Droplet microfluidics uses immiscible multiphase flows inside the microfluidic channels to generate and manipulate discrete droplets with volumes ranging from the nanoliter to the femtoliter (Teh et al. 2008). Each individual microfluidic droplet is separated by an immiscible phase and can be a vessel for reaction (Song et al. 2006). Microfluidic droplets have several advantages over conventional bioreactors, such as flasks, Petri dishes, and multi-well plates. The first is miniaturization: The droplets are confined to ultra-small volumes with the very efficient mass-heat transfer, low reagents cost, and significant surface effects (Kaminski et al. 2016). The second is high-throughput: By decreasing the volume of droplets to the ultra-small regime, many droplets can be generated and manipulated as reactors (Mashaghi et al. 2016). The third is monodispersity: the droplets are monodispersed at the diameters of nanometers to micrometers. This is important in chemical or biochemical reactions with uniform droplets (Guo et al. 2012). Effective and precise manipulation of droplets is critical for bioanalysis. There are various functional modules to manipulate and analyze the droplets, such as droplet generation, incubation, dividing, merging, and sorting (Shang et al. 2017). Droplet manipulation modules have been reviewed extensively and the most commonly used ones are shown in Table 1. Every single module can be used independently and multiple modules can also be integrated on a microfluidic chip for specific microbiology applications.
Manipulation Representative schematic Function and application in microbiology Generation Generate monodispersed droplets (Christopher and Anna 2007; Cramer et al. 2004; Thorsen et al. 2001): (1) microbial single-cell isolation (Martin et al. 2003); (2) massive single-cell encapsulation using droplets (Baret et al. 2009) Incubation Incubate droplets under suitable condition and time (Frenz et al. 2009; Köster et al. 2008): (1) microbial cell growth and enzyme expression on-chip or off-chip (Najah et al. 2014); (2) precisely control enzyme–substrate reaction time (Obexer et al. 2017) Dividing Divide the droplets into duplicated daughter droplets (Link et al. 2004): (1) multiplexed droplet generation for single-cell encapsulation (Abate and Weitz 2011); (2) dividing droplet arrays after cultivation droplets generation to create duplication of droplets library (Liu et al. 2009) Merging Merge droplets or add regents to droplets (Abate et al. 2010; Mazutis et al. 2009): (1) add substrate for controlled enzymatic reaction (Qiao et al. 2018); (2) co-cultivation (Guo et al. 2012); (3) virus–host interaction (Tao et al. 2015) Sorting Transport and subset the droplets of interest (Baret et al. 2009): (1) directed evolution of microbial strains or enzymes (Gielen et al. 2016); (2) microbial community screening (Najah et al. 2014)
Table 1. Functional modules of droplet manipulation and applications in microbiology
Surfactants are amphiphilic molecules that contain both hydrophilic and hydrophobic groups. In water–oil droplets emulsion, surfactant molecules absorb at the oil–water interface to prevent droplet coalescence (Fig. 2) (Baret 2012). The surfactant is crucial for droplets stability during droplet manipulation. Most droplet manipulations (except for droplet generation) are difficult without surfactants because of droplet coalescence. Thus, surfactants play a fundamental role in microfluidic droplet manipulation and are key to stabilizing droplets as a micro-bioreactor. Considerations in the selection of a surfactant include biocompatibility, stability, and molecular exchange. Biocompatibility is a prerequisite need because the enzyme activity and cell viability within the droplets should not be inhibited by the surfactant (Clausell-Tormos et al. 2008; Roach et al. 2005).
Figure 2. Schematic illustrates surfactant molecules to stabilize aqueous droplets. The amphiphilic molecules were added to the oil phase to maintain droplets as a stable micro-reactor
Stability is essential for droplets serving as bioreactors. In a droplet-based digital PCR reaction, the droplets must be stable after many cycles of PCR; the surfactant plays an important role in stability (Pinheiro et al. 2012). It is worth noting that the droplet is not a perfectly sealed container. Small molecules may be exchanged between closely located droplets over time. The proper selection of surfactant and oil can effectively suppress this process (Bai et al. 2010; Chowdhury et al. 2019; Gruner et al. 2016; Pan et al. 2014). The hydrophobic group of the surfactant molecule extends into the oil, and thus the oil phase is another important factor for the selection of surfactants. Hydrocarbon oils and fluorocarbon oils are two main types of continuous phase for w/o droplet emulsion. Of hydrocarbons, hexadecane is the most commonly used oil, and the corresponding surfactants are Span 80, Tween 20/80, etc. Fluorocarbon oils are widely used in cell culture, enzyme reaction and PCR because of their good gas permeability and the immiscibility of non-fluorinated molecules. Many fluorocarbons are commercially available, such as FC-40, FC-77, HFE 7500, and HFE 7100 from 3M. Many surfactants have been developed for fluorocarbon oil, such as Krytox (DuPont), DMP-PFPE (dimorpholinophosphinate perfluoropolyether), PEG-PFPE (block copolymer of polyethylene glycol and perfluoropolyether), LPG-PFPE (block copolymer of linear polyglycerol and perfluoropolyether), and fluorinated nanoparticles (F-NPs) (Baret 2012; Gruner et al. 2015; Holtze et al. 2008; Pan et al. 2014; Wagner et al. 2016). Of these, PEG-PFPE is the most commonly used surfactant because of its good biocompatibility and stabilization effect.
