2017 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 29. Drasar BS . Cultivation of anaerobic intestinal bacteria . J Pathol Bacteriol 1967 ; 94 : 417 – 27 . Google Scholar Crossref Search ADS PubMed WorldCat 30. Schaedler RW , Dubos R, Costello R. The development of the bacterial flora in the gastrointestinal tract of mice . J Exp Med 1958 ; 122 : 59 – 66 . Google Scholar Crossref Search ADS WorldCat 31. Starr SE , Killgore GE, Dowell VR. Comparison of Schaedler agar and trypticase soy-yeast extract agar for the cultivation of anaerobic bacteria . Appl Microbiol 1971 ; 22 : 655 – 8 . Google Scholar Crossref Search ADS PubMed WorldCat 32. Kozich JJ , Westcott SL, Baxter NT et al. Development of a dual-index sequencing strategy and curation pipeline for analyzing amplicon sequence data on the miseq illumina sequencing platform . Appl Environ Microbiol 2013 ; 79 : 5112 – 20 . Google Scholar Crossref Search ADS PubMed WorldCat 33. Schloss PD , Westcott SL, Ryabin T et al. Introducing mothur: open-source, platform-independent, community-supported software for describing and comparing microbial communities . Appl Environ Microbiol 2009 ; 75 : 7537 – 41 . Google Scholar Crossref Search ADS PubMed WorldCat 34. Cole JR , Wang Q, Fish JA et al. Ribosomal database project: data and tools for high throughput rRNA analysis . Nucleic Acids Res 2014 ; 42 :D633–42. Google Scholar OpenURL Placeholder Text WorldCat 35. Via M , Gignoux C, Burchard EG. The 1000 genomes project: new opportunities for research and social challenges . Genome Med 2010 ; 2 :3. Google Scholar OpenURL Placeholder Text WorldCat 36. Nurk S , Meleshko D, Korobeynikov A et al. MetaSPAdes: a new versatile metagenomic assembler . Genome Res 2017 ; 27 : 824 – 34 . Google Scholar Crossref Search ADS PubMed WorldCat 37. Gurevich A , Saveliev V, Vyahhi N et al. QUAST: quality assessment tool for genome assemblies . Bioinformatics 2013 ; 29 : 1072 – 5 . Google Scholar Crossref Search ADS PubMed WorldCat 38. Eren AM , Esen ÖC, Quince C et al. Anvi’o: an advanced analysis and visualization platform for ‘omics data . PeerJ 2015 ; 3 : e1319 . Google Scholar Crossref Search ADS PubMed WorldCat 39. Kim D , Song L, Breitwieser FP et al. Centrifuge: rapid and sensitive classification of metagenomic sequences . Genome Res 2016 ; 26 : 1721 – 9 . Google Scholar Crossref Search ADS PubMed WorldCat 40. Campbell JH , O’Donoghue P, Campbell AG et al. UGA is an additional glycine codon in uncultured SR1 bacteria from the human microbiota . Proc Natl Acad Sci USA 2013 ; 110 : 5540 – 5 . Google Scholar Crossref Search ADS PubMed WorldCat 41. Parks DH , Imelfort M, Skennerton CT et al. CheckM: assessing the quality of microbial genomes recovered from isolates, single cells, and metagenomes . Genome Res 2015 ; 25 : 1043 – 55 . Google Scholar Crossref Search ADS PubMed WorldCat 42. Aziz RK , Bartels D, Best A et al. The RAST server: rapid annotations using subsystems technology . BMC Genomics 2008 ; 9 :75. Google Scholar OpenURL Placeholder Text WorldCat 43. Karp PD , Latendresse M, Caspi R. The pathway tools pathway prediction algorithm . Stand Genomic Sci 2011 ; 5 : 424 – 9 . Google Scholar Crossref Search ADS PubMed WorldCat 44. Karp PD , Paley S, Romero P. The pathway tools software . Bioinformatics 2002 ; 18 : S225 – 32 . Google Scholar Crossref Search ADS PubMed WorldCat 45. Caspi R . MetaCyc: a multiorganism database of metabolic pathways and enzymes . Nucleic Acids Res 2006 ; 34 : D511 – 6 . Google Scholar Crossref Search ADS PubMed WorldCat 46. Segata N , Börnigen D, Morgan XC et