TY - JOUR
AU - Ruhdal, Jensen, Peter
AB - Abstract Strain development is frequently used to improve the performance and functionality of industrially important microbes. As traditional mutagenesis screen is especially utilized by the food industry to improve strains used in food fermentation, high-throughput and cost-effective screening tools are important in mutant selection. The emerging droplet-based microfluidics technology miniaturizes the volume for cell cultivation and phenotype interrogation down to the picoliter scales, which facilitates screening of microbes for improved phenotypical properties tremendously. In this mini review, we present recent application of the droplet-based microfluidics in microbial strain improvement with a focus on its potential use in the screening of lactic acid bacteria. lactic acid bacteria, strain improvement, droplet-based microfluidics, high-throughput screening, starter culture, classical mutagenesis INTRODUCTION Lactic acid bacteria (LAB) compose a group of gram-positive bacteria that are ubiquitously used as lactic starter in dairy fermentation. The starter culture industry shares a billion-sized growing global market per annum (de Vos 2011). As a food starter culture, LAB rapidly reduce pH in a given food matrix, which inhibits the growth of detrimental food-borne pathogens and spoilages. The metabolic activities of LAB, e.g. carbohydrate catabolism, proteolysis and lipolysis, dictate the formation of distinct flavor notes in different fermented foods. Other unique attributes of LAB have also extended the use to newer applications such as probiotics, bio-protection and plant material fermentation (Mozzi, Raya and Vignolo 2015). With multiple advanced genome-engineering tools, precise genetic alterations on LAB chromosomes and plasmids can be achieved within a few days by experienced specialists (Joutsjoki et al.2002; Levander and Svensson 2002; Papagianni 2012). Nevertheless, these GMO approaches have not been widely implemented in the starter culture industry due to the doubt of safe use from both the authority and the public (Derkx et al.2014). Traditional strategies are still dominating the strain improvement activity of industrial LAB, where mutagenesis screen, analog selection and laboratory evolution are common non-GMO tools applied in the field (Derkx et al.2014). Both analog selection and laboratory evolution have been proven as efficient tools for improving certain phenotypes. These methods rely on exposing cells to inhibitive analogs or stress conditions, which costs cellular fitness. Mutants with improved fitness compared to the general population can thereby be enriched on selection plates or through serial propagation with effectors. Successful implementation of either method relies on an increased cellular fitness that is genetically linked with a desired phenotype. Due to this constraint, the two methods are mainly applied in modifications of growth-relevant metabolic activities such as carbohydrate utilization, vitamin and cell wall synthesis and stress resistance (Burgess et al.2004; Papagianni 2012; Chen et al. 2015a,b; Kibenich, Kim Ib and Eric Johansen 2017). There are, however, many industrially relevant attributes of LAB that are not tightly associated with central metabolism, for which these two methods are not applicable (Johansen 2018). In contrast, whole-cell mutagenesis by chemical or physical mutagens has a more general application in non-GMO strain development. The use of mutagens can increase mutation frequencies within two orders of magnitude compared to spontaneous mutations (Bates et al.1989). Due to the random nature of mutagenesis, the use of this approach is only constrained by cellular genetic boundaries. The main drawback of random mutagenesis is though that phenotypes screening, without selective pressures, requires a large amount of screening activities to find desirable mutants. Fully automated liquid handling and colony picking have greatly accelerated the screening of mutagenesis libraries on miniaturized microliter plates (MTP). Nevertheless, for industrial-scale screening of LAB, automated screening on MTP is still a slow and costly process (Johansen et al.2015). Droplet-based microfluidics is an emerging technology, where picoliter-sized droplets are used as containers for biological experiments. These droplets are water dispersed in fluorinated oil, of which the interface is stabilized by biocompatible surfactants (Fig. 1A). This provides compartmentalization and each droplet can be considered as separate samples. Like in bioreactors or MTP wells, microbes can proliferate and accumulate metabolites in droplets. Due to the small size of droplets, the screening can achieve a throughput above 1000 events/second. The emergence of the droplet-based microfluidics has been providing a more efficient screening avenue for the strain improvement with the classical mutagenesis technique (Fig. 2). In this review, we will go through the recent progress of microbial strain optimization within the droplet-based microfluidics screening. Although the technique has not been widely applied for LAB, we think that the advances in other microbes are applicable in improvements of similar phenotypes for LAB, which could benefit the development of LAB-associated products in future. We