Detectors play an essential role in the application of droplets for microbiology. Different detection approaches allow the detection of droplets, either quantitatively or qualitatively. Many detection techniques can be incorporated for droplet analysis, such as imaging-based analysis (bright-field/fluorescence microscopy), laser-induced fluorescence, Raman spectroscopy, mass spectrometry, and absorption detection. These detectors all have distinct advantages and disadvantages (Table 2). In practical applications, it is important to select the appropriate detection method according to the intrinsic needs of the application.
Analysis tools Advantages Disadvantages Bright-field/fluorescence imaging (Girault et al. 2017; Pratt et al. 2019) Phenotypic identification; widely available microscopes; time-lapse analysis; parallel analysis of large-scale droplet array Low sensitivity; poor ability for molecules analysis Light absorption and scattering (Gielen et al. 2016) Cell density detection; label-free; high speed Low detection sensitivity Laser-induced fluorescence (Agresti et al. 2010) High analysis speed; high detection sensitivity; low equipment cost High dependence on fluorescent probes design; false-positive due to leakage of fluorescent probes Raman or infrared spectroscopy (Wang et al. 2017) Label-free; multiple parameter analysis long reading time compare with fluorescence; interference from matrices and cell status; high equipment cost Mass spectrometry (Holland-Moritz et al. 2020) High sensitivity; label-free; high molecular resolution Destructive for live cells; low throughput; high equipment cost
Table 2. A comparison of diferent droplet detection methods
Manipulation of droplets
Surfactants for microfluidic droplets
Detection of droplets
For this application, it is critical to obtain pure cultures of the uncultured microbes from the environment. Developing novel cultivation methods is an effective strategy for the isolation of uncultured microorganisms (Pham and Kim 2012). Because of the merit of single-cell, high-throughput, high-resolution, and low cost, etc., droplets have become a promising approach for the isolation of uncultured microbes (Huys and Raes 2018). Martin et al. (2003) first used monodispersed droplets for high-throughput microbial cultivation. They demonstrated that droplets were a suitable reactor for microbial cultivation starting from a single cell to a very high-cell density. Many novel droplet-based methods were later developed for microbial cultivation. For instance, ultra-small droplets, with a volume down to femtoliter volume, were generated and applied to bacterial quorum sensing. A single cell can also initiate the high-density behavior of quorum sensing (Boedicker et al. 2009). Here we introduce several promising droplet-based technologies for the isolation of uncultured microbes.
Targeted isolation for uncultured microbes: Fluorescence In Situ Hybridization (FISH) is a commonly used technique in microbiology with great specificity for microbial identification. A microfluidic device integrating FISH identification and droplets splitting modules has been developed for parallel high-throughput single-cell cultivation and identification (Fig. 3a) (Liu et al. 2009). First, 60 nl droplets were used to encapsulate a single cell and cultivate a mixture of Paenibacillus curdlanolyticus (P. curdlanolyticus, Pc196) and Escherichia coli. After cell growth, a droplet-splitting chip was used to split the droplets into two daughter droplets. One copy of the droplet was added to an agar plate and the other copy was used for FISH identification with a specific probe to Pc196. Finally, the droplet encapsulating the target species was selected from the agar plate, according to the results of FISH identification. This work presented an elegant strategy that combined droplets cultivation and genetic identification for targeted isolation of microbes. Other identification methods, such as 16S rRNA gene sequencing or matrix-assisted laser desorption/ionization–time of flight (MALDI–TOF), may also be incorporated into this strategy by splitting droplets to form duplicated copies of microbial populations (Liu et al. 2009). A similar strategy was adopted using a SlipChip device to split 2D droplet arrays for coupling high-throughput single-cell cultivation and 16rRNA PCR for quickly locating the target colonies. As an example, one of the most wanted taxa in the list of the Human Microbiome Project (HMP) was successfully isolated with the SlipChip (Ma et al. 2014). More recently, an integrated microfluidic device was developed, in which electrospray ionization mass spectrometry (ESI–MS) was used for microbial cell identification. Meanwhile, the study integrated microfluidic functional modules, including splitting, incubation, and sorting, into a single device, which they called mass activated droplet sorting (MADS) (Fig. 3b) (Holland-Moritz et al. 2020). Droplets were asymmetrically split into two daughter droplets: one was used for ESI–MS analysis and the target droplet could be selectively sorted according to the results of the ESI–MS data. This device has been successfully applied to the directed evolution of transaminase and also has great potential for uncultured microbe screening (Holland-Moritz et al. 2020).
