Microarray technology can be seen as the first imperative step in moving from macroscale cell culturing (mL) to microscale cell culturing (μL). The key advancement involves a robotic or manual arrayer that prints cells in either liquid or alginate gel droplets onto a microscope slide which is visualized under a microscope (Figure 2). The microarray results in parallel microcultures that can be screened in a highly automated and efficient manner for applications including effects of cell-culture miniaturization on phenotype, cell morphology, and growth and lag times; drug susceptibility; and cell-cell communication, making the microarray a diversely applicable tool (Ge et al., 2016). In one instance, Srinivasan et al. (2013) used microarray printing methods to print 1,200 individual cultures of Candida albicans “nano-biofilms” in 30 nL droplets. The “nano-biofilms” demonstrated their similarity to conventionally grown C. albicans biofilms in terms of morphology, architectural growth and phenotypic characteristics. The microdroplets were further subjected to antibiotic screening for drug susceptibility, showing 28 combinatorial synergistic antifungals, as well as susceptibility to three novel antifungals.

Gao et al. (2013) created a Microbe Observation and Cultivation Array (MOCA) (Figure 1), consisting of a 4 × 6 array of 1 μL culture droplets in a Petri dish. The Petri dish is plasma treated with a mask to exhibit hydrophilic spots in an equal array. When the cell culture is micropipetted onto the hydrophilic spots in the Petri dish, dominant surface tension forces govern the spontaneous entrapment of cells. The array of droplets is subsequently covered in mineral oil to prevent evaporation during the incubation period, which arguably, might introduce bias in the recovery of majority microaerophilic or anaerobic organisms, as opposed to strictly aerobic organisms (Tang, 2018). However, the parallel cultivation sampling in aqueous droplets is highly amenable to downstream (meta)-omic analyses and attempts to bridge the gap between large volumes of conventional culturing methods and the rapid miniaturization of -omic technologies (Klein and Macosko, 2017). Using this method, Gao et al. (2013) were able to culture eight distinct species from a unique marine location off the Oregon Coast; six belonging to Pseudoalteromonas spp., and the remaining two identified as Shewanella sp., and Colweillia piezophila.

Merging high-throughput efficiency of microfluidics with automation of microarrays, Jiang et al. (2016) created the Microfluidic Streak Plate (MSP) (Figures 1, 2B–E). In this technique, nanoliter, sessile droplets are “written” onto a Petri dish using a disc drive and microfluidic pen, creating a large array of droplets that are easily monitored optically. Much like the MOCA array, the droplets contain the cell inoculum which are immersed in mineral oil to prevent evaporation and coalescence of the droplets. In both cases, the nano- and microliter droplets present as discrete, “mini” agar plates, eliminating growth rate bias and interspecies competition through isolation and compartmentalization. Here, the MSP method was used for high-throughput microbial cell separation and cultivation of a soil community, targeting species that degrade polycyclic aromatic hydrocarbons (PAHs) in enzyme-based fluorescence assays. This method was particularly successful in the discovery and cultivation of several species with PAH-degrading capabilities and was also successful in isolating a previously unknown Blastococcus species. In another application, Zhou N. et al. (2019) modified the MSP method to cultivate the gut microbiome of Reticulitermes chinensis termites, leading to the successful cultivation of microbes that can metabolize lignocellulose, turning wood into biofuels. In addition, 18 novel operational taxonomic units (OTUs) were documented which were phylogenetically related to Bukholderia, Micrococcus, and Dysgonomas. The authors note the utility of this method in cultivating previously unknown microbiota, which ultimately lead to the discovery of commercially important enzymes (Zhou N. et al., 2019).

Fabrication of hydrophilic array Petri dishes negates the use of complex and expensive microfabrication processes, making the Petri dish array one of the simpler techniques for the general research community and increasing usability. Furthermore, microarrays are an efficient tool for accelerating discovery of novel metabolic processes which lead to novel drug discoveries against known ADROs. However, the array of droplets on a Petri dish presents its own disadvantages: liquid droplets are susceptible to disruption during experiments and furthermore, are limited to static interrogation (Coffey and Anderson, 2014). Although cells encapsulated in gel droplets provide a more robust environment during experiments and limit disruption, conditions for studying flow are still unmet and provide a challenge that is yet to be addressed in microarray technology (Srinivasan et al., 2013).