will, furthermore, address the problems that we consider critical for a wider application of the technique with regards to microbiology. Figure 1. View largeDownload slide Illustrations of droplet manipulation: (A) cell containing monodispersed droplets, (B) generation of droplets on polydimethylsiloxane chips and (C) sorting of droplets on polydimethylsiloxane chips Figure 1. View largeDownload slide Illustrations of droplet manipulation: (A) cell containing monodispersed droplets, (B) generation of droplets on polydimethylsiloxane chips and (C) sorting of droplets on polydimethylsiloxane chips Figure 2. View largeDownload slide Comparison of microbial screening with the droplet-based microfluidics, FACS and MTP. The figure is inspired from Autour and Ryckelynck (2017). *The throughput is dependent on the size of droplets. Figure 2. View largeDownload slide Comparison of microbial screening with the droplet-based microfluidics, FACS and MTP. The figure is inspired from Autour and Ryckelynck (2017). *The throughput is dependent on the size of droplets. Brief introduction of the droplets-based microfluidics for microbial library sorting Monodisperse water-in-oil droplets (Fig. 1A) can be generated by multiple methods, though a general approach is using an aqueous stream (containing cells and medium) and focusing it with fluorinated oil streams (containing surfactants). This is done at a T-junction on a chip termed as a droplet generator, which is traditionally casted in polydimethylsiloxane (Fig. 1B) (Mazutis et al.2014). The cell encapsulation in such droplet generating chips is influenced by the applied concentration of cells and the fact that the distribution follows the Poisson statistics (Collins et al.2015). By adjusting cell density in the disperse phase, the ratio of singe-cell-loaded droplets to multi-cell-loaded droplets can be manipulated. For screening, it is ideal to have a low cell density in the disperse phase, which results in a high ratio of single-cell-loaded to multi-cell-loaded droplets. This comes at a cost, as the screening load is significantly increased due to the introduction of more empty droplets. Empirically, we use a λ ≈ 0.5 (average of 0.5 cells in a droplet) for encapsulating LAB, which results in 30% singe-cell-loaded droplets with an appropriate balance between the number of empty and multi-cell-loaded droplets (Chen et al.2017). Due to the use of biocompatible chemicals during droplet generation, microbial cells can normally proliferate in droplets (Fig. 1A). The emulsions of droplets are normally stored in closed Eppendorf tubes or syringes to avoid evaporation. Compared to a laboratory-scale bioreactor, the cultivation of aerobes might be a problem due to limited gas exchange with the atmosphere over longer incubation periods in droplets (Kaminski, Scheler and Garstecki 2016). As LAB are facultative anaerobes, oxygen transmission is trivial for the cultivation in droplets. The subsequent sorting of droplets is normally performed through the fluorescence-activated droplet sorting (FADS) (Fig. 1C). The FADS utilizes a sorting chip that is embedded with two disproportionate-sized electrodes on sides of the sorting channel, which gives rise to an inhomogeneous electric field. Here, an integrated system with a laser illuminates each droplet, and records their fluorescent intensities on the chip. At values above a given threshold, the electrode activates and applies a dielectrophoretic force that pulls relevant single droplets into the collection channel, while the main population remains in the waste channel (Fig. 1C) (Abd Rahman, Ibrahim and Yafouz 2017). Collected droplets can thereafter be recovered and validated at laboratory scale. Due to the small size of droplets, both droplet generation and sorting are operated at a kHz range. For microbes such as yeasts and bacteria, a 30-minute screening of 100 000 cell-loaded droplets that is derived from randomly mutagenized libraries normally leads to identification of target mutants with improved secretion of metabolites or proteins. (Huang et al.2015; Chen et al.2017). Droplet sorting in microbial screening for enhanced phenotypes Confinement is the key feature in droplet-based screening, which provides opportunities to inspect microbial performance and analyze cell-to-cell interaction in a miniaturized compartment (Fig. 2). For LAB used as starter cultures, many important contributions in food fermentation are associated with secreted metabolites such as lactic acid, aromatic compounds, vitamins, exopolysaccharides and lantibiotic bacteriocins. Several studies have demonstrated the effectiveness of the droplet-based microfluidics in microbial library screening for improvement of metabolite and protein secretion (Wang et al.2014; Huang et al.2015). We will present them in this part. Screening of enhanced metabolite secretion LAB are well known for its capacity to raise micronutrient contents in fermented food, and a key example is the B-group vitamins. We previously explored the possibility to use droplet screening to find riboflavin over-secreting Lactococcus lactis mutants from random mutagenesis libraries. To minimize chances of accumulating harmful mutations, we applied mild chemical mutagenesis to L. lactis. Albeit the setup would result in a low frequent occurrence of variations, the high-throughput