Figure 3. Droplet-based devices for the cultivation of uncultured microbes. a Schematic illustrates the principle for parallel microbial cultivation and Fluorescence In Situ Hybridization (FISH) identification (Reprinted with permission from Liu et al. 2009. Lab Chip 9: 2153–2162. Copyright (2009) Royal Society of Chemistry). b Schematic of mass activated droplet sorting (MADS) system and the zoom-in microscopic images of the devices showing droplets operation (Reprinted with permission from Holland-Moritz et al. 2020. Angew Chem Int Edit 59: 4470–4477. Copyright (2020) John Wiley & Sons). c Schematic of Microfluidic streak plate (MSP) platform (Reprinted with permission from Chen et al. 2019. J Hazard Mater 366: 512–519. Copyright (2019) Elsevier). d Schematic of semi-automated droplet picker (Fig. 3D, and 3F reprinted with permission from Hu et al. 2020. Lab Chip 20: 363–372. Copyright (2020) Royal Society of Chemistry) e Left, heatmap of amplicon sequencing for comparison of the original community, pooled cells from agar plates, and pooled cells from MSP. Right, parallel comparison of isolated species and number of isolates for agar plate and MSP cultivations (Reprinted with permission from Jiang et al. 2016. Appl Environ Microb 82: 2210–2218. Copyright (2016) American Society for Microbiology). f MSP isolated microorganisms from deep-sea sediments samples
Microfluidic streak plate (MSP) cultivation: MSP is a simple method for high-throughput single-cell analysis and cultivation with nanoliter droplets (Jiang et al. 2016). This method uses microfluidic chips to generate nanoliter droplets, which can be arrayed on a standard Petri dish prefilled with the carrier oil (Fig. 3c). There are several distinguishing features of MSP compared to traditional agar plate cultivation: (ⅰ) it provides a much higher throughput than traditional agar plates with low consumption; (ⅱ) it avoids interspecific competition and eliminates biases due to differences in growth rates; (ⅲ) droplet storage on Petri dish prefilled with mineral oil allows long-term single-cell cultivation; (ⅳ) it allows microscopic observation during culture; (ⅴ) after cultivation, the droplets can be pooled for microbial community analysis using next-generation sequencing, or easily picked for scale-up cultivation; (ⅵ) liquid medium used by MSP avoids hydrogen peroxide production during gelation of agar, which is more favorable for the isolation of hydrogen peroxide-sensitive microbes (Tanaka et al. 2014).
The MSP method has been used to isolate several species with a high degradation efficiency of polycyclic aromatic hydrocarbons (PAH), including a previously unknown fluoranthene-degrading Blastococcus species (Fig. 3e). Later, Zhou et al. (2018) used the MSP platform under anaerobic conditions to separate several potentially new species and genera from the gut of Reticulitermes chinensis (a wood‐feeding termite). Recently, Chen et al. (2019) successfully incorporated chemotactic screening and MSP to isolate imidazolinone-degrading species from soil samples. The MSP platform was further improved for long-term cultivation to tackle the problem of rare marine microbe isolation (Hu et al. 2020). The improved MSP platform increased the recovery of slow-growing microbes and enabled long-term cultivation of up to five months. Furthermore, a semi-automated droplet picker was developed to facilitate the contamination-free recovery of the target droplets (Fig. 3d). During droplet picking, by comparison with the previous MSP platform, an inverted microscope was used to select droplets, based on morphology, and improved the isolation efficiency. The amplicon sequencing of the pooled cultures of MSP shows that MSP cultivation has a greater ability to reveal rare elements of the microbial biosphere than agar plate cultivation. With the improved MSP, about 15 potentially new species were successfully isolated from deep-sea sediments (Fig. 3f). Overall, these studies demonstrate that MSP can be applied to the microbial cultivation of samples from a variety of environments and can significantly improve the isolation efficiency and throughput of cultivation.
Microbial co-cultivation is a promising method for the discovery of natural microbial products and the isolation of uncultured microorganisms (He et al. 2015; Nai and Meyer 2018; Zhang and Wang 2016). The main challenge in microbiology is that most microbes cannot be obtained via pure culture in the laboratory. One reason is that the process of obtaining a pure culture may interrupt the physical and chemical interactions between microbes (Garcia 2016). Co-cultivation may offer an appropriate method to meet the communication needs of microbes. Many microfluidic-based co-cultivation methods have been developed, including nanochannel-based co-cultivation, membrane-based co-cultivation, and droplet-based co-cultivation (Burmeister et al. 2019; Kim et al. 2008; Park et al. 2011). The droplet-based co-cultivation system can simultaneously and rapidly analyze millions of microbes in droplets. Here we mainly describe the droplet-based co-cultivation system.