We have characterized micromachining methods as a separate category to bring light to the microfabrication methods available in novel micro-culturing designs. Characteristic features of this subgroup of miniaturized culturing systems include amenability to automation and the ability to culture devices in native environments. Microfabrication offers accuracy and precision, producing grid formats that are highly predictable and that can be automated through available software for growth scoring (Ingham et al., 2007). Furthermore, several materials have been explored for the manufacture of these devices, allowing researchers to control material bulk and surface properties relating to mechanical robustness, permeability, biocompatibility, and surface charge, all of which show strong variability at the microscopic level (Wondraczek et al., 2019). An impertinent benefit of such microdevices is their portability, making natural environments outside of the lab accessible with in situ incubation. Thus, we can now rely on the natural microcosm to support the growth of novel species that were previously uncultivable, without having to design and manipulate a lab-enriched growth media (Stewart, 2012).

In the early 2000s, Kaeberlein et al. (2002) brought about the idea that theoretically, one could isolate and culture microorganisms simply by using their natural milieu without knowing the specific components required to support the growth of a particular ecosystem. Hence, came the development of their diffusion chambers: agar matrix inoculated with an environmental sample, sandwiched between two membranes having 0.03 μm pores (Bollmann et al., 2007). After inoculating the chamber, it is re-installed back into the same environmental sample for the incubation period. The diffusion chambers allow natural external growth factors from the surrounding environment to diffuse through the porous membranes which would otherwise be missing in standardized laboratory media, whilst restricting movement of captured microorganisms. The idea proved promising, exhibiting a ∼300-fold increase in isolate recovery compared to the Petri dish (Kaeberlein et al., 2002). Utilizing the idea of the diffusion chamber, a trapping device was similarly developed, whereby the bottom membrane was increased to a pore size of 0.2–0.6 μm. In contrast to the diffusion chamber, the trapping device was filled with sterile agar before being placed into a soil environment. Mobile species then colonize the agar, allowing for the cultivation of a number of rare species, including filamentous actinobacteria, well known as an important source of antibiotics (Gavrish et al., 2008).

Success of the diffusion chamber led to the development of the isolation chip, or “ichip” (Figure 1); a miniaturized array format of many diffusion chambers that would increase throughput and streamline the isolation and passaging process (Nichols et al., 2010). Dilution-to-extinction of the environmental sample is used to inoculate the ichip with an average of one cell per chamber, allowing for the growth of, preferably, monocultures such that one does not need to pick and isolate cultures by hand (Berdy et al., 2017). The ichip has been used extensively in a variety of environmental samples (seawater and soil) (Nichols et al., 2010), as well as in humans (mouth) (Sizova et al., 2012), allowing for parallel cultivation and isolation of a diverse assortment of phylogenetically novel species that have previously never been cultured with conventional methods.

The diffusion chips and ichips described have been used on several occasions within the last few years to successfully cultivate rare environmental species, leading to the discovery of novel antibiotics including Novo10, and Neocitreamicin I and II (Sherpa et al., 2015; Lodhi et al., 2018). The isolation and passaging of a novel β-proteobacteria, Eleftheria terrae (Ling et al., 2015) led to the discovery of antibiotic compound teixobactin (Figure 3), which proved to have excellent antimicrobial activity against Gram-positive organisms, including drug-resistant strains. Remarkably, mutants of Staphylococcus aureus and Mycobacterium tuberculosis showed no acquired resistance to teixobactin. This discovery paints the picture that the fight against ADROs is far from over, and the ichip is a fine example of how we can harness the power of microorganisms for advances in medicine and biotechnology.