feature of droplet screening would allow easy identification of over-secreting mutants. After two iterative rounds of mutagenesis and droplet screening, we isolated a group of mutants, in which the best one secreted 4 times and 3.5 times more riboflavin compared to the starting strain in both the chemical-defined medium and bovine milk, respectively. We observed a high reproducibility for the strain improvement process. In each round, approximately four variations were introduced onto the chromosome of L. lactis by mutagenesis. To address the issue of screening low-error-rate library, we set strict sorting criteria, where top 0.05% cell-loaded droplets with the highest arbitrary units of fluorescence were sorted. A 30% true positive rate of riboflavin over-secreting phenotype was found in the sorted droplets (Chen et al.2017). Due to the non-GMO nature of the development process, these strains have potential use for production of riboflavin-fortified foods. Instead of detecting secreted metabolites in droplets, the production can also be indirectly interrogated by detecting intracellular fluorescent read-outs of metabolites with the assumption that accumulation of extracellular metabolites is proportionate to its intracellular concentration (Binder et al.2012). Therefore, screening of bulk culture can be performed on fluorescence-activated cell sorting (FACS). In a parallel screening experiment for riboflavin over-secreting mutants, Wagner et al. found that mutants isolated with droplet screening yielded two times more extracellular riboflavin compared to mutants isolated through FACS without encapsulation (2018). The difference corroborates that the compartmentalization feature of droplets screening provides a better selection on secretion phenotypes compared to FACS (Fig. 2). Another important phenotype of LAB in industrial use is acidification of food by accumulation of lactic acid (Ardö and Nielsen 2014). LAB with a fast acidification phenotype are expected to reduce batch fermentation time or amounts of starter culture used for inoculation in milk fermentation (Chen et al.2015a). A coupled-enzymatic assay has been designed to quantify lactic acid secretion of microbes in droplets (Wang et al.2014) (Fig. 3). In this assay, lactic acid is first oxidized by lactate oxidase to generate H2O2. In the second reaction, the fluorescein resorufin (ex/em: 571/585 nm) is generated from oxidation of Amplex UltraRed (Thermo Fisher Scientific, Waltham, Massachusetts, US) (N-acetyl-3,7-dihydroxyphenoxazine) with H2O2 under the catalysis of horseradish peroxidase (Wang et al.2014). Figure 3. View largeDownload slide The coupled-enzymatic assay to quantify lactic acid in droplets. Figure 3. View largeDownload slide The coupled-enzymatic assay to quantify lactic acid in droplets. Wang et al. validated the droplet screening of lactic acid production using an artificial Escherichia coli library, where high-lactate-producing E. coli strains were spiked into a bulk of low-lactate-producing strains at a ratio of 1:10 000 (Wang et al.2014). With three iterative rounds of droplet screening, the high-lactate-producing strains were enriched 5800 times. Even though the setup was not applied for a real mutagenesis library, the magnitude enrichment implies the effectiveness of droplet screening for enriching the lactic acid producing phenotype. Screening of enhanced enzyme activity and protein secretion One of the most pioneering applications using droplet screening in the biotechnology area is for enzyme engineering. The size of enzyme activity assay with picoliter droplets is at least six orders of magnitudes lower compared to the screening on MTP, which tremendously saves the cost of expensive substrates, buffer reagents and plastic consumables (Agresti et al.2010). The screening of enzyme activities in droplets is also facilitated by direct use of many fluorescence-based enzymatic assay system originally applied for MTP screening (Wang et al.2014; Huang et al.2015). The droplet screening has been extensively used for improving microbial glycoside hydrolases such as lactase and amylase. Lactase (β-galactosidase) is an important glycoside hydrolase for hydrolysis of lactose. It is commercially produced from yeast or molds, but some attributes of LAB lactase have also attracted attention for the industrial application (Henriksen et al.2016). For screening of the cytoplasmic enzymes such as lactase, it is critical to ensure the recovery of live bacteria after sorting. Nir et al. screened the microbial β-galactosidase activity in hydrogel droplets (1990). They entrapped single cells in agarose microbeads to form micro-colonies. The treatment of organic solvents allowed diffusion of the fluorogenic substrate fluorescein-di-beta-D-galactopyranoside into cytoplasm for reaction. Live cells with high lactase activity could be recovered after screening due to the survival of microbes in the core of micro-colonies entrapped in the droplets when appropriate permeabilization strength was used. Besides cytoplasmic enzymes, LAB also possess several industrially important exoenzymes that are secreted or displayed on the cell wall surface, e.g. amylase and proteinase. Amylase secretion is mainly observed for the Lactobacillus species. The expression of amylases by Lactobacillus is important for