al. PhyloPhlAn is a new method for improved phylogenetic and taxonomic placement of microbes . Nat Commun 2013 ; 4 :2304. Google Scholar OpenURL Placeholder Text WorldCat 47. Park J , Kerner A, Burns MA et al. Microdroplet-enabled highly parallel co-cultivation of microbial communities . PLoS One 2011 ; 6 :e17019. Google Scholar OpenURL Placeholder Text WorldCat 48. Abulencia CB , Wyborski DL, Garcia JA et al. Environmental whole-genome amplification to access microbial populations in contaminated sediments . Appl Environ Microbiol 2006 ; 72 : 3291 – 301 . Google Scholar Crossref Search ADS PubMed WorldCat 49. Wexler AG , Goodman AL, Voigt CA et al. An insider’s perspective: bacteroides as a window into the microbiome . Nat Microbiol 2017 ; 2 : 17026 . Google Scholar Crossref Search ADS PubMed WorldCat 50. Koropatkin NM , Cameron EA, Martens EC. How glycan metabolism shapes the human gut microbiota . Nat Rev Microbiol 2012 ; 10 :323–35. Google Scholar OpenURL Placeholder Text WorldCat 51. El KA , Armougom F, Gordon JI et al. The abundance and variety of carbohydrate-active enzymes in the human gut microbiota . Nat Rev Microbiol 2013 ; 11 : 497 – 504 . Google Scholar Crossref Search ADS PubMed WorldCat 52. Everard A , Belzer C, Geurts L et al. Cross-talk between Akkermansia muciniphila and intestinal epithelium controls diet-induced obesity . Proc Natl Acad Sci USA 2013 ; 110 : 9066 – 71 . Google Scholar Crossref Search ADS PubMed WorldCat 53. Belzer C , Chia LW, Aalvink S et al. Microbial metabolic networks at the mucus layer lead to diet-independent butyrate and vitamin B12 production by intestinal symbionts . MBio 2017 ; 8 :e00770–17. Google Scholar OpenURL Placeholder Text WorldCat 54. Ventura M , O’Flaherty S, Claesson MJ et al. Genome-scale analyses of health-promoting bacteria: probiogenomics . Nat Rev Microbiol 2009 ; 7 : 61 – 71 . Google Scholar Crossref Search ADS PubMed WorldCat 55. Hotel AC , Cordoba A. Health and nutritional properties of probiotics in food including powder milk with live lactic acid bacteria . Prevention 2001 ; 5 : 1 – 10 . 56. Shreiner AB , Kao JY, Young VB. The gut microbiome in health and in disease . Curr Opin Gastroenterol 2015 ; 31 : 69 – 75 . Google Scholar Crossref Search ADS PubMed WorldCat 57. Lopetuso LR , Scaldaferri F, Petito V et al. Commensal clostridia: leading players in the maintenance of gut homeostasis . Gut Pathog 2013 ; 5 : 23 . Google Scholar Crossref Search ADS PubMed WorldCat 58. Segata N , Waldron L, Ballarini A et al. Metagenomic microbial community profiling using unique clade-specific marker genes . Nat Methods 2012 ; 9 : 811 – 4 . Google Scholar Crossref Search ADS PubMed WorldCat 59. Iqbal J , Hussain MM. Intestinal lipid absorption . Am J Physiol Metab 2009 ; 296 : E1183 – 94 . Google Scholar OpenURL Placeholder Text WorldCat 60. Daniel H , Gholami AM, Berry D et al. High-fat diet alters gut microbiota physiology in mice . ISME J 2014 ; 8 : 295 – 308 . Google Scholar Crossref Search ADS PubMed WorldCat 61. Hildebrandt MA , Hoffmann C, SA S–M et al. High-fat diet determines the composition of the murine gut microbiome independently of obesity . Gastroenterology 2009 ; 137 : 1716 – 1724.e2 . Google Scholar Crossref Search ADS PubMed WorldCat 62. Koeth RA , Levison BS, Culley MK et al. γ-Butyrobetaine is a proatherogenic intermediate in gut microbial metabolism of l-carnitine to TMAO . Cell Metab 2014 ; 20 : 799 – 812 . Google Scholar Crossref Search ADS PubMed WorldCat 63. Koeth RA , Wang Z, Levison BS et al. Intestinal microbiota metabolism of l-carnitine, a nutrient in red