DOI:https://doi.org/10.65281/709938
Hanlin Wang1 *
School of Medical Devices, Shanghai University of Medicine & Health Sciences, Shanghai 201318, China;
Corresponding Email： wanghl@sumhs.edu.cn
Abstract
Single-cell printing has become an important technology for automated cell isolation in applications such as monoclonal cell line development, antibody discovery, and single-cell functional analysis. However, existing approaches often rely on oil-based microfluidics, fluorescence labeling, or complex instrumentation, which limit downstream compatibility, increase operational complexity, and constrain accessibility. Here, we report a compact, oil-free single-cell printing platform that integrates piezoelectric drop-on-demand droplet generation with real-time, image-based morphological gating for label-free cell identification. In contrast to conventional image-guided systems that rely on simple thresholding, the proposed method employs multi-parameter descriptors, including cell size, circularity, and solidity, to classify droplet occupancy and selectively deposit verified single-cell droplets into standard microwell plates. The system enables continuous generation of highly monodisperse aqueous droplets and automated closed-loop control of droplet sorting and deposition. Using Chinese hamster ovary (CHO) cells as a model system, the platform achieves single-cell encapsulation purities of up to 98.2% and a full-plate average of 94.3%, while maintaining high cell viability across operating conditions. The printed single cells further demonstrate robust long-term proliferation, forming clonal colonies over 20 days of culture. These results establish an accessible and efficient single-cell printing strategy that combines oil-free operation, label-free detection, and high-purity isolation, offering broad potential for applications in biopharmaceutical development, rare-cell analysis, and single-cell-based screening workflows.
Introduction
Monodisperse small droplet dispensing is an area of great research interest due to its wide-ranging applications in single-cell analysis1, drug screening2, and tissue engineering3. Currently, the most extensively employed use of monodisperse droplet dispensing is the isolation of single cells, enabling assays critical for generating monoclonal antibodies4, elucidating disease mechanisms in cancer research5, and advancing personalized medicine at the patient’s cellome level6. For all these applications, the essential first step is the effective separation of individual living cells from a cell suspension. However, most existing single-cell encapsulation technologies depend heavily on oil-based systems to generate continuous droplets. This oil dependency limits direct transfer of single cells into common thermocyclers for PCR, prevents straightforward extraction of individual cells for downstream analysis, and restricts compatibility with analytical methods other than PCR-based assays7.
To overcome these limitations, oil-free, in-air droplet generation technologies have been developed for single-cell dispensing, including flow cytometry methods such as fluorescence-activated cell sorting (FACS8) and various single-cell printers (SCPs9). Although these in-air droplet generation platforms offer flexibility regarding the choice of target reaction vessels (e.g., microwell plates, PCR tubes), minimize excess sample volume around selected cells, and support automated single-cell handling, they typically remain expensive, bulky, or inconvenient to use. Therefore, there is a clear demand for an affordable, compact, and user-friendly in-air droplet dispensing technology suitable for single-cell encapsulation, accessible to a wide range of biomedical laboratories.
Existing single-cell printers are commonly equipped with high-profile optical detection modules, precision motorized stages, and custom-fabricated chips, which together drive up the instrument cost and footprint10. Apart from the hardware, many platforms require fluorescent labelling of target cells prior to sorting, which introduces extra sample preparation, may perturb cell physiology, and precludes truly label-free workflows11. In addition, the detection algorithms adopted by most image-guided dispensers rely on simple intensity thresholding or single-parameter blob counting, which are vulnerable to debris, morphological irregularities, and variable suspension backgrounds. As a result, the single-cell purity reported for image-guided SCPs has historically remained in the range of 80%–90%, leaving a meaningful space for improvement in both hardware compactness and detection fidelity12.
Apart from the isolation step itself, the practical value of any single-cell dispensing platform largely depends on the downstream fate of the dispensed cells. In antibody discovery, cell-line development, and single-cell functional screening, the isolated cells must be subsequently cultured, expanded into clonal populations, and analyzed by secretion assays, phenotypic imaging, or molecular profiling13. This requires the dispensing process to preserve not only acute membrane integrity but also long-term proliferative capacity14. Similar considerations apply to emerging clinical applications involving rare-cell recovery, such as the isolation of circulating tumour cells (CTCs) for non-invasive cancer monitoring and the retrieval of patient-derived cells for personalized therapy development, where each recovered cell must remain viable and traceable for subsequent molecular or functional readout15. However, handling-induced mechanical stress, residual oil carryover, and encapsulation inefficiency remain persistent bottlenecks that compromise both assay fidelity and clonal outgrowth.
In this work, a compact, oil-free, in-air single-cell printer is developed and applied for high-purity single-cell isolation and downstream clonal expansion. The platform integrates piezoelectric drop-on-demand droplet ejection with a real-time, image-based morphological gating module, and performs three coordinated functions: (i) continuous generation of monodisperse aqueous droplets from a cell suspension through a hydrophilically treated glass–polymer dispensing chip; (ii) real-time classification of droplet occupancy using multi-parameter morphological descriptors, including cell size, circularity, and solidity, extracted from images acquired at the nozzle outlet; and (iii) automated deposition of verified single-cell droplets into predefined wells of a standard 384-well microplate, with empty and multi-cell droplets diverted to a waste reservoir. The droplet monodispersity, printing accuracy, and single-cell encapsulation purity are systematically characterized across a range of driving voltages. Chinese hamster ovary (CHO) cells, one of the most widely used mammalian cell lines in biopharmaceutical research and monoclonal antibody production, are chosen to validate the platform. The results demonstrate that the printer generates monodisperse droplets with a coefficient of variation (CV) of 0.350%, achieves single-cell purities of up to 98.2% and a full-plate average of 94.3%, preserves post-dispensing cell viability across the operating voltage range, and supports long-term clonal expansion of dispensed single cells into confluent colonies over 20 days. This work provides a practical and accessible tool for oil-free, label-free single-cell isolation and its downstream biomedical applications.