starch ingestion in the human and animal gastrointestinal tracts, and the fermentation of starchy food (Lee, Gilliland and Carter 2001; Oguntoyinbo and Narbad 2015). Screening for more efficient amylase over-secreting Lactobacillus will enhance starch hydrolysis for assimilation. Amylase secretion is screened by examining formation of halos on chromogenic amylose agar plates, or with MTP-based colorimetric assays at low throughput (Castro et al.1993). Fluorophore-labeled substrates for amylase activity assay are also commercially available (BODIPY starch, ThermoFisher Scientific, Waltham, MA) that can be adapted for high-throughput screening of microbial amylase secretion in droplets. Huang et al. screened a UV-mutagenized yeast library for better amylase secreting Saccharomyces cerevisiae in droplets using BODIPY starch as the indicator (2015). They screened 105 cell-loaded picoliter droplets in a few hours and collected 970 candidates, in which 122 mutants showed at least a 30% increase in amylase secretion compared to the starting strain in test-tube fermentation. Cell-wall-anchored proteinases are another important LAB exoenzyme. Milk protein is mainly composed of casein, which cannot be used by LAB as a nitrogen source without enzymatic degradation. Proteolytic degradation at the cell envelope by the proteinase lactocepin is, therefore, essential for the acidification in bovine milk, as the growth of LAB relies on peptide and amino acids released from casein digestion (Juillard et al.1995). Proteolysis is, furthermore, important for the taste and flavor formation in cheese maturation. During cheese ripening, casein-derived free amino acids and aroma compounds generated from the amino acids catabolic activity of LAB dictates the organoleptic quality of cheeses. High proteolytic starter cultures are sometimes not appreciated for cheese ripening due to their concurrent generation of bitter peptides from the plasmin and chymosin-derived peptides as these give an unfavorable bitter taste in cheeses (Ardö and Nielsen 2014). Genetic and biochemical characterization has clarified the fact that the caseinolytic specificity of the lactococcal lactocepin dictates the formation of bitter peptides during cheese aging (Ardö and Nielsen 2014). Protein engineering by amino acid substitution is one option to alter the substrate specificity and improve the proteolysis property of the proteinase (de Vos and Siezen 2005). For scouting proteinase variants, dye-labeled peptides enable the fluorescence-based assay of specificity in a high-throughput manner (Jones et al.1997). Ng et al. investigated the substrate specificity of single-cell proteinase in droplets (2016). The experiments comprised of multiplex substrates, which enabled a simultaneous screening of four cleavage sites of the cellular proteinase. They designed four oligopeptides substrates with different cleavage sites, where quenched fluorophores with distinct excitation/emission wavelengths were attached in an uncleaved form. The screening was conducted by inspection under fluorescence microscopy at low throughput. However, the concept can be adapted to the evolution of LAB proteinases with the droplet-based screening by using multiple read-out systems. It could help the use of mutagenesis to eliminate unfavorable cleavage capabilities of the lactococcal proteinase but retain the useful specifications. Screening of antimicrobial properties As starter culture, LAB does not only contribute to the acidification and flavor formation of fermented food products but also provides significant antimicrobial properties against food-borne spoilages and pathogens during food preservation (Mozzi, Raya and Vignolo 2015). The fast growth of LAB in food matrix helps reach an environment with low nutrients, pH and redox potentials, which is not favorable for most other microbes. In addition, LAB can also secrete a variety of metabolites functioning as antagonistic effectors (Klaenhammer 1988). The confinement feature of droplets provides possibilities to investigate and screen bacterial interaction and antimicrobial functions at a single-strain level, where the antimicrobial target is typically labeled with fluorescence markers, e.g. fluorescence proteins to indicate the growth in droplets (Terekhov et al.2017). Terekhov et al. screened the human microbiota for microbial antagonists against Staphylococcus aureus in droplets (2017). The most enriched antagonists belonged to the Streptococcus and Pseudomonas species. The antagonistic effects of the isolates were confirmed by conventional inhibitory assays. Remarkably, the antagonistic role of Pseudomonas aeruginosa against S. aureus was not reported previously, which indicates that the high-throughput droplet screening can be an effective tool for identification of new protective microbes. Challenges for general applications of the droplet-based microfluidics screening in lactic acid bacteria improvement The technological advancement of the droplet-based microfluidic screening has hitherto been validated for a limited number of applications in strain improvement, as the lack of commercial devices hinders the accessibility for non-specialists. The situation has recently been changed by the emergence of several dedicating start-up companies that supplies standard microfluidics chips, modular