Park et al. (2011) were the first to report that microfluidic droplets could be used for microbial co-cultivation. Symbiotic bacterial strains were stochastically encapsulated in microfluidic droplets based on a Poisson distribution and each bacterial strain could not live alone (Fig. 4a). When two cross-feeding strains were encapsulated in the same droplet, they both can live and prospered. Over 1000 droplets were stored in the microfluidic chip for simultaneous analysis (Fig. 4b). This work demonstrated that microfluidic droplets could be applied to the massive co-cultivation of microbes. Cross-kingdom communication between yeast and bacteria was also explored with this strategy by Jarosz et al. (2014), who demonstrated that few bacterial cells could induce the metabolic transformation of yeast (Jarosz et al. 2014). This strategy is also used for microbial secondary metabolite screening.
Figure 4. Microfluidic droplet-based co-cultivation and its application. a The workflow of droplet co-cultivation. b The cultivation results of microbial droplets co-cultivation with W- and Y- Strains (Adapted from Park et al. 2011. PLOS ONE 6: e17019. followed the terms of the Creative Commons Attribution License). c Schematic of using a co-cultivation strategy for bear oral microbiome screening. d The microfluidic droplets with the encapsulation of Staphylococcus aureus (S. aureus) and oral microbes after cultivation (scale bar 50 μm) (Fig. 4C and 4D reprinted from Terekhov et al. 2018. Proc Natl Acad Sci USA 115:9551–9556. followed Creative Commons Attribution-NonCommercial-NoDerivatives License 4.0 (CC BY-NC-ND))
A so-called 'Syntrophic Co-culture Amplification of Production phenotype' (SnoCAP) strategy was established by Saleski et al. (2019), who used a cross-feeding screening strategy to screen the Escherichia coli (E. coli) mutagenesis library and isolated a high isobutanol production strain. In another report, the oral microbiome of a Siberian bear and Staphylococcus aureus were co-encapsulated into microfluidic droplets for high-throughput screening of microbes with antimicrobial activity against S. aureus (Fig. 4c, d) (Terekhov et al. 2018). A Bacillus strain with antimicrobial activity against S. aureus was successfully isolated, which demonstrated a promising potential for antibiotics discovery and microbial community analysis.
Bacterial multi-drug resistance is rapidly increasing and has become a global crisis of public health. The speed of new antibiotic discovery has slowed (Lewis 2013; Roca et al. 2015). One effective strategy is accurate and timely antibiotic prescriptions for patients, but these rely heavily on rapid antibiotic resistance evaluation (Laxminarayan et al. 2013). However, current gold-standard antibiotic susceptibility tests (ASTs) are time-consuming, heavily depend on bacterial enrichment in media with antibiotics, which may take a few days to undertake. One way to shorten the time of ASTs is to decrease the reaction volume for bacterial growth. Microfluidic-based methods are a promising tool for rapid ASTs because of their miniaturization and automation (Zhang et al. 2020). Various microfluidics-based methods have been developed for ASTs, such as microchamber, microchannel, and droplet-based tools (Hassan and Zhang 2020). Of these, droplet-based methods are commonly used because they generate large-scale droplets, for high-throughput and rapid analysis, and ultra-small volumes for highly sensitive detection.
An early droplet-based ASTs platform was developed by Boedicker et al. (2008). In this platform, 50 nl droplets were used to encapsulate cells, antibiotics, and medium. Teflon tubing was used for storing droplets (Fig. 5a). Multiple drugs and different drug concentrations could be tested within seven hours. An integrated microfluidic device was further developed by Kaushik et al. (2017) with picoliter droplets, including droplet generation, incubation, and detection (Fig. 5b). The ASTs could be performed within one hour by decreasing the volume of the droplet to 20 pl, but only one drug and fixed concentration could be tested per assay. This method is an endpoint detection, and the kinetic information during cell growth was lost. Another type of rapid ASTs is the stationary nanoliter droplet array reported by Shemesh et al. (2014) who used microwells to confine the droplets, cell growth could be successively monitored with microscopy. The drug gradient can be formed in this device, and the growth of the cells can be monitored with a microscope over time (Fig. 5c). We reported the generation of nanoliter droplets array with a linear drug gradient using the MSP technique (Jiang et al. 2016). The droplets were written onto a 9 cm petri dish following a spiral track. A linear antibiotic drug gradient could be generated in an array of over 2000 droplets to obtain a high-resolution dose–response curve to determine the antibiotic resistance pattern of the pathogen (Jiang et al. 2016).