meat, promotes atherosclerosis . Nat Med 2013 ; 19 : 576 – 85 . Google Scholar Crossref Search ADS PubMed WorldCat 64. Romano KA , Vivas EI, Amador-noguez D et al. Intestinal microbiota composition modulates choline bioavailability . MBio 2015 ; 6 : 1 – 8 . Google Scholar Crossref Search ADS WorldCat 65. Wang Z , Klipfell E, Bennett BJ et al. Gut flora metabolism of phosphatidylcholine promotes cardiovascular disease . Nature 2011 ; 472 : 57 – 65 . Google Scholar Crossref Search ADS PubMed WorldCat 66. Kashtan N , Roggensack SE, Rodrigue S et al. Single-cell genomics reveals hundreds of coexisting subpopulations in wild Prochlorococcus . Science 2014 ; 344 : 416 – 20 . Google Scholar Crossref Search ADS PubMed WorldCat 67. Mußmann M , Hu FZ, Richter M et al. Insights into the genome of large sulfur bacteria revealed by analysis of single filaments . PLoS Biol 2007 ; 5 : 1923 – 37 . Google Scholar Crossref Search ADS WorldCat 68. Bassis CM , Erb-Downward JR, Dickson RP et al. Analysis of the upper respiratory tract microbiotas as the source of the lung and gastric microbiotas in healthy individuals . MBio 2015 ; 6 :e00037–15. Google Scholar OpenURL Placeholder Text WorldCat 69. Byrd AL , Belkaid Y, Segre JA. The human skin microbiome . Nat Rev Microbiol 2018 ; 16 : 143 – 55 . Google Scholar Crossref Search ADS PubMed WorldCat 70. Baret JC , Miller OJ, Taly V et al. Fluorescence-activated droplet sorting (FADS): efficient microfluidic cell sorting based on enzymatic activity . Lab Chip 2009 ; 9 : 1850 – 8 . Google Scholar Crossref Search ADS PubMed WorldCat 71. Cao Z , Chen F, Bao N et al. Droplet sorting based on the number of encapsulated particles using a solenoid valve . Lab Chip 2013 ; 13 : 171 – 8 . Google Scholar Crossref Search ADS PubMed WorldCat 72. Zelezniak A , Andrejev S, Ponomarova O et al. Metabolic dependencies drive species co-occurrence in diverse microbial communities . Proc Natl Acad Sci USA 2015 ; 112 : 6449 – 54 . Google Scholar Crossref Search ADS PubMed WorldCat 73. Saleski TE , Kerner AR, Chung MT et al. Syntrophic co-culture amplification of production phenotype for high-throughput screening of microbial strain libraries . Metab Eng 2019 ; 54 : 232 – 43 . Google Scholar Crossref Search ADS PubMed WorldCat 74. Mittelbach G . Community Ecology . Sunderland, MA: Sinauer Associates , 2012 , 46 – 50 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 75. Zengler K , Zaramela LS. The social network of microorganisms—how auxotrophies shape complex communities . Nat Rev Microbiol 2018 ; 16 : 383 – 90 . Google Scholar Crossref Search ADS PubMed WorldCat 76. Mee MT , Collins JJ, Church GM et al. Syntrophic exchange in synthetic microbial communities . Proc Natl Acad Sci USA 2014 ; 111 : E2149 – 56 . Google Scholar Crossref Search ADS PubMed WorldCat 77. Pande S , Merker H, Bohl K et al. Fitness and stability of obligate cross-feeding interactions that emerge upon gene loss in bacteria . ISME J 2014 ; 8 : 953 – 62 . Google Scholar Crossref Search ADS PubMed WorldCat Author notes James Y. Tan and Sida Wang contributed equally to this work. © The Author(s) 2020. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)
TI - Co-cultivation of microbial sub-communities in microfluidic droplets facilitates high-resolution genomic dissection of microbial ‘dark matter’
JF - Integrative Biology
DO - 10.1093/intbio/zyaa021
DA - 2020-11-18
UR - https://www.deepdyve.com/lp/oxford-university-press/co-cultivation-of-microbial-sub-communities-in-microfluidic-droplets-cn0fSZ07hx
SP - 263
EP - 274
VL - 12
IS - 11
DP - DeepDyve
ER -