devices and other accessories (Holtze, Weisse and Vranceanu 2017). While these companies provide an essential service, several technological challenges remains unsolved before the technology should be considered mature. We will elaborate on these points in the following part. Development of sensitive and biocompatible assays A sensitive and non-invasive indication system is critical for a successful droplet sorting of live cells. In the screening of riboflavin over-secreting L. lactis mutants, we harnessed the autofluorescence feature of riboflavin for the indication of riboflavin production in droplets. Also, the culture medium used for screening was chemically defined, which also resulted in a low background fluorescence noise for droplet detection. Therefore, we achieved an efficient sorting and a high true positive rate of riboflavin overproduction (Chen et al.2017). Autofluorescence is a scarce feature among primary microbial metabolites, and, therefore, assay development is a prerequisite for application of the FADS. Wang et al. have demonstrated the feasibility of the oxidase Amplex system for assaying metabolites in droplets with the FADS (2014). Due to the issue of dye leaching, the detection requires injection of the assay mix into droplets after fermentation is finished, which requires critical design and manipulation of injection devices (Abalde-Cela et al.2015). One alternative solution is to use a riboswitch or synthetic RNA/DNA-based aptamers to construct metabolite-responsive fluorescent sensors for detection (Fig. 3), as a wealth of metabolite-responsive riboswitch sequences are available (Mccown et al.2017). Albeit native riboswitches are usually an off-switch, they can be inverted by engineering or in combinational use with the hammerhead ribozymes to become an on-switch sensor (Muranaka et al.2009; Meyer et al.2015). Even without prior knowledge about sequences of metabolite-binding nucleic acid, synthetic aptamer sequences can be discovered by the systematic evolution of ligands by exponential enrichment procedure (Ellington and Szostak 1990), which facilitates the general application for detecting different compounds in droplets. Riboswitch/aptamer sensors can be used for construction of whole-cell biosensors, in which a reporter gene (gfp) is typically transcribed by metabolite-responsive promoters in another microbe (Fig. 4). In a parallel screening of riboflavin over-secreting microbes, Meyer et al. designed a whole-cell biosensor, in which a riboflavin responsive GFP expression cassette was generated by linking a flavin mononucleotide riboswitch with a hammerhead ribozyme in between the promoter and the gfp gene (2015). The sensoring cell was engineered to use an orthogonal carbon source to circumvent the competition with the target microbe during the co-cultivation in droplets. By co-capsulation of the library and sensor cells, they successfully isolated several riboflavin overproducers from 30 000 candidates with the FADS. For the use of whole-cell sensors, transport activities should be considered for the analysis of low abundant metabolites in droplets (Fig. 4A). If the cross-membrane diffusion of an analyte into the sensor cells is not efficient, the cell membrane could be a barrier to lift the detection threshold with the whole-cell sensors compared to the direct measurement of metabolites in droplets. Instead, the riboswitch/aptamer sequences are used for the construction of in vitro biosensors. Abatemarco et al. designed in vitro responsive Spinach sensors, in which the ligand-binding aptamer sequences were grafted into the modular Spinach domain (2017). Upon binding with analytes, the Spinach sequence incorporates the dye molecule DFHBI and becomes fluorescent (Strack, Song and Jaffrey 2013). Due to the direct contact between sensors and analytes in droplets, these Spinach-aptamer sensors showed a very sensitive quantification of the metabolites secreted by microbes in droplets (Abatemarco et al.2017). With the sensors, the authors used the FADS to improve the tyrosine secretion in yeast by 28-folds (Abatemarco et al.2017). Due to the in vitro use of these aptamer-based sensors, thorough tests of polynucleotide stability and optimal in vitro assay condition should be conducted before the application in droplets (Abatemarco et al.2017). Figure 4. View largeDownload slide Illustrations of analytes sensing using a riboswitch-based whole-cell biosensor: (A) analytes transport by a host strain bearing biosensors and (B) sensing analytes by the riboswitch-guide reporters. If no analytes are present, expression of reporters is not actuated due to a hairpin structure containing a sequestered ribosome-binding site. When analytes are available, the ribosome-binding site is released, and the expression of GFP occurs (Muranaka et al.2009). Figure 4. View largeDownload slide Illustrations of analytes sensing using a riboswitch-based whole-cell biosensor: (A) analytes transport by a host strain bearing biosensors and (B) sensing analytes by the riboswitch-guide reporters. If no analytes are present, expression of reporters is not actuated due to a hairpin structure containing a sequestered ribosome-binding site. When analytes are available, the ribosome-binding site is released, and the expression of GFP occurs (Muranaka et al.2009). Colorimetric