Picoliter droplet array based on bioinspired microholes for in situ single-cell analysis

The division of aqueous samples into microdroplet arrays has many applications in biochemical and medical analysis. Inspired by biological features, we propose a method to produce picoliter droplet arrays for single-cell analysis based on physical structure and interface. A 0.9 pL droplet array with an RSD (relative standard deviation) less than 6.3% and a density of 49,000 droplets/cm2 was successfully generated on a PDMS chip (polydimethylsiloxane) from a micromachined glass mold. The droplet generation principle of the wetting behavior in the microholes with splayed sidewalls on the PDMS chip by liquid smearing was exploited. The feasibility of the picoliter droplets for bacterial single-cell analysis was verified by the separation of mixed bacteria into single droplets and isolated in situ bacteria propagation. Researchers in China have developed a method to rapidly and reliably separate a solution into an array of picoliter-scale droplets. A team led by Jia Zhou of Fudan University first made a negative image of the surface of a lotus leaf. This served as the basis for a micromachined glass mold which was then used to manufacture PDMS chips. The structure of these chips—a series of microholes with angled walls—causes liquid smeared across the surface to form droplets within the holes. The droplets have an average volume of 0.9 pL. The team tested their approach by using it to isolate single-cell droplets of bacteria from a mixed solution. The ability to rapidly, reliably, and affordably generate picoliter microdroplets will prove valuable for studies in single-cell genomics and proteomics, as well as other fields.


Introduction
The compartmentalization of aqueous samples into millions of independent microdroplet reactors has revolutionized chemistry and biology by enabling single-cell analysis 1,2 , high throughput screening 3 , and molecular diagnostics [4][5][6] . Shrinking the droplet size minimizes the consumption of precious samples (such as rare mutant alleles and noncultivable bacteria) 7,8 . For a given volume of sample with a smaller droplet size, the much greater number of droplets improves the statistical significance of the experimental results. With a few exceptions [9][10][11][12][13] , most microdroplet generators are continuous microchannels [14][15][16][17] , sealed chamber chips 18 , and chemically prepatterned surface chips 19,20 . Continuous microchannel and sealed chamber chips produce highly monodispersed droplet streams but require sophisticated precision pumps and surfactants to prevent coalescence 21,22 . The performance of droplet production may vary with the precise conditions of the equipment and the parts included in the equipment, such as sealed chips with necessary extra bonding or sealing. Compared with the above methods, the formation of droplet arrays on chemically prepatterned surfaces provides a high density and open structure with great potential for high-throughput applications, relying on a surface treatment to alter contact angles 23 . However, such treatment complicates the manufacturing process and increases the chemical complexity of the chip. As the droplet size decreases, the surface energy becomes more substantial, so reinforced treatments are required to form droplets below the picoliter range 24 .
Consequently, Jackman 25 and Thalladi 26 fabricated large arrays of microwells with a simple physical structure (without chemical surface treatment) 27-29 on polydimethylsiloxane (PDMS) and formed 2D arrays of microdroplets for genome sequencing. Embedding the liquid samples into these microwell arrays, however, adds further complications to experimental manipulations, such as requiring a bulk solution. From the angle of minimizing sample volume, such a loading method in the single-side-opened cylindrical microwells is not well matched to the actual biological demands. The issue hinders the potential for future applications requiring a minimum sample for rapid distribution of small, uniform volumes of solutions or suspensions into spatially well-defined microreactors. Herein, we report a rapid generation method of picoliter droplet arrays based on physically structured microholes, inspired by a reversed mold from a lotus leaf. The picoliter droplet array was applied for in situ bacterial propagation for single-cell analysis.

