Introduction

Isolation, identification, quantification, and clonal cultivation of microorganisms are crucial from fundamental microbiology to infectious disease diagnosis to industrial microbiology1,2. Traditional culture-based methods, particularly plate culturing, have long been considered the “gold standard” for identifying and characterizing microorganisms3,4,5. These techniques provide a reliable, well-established framework for studying microbial communities, enabling researchers to isolate, identify, and quantify bacteria, fungi, and other microorganisms from various environmental, clinical, or industrial samples. Up to date, the plate culturing methods remain an essential tool in microbiology, and form the basis of diagnostic protocols in clinical laboratories and are crucial for the discovery and characterization of new microbes in environmental or industrial settings. However, despite their widespread use, the traditional plate culturing methods face several critical limitations: (i) prolonged incubation times (18–72 h) delay diagnostic and industrial workflows; (ii) labor-intensive serial dilutions and manual spreading limit scalability; and (iii) interspecies competition in mixed samples obscures rare or slow-growing taxa6. These challenges underscore the need for more efficient, rapid, and versatile platforms that can overcome the limitations of conventional plate culturing.

To improve the efficiency and scalability of microbial detection and analysis, a variety of robotic systems have been developed for automated microbial culturing7,8,9. These robotic systems can prepare, inoculate, and process culture plates with minimal human intervention, and significantly reduce the labor required for routine microbial analyses, enabling laboratories to handle large numbers of samples efficiently. However, the robotic systems primarily enhance existing workflows but do not address fundamental issues such as the long incubation times required for colony formation or the difficulty of isolating single cells from complex communities.

Recently, droplet microfluidic technology has gained recognition as a powerful tool for microbiological research10,11,12. Due to its ability to encapsulate individual bacteria in microscale droplets and facilitate microconfined growth, droplet microfluidics has opened new avenues to various microbiological applications, including early detection of bacteria13,14, isolation of rare and uncultured microbes15,16,17,18, precise quantitation of bacteria19,20, characterization of the heterogeneity in bacterial populations21,22, selection of improved strains (e.g., microorganisms genetically engineered or adaptively evolved for enhanced industrial traits)23,24, and exploration of microbial interactions25,26,27. Furthermore, due to its capability of analyzing massively large numbers of individual droplets, droplet microfluidics also enables high-throughput screening of genetically engineered strains over large pools of cells or populations and examination of phenotypic and genetic variabilities at the level of cells or small populations. Although droplet microfluidics, as a versatile and powerful tool, has addressed some challenges, it introduces new barriers. For example, in droplet microfluidics platforms, the generated droplets are often tightly packed, which can lead to the unintended fusion of neighboring droplets during microbial cultivation, thereby compromising microorganism isolation. Although various surfactants are available to stabilize droplets28, achieving a balance between high droplet stability and low microbial cytotoxicity remains difficult. Additionally, the limited capability of substance exchange within droplets throws up difficulties in long-term cultivation, especially for metabolically active microorganisms. The free movement of droplets also poses challenges for time-lapse monitoring of large numbers of individual droplets, which is critical for studying temporal processes. Meanwhile, the droplet microfluidics platforms are primarily designed for cultivating suspended-growth microorganisms and are not well-suited for bacterial surface-adherent cultivation. Finally, most droplet microfluidic systems require expensive bulky pumping equipment and complex fluid operations, which are challenging for non-experts and smaller laboratories.

Microchamber-based microfluidic technology represents another innovative approach to microbial detection and quantification29,30,31,32. Similar to the droplet microfluidic technology, the microchamber-based microfluidic technology allows for the detection and analysis of individual bacterial cells in a “digital” manner by partitioning samples into numerous microscale compartments. However, unlike the droplet microfluidic technology, the microchamber-based microfluidic technology encapsulates individual bacteria in fixed and stable microcompartments, avoiding the risk of droplet coalescence during incubation and offering remarkable benefits for studying temporal processes of large numbers of individual bacteria. In addition, the microchamber-based microfluidic technology does not require specialized and complicated microfluidics setups for droplet generation, making it more accessible to unskilled users. Despite these obvious advantages, existing microchamber-based digital microbial assays still face a number of challenges, such as difficulty to recover the detected bacteria for further analysis and limited flexibility to adjust the growth conditions for studying how bacteria respond to different environmental conditions.

To bridge these gaps, we developed the digital plating (DP) platform—a hybrid system integrating the simplicity of conventional agar culturing with the precision of digital single-cell compartmentalization. The DP platform is based on a solid medium-covered PicoArray device, in which a bacterial suspension is partitioned into numerous picoliter microwells via a pre-degassing-induced vacuum33,34, followed by coverage with a solid agar medium sheet for incubation and analysis. It not only provides a “digital” means of microbial analysis but also recalls the key advantages of traditional culturing techniques. Most importantly, the replaceability of the covering solid medium sheet al.lows dynamic control and flexible regulation of microbial growth conditions via agar replacement, which gives access to several new possibilities – from precise selection of individuals with desired properties to rapid antibiotic susceptibility testing (AST) to cultivation of uncultivable microbes. The DP aims to achieve three objectives: (1) accelerate microbial detection to ≤ 8 h via microconfinement-enhanced metabolite accumulation; (2) enable high-resolution isolation of individual cells from complex communities without prior dilution; and (3) provide a flexible microenvironment for phenotypic screening (e.g., antibiotic susceptibility) through replaceable agar sheets.