Figure 5. Microfluidic droplet-based platform for rapid antibiotic susceptibility tests (ASTs). a Schematic of a droplet array-based gradient AST system. Methicillin-susceptible Staphylococcus aureus (MSSA) was tested with serial gradients of antibiotics include LVF (levofloxicin), OXA (oxicillin), VCM (vancomycin), and AMP (ampicillin) (Reprinted with permission from Boedicker et al. 2008. Lab Chip 8: 1265–1272. Copyright (2008) Royal Society of Chemistry). b An integrated microfluidic chip illustrates the principle and workflow for droplet-based high-resolution dose–response profiling (Reprinted with permission from Kaushik et al. 2017. Biosens Bioelectron 97: 260–266. Copyright (2017) Elsevier). c Stationary nanoliter droplets array fixed with microwell illustrates the workflow for AST (Reprinted with permission from Shemesh et al. 2014. Proc Natl Acad Sci USA 111: 11293–11298. Copyright (2014) National Academy of Sciences, USA.). d Schematic illustrates a chip-free platform of multichannel dynamic interfacial printing (MC-DIP) for AST (Reprinted with permission from Liao et al. 2017. ACS Appl Mater Interfaces 9: 43545−43552. Copyright (2017) American Chemical Society)
Several chip-free droplet generation methods have been applied to antibiotic susceptibility testing, which avoids the complex microfabrication process and high-cost of microfluidic devices. Liao et al. (2017) developed a platform, called multichannel dynamic interfacial printing (MC-DIP), which generates droplets via a vibrating capillary at the oil surface. The MC-DIP can combine a 96-well plate and microscopy for parallel and large-scale ASTs (Fig. 5d). Another example of single-cell resolution ASTs with a chip-free platform is performed using interfacial nanoinjection (INJ) and fluorescence-activated cell sorting (FACS) with flow cytometry (Yun et al. 2019). They used FACS to isolate single-cell in 384-well plates and perform nanoliter-scale ASTs of single microbial cells by displacing nanoliter reagents using INJ to be mixed with single cells. In FACS-INJ based ASTs, microbial cells were sorted into 384-well plates with FACS, and then 100 nl droplets with media, resazurin, and antibiotics were generated with the capillary; finally, the growth of the cells was successively monitored with a fluorescence reader. Compared with traditional plate assays (200 μl), FACS-INJ can provide statistical analysis of antibiotic resistance for a microbial population at single-cell resolution with much lower reagent consumption.
Enzymes are very important in biological, industrial, and environmental studies. Diverse bio-active compounds are produced by microbes and they are an important resource for enzyme discovery (Chiu and Stavrakis 2019). Microbial enzyme screening is a simple and efficient method for discovering enzymes. There are three main methods for high-throughput microbial enzyme screening (Table 3) (Autour and Ryckelynck 2017). The first is microtiter plates (MTPs) screening, which can achieve a screening throughput of up to 105 assays per day with liquid-handling robots or colony pickers robots (Autour and Ryckelynck 2017; Mayr and Bojanic 2009). The second is in vitro compartmentalization-based fluorescence-activated cell sorting (IVC-FACS), which takes advantage of the compartmentalization of microfluidic droplets and the ultra-high sorting speed of FACS to screen enzymes with a throughput up to 108 assays per hour (Griffiths and Tawfik 2006; Sciambi and Abate 2015). The third is microfluidic droplet-based screening, such as fluorescence-activated droplet sorting (FADS), Raman-activated droplet sorting (RADS), mass-activated droplet sorting (MADS) (Baret et al. 2009; Holland-Moritz et al. 2020; Wang et al. 2017). FADS can achieve a screening throughput of up to 108 assays per hour (Baret et al. 2009; Sciambi and Abate 2015) and is more commonly used than RADS and MADS. In this section, we mainly focus our discussion on FADS screening.
Screening parameters Microtiter plates (MTP) In vitro compartmentalization-based fluorescence-activated cell sorting (IVC-FACS) Fluorescence-activated droplet sorting (FADS) Throughput Up to 105/day Up to 108/hour Up to 108/hour Time Long Short Short Cost High Low Low Suitable for multi-step assays Yes No Yes
Table 3. Comparison of three high-throughput microbial enzyme screening methods
While the throughput of MTPs is much lower than IVC-FACS and FADS, the screening cost per assay is much higher. In comparison to FADS, IVC-FACS relies on the generation of a double emulsion, and the droplet content of IVC-FACS is difficult to modify after droplets generation (Griffiths and Tawfik 2006) (Table 3). After the droplet generation, the droplet content of FADS can be modified with droplets fusion. This is more effective in multi-step screening assays than in FACS. FADS uses microfluidic droplets as a reactor and the DEP force as a sorting force, laser-induced fluorescence is used for detection and analysis. The droplet volume used in FADS ranges from several picoliter to hundreds of picoliter. To date, FADS has successfully applied for the screening of β-galactosidase, horseradish peroxidase, sulfatase, cellulase, aldolase, lipase, and others (Agresti et al. 2010; Baret et al. 2009; Fenneteau et al. 2017; Kintses et al. 2012; Obexer et al. 2017; Ostafe et al. 2014; Qiao et al. 2018). The operation of FADS can be divided into the following steps (Baret et al. 2009):
1) Droplet generation. Microbial cells, medium, and fluorogenic enzyme substrate, were encapsulated within the droplets with microfluidic chip (if the fluorogenic substrate is not suitable to be incubated with microbial cells for a long time, the substrate should be added to droplets with droplet fusion device after droplet incubation).
2) Droplet incubation. Droplets were incubated off-chip to allow enzyme expression and secretion.