Results and discussion
Inspired by the lotus leaf mold, we successfully developed a microdroplet array using a glass mold. The droplet array is shown in Fig. 1.
The distribution of the microdroplet array formed from the glass chip is shown in Fig. 1a. Compared with the distribution of droplet volume from the lotus chip, it is much narrower and indicates that approximately 60% of the droplet volume ranges from 0.9 to 1 pL, as shown in Fig. 1b. Unlike using cylindrical microwells, we observed the reliable formation of the microdroplet array (see Supporting Information, video) followed by the smearing process. The size of such a microdroplet array, with the relative standard deviation (RSD) of the volume smaller than 6.3%, demonstrates that the volume of microdroplets smearing from the glass chip approaches homogeneity.
Glass molds with different sizes of microholes (10,20,45, and 80 μm in bottom diameter) with similarly splayed sidewalls were studied to improve the controllability of the microdroplet array. The gaps between the microholes of the glass chip in the top edge were all 35 μm, and the depth of etching was approximately 10 μm. Figure 2a shows pictures of the droplet array from each glass chip. The corresponding relationship between the different diameters of the microholes on the glass chip and the volume of the smearing droplet is shown in Fig. 2b. The volumes of droplets produced in the different diameters of microholes on the glass chip are 0.9, 3.7, 6.7, 29.5 pL, which increases with the increasing diameter nonlinearly within this range.
The mechanics of droplet formation in bionic microholes with splayed sidewalls can be discussed as follows.
As shown in Fig. 3, φ a and φ r represent the advancing angle and the receding angle on the surface of the PDMS, respectively. The previous wettability characterization shows that the angle φ a is much larger than the angle φ r On the other hand, when liquid recedes from the splayed sidewall, the corrugated surface results in an altered contact line of the droplet. The contact line continues to move, and it forms a narrow bridge between the microhole and the bulk aqueous solution, as shown in Fig. 3b. Finally, the narrow bridge is ruptured to form a separated microdroplet (see Fig. 3c). The sidewall of the bioinspired microhole has the same effect as the chemically prepatterned surfaces to pin the droplets to the three-phase contact line 30,31 , as seen in Supporting Information, Fig. S2. After comparing the maximum radius of the droplets from the top view of the glass chip, we also found that a larger radius (R) of the microhole is associated with a larger droplet volume. According to the geometric analysis (Supporting Information, Fig. S3), we estimated the volume of the droplet stayed in different diameters of microholes with and without considering the effect of the receding angle on the splayed sidewall, as shown in Table 1.
It is clear that the receding angle is of great significance in defining the volumes of the droplet from the glass chips featuring diameters and heights close to the lotus chip (chips 1 & 2), while it has little effect on those from glass chips with larger aspect ratios of the microholes (chips 3 & 4). This can be attributed to the large diameter, which means a large distance between the two splayed sidewalls of the microholes so that the influence of the receding angle on the splayed sidewall becomes small. Regardless,  b Schematic cross-section of the droplet formation process on the glass chip. c Top view of the droplet formation process the volume of the microdroplets increases with increasing radius of the microholes. However, we also observed in our experiments that the microdroplet array generated in the glass chip with a diameter of 80 μm was occasionally unstable, i.e., exhibiting small satellite droplets (Supporting Information, Fig. S4). Such results show that further study of the aspect ratio of the splayed microholes is necessary to obtain stable and reliable larger droplets volumes (~>30 pL).