assays are sometimes preferred over fluorometric assays if background fluorescence is high in culture medium. Especially for LAB, the fastidious nutritional demand requires a supplement of complex substrates in the culture medium, which interferes with fluorescent signals. Absorbance-based colorimetric assays are generally incompatible with the droplet-based microfluidics due to the low sensitivity, which is caused by a very short path length of light through droplets. However, under some uncommon circumstances, it can also be applied for droplet sorting. Gielen et al. reported the successful use of a formazan color dye WST-1 for measurement of enzyme activities on a droplet screening system equipped with LED lights and photodetectors (2016). To improve the sensitivity of photodetection in droplets, they used large-sized droplets (180 pL), 100 times concentrated dye (2 mM) and a longer time for reaction (3 hours) to increase the assay sensitivity in droplets. With such compromises, they were still able to sort the droplets at a high frequency (300 Hz) with a low false positive rate. Leaching of metabolites and fluorescent dyes Unlike MTP with rigid plastic barriers, molecules in droplets are segregated by surfactants at the water–oil interface. The small size of droplets dictates a high surface area to volume ratio. Solute leakages and droplets cross-communication is possible for relatively hydrophobic or small, uncharged molecules (Kaminski, Scheler and Garstecki 2016). We previously attempted to investigate the acidification of L. lactis in droplets with pH-responsive fluorescent dyes (data not shown). We could detect high fluorescent signals in acid droplets with lactic acid. Once they were mixed with pH-neutral droplets, the fluorescent signals were decayed. At a low pH, lactic acid becomes protonated (uncharged) and tends to diffuse into neighboring neutral droplets with high pH. Therefore, the real screening of acidification in droplets did not succeed. Solutes diffuse into neighboring droplets by droplets adhesion or through the fluorinated oil phase. The adhesion is caused by physical connection of droplets to form surfactant bilayers essentially due to a low solubility of some surfactants in the fluorinated oil. If chemical potentials of adhesive droplets are mismatched, small molecules, e.g. water, move from the droplets with high potentials to the ones with low potentials until equilibrium is reached (Thiam, Bremond and Bibette 2012). Diffusion through the oil phase is also caused by the mismatch of chemical potentials among the droplets, but the mechanism is different. It is characterized by the partition of solutes into the neighboring droplets through the continuous phase, in which a finite solubility of solutes takes places. Albeit the solubility of organic solutes in fluorinated oil is low, the presence of amphiphilic surfactants in oil increases their solubility (Gruner et al.2016). This kind of leaching is commonly observed for secreted microbial metabolites in the droplets, e.g. uncharged organic acids, volatile compounds and alcohols, during incubation. It is also a critical issue for assays, in which low-molecular-weight fluoresceins are used as indicators. In the oxidase Amplex assay system, the generated fluorescent resorufin quickly diffuses into empty droplets through the continuous phase. This leaching illuminates neighbor droplets and interferes with screening. Therefore, the fermentation and the assay steps are sequenced, in which reaction mix is afterward injected into the droplets for assay of metabolites (Wang et al.2014). The screening also gets finished within a few hours before a significant cross-contamination occurs. Some measures can help attenuate the diffusion rate such as using protective additives or reducing the amount of surfactants applied (Kaminski, Scheler and Garstecki 2016). Instead of using the conventional ploy-ethylene glycol-perfluoropolyether surfactant, the advancement of new surfactant development is also expected to solve the problem of droplet cross-talk in future applications (Pan et al.2014). CONCLUDING REMARKS The predominant characteristics of FACS- and MTP-based screening are miniaturization and compartmentalization, respectively (Fig. 2). Combination of these two features in the droplet-based microfluidics technique enables the screening processes for improved microbial secretion of value-added metabolites and proteins, both high-throughput and low-cost. The technique is especially of value for optimization of industrial production strains, where classical improvement approaches are still heavily implemented. For example, the use of classical mutagenesis/selection and other non-GMO tools is preferred for improvement of performance and functionality of food-associated LAB, even though the use of modern metabolic engineering or synthetic biology approaches for strain improvement is booming. Mild mutagenesis minimizes accumulation of irrelevant mutations, which is appreciated for strain development. Augmentation of screening load due to the low mutation rates, however, is a major obstacle for finding target phenotypes. With advancements of the droplet-based microfluidics, isolation of desirable mutants from low-error-rate mutagenesis libraries becomes affordable. Development of feasible fluorescence-based