3) Re-loading droplets. The incubated droplets were re-injected into the microfluidic chip.
4) Droplet sorting. Droplet fluorescence was detected, and the positive droplets were selected for further experiments.
The selection of a proper fluorogenic enzyme substrate is extremely important in FADS screening. Two aspects need attention in the selection of a fluorogenic substrate. The first is that the substrate must be compatible with the droplet system and does not lead to undesired droplet coalescence. The second is that the substrate must be effectively retained within the droplets with limited droplet cross-contamination (Skhiri et al. 2012). Several methods have been developed to slow down the leakage and diffusion of small molecules, including fluorogenic substrates among droplets, such as modifying the fluorogenic substrates with higher hydrophilicity, improving the viscosity of the droplets by adding bovine serum albumin (BSA) or carboxymethyl cellulose (CMC) and using nanoparticles as a surfactant to avoid surfactant micelles transport of the substrate between droplets (Courtois et al. 2009; Fenneteau et al. 2017).
Obexer et al. (2017) developed a fully automated FADS device that integrated the modules of droplet generation, incubation and sorting on a single chip for directed evolution of a previously computationally designed enzyme into a highly active enzyme which requires screening of ~ 108 protein variants (Fig. 6). An integrated device was designed to realize high-throughput screening of the aldolases plasmid library in a fully automated manner. The device can sort up to 107 library within 2 h with few manual operations. First, the E. coli cells, substrate, and cell lysis reagents were encapsulated within the droplets with a T-shaped droplet generation module. After droplet formation, the droplets entered the incubation module in which E. coli cells were lysed to release the plasmid encoding a variant of aldolase for cell-free protein expression. The activity of the aldolase was evaluated via the degradation of a preloaded fluorogenic substrate. Finally, the positive droplets with high fluorescence intensity were sorted with a droplet sorting module. After FADS, the activity of the screened aldolase improved 30 fold to give a > 109 rate enhancement compared with the uncatalyzed retro-aldol reaction. This study demonstrates the feasibility and superiority of droplet microfluidics by combining different functional modules.
Figure 6. Schematic representation of the integrated microfluidic chip for the directed evolution of aldolases. a Schematic showing the use of aqueous droplets as bioreactors for enzyme directed evolution. b Schematic of the integrated chip, the droplets generation module is shown in blue, the droplets incubation module is shown in red, and the droplet sorting module is shown in green. "-" indicates the negative electrode and " + " is the positive electrode. c Zoom-in schematics and images of the functional modules of the fully automated FADS system (Reprinted with permission from Obexer et al. 2017. Nat Chem 9: 50–56. Copyright (2017) Springer Nature)
There are two commonly used bulk sequencing strategies: shotgun sequencing and amplicon sequencing (Rausch et al. 2019). Shotgun sequencing, which captures the entire DNA sequence within a sample but generally only represents the most abundant genes, has high costs and bioinformatic demands (Sharon and Banfield 2013). Amplicon sequencing enriches specific microbial groups or genes before sequencing using PCR (Mamanova et al. 2010). Thus, it offers higher resolution for the target genes, functions, or microbial species (Lundberg et al. 2013; Singer et al. 2016). However, bias can be created during the PCR amplification step and this is difficult to remove (Boers et al. 2015). Droplet microfluidics is advantageous for massive encapsulation of single molecules or microbial cells and offers individual and independent reaction containers. The use of droplets can effectively reduce biases caused by inter-sample interference (Hori et al. 2007; Shao et al. 2011).
Boers et al. (2015) introduced micelle PCR (micPCR) for microbiota profiling (Boers et al. 2015). Conventional NGS-based amplicon sequencing of bacterial 16S rRNA gene is widely used for phylogenetic analysis of a complex microbial community (Boers et al. 2015). However, the presence of multiple PCR targets in a single amplification reaction may lead to amplification chimeras and bias towards the amplification of a subset of genes due to the competitive nature of PCR. micPCR separates template DNA molecules into a large number of water-in-oil emulsions, which reduces chimera formation and PCR competition within the same sample. The results show that micelle PCR reduces chimera formation by a factor of 38 compared with bulk PCR, resulting in improved microbial diversity estimates and generating the robust and accurate 16S microbiota profiles required for comparative studies.
In addition to amplifying 16S rRNA gene in droplets only for microbiota profiling, Spencer et al. (2016) developed Emulsion, Paired Isolation, and Concatenation PCR (epicPCR) for linking the phylogenetic markers with functional genes from millions of cells in a single experiment (Fig. 7). This method includes the following steps. (1) Single cells are encapsulated in ~ 500 million droplets at a ratio of < 1 cell in 100 drops on average. Acrylamide monomers in the droplets polymerize into polyacrylamide beads with the addition of a catalyst. (2) Cells are lysed in gel beads and the released DNA is trapped inside the beads. (3) Hydrogel beads are then re-emulsified to perform fusion PCR, amplifying the 16S ribosomal RNA gene and a separate target gene. The overlapping of the primers of the two targets will create concatemers containing both the 16S rRNA gene and the target functional gene. (4) After demulsification, fusion amplicons are pooled for a bulk nested PCR followed by NGS. The epicPCR method was demonstrated by linking the dissimilatory sulfate reductase gene dsrB with the 16S rRNA gene and detected both known and novel sulfate reducers in a freshwater lake sample (Spencer et al. 2016). EpicPCR can be extended to determine the host cells of theoretically any target functional gene, providing a throughput of millions of cells, with only the cost of a single sequencing library preparation. This technique has been successfully applied to the study of bacterial hosts of antibiotic resistance genes in wastewater treatment plants and the diversity of sulfate-reducing prokaryotes (SRP) in Tibetan saline lakes (Hultman et al. 2018; Qin et al. 2019).