Experimental

Fabrication of PicoArray devices

The PDMS PicoArray device containing an array of 113,137 hexagonal microwells was fabricated using the conventional soft lithography as described previously33,34. Briefly, SU-8 3010 and 3050 negative photoresists (MicroChem Corp., Newton, MA) were patterned onto separate silicon wafers to create two molds for the channel layer and microwell layer, respectively. Typical dimensions of the molds are: main channel = 52 mm (length) × 80 μm (width) × 60 μm (height); loading microchannel = 17.9 mm (length) × 30 μm (width) × 20 μm (height); gap of neighboring channels = 48 μm; microwell = 70 μm (diagonal) × 40 μm (height). Subsequently, a thoroughly degassed PDMS prepolymer, consisting of silicone elastomer (Sylgard 184) and curing agent (10:1, w/w), was poured onto the prepared SU-8 molds. After curing at 90 °C for 1 h, the molded PDMS slabs were carefully peeled off from the molds. Meanwhile, a channel layer was prepared with a channel mold according to the same procedure and an inlet port was created on the PDMS channel layer with a punching tool. Finally, the PDMS channel layer and the PDMS microwell layer were face-to-face aligned and conformally contacted to form a reversible seal for subsequent experiments.

Preparation of covering agar solid media sheets

Covering agar medium sheets were prepared as follows: 2.5 g LB broth powder (CM158, Beijing Land Bridge Technology, China) and 1.5 g agar powder (Biowest, Spain) were dissolved in 1000 mL water and autoclaved. After cooling to 60 °C, the appropriate reagents (e.g., dye, antibiotics, specific metabolic indicator, etc.) were mixed thoroughly into the agar solution depending on the experimental purposes. Next, the mixture was poured into a sterilized PDMS chamber mold with dimensions of 76 mm × 26 mm × 1 mm and covered with a sterilized plastic sheet. Then, a glass slide and a weight were placed onto the plastic sheet. After solidification at room temperature, the PDMS chamber mold was removed to obtain an agar solid media sheet.

Preparation of bacterial suspensions

The bacterial species used in this work were Escherichia coli (E. coli) JM109, green fluorescent protein (GFP)-tagged E. coli BL21, Staphylococcus aureus (S. aureus) ATCC 43,300, and Salmonella enterica 14,028, which were purchased from China General Microbiological Culture Collection Center. The novel Bacillus strain used in this work was collected from slugs at the campus of Southwest University, Chongqing, China (see Section S1 and Fig. S1 in the Supplementary Information for more details). Prior to experiments, all bacteria were inoculated from frozen glycerol stocks into tubes. The frozen stock of each species was stored at − 80 °C and thawed at room temperature (25 °C) before use. After thawing, the stock was inoculated into a liquid medium and stabilized for over an hour in a shaking incubator at 37 °C. The stabilized bacteria were seeded in an agar plate containing a suitable medium. The agar plates were incubated at 37 °C for 12 − 24 h until colony formation was visible. Then, a liquid subculture was performed by picking up one colony (or a piece of a colony) with a sterilized inoculating loop, transferring it to a liquid medium, and incubating it at 37 °C overnight in a shaking incubator. The subculture solution was diluted with normal saline to a desired concentration.

Antibiotic

Ampicillin sodium salt was purchased from Shanghai Yuanye Bio-Technology Co., Ltd. The stock solution of ampicillin sodium was prepared by dissolving ampicillin sodium salt in distilled water at 100 mg/mL. The stock solutions were sterilized by filtering through 0.22-µm sterile nylon syringe filters and were stored at − 20 °C until use. The final concentrations of ampicillin sodium used for AST were between 0 and 400 µg/mL.

PCR amplification and sequencing of the 16 S rRNA gene

Genomic DNA samples were prepared from the subculture solutions inoculated from the recovered microcolonies using the MiniBEST Bacteria Genomic DNA Extraction Kit Ver. 3.0, Takara, Japan). The 16 S rRNA gene was PCR-amplified using the universal bacterial primer pair 27 F/1492R (Table S1 in the Supplementary Information). The PCR products were sent to a sequencing service provider (Sangon Biotech Co., Ltd., Shanghai, China) for the determination of the nucleotide sequence of the 16 S rRNA gene fragment. The sequences were then aligned against the NCBI database using BLAST to identify the closest phylogenetic matches and a phylogenetic tree was constructed using MEGA X software (version 10.2.6)35.

Images and data analysis

Bright-field and fluorescent images were acquired by an upright fluorescence microscope (CX40, Sunny Optical Technology Co., Ltd., China) equipped with a CMOS camera (OD230R, SOPTOP, China). Image J (https://imagej.net) was used to recognize and count the microcolonies formed in the PicoArray devices. The detailed procedure is as follows: (i) Load an image which is needed to be analyzed. To do this, select: File > open. (ii) Convert the image to grayscale. To do this, select: Image > Type > 8-bit. (iii) Adjust threshold to highlight the picowells containing microcoloies. To do this, go to Image > Adjust > Threshold. (iv) Binary conversion after thresholding is done. To do this, go to Process > Binary > Make Binary. (v) Count the highlighted picowells. To do this, go to Analyze > Analyze Particles. All quantitative data are expressed as means ± standard deviation (SD). A one-way analysis of variance (ANOVA) was employed to statistically analyze the collected data. Subsequently, Post hoc comparisons were conducted using Tukey’s test. P-values < 0.05 were considered statistically significant.