assays for detection and new surfactants for stabilizing droplets and produced metabolites thereof are still the two most important tasks before we can fully implement the droplet-based microfluidics in the modern strain improvement process. Nevertheless, we can see initiatives have been being taken to mature the technique, and in some paradigms, it has been proved as an efficient primary enrichment tool that minimize the workload for the MTP screening, which reduces the cost of time and expenditure in a large screening activity (Wang et al.2014; Huang et al.2015; Chen et al.2017). FUNDING This work was supported from the Novo Nordisk Foundation by the grant NNF12OC0000818. Conflict of interest. None declared. REFERENCES Abalde-Cela S , Gould A , Liu X et al. High-throughput detection of ethanol-producing cyanobacteria in a microdroplet platform . J R Soc Interface 2015 ; 12 , DOI: 10.1098/rsif.2015.0216 . Abatemarco J , Sarhan MF , Wagner JM et al. RNA-aptamers-in-droplets (RAPID) high-throughput screening for secretory phenotypes . Nat Commun 2017 ; 8 , DOI: 10.1038/s41467-017-00425-7 . Abd Rahman N , Ibrahim F , Yafouz B . Dielectrophoresis for biomedical sciences applications: a review . Sensors 2017 ; 17 : 1 – 27 . Google Scholar Crossref Search ADS Agresti JJ , Antipov E , Abate AR et al. Ultrahigh-throughput screening in drop-based microfluidics for directed evolution . P Natl Acad Sci USA 2010 ; 107 : 4004 – 9 . Google Scholar Crossref Search ADS Ardö Y , Nielsen EW . Biochemical, Chemical and Physical Processes in Cheese During Manufacture and Ripening: Lecture Notes for MSc Course on Cheese Technology . Department of Food Science, University of Copenhagen , Denmark . 2014 . Autour A , Ryckelynck M . Ultrahigh-throughput improvement and discovery of enzymes using droplet-based microfluidic screening . Micromachines 2017 , DOI: 10.3390/mi8040128 . Bates H , Randall SK , Rayssiguier C et al. Spontaneous and UV-induced mutations in Escherichia coli K-12 strains with altered or absent DNA polymerase I . J Bacteriol 1989 ; 171 : 2480 – 4 . Google Scholar Crossref Search ADS PubMed Binder S , Schendzielorz G , Stäbler N et al. A high-throughput approach to identify genomic variants of bacterial metabolite producers at the single-cell level . Genome Biol 2012 ; 13 , DOI: 10.1186/gb-2012-13-5-r40 . Burgess C , O’Connell-Motherway M , Sybesma W et al. Riboflavin production in Lactococcus lactis: potential for in situ production of vitamin-enriched foods . Appl Environ Microb 2004 ; 70 : 5769 – 77 . Google Scholar Crossref Search ADS Castro GR , Ferrero MA , Méndez BS et al. Screening and selection of bacteria with high amylolytic activity . Acta Biotechnol 1993 ; 13 : 197 – 201 . Google Scholar Crossref Search ADS Chen J , Shen J , Ingvar Hellgren L et al. Adaptation of Lactococcus lactis to high growth temperature leads to a dramatic increase in acidification rate . Sci Rep 2015a ; 5 , DOI: 10.1038/srep14199 . Chen J , Shen J , Solem C et al. A new type of YumC-like ferredoxin (flavodoxin) reductase is involved in ribonucleotide reduction . mBio 2015b ; 6 , DOI:10.1128/mBio.01132-15 . Chen J , Vestergaard M , Jensen TG et al. Finding the needle in the haystack-the use of microfluidic droplet technology to identify vitamin-secreting lactic acid bacteria . mBio 2017 , DOI: 10.1128/mBio.00526-17 . Collins DJ , Neild A , deMello A et al. The Poisson distribution and beyond: methods for microfluidic droplet production and single cell encapsulation . Lab Chip 2015 ; 15 : 3439 – 59 . Google Scholar Crossref Search ADS PubMed Derkx PMF , Janzen T , Sorensen KI et al. The art of strain improvement of industrial lactic acid bacteria without the use of recombinant DNA technology . Microb Cell Fact 2014 ; 13 , DOI: 10.1186/1475-2859-13-S1-S5 . de Vos WM . Systems solutions by lactic acid bacteria: from paradigms to practice . Microb Cell Fact 2011 ; 10 , DOI: 10.1186/1475-2859-10-S1-S2 . de Vos WM , Siezen RJ . Engineering pivotal proteins for Lactococcal proteolysis . In: Andrews AT, Varley JBT-B of MP Varley JBT-B of MP (eds). Woodhead Publishing Series in Food Science, Technology and Nutrition . Cambridge , Woodhead Publishing , 2005 , 56 – 71 . Ellington AD , Szostak JW . In vitro selection of RNA molecules that bind specific ligands . Nature 1990 ; 346 : 818 – 22 . Google Scholar Crossref Search ADS PubMed Gielen F , Hours R , Emond S et al. Ultrahigh-throughput–directed enzyme evolution by absorbance-activated droplet sorting (AADS) . P Natl Acad Sci USA 2016 ; 113 : E7383 – 9 . Google Scholar Crossref Search ADS Gruner P , Riechers B , Semin B et al. Controlling molecular transport in minimal emulsions . Nat Commun 2016 ; 7 , DOI: 10.1038/ncomms10392 . Henriksen HV , Ernst S , Wilting R et al. Method for producing a low-lactose dairy product . European Patent, EP2234501B1 , 2016 . Holtze C , Weisse SA , Vranceanu M . Commercial value and challenges of drop-based microfluidic screening platforms-An opinion . Micromachines 2017 ; 8 , DOI: 10.3390/mi8060193 . Huang M , Bai Y , Sjostrom SL et al. Microfluidic screening and whole-genome sequencing identifies mutations associated with improved protein secretion by yeast . Proc Natl Acad Sci USA 2015 ; 112 : E4689 – 96 . Google Scholar Crossref Search ADS PubMed Johansen E . Use of natural selection and evolution to develop