Figure 7. Schematic principle of Emulsion, Paired Isolation, and Concatenation PCR (epicPCR) for the association of functional genes with phylogenetic status at the single-cell level. A Microbial cells in acrylamide suspension are mixed into emulsion oil. The emulsion droplets are polymerized into polyacrylamide beads containing single cells. The emulsion is broken, and the cells in the polyacrylamide beads are treated enzymatically to expose the genomic DNA by destroying cell walls, membranes, and protein components. B Polyacrylamide-trapped, permeabilized microbial cells are encapsulated into an emulsion with fusion PCR reagents. C Fusion PCR first amplifies a target gene with an overhang of 16S rRNA gene homology. With a limiting concentration of overhang primer, the target gene amplicon will anneal and extend into the 16S rRNA gene, forming a fusion product that continues to amplify from a reverse 16S rRNA gene primer. D The fused amplicons only form in the emulsion compartments where a given microbial cell has the target functional gene. e After breaking the emulsion, the fused amplicons are prepared for next-generation sequencing. The resulting DNA sequences are concatemers of the functional gene and the 16S rRNA gene of the same cell (Reprinted from Spencer et al. 2016. ISME J 10: 427–436. followed Creative Commons Attribution 4.0 International License)
Droplet microfluidics also enables the enrichment of specific functional gene clusters at high-throughput (Eastburn et al. 2015). Xu et al. (2020) developed the Microfluidic Automated Plasmid Library Enrichment (MAPLE) workflow for isolating and sequencing microbial biosynthetic gene clusters (Fig. 8). Metagenomic libraries are usually used for screening and recovering biosynthetic gene clusters. They are made of bacterial artificial chromosome (BAC) or fosmid plasmids that carry hundred-kilobase-length fragments of metagenomic DNA and transformed into E. coli host cells. Traditional plate-based screening methods involve multiple rounds of hundreds of PCR reactions and gel electrophoresis. This is laborious and expensive. MAPLE performs the screening steps using droplet microfluidics with a millionth the normal volume, and at screening throughputs of thousands per second. It comprises four steps.
Figure 8. Microfluidic Automated Plasmid Library Enrichment (MAPLE) workflow and associated microfluidic devices. a Overview of MAPLE workflow. b The droplet maker device used for encapsulating single cells. Top, device schematic; middle, enlarged view of the cross junction where droplets are generated; bottom, an image of a single bacterium (red arrow) in the droplet before culture (left) and resulting colony after incubation (right). Scale bar = 20 µm. c Droplet merging device. Left, device schematic. Right, inserts showing magnified views of the three numbered regions. Insert 1, reinjection of close-packed droplets containing cell colonies spaced out by oil flow. Insert 2, pairing of colony droplets (orange) with PCR reagent droplets (blue) at a ~1:1 ratio. Insert 3, the entrance of droplet pairs into merging zone for electro-coalescence. Scale bar = 100 µm. d Droplet sorter device used for sorting fluorescently positive droplets. Top, device schematic. Bottom, inserts showing the junction where droplets are sorted. If a droplet passing the laser (light spot) has a fluorescence signal exceeding the threshold, then the electrode (yellow bar) activates, applying a dielectrophoretic force to pull it into the 'sorted' channel. Scale bar = 100 µm (Reprinted with permission from Xu et al. 2020. Nucleic Acids Res 48: e48. Copyright (2020) Oxford University Press)
1) Single-cell colony formation in droplets. Individual E. coli cells are encapsulated and cultured in millions of picoliter droplets, where single cells expand into pico-colonies within each droplet.
2) Target detection by droplet PCR. Each colony-containing droplet is then fused with a droplet carrying primers and PCR reagents using a droplet-merge device followed by thermocycling to identify colonies carrying sequences of interest.
3) Plasmid isolation by droplet sorting. PCR-positive drops become fluorescent either by TaqMan assay or SYBR Green staining and are sorted out using a droplet sorter device.
4) Sequencing of isolated plasmids. The DNA is recovered from sorted droplets by breaking emulsion, and sequence information is obtained by NGS. MAPLE enables efficient enrichment of the target genome, facilitating deep sequencing coverage and reassembly of the target gene cluster, and promoting the discovery of functional-related genes and biosynthetic pathways from a highly diverse metagenomic sample.