During the mid-2000s, Ingham et al. (2007) designed the micro-Petri dish: a disposable culture chip microengineered out of porous aluminum oxide (PAO) (Figure 1). The innovation behind these chips holds the same sentiment as for the ichips; however, with more focus placed on the ability to manipulate the cultivated microbes for downstream screening purposes in situ, including automation for ultra-high throughput efficiency. PAO is reported to be stable in a wide range of temperatures, biocompatible and inert to many solvents, allowing for the growth of a vast number of species (La Flamme et al., 2007). Furthermore, the PAO substrate contains natural pores approximately 200 nm in diameter; hence, diffusion of external factors is not limited, whilst the microbes are contained within rigid wells of the chip, with each well measuring 7 × 7 μm. The micro-Petri dish was used to show high throughput screening for (1) phenotypic variances; and (2) enzyme-based metabolic phenotypes using fluorescent assays for the targeted recovery of selected types. Isolation and recovery of species from environmental samples (Rhine river water in the Netherlands) resulted in screening more than 200,000 isolates for the targeted selection of species that perform organic phosphate degradation (Ingham et al., 2007). In comparison to standard plate culturing, the micro-Petri dish exhibited > 10-fold increase in culturability. An observable challenge with the micro-Petri dish was the contamination of monocultures via motile species or overgrowth of species; simply remedied by increasing the well size to 20 × 20 μm, thereby increasing the chamber volume. The micro-Petri dish chips result in a fast and ultra-high throughput screening system that exhibits a low degree of bias with high culturing density.

Toward the end of the decade, Jason B. Shear’s group developed a novel miniaturized device for bacterial culturing, termed “lobster traps” (Connell et al., 2010). The microbial traps are 3D-printed microcontainers, having picolitre-sized cavities (Figure 1). The walls of the lobster traps are made of photo-crosslinked Bovine Serum Albumin (BSA), which exhibits high permeability to external nutrients, metabolic waste, and other biologically relevant molecules. Although the lobster traps have not been tested for the cultivation of “unculturable” species particularly, their use has been demonstrated by growing low count/high density cell aggregates for quorum sensing studies (Connell et al., 2014). The high permeability to small molecules makes the lobster trap an enticing idea for use in the natural environment as a novel method for in situ incubation. The first design of the lobster traps had limited versatility; growth of a monoculture was reliant on a single motile bacteria swimming into the trap before the trap entrance was pinched off by increasing the temperature of the surrounding medium, causing the BSA walls to swell and irreversibly crosslink- a nifty feature in capturing bacteria. Subsequent designs led to printing the microchamber in situ, surrounding anchored, thermally set gelatin encapsulating a monoclonal culture. This results in a nested lobster trap (Connell et al., 2013), whereby one could study the intimate interactions between multiple species of cells. With this new method of printing, the group was able to show that a small aggregate of S. aureus exhibits heightened antibiotic resistance to β-lactams when surrounded by a culture of Pseudomonas aeruginosa encased within the lobster trap. Still, the group notes of potential shortcomings of the printing method, needing expensive, specialized equipment for the laser-printing, and which results in an opaque casing, making optical observations of the growing cells difficult. Furthermore, the direct writing method is not high throughput, hindering the ability to screen, isolate and culture “unculturable” environmental species (Connell et al., 2016).

For applications more specific to the human microbiome, Ismagilov’s group designed the SlipChip (Figure 1); a microfabricated device using standard photolithography and wet chemical etching techniques on soda-lime glass plates (Du et al., 2009). Two microfabricated glass plates sit atop each other, containing channels and wells in the nanoliter range that overlap when the top plate is slipped into the correct configuration. Although the design of the SlipChip lends itself well to multiplexed arrays for applications in diagnostics, parallel analytics and evaluation of contamination in various samples, a more interesting application of the SlipChip was its use in the gene-targeted isolation of a microorganism from the Human Microbiome Project’s “Most Wanted” taxa (Ma et al., 2014). A clinical sample from a human cecum was used as the inoculum, whereby stochastic confinement results in isolation of microorganisms and subsequently, pure colonies. Separating the SlipChip after cultivation splits the target colony in two; one half is used for scalable culturing whilst the other is used for quantitative PCR (qPCR) validation. This approach places emphasis on genetically targeted isolation efforts, such that minimal effort and resources are wasted isolating and cultivating off-target species. Moreover, we reinforce the notion that microdevices used for cell culturing can and should be used for complementary metagenomic studies (Rettedal et al., 2014; Versluis et al., 2019), capitalizing on the advantages of both microbiology and engineering fields to advance the discovery of novel microbial species (Emerson et al., 2017).