Results

Concept and workflow of the DP platform

An ideal microbial assay platform should be user-friendly (requiring no expensive instruments and well-trained personnel), rapid (producing results within a time frame suitable for point-of-care testing), precise (enabling sensitive detection of bacteria in a quantitative manner), reliable (with a low probability of false positive or false negative results), flexible (easily accommodating a wide range of microbiological applications), and cost-effective (readily accessible to any global setting). To this end, we propose the “digital plating” (DP) technique. The principal strategy of our DP technique is to integrate a powerful digital assay format with the conventional plating principle for rapid, precise, reliable, and cost-effective microbial detection and analysis. Figure 1A illustrates the concept of our DP technique. This technique involves two principal components: a PicoArray device for stochastic compartmentalization of single bacteria to enable digital assays and a covering agar sheet that serves as a solid culture medium to support bacterial growth. The DP workflow comprises four major steps: sample discretization, assembly of the agar/chip, device incubation, and microscopic examination (Fig. 1B and Movie S1 in the Supplementary Information). To facilitate sample discretization, a pre-degassing-driven self-pumping mechanism is used to partition the bacteria suspension into a large number of picowells and stochastically encapsulate single bacteria into compartments33,34; after compartmentalization, an agar-based solid medium sheet is conformally attached to the device to seal all picowells, which provides a physical barrier to trap motile cells within the picowells; next, the device is incubated and the growth of bacteria in microwells is examined microscopically to determine phenotypic characters (see Fig. S2 and Section S2 in the Supplementary Information for detailed procedures and Section S3 for detailed post-experimental handling). Due to its digital format, this technique offers several main advantages: (i) eliminating the interspecies competition and biases due to growth rate differences, which facilitates recovery of rare or slow-growing microorganisms from complex ecosystems; (ii) allowing the study of the heterogeneity of bacteria that are masked in ensemble measurements; (iii) absolutely quantifying the concentration of viable bacteria, which enables highly sensitive detection of viable bacteria over a wide range of concentrations in a quantitative manner; (iv) facilitating timely detection of pathogenic bacteria by leveraging microconfinement-induced accelerated accumulation of detectable metabolites within individual compartments or by microscopically examining the growth morphology of microcolonies in isolated compartments. Meanwhile, the agar sheet provides culture conditions analogous to those in the conventional plating techniques, thus allowing smooth and seamless transfer of well-established characterization and analysis methods from conventional plating techniques to this platform and easy interpretation of the obtained results. More importantly, the replaceability of the covering agar sheet offers more flexibility for digital microbial assays. For instance, if an agar sheet containing a selective medium is applied on the PicoArray device, the DP technique can isolate a particular strain of microorganisms; if an agar sheet containing a differential medium is applied on the PicoArray device, the DP technique can identify and differentiate closely-related microorganisms; if an agar sheet containing a certain antibiotic is applied on the PicoArray device, the DP technique can analyze the response of bacteria to the antibiotic. Furthermore, our DP platform can sequentially combine the PicoArray with multiple different medium sheets, i.e., allowing a wide range of culture conditions to be applied on a single device. Thanks to its similarities to conventional plate methods, our DP platform allows in situ clonal culturing of single bacteria isolated from heterogeneous samples without requiring extra sorting and inoculation steps, and the cultivated pure microbes can be recovered directly for further off-chip culture or genotyping.

Fig. 1
figure 1

Overview of the DP (digital plating) technique. (A) Schematic of the DP technique for fundamental microbiological practices. (B) Work flow of the DP technique.

Full size image

Characterization of the DP platform

A key premise for digital bioassays is random partitioning of individual biospecimens into massive numbers of equal-volume small compartments. To achieve precise and reliable digital bioassays, the homogeneity of partitions is a critical factor36,37. Here, we first quantified the variance of partition volumes created by the self-discretization based on the combination of pre-degassing pumping and capillary force-assisted dewetting38. To investigate the distribution of the partitioned volumes across the device, the fluorescence microscopy images of the partitioned Rhodamine B droplets from different positions on the device were captured using a CCD camera. Due to the excellent water solubility of Rhodamine B, it is reasonable to assume that the fluorescence intensity in its aqueous solution is uniformly distributed. Consequently, the integrated fluorescence intensity of each Rhodamine B droplet is directly proportional to its volume, thereby enabling the uniformity of integrated fluorescence intensities across partitioned droplets to serve as an indicator of their volume uniformity. The captured images of droplets from different areas on the device were analyzed using an image analysis software (ImageJ) to determine the total fluorescence intensity of each droplet. From six random areas of the device, 3,200 droplets were selected for measurement and calculation. The calculated total fluorescence intensities for all these droplets were then plotted into a volume distribution histogram, showing that the coefficient of variation (CV) of the total fluorescence intensities measured from 3,200 randomly selected droplets is about 4.78% (Fig. 2A). This demonstrates that the PicoArray device, combined with the self-discretization mechanism, achieves highly uniform discretization for digital microbial assays.

Fig. 2
figure 2

Characterization of the key features of the DP platform. (A) Quantitative evaluation of the uniformity of the droplets partitioned into microwells of PicoArray device. (B) Microscopic images of a PicoArray device before and after covering of solid agar sheet showing the diffusion of red dye molecules from agar sheet to microwells. (C) Time-lapse microscopic images of agar sheet-covered PicoArray device showing the development of bacterial microcolonies within picowells. Red arrows indicate overgrown microcolonies.