Considering the differences of the microholes with splayed walls and straight walls, i.e., cylindrical wells, droplet formation using glass chips with a bottom diameter and depth of 10 μm for both was compared. The results demonstrated that the droplets could be stably generated in 100% of the 736 splayed microholes using chip 1, as shown in Fig. 1a. However, only liquid films were formed on the surface of cylindrical holes, and more than 40% of them broke within 8 s, which means that at least 40% of such cylindrical holes could not be filled with liquid by smearing. The liquid "film" actually cracked above the cylindrical microwells (similar behavior was also observed by other studies 32 ). Please refer to Fig. S5 (Supporting Information). The main difference between cylindrical and splayed holes is that in cylindrical holes, the liquid forms an unstable film that quickly ruptures, whereas the splayed holes yield stable droplets. Such results demonstrate that droplet formation on cylindrical wells is not stable by smearing. Similar results were also reported by ref. 25 .
A preliminary study of the repeatability of droplet forming using glass chips was also carried out. In our previous research, we found that chip 1 (with 10 μm as the bottom diameter of the microholes, similar to the microstructure of the bulges of the lotus leaf) showed optimum behavior to form a droplet array. Therefore, we used 10 chip 1 to show the level of repeatability in the drop size of the approach. All ten chips could produce droplet arrays well. When the velocity of smearing ranged from 0.97 to 1.39 mm/s, the average droplet volume was 1.15 pL using three chips, with an RSD of approximately 7.45%. The droplet sizes ranged from 0.9 to 6.1 pL at speeds from 0.08 to 2.15 mm/s. Such results indicate that further study of the effect of the velocity of smearing on the coefficient of variation of droplet volume is necessary.
The controllable and rapid formation of a microdroplet array is significant for quantitative particle separation 33,34 . We focused on the application in single-cell analysis, which has enormous potential in various fields, such as single-cell sequencing, cell-based detection assays, and protein expression, as well as microsensors 35 . However, facile separation of the single-cell level array from the cell culture medium is still a challenge at present 36 . In this paper, single-cell separation experiments were performed on S. aureus. Moreover, E. coli was added to the test sample to demonstrate the single-cell separation and in situ propagation in the case of mixed strains.
Through the staining process (Fig. 4a), we obtained the single-cell array on the glass chip (Fig. 4b). The number distribution of bacteria suggests that approximately 29% of the microholes contain only one bacterium (see Supporting Information, Fig. S6). The isolated bacteria in the droplet array were cultured in situ without mutual interference from the surroundings for further individual research 36,37 . Figure 5 demonstrates typical examples of single-cell proliferation utilizing the smearing of cell culture medium with mixed strains. It can be seen that there are mixed bacteria (those labeled with "Mixture"), nothing (those labeled with "0"), and E. coli and S. aureus in the microholes. Such results indicate that these isolated bacteria have been successfully proliferated for several generations, increasing the number of the same species for more biological applications. For example, the individual differences in bacteria may cause disequilibrium in the microholes, which is also one of the important concerns of biologists for exploring the differences between cells. Due to the Poisson distribution, we obtained not only isolated single bacteria but also a combination of multiple bacteria. Moreover, the randomly distributed combination of bacteria can bring greater convenience to the interaction relationship between high-throughput bacteria of various groups, which is one of the most attractive topics in biology 38 . Therefore, the rapid generation method of picoliter microdroplets based on the simple The black font datum is the most accurate estimation of the volume with the multiple methods physical interface is facile, sample-effective, and low-cost, which will be of great potential for the development of microdroplet arrays for biological analysis, such as singlecell genomics studies, aptamer selection, and nucleic acid analysis. The droplets produced by our alternative, simple methods are unconventional and can be implemented without using external, bulky laboratory devices such as pumps 39 . This method provides a new technical means for chemists, biologists, medical scientists, etc. to work in a wider range of applications.