Metagenomics has introduced the power of genomic analysis to entire communities of microbes to depict the structure and potential metabolism pathway of a microbial community with high resolution. Metagenomics allows the study of all of the genomes in a community without the isolation and cultivation steps. However, difficulties in sequence assembly and function annotation make it difficult to link species to their functions. As a complementary technique, single-cell genome sequencing allows the identification of metabolic features of individual species and enables the de-convolution of genetic heterogeneity in diverse cell populations (Woyke et al. 2017). In general, single-cell sequencing includes three basic steps: 1) Single-cell isolation by micromanipulation (Chiou et al. 2005), flow cytometry, (Rinke et al. 2014) or microfluidics (Marcy et al. 2007b); 2) Whole-genome amplification (WGA); and 3) NGS and bioinformatic analysis of single-cells genomes.
Flow cytometry enables high-throughput single-cell sorting with multiple dimensions and high flexibility. It has been coupled with multiple displacement amplification (MDA) for routine single-cell sequencing of microorganisms (Rinke et al. 2013). We recently reported an improved single-cell MDA pipeline using the interfacial nanoinjection (INJ) technology coupled with fluorescence-activated cell sorting (FACS) (Fig. 9) (Yun et al. 2019). FACS sorted single cells from deep-sea microbial samples were inserted into each well of a 96-well plate and the INJ was used to add nanoliter droplets of MDA reagents with high precision for single-cell WGA. By real-time monitoring of single-cell MDA reaction, early positive reactions were selected for NGS and phylogenetic identification. This presented good coverage of the genomes with significantly lower contamination levels compared to single-cell amplification in large volumes. Using deep-sea sediment samples from the Southwest Indian Ocean, the INJ-FACS pipeline identified a large number of carbohydrate-active enzymes (CAZymes) in all single-cell genome assemblies, indicating their role in carbon cycling in deep oceans (Yun et al. 2019).
Figure 9. Single-cell whole-genome amplification and sequencing using fluorescence-activated cell sorting (FACS) coupled with interfacial nanoinjection (INJ) technique
The droplet microfluidic device allows single-cell sequencing to be performed at much higher throughput. For example, Hosokawa et al. (2017) introduced a single droplet multiple displacement amplification (sd-MDA) method, which enabled massive simultaneous amplification of single-cell genomes while maintaining sequence accuracy and specificity. A massive parallel single-cell genomic sequencing (SiC-seq) workflow was developed by adding barcodes to all fragments in droplets (Lan et al. 2017), which includes the following steps (Fig. 10).
Figure 10. Microfluidic and biochemical workflow to generate a single-cell genomic sequencing (SiC-seq) library. a Generating barcode droplets by encapsulating random DNA oligos at limiting dilution and amplification by in-droplet PCR (SYBR-stained for visualization). b Cells are encapsulated at limiting dilution with molten agarose to generate agarose microgels, each contains a single cell. c The single-cell genomes are purified through a series of bulk enzymatic and detergent lysis steps. d Microgels are re-encapsulated in droplets containing tagmentation reagents. e The droplets containing tagmented genomes are merged sequentially with PCR reagents and barcode droplets at a 1:1 ratio, followed by PCR to splice barcodes to genomic fragments (Reprinted with permission from Lan et al. 2017. Nat Biotechnol 35: 640–646. Copyright (2017) Springer Nature)
1) Single-cell isolation: Single cells are encapsulated in agarose droplets using a two-stream co-flow droplet maker device, which merges a cell suspension stream with a molten agarose stream. After cooling to solidify the agarose microgel beads, the beads are transferred from oil to aqueous carrier phase.
2) Single-cell lysis: The microgel beads are incubated in a mixture of lytic enzymes overnight, and then in a mixture of detergents and proteases for 30 min to digest the cell wall, lipids, and proteins, preserving only genomic DNA.
3) Fragmentation of the genomic DNA: The microgel beads are re-encapsulated in the Nextera reaction in separate droplets, and the genomic DNA in the beads is fragmented by transposases with PCR handles added.
4) Droplet barcoding: Using a droplet merging device, each microgel-containing droplet is fused with droplets containing PCR reagents and a barcode droplet; the sample is then thermal-cycled, introducing the cell-specific barcode and sequencing adaptor sequences onto the genomic fragments through the PCR handles.
5) NGS: Droplets are demulsified to pool all the barcoded DNA for sequencing, and the sequencing reads are filtered and grouped by barcode, providing single-cell genomic sequence data. The SiC-seq method yields ~ 50, 000 cells by sequencing a synthetic microbe community and is used to study the distributions of antibiotic resistance genes, virulence factors, and phage sequences from environmental samples. This shows the microbial genetic heterogeneity at the population level. Through a combination of targeted sequencing, single-cell sequencing, and metagenomics, we can expect our understanding of microbial diversity and their ecological and environmental roles to be greatly expended (Xu and Zhao 2018).