new starter cultures for fermented foods . Annu Rev Food Sci Technol 2018 ; 9 : 411 – 28 . Johansen E , Øregaard G , Sørensen KI et al. Modern approaches for isolation, selection, and improvement of bacterial strains for fermentation applications . In: Holzapfel WBT-A in FF and B (ed). Woodhead Publishing Series in Food Science, Technology and Nutrition . Cambridge , Woodhead Publishing , 2015 , 227 – 48 . Jones LJ , Upson RH , Haugland RP et al. Quenched BODIPY dye-labeled casein substrates for the assay of protease activity by direct fluorescence measurement . Anal Biochem 1997 ; 251 : 144 – 52 . Google Scholar Crossref Search ADS PubMed Joutsjoki V , Luoma S , Tamminen M et al. Recombinant Lactococcus starters as a potential source of additional peptidolytic activity in cheese ripening . J Appl Microbiol 2002 ; 92 : 1159 – 66 . Google Scholar Crossref Search ADS PubMed Juillard V , Bars DLE , Kunji ERS et al. Oligopeptides are the main source of nitrogen for Lactococcus lactis during growth in milk . Appl Environ Microb 1995 ; 61 : 3024 – 30 . Kaminski TS , Scheler O , Garstecki P . Droplet microfluidics for microbiology: techniques, applications and challenges . Lab Chip 2016 ; 16 : 2168 – 87 . Google Scholar Crossref Search ADS PubMed Kibenich A , Kim Ib S , Johansen E . Texturizing lactic acid bacteria strains. US Patent, US20170009198A1 2017 . Klaenhammer TR . Bacteriocins of lactic acid bacteria . Biochimie 1988 ; 70 : 337 – 49 . Google Scholar Crossref Search ADS PubMed Lee HS , Gilliland SE , Carter S . Amylolytic cultures of Lactobacillus acidophilus: potential probiotics to improve dietary starch utilization . J Food Sci 2001 ; 66 : 338 – 44 . Google Scholar Crossref Search ADS Levander F , Svensson M . Enhanced exopolysaccharide production by Metabolic engineering of Streptococcus thermophilus. Society 2002 ; 68 : 784 – 90 . Mazutis L , Gilbert J , Ung WL et al. Single-cell analysis and sorting using droplet-based microfluidics . Nat Protoc 2013 ; 8 : 870 – 91 . Google Scholar Crossref Search ADS PubMed Mccown PJ , Corbino KA , Stav S et al. Riboswitch diversity and distribution . RNA 2017 ; 23 : 995 – 1011 . Google Scholar Crossref Search ADS PubMed Meyer A , Pellaux R , Potot S et al. Optimization of a whole-cell biocatalyst by employing genetically encoded product sensors inside nanolitre reactors . Nat Chem 2015 ; 7 : 673 – 8 . Google Scholar Crossref Search ADS PubMed Mozzi F , Raya RR , Vignolo GM . Biotechnology of Lactic Acid Bacteria: Novel Applications . 2nd edn . Chichester : John Wiley & Sons Ltd , 2015 . Muranaka N , Sharma V , Nomura Y et al. An efficient platform for genetic selection and screening of gene switches in Escherichia coli . Nucleic Acids Res 2009 ; 37 , DOI: 10.1093/nar/gkp039 . Ng EX , Miller MA , Jing T et al. Single cell multiplexed assay for proteolytic activity using droplet microfluidics . Biosens Bioelectron 2016 ; 81 : 408 – 14 . Google Scholar Crossref Search ADS PubMed Nir R , Yisraeli Y , Lamed R et al. Flow cytometry sorting of viable bacteria and yeasts according to β-galactosidase activity . Appl Environ Microb 1990 ; 56 : 3861 – 6 . Oguntoyinbo FA , Narbad A . Multifunctional properties of Lactobacillus plantarum strains isolated from fermented cereal foods . J Funct Foods 2015 ; 17 : 621 – 31 . Google Scholar Crossref Search ADS Pan M , Rosenfeld L , Kim M et al. Fluorinated pickering emulsions impede interfacial transport and form rigid interface for the growth of anchorage-dependent cells . ACS Appl Mater Inter 2014 ; 6 : 21446 – 53 . Google Scholar Crossref Search ADS Papagianni M . Metabolic engineering of lactic acid bacteria for the production of industrially important compounds . Comput Struct Biotechnol J 2012 , DOI: 10.5936/csbj.201210003 . Strack RL , Song W , Jaffrey SR . Using Spinach-based sensors for fluorescence imaging of intracellular metabolites and proteins in living bacteria . Nat Protoc 2014 ; 9 : 146 – 55 . Google Scholar Crossref Search ADS PubMed Terekhov SS , Smirnov IV , Stepanova AV et al. Microfluidic droplet platform for ultrahigh-throughput single-cell screening of biodiversity . P Natl Acad Sci USA 2017 ; 114 : 2550 – 5 . Google Scholar Crossref Search ADS Thiam AR , Bremond N , Bibette J . From stability to permeability of adhesive emulsion bilayers . Langmuir 2012 ; 28 : 6291 – 8 . Google Scholar Crossref Search ADS PubMed Wagner JM , Liu L , Yuan S et al. A comparative analysis of single cell and droplet-based FACS for improving production phenotypes: riboflavin overproduction in Yarrowia lipolytica . Metab Eng 2018 ; 47 : 346 – 56 . Google Scholar Crossref Search ADS PubMed Wang BL , Ghaderi A , Zhou H et al. Microfluidic high-throughput culturing of single cells for selection based on extracellular metabolite production or consumption . Nat Biotechnol . Massachusetts, US : Thermo Fisher Scientific, Waltham , 2014 ; 32 : 473 – 8 . Google Scholar Crossref Search ADS PubMed © FEMS 2018. 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 - Droplet-based microfluidics as a future tool for strain improvement in lactic acid bacteria
JF - FEMS Microbiology Letters
DO - 10.1093/femsle/fny258
DA - 2018-12-01
UR - https://www.deepdyve.com/lp/oxford-university-press/droplet-based-microfluidics-as-a-future-tool-for-strain-improvement-in-eSuMaPRpy3
VL - 365
IS - 23
DP - DeepDyve
ER -