Full size image

Furthermore, for our DP platform, effective diffusion of molecules from the covering agar sheet to the picowells is crucial for nutrient delivery and microenvironment regulation. To investigate whether the nutrient or drug molecules could diffuse from the covering agar sheet into the picowells, we conducted a simulation experiment by attaching a red dye-stained agar sheet (1% w/v dye concentration) to the PicoArray device and subsequently monitoring the color change in the picowells of the device. After attaching for 15 min and then peeling off the agar sheet, the color of the solution in all picowells of the device changed to red (Fig. 2B and Fig. S3 in the Supplementary Information), confirming that the dye molecules within the agar sheet could effectively and homogeneously diffuse into the picowells of the device. Such diffusion-induced molecule transport allows us to use the agar sheet for supplying nutrients to microbes confined in picowells and chemically regulating the microbial growth microenvironment in picowells, which offers unparalleled flexibility for various microbiological studies, such as monitoring microbial response upon perturbation, screening novel antibiotics, and culturing the uncultured microorganisms. Furthermore, this diffusion also allows reverse molecule transport39, i.e., from picowells to the covering agar sheet, thus enabling facile detection of metabolites and signaling molecules secreted by bacteria confined in picowells. That is, this platform allows us to perform biochemical assays to identify microcolonies formed in picowells if the covering agar sheet contains differential media.

Our DP method is based on the fact that when single bacteria partitioned into picowells are incubated, each of the microconfined single viable bacteria will proliferate forming a visible and independent colony. This allows for precise and absolute quantification of bacteria in samples. To achieve accurate enumeration, it is essential to culture individual bacterial cells until their colonies are visible, while also ensuring that each colony remains independent, avoiding fusion with neighboring colonies. To determine the optimal time point for microcolony counting using the PicoArray device, we examined the effect of cultivation time on the visibility and independence of microcolonies formed in picowells. A time-lapse sequence of E. coli cultured in the PicoArray device covered with a Luria-Bertani (LB) agar sheet is shown in Fig. 2C. In these images, dark spots represent microwells where bacteria grew and formed microcolonies, while blank spots indicate microwells that either initially lacked cells or where the cells failed to form colonies, likely due to low growth rates. Most microcolonies were nearly invisible before 5 h. After incubation for 6 h, most microcolonies were discernible and confined to individual picowells without fusion between neighboring colonies. However, after 7 h of incubation, there was a noticeable overgrowth of the formed microcolonies (as indicated by red arrows in Fig. 2C), which spread into adjacent picowells. Thus, to ensure a precise enumeration, counting of E. coli microcolonies should be performed between 6 and 7 h of incubation. It is worth noting that the optimal counting time varies across different microorganisms due to significant differences in their growth rates. Furthermore, it must be clearly recognized that the validity of the DP platform for viable bacterial quantification fundamentally hinges on microbial cultivability under defined conditions. This platform faces inherent constraints when analyzing samples harboring unculturable or fastidious microbial populations, thereby compromising its capacity to achieve accurate viable cell enumeration. To mitigate potential underestimation of viable bacteria, two alternative approaches can be employed: extending incubation duration and modifying culture media6.

In the DP platform, picowells are filled with nutrient-rich medium, while the overlying agar sheet ensures oxygen diffusion and sustained nutrient supply for non-fastidious microorganisms, mirroring conventional culture environments. Thus, the DP platform preserves bacterial viability equivalent to conventional methods, enabling precise enumeration of viable bacteria through digital quantification—a significant advancement over traditional plate counting, which is prone to colony merging and underestimation. To demonstrate the capability of our DP platform to rapidly and accurately quantify bacteria, we loaded serially diluted samples over a range of starting cell densities from a previously quantified Salmonella stock with Mueller-Hinton broth into the device, followed by discretization and incubation. After 6 h of incubation, all devices were examined by microscopy to count the positive microwells containing microcolonies in each chip. Figure 3A shows representative micro-images of PicoArray devices loaded with different concentrations of Salmonella after incubation. As expected, reducing the bacterial input concentration resulted in fewer picowells occupied by microcolonies. The proportion of positive microwells in each device was then counted, and the cell concentration in colony-forming units per milliliter (CFU/mL) was calculated using Poisson statistical analysis (see Section S4 in the Supplementary Information for detailed derivation of the Poisson model)36,37. This count was plotted against the CFU/mL measured in parallel by conventional colony counting on agar plates. As shown in Fig. 3B, the Salmonella concentrations measured using the PicoArray-based digital CFU method correspond very well to those obtained via the conventional plate counting method (R2 = 0.9992) in the entire 5 log range of the measured sample concentrations, demonstrating the capability of the PicoArray-based digital CFU method to precisely quantify microbes. Importantly, the entire process for the digital CFU method took only 6 h, significantly shorter than that with the conventional plating method, which is beneficial to rapid diagnosis and prompt treatment of infectious diseases. For the conventional plating method, while fast-growing bacteria may form visible microcolonies on agar plates within 6–8 h, their direct enumeration at this stage remains error-prone and labor-intensive. The DP platform overcomes these limitations by digitizing the detection process, enabling rapid, bias-free quantification even for low-abundance species. This capability is particularly critical for clinical diagnostics, where early pathogen detection directly impacts therapeutic outcomes40.

Fig. 3
figure 3

Digital enumeration of viable bacteria via the DP platform. (A) Bright-field microimages showing the representative results of growth of different concentrations of Salmonella in the PicoArray device after 6 h of incubation at 37 °C. (B) The Salmonella concentration measured using the DP platform vs. the concentration determined using the plate culturing (gold standard). All measurements were performed in triplicate (n = 3). Error bars represent standard deviation (SD).