These novel tools for isolating and culturing microorganisms allow for exploring high-grade questions; however, they are not without their challenges. Compartmentalization of isolates is likely to introduce bias when microorganisms rely on synergistic proximity to their neighbors. For example, culturing hydrogen-producing bacteria and methanogenic archaea axenically might prove challenging for these same reasons (Berdy et al., 2017). Moreover, micro-technologies have yet to evolve to function optimally in extreme conditions; arid conditions dry out gelling agents, whereas use in sedimental aquatic environments results in anoxic conditions, limiting the number of cultivable cells. Other challenges lie in the manufacture of microfabricated devices: microfabricated chips require expensive machinery and starting materials, the expertise in know-how, as well as the diligence in precise, careful, and clean manufacturing (Whitesides, 2006). Further, many of the designed microfabricated chips are non-reusable; hence, resources may be saved in miniaturization, but manufacturing is a labor- and time-intensive process. These challenges all pose a barrier to scale-up production but help to identify where more work is needed in the field.

Although there is a plethora of applications for which microfluidic chips are used, including detection and identification of microorganisms, antimicrobial susceptibility testing, microbial physiology and cell-cell dynamics, as well as applications within bacterial sensing and synthetic engineering, we will emphasize the use of droplet microfluidics for single-cell encapsulation, as we perceive this technology to be the most amenable to culturing “unculturable” and rare species, and is readily available to be integrated beyond the academic realm.

Due to the characteristic feature of laminar flow within microfluidic channels, multiphase flow is achieved that enables the generation of monodisperse droplets, commonly referred to as droplet microfluidics. Droplet microfluidics has become of particular interest for cell studies because each droplet behaves as an isolated bioreactor. With droplets ranging from pico-to-microliters, the characteristic length scales of the culturing environment is comparable to that of prokaryotic and eukaryotic cells, therefore achieving quick diffusion of gases, nutrients, metabolic waste, and the like (Aleklett et al., 2018). Droplet microfluidics are further defined by single or double emulsion droplets: single emulsion droplets refer to cells being encapsulated in an aqueous phase surrounded by an oil phase (w/o) (Figure 1); in contrast, double emulsion droplets are termed water-in-oil-in-water (w/o/w) systems, whereby cells are encapsulated in a “core” aqueous phase, surrounded by a lipophilic membrane which are dispersed in aqueous solution. Addition of a surfactant to the continuous external aqueous phase maintains stability of the droplets and inhibits coalescence (Figure 1).

Several research groups have taken advantage of microfluidics to develop novel ways in which to study microbial phenomena (Mahler et al., 2021); high-throughput procedures allow for short experimental times and significantly lowered costs. For example, Beneyton et al. (2016) used a microfluidic platform to produce nanoliter-range droplets (∼10–18 nL) at a rate of 80–90 droplets per second to screen for the enzymatic production of α-amylase by filamentous fungi, Aspergillus niger. In using a microfluidic platform instead of the conventional robotic microtiter plate-based platform with liquid handling systems, the group observed that screening 10⁴ variants of A. niger took less than 24 h and cost only $14. In comparison, they estimated that the same assay using a microtiter-plate would have taken more than 16 days and cost upwards of $8,000, based only on consumables (Beneyton et al., 2016).

In the case of single emulsions, droplets are not limited to a liquid, aqueous phase; agar droplets have also been used extensively to constrain cells. Harnessing a temperature-controlled water bath with which to bathe the syringe containing the agar ensures that the agar remains melted during the encapsulation process (Lin et al., 2011). In one application for cultivation from a marine environment, Alkayyali et al. (2021), designed the Microbe Domestication Pod (MD Pod), used to hold single-emulsion agarose beads which encapsulated marine samples for easy isolation. Once encapsulated, the gel beads are injected into the Pod, then deployed back into the environment for in situ cultivation. Using representative bacteria isolated from temperate marine sediment samples, the team found that the encapsulation and in situ cultivation led to higher metabolic activity of Psychrobacter aquimaris and Bacillus licheniformis but resulted in loss of viability of Marinomonas polaris. Loss of viability is attributed to the higher melting temperatures (∼45°C) required for agarose during the encapsulation process to prevent gelling (Alkayyali et al., 2021). This increased temperature is tolerable to mesophilic species, but proved detrimental to psychrotolerant species, which limits the usefulness of agarose in cultivating temperature-sensitive microorganisms (Russell, 2003). Furthermore, due to the hydrophilic nature of agarose, it has been reported that agarose microbeads are susceptible to significant swelling (Lin et al., 2011), which further alters the diffusion properties of metabolites and waste, inevitably affecting the growing cells (Chen et al., 2015).