Discussion
Inspired by a reversed mold from a lotus leaf, we proposed a method based on the physical structure for picoliter microdroplet generation and application in single-cell analysis, which avoids the complicated chemical treatment of the chip surface and complex operating equipment. In the experiment, by adjusting the size of the microholes with splayed walls, we successfully obtained a microdroplet array with volumes from 0.9 to 29.5 pL. The micromachined layout ensures an accurate droplet position and arrangement. We preliminarily generated a single bacterium array from a bulk aqueous solution containing multiple bacteria and propagated the bacterium in situ. Staining of the isolated bacteria showed that they had been successfully proliferated for several generations, indicating that the method has excellent biocompatibility for biological analysis, combinatorial chemistry, and related research. The presented strategy based on the rapid molding of the simple physical structure is simple, feasible, and sample-effective for tiny microdroplet generation.

Materials and equipment
Fresh lotus leaves were picked in the pond during midsummer (Shanghai, China). Sylgard prepolymer and curing agent (Sigma-Aldrich) were used to fabricate the

Generation of picoliter droplet array
After analysis of the picoliter droplet array generated from a piece of lotus leaf, a PDMS chip reversed from a micromachined glass mold with a structure mimicking that of a lotus leaf was proposed.

Picoliter droplet array from lotus leaf
A typical replication process 40 was applied to obtain a reproducible PDMS molding chip from a piece of lotus leaf, as shown in Fig. 6a. First, a 20:1 mixture of Sylgard prepolymer and curing agent was poured on a fresh lotus leaf to make a first negative replica. Then, we obtained a positive hardened replica by curing NOA 81 from the first negative replica. Ultimately, this NOA mold was used to manufacture many PDMS chips, namely, lotus chips, for the following experiments.
Smearing is a simple and fast way to handle droplets 41,42 , as shown in Fig. 6b. The lotus chip was first placed on a clean coverslip and showed exceptional adhesion even without glue. After adding the aqueous solution on the surface of the lotus chip, droplets were formed by prompt smearing on its surface using another coverslip. Then, the newly compartmentalized microdroplets were shrunk by dripping several drops of mineral oil on the preparation, which prevented the evaporation of the droplets. The oil also washed away the excess solution remaining on the surface of the lotus chip and reduced the specific surface area of the aqueous samples. Therefore, we idealize the droplets as spheres to approximate the volume 43 . The radius (r) of the droplet profile was measured under a microscope. The volume (V) of the droplets was calculated by the formula of a sphere, i.e., V = (4r 3 )/3. Finally, a third coverslip was placed over the preparation to provide additional protection against evaporation and airborne contamination. Figure 6c shows that the microdroplets are scattered randomly across the lotus chip. The distribution of droplet volume is wide and indicates that approximately 30% of the volume of the droplets ranges from 0.1 to 0.15 pL (Fig. 6d).

Picoliter droplet array from a micromachined glass mold
To further increase the controllability of the microdroplets and significantly promote promising applications, micromachining would be one of the feasible approaches. Through characterization by SEM and 3D microscopy, we obtained the surface morphology and cross-section of the lotus chip (see Supporting Information, Fig. S1). Unlike the cylindrical microwells proposed by Jackman et al. 25 , the lotus chip exhibits a corrugated array of smooth microholes. After fitting the cross-sectional profiles, the concrete microhole has a diameter of approximately 20 μm and a depth of approximately 10 μm. Therefore, we replaced the NOA mold with a micromachined one, i.e., a piece of glass with etched micropillars, which imitated the lotus leaf with the specific feature of a gradually larger diameter from bottom to top of the microbulges, i.e., splayed walls. The fabrication process of the PDMS chip from the glass mold, namely, glass chip, was the same as for the NOA mold. The schematic of the micromachined glass mold and the glass chip is illustrated in Fig. 7. Glass chips were applied to produce droplet arrays in later experiments, if not specially defined. H and R represent the depth of etching and the bottom radius of the microhole, respectively. The process of forming microdroplet arrays using the glass chip is similar to the lotus chip, as shown in Fig. 6a.

Separation and propagation of a single bacterium
We performed a single-cell separation experiment using S. aureus at a suitable concentration of approximately 1.5 × 10 8 copies/ml according to the McFarland Equivalence Turbidity Standard. We used the nutrient broth medium for propagating the two bacteria strains in a 37°C incubator for 20 h. To exhibit the results of single-cell analysis intuitively, we stained the bacteria without adding oil. The most significant difference from the normal staining process was that before each step of the staining, the glass chip needed to be baked to accelerate the evaporation of deionized water and facilitate the reentry of the stain into the microholes more effectively. We applied the glass chip in single-cell analysis, where the bacteria were isolated and trapped in the microholes. Additionally, we used S. aureus and E. coli to verify the picoliter droplets for single-cell analysis under non-single-species conditions.