Full size image

Isolation of single bacteria from a heterogeneous sample

In nature, microorganisms typically grow as mixed populations rather than as single species. Therefore, isolating individual microorganisms from complex microbial communities is an essential prerequisite for downstream phenotypic and genomic characterization, as well as for scaled-up culture. Due to its ability to compartmentalize single cells into massively large numbers of subunits for clonal growth, our DP platform is ideal for the isolation of individual bacteria from heterogeneous samples. To demonstrate the DP platform’s capacity for single-cell separation and cultivation, we prepared a simple two-strain microbial community consisting of S. aureus and green fluorescent protein (GFP)-expressing E. coli (BL21). First, both strains were cultivated overnight, diluted 10,000 times in LB broth, respectively, and then mixed in a ratio of 1:1 (v/v) to obtain a microbial mixture. After discretization in the device and incubation at 37 ºC for 8 h (), three distinct populations of microwells were observed via microscopy (Fig. 4A). One population of microwells was empty microwells, which was predicted to represent compartments containing no bacteria. The remaining microwells contained microcolonies with two distinct morphologies: densely packed microcolonies and thin microcolonies with fewer cells. Given that E. coli showed slower growth under the used conditions, the population of microwells containing thinner microcolonies was predicted to represent compartments containing E. coli strain. The population of microwells containing denser microcolonies was assigned to correspond to compartments containing the best-growing bacteria, here, S. aureus. This hypothesis was confirmed through fluorescence microscopy. Under 523 nm excitation light, only the thinner microcolonies exhibited green fluorescence, indicating the presence of GFP-expressing E. coli, while the denser microcolonies showed no fluorescence, confirming them as S. aureus (Fig. 4A). Based on the total number of picowells and the count of microwells occupied by fluorescent and dense microcolonies, we calculated the numbers of S. aureus and E. coli using a method similar to digital PCR41. As shown in Fig. 4B, the calculated numbers of S. aureus and E. coli corresponded closely with the expected number of S. aureus and E. coli (estimated via agar plating). These results suggest the potential of the DP platform for isolating bacterial species in mixed communities. Future studies will assess its performance across diverse microbial samples.

Fig. 4
figure 4

Isolation and cultivation of bacterial cells from a microbial mixture consisting of S. aureus and GFP-expressing E. coli with the DP platform. (A) Typical bright-field, fluorescence, and merged microimages of microcolonies grown in a PicoArray device after 8 h of incubation at 37 °C. Initial concentrations of S. aureus and E. coli ~ 6 × 104 CFU/mL. (B) Enumeration of S. aureus and GFP-expressing E. coli in the microbial mixture using the DP platform vs. the agar plating.

Full size image

Enrichment and identification of specific bacterial species from mixed-species microbial communities

In addition to enabling spatial separation of different bacteria in a single sample, our DP platform can also be combined with selective and/or differential media to enrich and identify specific bacteria in complex samples. To demonstrate the selective-medium-assisted enrichment capability of the DP platform, we prepared a microbial mixture consisting of S. aureus and GFP-expressing E. coli and loaded it into a PicoArray device. When the device was covered with an agar sheet containing LB medium and incubated at 37 ºC for 8 h, both bacterial strains grew into microcolonies (see the left column in Fig. 5A-i). The two species were easily distinguished by comparing bright field and fluorescent microimages: one population exhibited green fluorescence, while the other did not. In contrast, when the device was covered with an agar sheet made with NaCl-based selective medium (containing 7.5% NaCl) and incubated at 37 ºC for 8 h, only non-fluorescent microcolonies (S. aureus) grew in the NaCl-based medium (see the right column in Fig. 5A-i), confirming selective inhibition of E. coli under high-salt conditions. This aligns with the known halo-tolerance of Staphylococci and the sensitivity of Enterobacteriaceae to high salt concentrations42. While certain E. coli strains exhibit halo-tolerance under specific conditions43, the strain used here (BL21) showed no growth at 7.5% NaCl, consistent with typical Enterobacteriaceae behavior. Additionally, the number of non-fluorescent microcolonies in the PicoArray device covered with LB medium was nearly equal to the total number of microcolonies formed in the device covered with NaCl-based medium (Fig. 5A-ii), further confirming the selectivity of the NaCl-based medium. These results demonstrate that the DP platform, when coupled with selective media, can effectively enrich target microorganisms from complex microbial communities with high specificity.

Fig. 5
figure 5

Enrichment and identification of bacteria in complex samples by coupling the DP with selective and differential medium. (A) Enrichment of S. aureus from a microbial mixture consisting of S. aureus and GFP-expressing E. coli by coupling the DP platform with a NaCl-based selective medium. (i) Representative bright-field, fluorescence, and merged microimages of microcolonies formed in two PicoArray devices after being covered with LB medium agar sheet and selective medium agar sheet (incubated at 37 ºC for 8 h), respectively. Initial concentrations of S. aureus and E. coli ~ 3 × 105 CFU/mL and ~ 5 × 104 CFU/mL (ii) Enumeration of the fluorescent and non-fluorescent microcolonies formed in the PicoArray devices with LB medium agar sheet and selective medium agar sheet, respectively. (B) Representative bright-field microimages of microcolonies formed in three PicoArray devices after being loaded with S. aureus, E. coli, and S. aureus/E. coli mixture and incubated with MCA sheets at 37 ºC for 8 h.

Full size image

In addition, we also performed an experiment to demonstrate the differential-medium-assisted identification ability of our DP platform. Here, MacConkey agar (MCA), one of the earliest and most common differential media, was chosen to demonstrate the applicability of “DP + differential medium” to distinguishing specific bacterial species from other bacterial types. The experiment was performed by discretizing a S. aureus suspension, an E. coli suspension, and a mixture of the two in three PicoArray devices, respectively, covering each device with an MCA sheet, and then incubating them at 37 ºC for 8 h. As shown in Fig. 5B, the microcolonies formed in three devices displayed different appearances in color: the E. coli-loaded device exclusively exhibited red microcolonies, the S. aureus-loaded device displayed only colorless microcolonies, and the device loaded with the mixture of both species presented both red and colorless microcolonies. This differentiation is attributed to the composition of the MCA medium, which contains lactose and neutral red (pH indicator). E. coli can ferment lactose to produce lactic acid, thereby decreasing the pH of the agar and turning the indicator (neutral red) pink44. In contrast, S. aureus cannot utilize lactose under standard conditions, resulting in colorless colonies due to the absence of pH alteration and indicator color change45. Although rare S. aureus strains may exhibit this trait46. In our experiments, no lactose fermentation was observed for the tested S. aureus strain (ATCC 25923).