As previously noted, culturing of many microbial species is impeded by the inability to mimic their natural environment—this may be especially true for obligate anaerobes, requiring anaerobic chambers and special treatment by pre-reducing all liquid media that is to be used in contact with anaerobes. These cumbersome efforts may be wasted during inefficient transport to and processing of samples within the anaerobic chamber, resulting in a loss of low-abundant species. Microfluidics may be used to address some of these issues, as demonstrated by Villa et al. (2019). The group developed the MicDrop, a droplet microfluidic platform that was used to culture human gut microbiota successfully in an anaerobic chamber. The water-in-oil droplets resulted in 2.6 times higher diversity than when samples were grown in mixed conditions and were simultaneously combined with molecular techniques to validate growth of isolates in the microfluidic droplets. This platform demonstrates the utility of droplet microfluidics in traditionally difficult culturing settings (Villa et al., 2019).

Much like micromachined devices, a major challenge in using microfluidics is accessibility to expensive fabrication equipment. Most microfluidic devices are produced by casting a mold of poly(dimethylsiloxane) (PDMS); however, the mold must first be designed and manufactured with expensive and brittle silicon wafers (Weibel et al., 2007). Microelectromechanical systems (MEMS) microfabrication techniques offer several ways to etch, emboss or lithograph silicon molds. One of the more frequently used fabrication methods now is soft photolithography; a technique developed by Xia and Whitesides (1998). Soft lithography, although still making use of silicon wafers for initial fabrication, uses photo-crosslinkable polymers to create a master mold of the microfluidic device required. The designed master can subsequently be used repeatedly to cast microfluidic devices in PDMS, a quick and straightforward process to do in any lab. PDMS, in and of itself, has attractive properties for use in microfluidics due to its optical transparency, biocompatibility, permeability to gases, and low cost. However, fabrication of the device mold still requires use of a clean room and the extensive “know-how,” an obstacle for many researchers. For those who do not have access to such resources, one can buy commercialized microfluidic chips, significantly reducing the amount of time designing and fabricating one’s own chips. However, commercialized chips may have limited applicability for explorative studies, particularly for environmental samples. To increase accessibility to and personalization of the fabrication process, several public foundries have been developed at universities that allow users to send in their own designs for microfabricated chips, negating the investment of time and capital for users who do not want to be specifically trained in the process (Weibel et al., 2007).

In a similar fashion to single emulsion droplets, double emulsion droplets are generated with hydrodynamic pressure flow and co-flowing geometry within microfluidic channels. In contrast is the interphase at the channel junction which consists of three phases. The innermost phase, or core, is an aqueous phase containing the cell inoculum. The middle phase, hydrophobic in nature, may be polymeric or oleophilic, forming double emulsion droplets which are suspended in a continuous aqueous (hydrophilic) phase, including a surfactant to stabilize the droplets in solution. The addition of a membrane to house the encapsulated cells imparts unique functionality to the droplets, such as mechanical robustness for long-term studies, optical transparency for microscopy, and semi-permeability such that diffusion of chemical species can be selectively controlled (Raj and Chakraborty, 2020).

High-throughput assays may be achieved with droplet microfluidics; whereby double emulsion droplets compartmentalize chemical reactions into nanoliter-scale bioreactors. The chemical assay may then be complemented with well-established sorting methods, such as fluorescence-activated cell sorting (FACS), for the discrimination of successful assay products. Zinchenko et al. (2014) has demonstrated the powerful utility of this platform, screening for enriched cellular clones that produce catalytically active enzymes from as many as 10⁶ low-active variants, all encapsulated in 10 μm droplets. Moreover, the group was able to show that the assay can be heat inactivated to stop catalytic function, and furthermore, the droplets can be successfully frozen (−80°C) and subsequently thawed for discontinuous workflows. Discontinuous workflows could prove hugely advantageous for breaking up long workflows into shorter time-frames, allowing for more flexibility for researchers.