Rapid phenotypic AST

Another important application scenario of the DP platform is AST, a fundamental mission of the clinical microbiology laboratory, which is paramount for effective infection management. In particular, a rapid and precise AST assay can assist clinicians to choose appropriate antimicrobial therapies for various pathogen infections and avoid unnecessary overprescription. Thanks to its ability to miniaturize bacterial cultures to picoliter levels which can dramatically alter early colonization and self-regulated quorum signalling47, the DP platform enables faster measurement of the ability of individual bacteria to form microcolonies at different antibiotic concentrations. This is achieved through microconfinement-induced rapid increases in cell density or the accumulation of detectable metabolites, thus enabling rapid AST assays. To demonstrate the applicability of the DP platform to AST assays, we used E. coli as a model bacterium and ampicillin sodium as a model antibiotic to perform AST experiments. Multiple PicoArray devices were run in parallel, each loaded with the same concentration of E. coli suspensions and covered with an agar sheet containing a different concentration of ampicillin sodium. In addition, a control device was run in parallel, which was loaded with the same bacterial concentration as the others but covered with an agar sheet without any antibiotic. This control allows us to estimate the number of cells initially loaded into each DP chip. After 6 h incubation at 37℃, the chips were imaged via microscopy. It was observed that the bacterial growth was inhibited in the presence of antibiotics. As shown in Fig. 6A, the number of microwells containing microcolonies decreased with increasing concentrations of ampicillin sodium, and no cells survived at concentrations above 200 µg/mL. The percentages of surviving cells at 6 h were plotted against the concentrations of ampicillin sodium, exhibiting a monophasic dose-response curve (Fig. 6B). The minimum inhibitory concentration (MIC) value, obtained by fitting the bacterial growth data to a Gompertz model48, was 196.9 µg/mL, which closely matches the MIC determined by the broth microdilution method (193.4 µg/mL) (Fig. S4 in the Supplementary Information). Compared to conventional AST methods, our DP platform not only consumes fewer reagents (~ 20 µL) but also allows faster antibiotic susceptibility assays (< 6 h). More importantly, it can characterize physiological responses of individual cells to different antibiotic concentrations, which enables the quantitative phenotyping of heterogeneous resistance, or heteroresistance, with single-cell resolution21. These advantages make it an excellent alternative to standard phenotypic AST with potential applications in clinical diagnostics and point-of-care testing, enabling rapid clinical decision-making, improvement of infectious disease management, and facilitation of antimicrobial stewardship.

Fig. 6
figure 6

Phenotyping antibiotic resistance using the DP platform. (A) Representative bright-field microimages showing E. coli microcolonies grown in the PicoArray chips under different concentrations of ampicillin sodium after 6 h of incubation at 37 ºC. Initial concentration of E. coli ~ 1 × 105 CFU/mL. (B) Dose response curve for E. coli viability after treatment with different concentrations of ampicillin sodium. Calculation of MIC using a Gompertz function fit. Blue vertical dashed line shows the position of the MIC. Each test was performed in triplicate (n = 3). Error bars represent standard deviation (SD). (C) Representative bright-field microimages showing the morphology changes of E. coli under antibiotic stress.

Full size image

In addition to digital AST, the DP platform also provides a unique ability to track the temporal evolution of the bacterial colonies within each picowell. This is due to its confinement of isolated droplets into static, spatially-defined arrays, enabling the indexing and time-lapse microscopy monitoring of droplet contents over time. This feature allows for the exploration of the progeny of individual cells as they respond to antibiotic stress. For example, E. coli cells exposed to ampicillin sodium underwent adaptive morphogenesis, as shown in Fig. 6C. In the absence of antibiotics, nearly all bacteria within the picowells were in their planktonic state, actively swimming within the droplets (Movie S2) and exhibiting the characteristic size and morphology of E. coli in culture (see the top row in Fig. 6C). Conversely, at sub-MIC concentrations (100 µg/mL) of ampicillin sodium (see the bottom row in Fig. 6C), the cells exhibited morphological changes, elongating into filamentous forms, and became immobile (Movie S3). This filamentation may represent a survival strategy where bacteria exposed to ampicillin sodium delay division to avoid lethal damage. This behavior aligns with prior reports of antibiotic-induced filamentation as a stress adaptation49, though further studies are needed to distinguish active regulation from passive physiological disruption. The DP platform’s ability to follow morphological changes over time in a large number of individual cells allows the comparison of different bacterial strains’ responses to antibiotics with varying mechanisms of action. This capability is invaluable for elucidating the biological mechanisms that enable cells to survive antibiotic stress at sub-MIC concentrations.

Investigating microbial interactions

Understanding microbe-microbe interactions is critical to predict microbiome function and to construct communities for desired outcomes. However, investigating these interactions poses a significant challenge due to the lack of suitable probing tools. The DP platform, with its ability to compartmentalize single cells and its assemblability, provides an ideal tool for studying microbial interactions. As a proof of concept, we used the DP platform to explore the interaction between Bacillus (an unreported strain isolated from the mucus layer of a slug by our group) and Salmonella. As depicted in Fig. 7A, 10 µL of Bacillus suspension was pipetted onto the center of a solidified agar medium sheet and evenly spread across the surface using a glass rod spreader. Then, the inoculated sheet was incubated at 37 °C for 3 h. Next, 107 CFU/mL of Salmonella suspension was prepared and loaded into a PicoArray device. After discretization, the device was overlaid with the agar medium sheet inoculated with Bacillus. Following incubation at 37 °C for 5 h, the Bacillus cells grew to dense microcolonies and there are clear zones around the microcolonies, indicating inhibition of Salmonella growth (Fig. 7B). This suggests an antagonistic relationship between Bacillus and Salmonella, where Bacillus exerts an inhibitory effect on Salmonella. The “digital” feature of the DP platform allows for precise, quantitative assessment of Bacillus antagonism against Salmonella. This was easily accomplished by dividing the number of microwells in the inhibition zone induced by a Bacillus microcolony by the number of microwells covered by the Bacillus microcolony (more details see Section S5 in the Supplementary Information). Furthermore, by measuring and comparing the antagonism levels of different Bacillus microcolonies (Fig. S5), we can also characterize the heterogeneity in cell-cell interactions. By quantifying spatial and temporal variations in microbial interactions (e.g., antagonism or synergism), the DP platform enables the identification of subpopulations with unique metabolic or competitive traits. For instance, in environmental samples, microbes exhibiting localized antagonism (e.g., inhibition zones around colonies, as in Fig. 7B) are prioritized candidates for antibiotic discovery, as such behaviors often correlate with secondary metabolite production50,51. This approach aligns with ecological principles where antibiotic synthesis is a competitive strategy in densely populated niches52.

Fig. 7
figure 7

Investigation of microbial interactions using the DP platform. (A) Workflow for investigating microbial interactions based on the DP method. (B) Representative bright-field microimages showing the inhibition zones against Salmonella formed around the Bacillus colonies after 5 h of co-incubation at 37 ºC. Initial concentration of Salmonella ~ 1 × 107 CFU/mL.

Full size image

Recovery of target microorganisms

A recurring desire of microbiologists is to be able to extract individual colonies of interest for further culture or analysis. This operation is particularly difficult in most microfluidic approaches and remains a blocking point for the adoption of microfluidics by biologists. Thanks to its reversible assembly, our DP platform can be readily used for recovering isolated target microorganisms for further characterization and scaling-up cultivation. To demonstrate the ability of the DP platform to recover target bacterial species, we prepared a two-strain bacterial suspension consisting of S. aureus and E. coli, loaded it into a PicoArray chip, and covered the chip with an agar sheet containing selective medium. The protocol was similar to that used in the experiment to isolate specific bacteria by coupling the DP chip with selective media (see the subsection of “enrichment and identification of specific bacterial species from mixed-species microbial communities”). To ensure microcolonies reached sufficient biomass for visible identification and manual recovery, a relatively extended incubation (~ 10 h) at 37℃ was performed. Note that, to eliminate the cross-contamination risk during recovery, the bacterial suspension must be diluted enough so that the probability of two single cells located in neighboring microwells or sharing the same microwell is extremely low. We randomly picked three microcolonies from the chip using sterile toothpicks and transferred them to three tubes of liquid medium for scale-up cultivation (Fig. 8A). To identify the recovered microorganisms, genomic characterization was performed by PCR amplification and 16 S rRNA gene sequencing. As expected, 16 S rRNA sequencing revealed that all three recovered clones were S. aureus (Fig. 8B), indicating that the recovered cultures are free from contamination and accurately represent the microorganisms enriched by the selective medium. These results suggest that isolation and recovery of target bacterial strains can be conducted effectively with the DP platform.

Fig. 8
figure 8

DP platform for in situ clonal culture and microcolony extraction. (A) Workflow of recovering microcolonies from the PicoArray chip for genetic characterization. (B) Neighbour-joining phylogenetic tree based on 16 S rRNA gene sequences of the recovered clones and the closest type strains sequences in the NCBI database. The red triangle represents the sequences of the recovered.

Full size image

Discussion

We present here an attach-and-use digital plating (DP) platform for microbiological analysis. The DP platform, integrating digital assay formats with traditional plate culturing principles, demonstrates significant advantages in terms of efficiency, accuracy, and versatility for microbial detection and analysis. One of its most outstanding features is the ability to encapsulate single cells or small colonies in massively large numbers of compartments for growth, monitoring, and selection. This provides several benefits for microbiological detection and analysis, including: (i) confining bacterial growth in ultra-small volumes, which allows rapid accumulation of metabolic products or secreted molecules, thus significantly reducing the time required for microbial detection53. This rapid turnaround is particularly beneficial in clinical settings where timely identification of pathogens can be critical for patient outcomes; (ii) isolating individual bacteria from complex communities, which eliminates cultivation bias caused by competition and inhibitory effects, thus facilitating the recovery of rare or slow-growing microorganisms from complex ecosystems54; and (iii) massively analyzing large numbers of individual bacteria, which enables examination of phenotypic and genetic variability at the single-cell or small population level and high throughput screening of functional microorganisms55.