In applications requiring more robust microcapsules for use in diverse environments, a polymeric membrane provides the means to constrain the cells in a tough shell, without impeding permeability of small chemical species. For example, Barlow et al. (2017) used diblock co-polymer poly(ethylene glycol)-b- poly(D,L-lactic acid) (mPEG-PGDLLA) to generate double emulsion polymersomes with encapsulated Bacillus subtilis for the remediation of elemental selenium from wastewater. The polymersomes demonstrated that encapsulated B. subtilis remained viable and produced robust biofilm within the microcapsules, reducing soluble selenite (Na₂SeO₃) into less toxic elemental selenium (Se) (Barlow et al., 2017). In this case, the polymeric membrane is biodegradable, therefore, harvesting the microcapsule contents becomes simple. The polymersomes are versatile and allow for remediation of polluted areas that are not amenable to traditional wastewater treatment.

In our own novel application for double emulsion microdroplets, the Niepa group has designed nanocultures: nanoliter-sized capsules that contain nutrient broth and grow microorganisms inside planktonically, with each capsule serving as its own, miniaturized flask culture (Niepa and Davidson, 2020; Usman et al., 2021). An example of this is shown in Figure 4, whereby nanocultures were generated with P. aeruginosa (PAO1) and growth dynamics of the monoculture were observed in real-time over a course of 20 h (Usman et al., 2021). Planktonic growth of the PAO1 cells simulated accurately the same planktonic growth that is commonly seen in macroscale flask cultures. Furthermore, the nanocultures shrank significantly over the course of exponential growth, which can be attributed to resource consumption and subsequent osmosis to maintain osmotic equilibrium over the capsule membrane (Chang et al., 2015). It is possible then, to realize that nanocultures present an accessible tool for long-term studies, whereby cells may change phylogenetically due to various stressors including loss of space and competition for resources. Population heterogeneity may be introduced in this way, such as with persister cells, and their microbial dynamics assessed under a microscope. The nanocultures demonstrate their use in applications for observing growth of microbial consortia in a high throughput manner for spatial and temporal dynamics.

Microcapsules provide an ideal environment for the study of microbial communities that can be finely tuned and controlled to study the effects of independent stimuli, making it easy to decouple between physico-chemical dynamics. Furthermore, the polymeric membrane can be manipulated to satisfy the specific design requirements for varying applications. For example, the size of the nanocultures is controlled by both physical and chemical means: changing flowrates of the liquid phases present physical means to change capsule size during the encapsulation process, or a difference in osmotic pressures may be used to either draw water in or out of the capsules after their collection (Usman et al., 2021). This changes the concentration of chemical species within the capsule, and further dictates the success at which inter-cellular communication occurs. Although seemingly trivial, many cell functions are governed by community signaling and synergistic growth of recalcitrant species may heavily rely on such signaling (Park et al., 2011; Cai et al., 2019). Moreover, materials such as PDMS allow for chemical functionalization to control diffusive and mechanical properties of the nanocultures. In an explorative study, Manimaran et al. (2020) designed a new polymeric biomaterial for the use of cell encapsulation. Like commercial PDMS, the novel polymer is a mix of vinyl and hydromethylsiloxane polymers with the addition of N, N-dimethylallylamine (DMAA). The addition of DMAA into the polymer prevents tight crosslinking, subsequently resulting in a polymer network with a larger free volume. This should, in turn, increase the permeability of the membrane—these studies are ongoing.

One of the characteristics of PDMS is its mechanical robustness and elasticity, exhibiting a Young’s modulus of ∼0.5–3 MPa (Wang et al., 2014). Although beneficial for creating robust capsules for cell encapsulation, it becomes a challenge for downstream processing, which includes breaking the capsules open to retrieve the contents for further study. Therefore, one of the key benefits of functionalizing the polymeric membrane is the ability to reduce the Young’s modulus, therefore, creating microcapsules that are more brittle and that require less shear force to lyse the capsules, done simply with sonication or mechanical bead beating.

As discussed previously, a major improvement in the miniaturization of culturing is the ability to incubate samples in situ, as with the ichip. Hence, it is important that the nanocultures described here comply with this design requirement too, to study the effect of unknown metabolites on environmental microcosms (Barkal et al., 2016). A challenge presents itself here in the retrieval of nanocultures after their dissemination into the environment, due to the free-floating nature of each nanoculture. To this end, the PDMS membrane may be functionalized with ferrimagnetic iron oxide (Fe₃0₄), thereby allowing the nanocultures to be collected by simply moving a magnet nearby the sample. Therefore, the nanocultures may be suspended freely without confining them to a substrate, which might result in less effective diffusion of metabolites across the membrane.