Another key feature provided uniquely by the DP platform is the flexibility offered by its assemblability, which allows for the replacement of the covering agar medium sheet to introduce multiple detection or stimuli reagents to microorganisms within picowells, thereby creating well-controlled and variable culture environments. This feature gives access to two new possibilities: (i) tracking the evolution of a population in controllably changing chemical environments, and (ii) performing more complex experimental protocols. Furthermore, this feature makes the DP platform compatible with commercial ready-to-use agar plates (more details see Section S6 and Fig. S6 in the Supporting Information). This compatibility offers several benefits: (i) facilitating seamless integration of the DP chip with existing laboratory protocols which allows researchers to leverage commercially available resources without the need for extensive modifications; (ii) eliminating the need for end-users to prepare the solid agar media themselves, which significantly speeds up lab processes; and (iii) ensuring high reproducibility and reliability since commercial ready-to-use agar plates are manufactured under controlled conditions.

Compared to prevalent droplet microfluidic methods which are limited to cultivating suspended growth microorganisms and are not conducive to observing and examining the growth morphology of microcolonies12, the DP platform combines the advantages of both solid culture medium surface-adherent cultivation and liquid culture medium suspension cultivation. This makes it suitable for cultivating a wider range of microorganisms. Moreover, the DP platform allows the expansion of single bacteria to bigger microcolonies since the reversible assembly of agar/chip allows overgrown microcolonies to spread to adjacent picowells, potentially increasing the sensitivity of biological readouts. In addition, downstream harvesting of selected microcolonies is more straightforward from discrete microcompartments than from droplets.

In addition, compared to commercial automation microbial detection systems (e.g., BacT/ALERT® 3D, BacT/Alert® Virtuo), which excel in standardized microbial culture workflows56, the DP platform prioritizes single-cell phenotyping and microenvironment flexibility. However, it currently lacks the automation and regulatory validation required for direct clinical deployment. Conversely, its low cost and compatibility with conventional agar media make it a versatile tool for research settings exploring microbial heterogeneity or uncultured species.

As summarized in Table S2, the DP platform occupies a unique niche between conventional plating and advanced microfluidic systems. Its ability to combine agar-based workflows with single-cell resolution addresses critical needs in clinical diagnostics (e.g., rapid AST) and environmental microbiology (e.g., isolating uncultured species). While previous studies have also combined agar/agarose pads with microscopy for the real-time observation of bacterial growth with single-cell resolution in a controlled environment57,58,59, it is extremely time-consuming and labor-intensive due to the small size of individual bacteria and narrow field of view under high microscopic magnification. Unlike these prior studies, our DP platform does not involve direct observation of individual bacterial cells. Instead, it monitors the microcolonies that form after a defined period of proliferation from single cells. As a result, the need for high-magnification microscopy is eliminated, making it much easier to locate and observe the target cells. Furthermore, the prior studies could only achieve dynamic environmental control by stacking multiple pads. In contrast, our DP platform offers greater flexibility, enabling dynamic control either through the stacking of multiple pads or via direct pad replacement. This agar pad replacement offers two unique advantages: (i) direct use of off-the-shelf commercial agar plates, eliminating custom pad fabrication; (ii) rapid full-environment switching (complete media exchange in < 30 s).​ While throughput remains lower than droplet microfluidics, the DP platform’s user-friendly operation, compatibility with commercial agar plates, and low operational costs make it particularly suited for laboratories prioritizing flexibility and cost-efficiency over ultra-high throughput.

In recent years, a few innovative technologies have emerged to address critical challenges in microbial analysis. For colony counting, the Geometric Viability Assay (GVA) developed by Meyer and co-workers eliminates serial dilutions by leveraging cone geometry60, achieving rapid CFU quantification with minimal waste. However, GVA’s 3D growth format limits compatibility with conventional agar-based workflows. In contrast, our Digital Plating (DP) platform integrates digital compartmentalization with replaceable agar sheets, enabling direct use of selective/differential media and single-cell recovery—features essential for clinical and environmental applications. For rapid antibiotic susceptibility testing (AST), single-cell imaging platforms like fASTest61 achieve phenotypic classification in < 10 min but require specialized microfluidics and high-resolution microscopy, restricting accessibility. DP circumvents such instrumentation by employing antibiotic-laden agar sheets. Furthermore, DP enables simultaneous AST for multiple antibiotics via replaceable agar sheets, a feature critical for combinatorial drug testing or resistance mechanism studies. DP’s agar-based format also supports long-term incubation (> 24 h), facilitating AST for slow-growing or biofilm-associated pathogens. While DP’s turnaround time is longer than fASTest, its agar-based design aligns with entrenched laboratory protocols, offering a balance between innovation and practicality. Together, these comparisons highlight DP’s unique niche: bridging digital quantification with traditional microbiological methods to support diverse research and diagnostic needs without relying on complex instrumentation.

While the DP platform offers significant advantages in flexibility and single-cell resolution, it is important to acknowledge its current limitations. First, manual handling during agar sheet replacement and microcolony recovery introduces variability, particularly in high-throughput workflows. Second, the PDMS-based microwell array chips may limit large-scale deployment of the DP platform due to the difficulty of PDMS-based devices in mass production. Third, it currently lacks the regulatory validation required for direct clinical deployment. To address these challenges, future iterations will focus on automating key steps (e.g., robotic agar-sheet covering, machine learning-based colony recognition, and robotic colony picking), expanding compatibility with clinical specimens (e.g., sputum, blood), and developing scalable manufacturing processes of microwell array chips.

In conclusion, the technical innovations of the DP platform will open new possibilities to various fields of microbiology, from defining single-cell phenotypic and genetic heterogeneity to investigating spatiotemporal dynamics of microbial communities, from precise quantitation of microbiota to systematically deciphering microbial interactions, from isolating rare and uncultured microbes to selectively extracting improved microbial strains. We believe that, attributed to its unique advantages, the DP platform will serve as a versatile and powerful tool for microbiologists to explore the world of microorganisms.