- Materials and methods
Fig.1 Schematic overview of the Soundpen™ CB. The system begins with the introduction of a cell suspension into the platform, where real-time detection and acoustic actuation enable precise, on-demand droplet formation. Single cells are encapsulated into aqueous droplets and accurately dispensed into downstream collection formats such as culture dishes or plates, supporting high-throughput and label-free single-cell handling.
2.1 Droplet separation and detection algorithms
The single-cell printer developed in this work (Soundpen™ CB) performs three coordinated functions: (i) continuous generation of uniform aqueous droplets from a cell suspension, (ii) real-time recognition of droplet occupancy to distinguish single-cell droplets from empty or multi-cell droplets, and (iii) automated deposition of single-cell droplets into designated microwells while diverting the rest into a waste reservoir. Fig. 1(a) illustrates the overall workflow and principal components of the platform, including the sample chamber, piezoelectric actuator, nozzle, and downstream single-cell encapsulation process.
The core of the printer is a disposable dispensing chip driven by a piezoelectric actuator. The chip consists of an upper sample chamber connected to a tapered microchannel that narrows toward the nozzle outlet, a geometry that promotes smooth convergence of the suspension flow and facilitates reproducible droplet pinch-off. The chip is fabricated from glass and polymer using standard microfabrication processes, and the inner surface of the microchannel is rendered highly hydrophilic by oxygen-plasma treatment followed by passivation to ensure stable fluid handling and to prevent nozzle fouling over typical experimental durations of 30–60 min. A region of interest (ROI) for optical monitoring is defined directly at the nozzle exit, allowing real-time inspection of each droplet for cell size, morphology, and circularity. By tuning the electrical driving signals applied to the piezoelectric actuator, droplets in the range of 100–600 pL can be reproducibly generated. The dynamics of droplet formation is further characterized by high-speed imaging, which reveals a consistent sequence of necking, elongation, and pinch-off events (Fig. 2b).
The ROI is continuously monitored by an image-based detection module that classifies each droplet in real time. Prior to the dispensing experiment, a pre-profiling procedure is performed on the sample cells to establish cell-specific gating thresholds. Several hundred representative cells are imaged under identical optical conditions, and their morphological features are extracted through edge enhancement and threshold segmentation. From these measurements, distributions of cell size, circularity, and solidity are obtained, and three gating thresholds are defined accordingly: the size range is set between the 10th and 90th percentiles (P10–P90) of the profiling dataset to exclude debris and abnormally large aggregates, the circularity threshold is fixed at 0.85 to reflect the near-spherical geometry of single suspended cells, and the solidity threshold is fixed at 0.95 to ensure compact morphology without concavity. During operation, each droplet passing through the ROI is analyzed in real time by the same image-processing pipeline. For every segmented object, three morphological descriptors are computed on the fly: the equivalent diameter , where A is the projected area; the circularity , where P is the perimeter, and the solidity , where is is the area of the convex hull. Only when all three descriptors simultaneously satisfy their pre-profiled acceptance windows, the droplet is classified as a single-cell droplet (N = 1). In this case, the droplet is deposited into the current well of the microwell plate and the nozzle advances to the next position. If any descriptor fails to meet its threshold (N ≠ 1), the droplet is diverted into the waste reservoir and the nozzle remains aligned with the same well until a valid single-cell droplet is obtained. Once a droplet containing a single cell is successfully deposited, the cycle restarts, and the procedure is repeated until the entire plate is filled with single-cell droplets.
2.2 Cell culture
CHO-K1 cells (CL-0058, Procell Life Science & Technology, Wuhan, China) were maintained in CHO-K1 cell-specific medium (CM-0058, Procell) supplemented with 10% fetal bovine serum (164210-50, Procell), 1% penicillin–streptomycin solution (PB180120, Procell), and 2 mM L-glutamine (PB180420, Procell) under standard incubation conditions (37 °C, 5% CO₂). Cells were harvested once they reached approximately 80% confluence, then washed, trypsinised with 0.25% trypsin-EDTA (PB180226, Procell), centrifuged at 1,000 rpm for 5 min, and re-suspended to obtain a cell suspension at the desired concentration. Cells were routinely passaged by trypsinisation every 2 to 3 days to maintain healthy growth.
2.3 Chip cleaning and sterilization
Once the disposable dispensing chip was fitted into the platform, the following procedure was performed to ensure aseptic conditions prior to operation. First, 15 mL of 70% (v/v) ethanol was pipetted into the sample reservoir, and the dispenser was driven in high-frequency mode until the solution was completely ejected. Next, 15 mL of sterile phosphate-buffered saline (PBS) was loaded and dispensed in the same manner to remove residual ethanol. Finally, 20 mL of the prepared CHO cell suspension was introduced into the reservoir, and continuous dispensing was carried out until cells were clearly observed passing through the nozzle region under the camera view. Each chip was used only once and discarded after the experiment to avoid cross-contamination between samples.
2.4 Cell printing and growth
Droplets classified as single-cell droplets by the morphological gating module were printed into individual wells of a 384-well flat-bottom microplate (Corning, USA), with each well pre-filled with 20 μL of culture medium. After printing, each well was immediately inspected under an inverted microscope, and the actual number of cells per well was recorded. The single-cell purity was defined as the ratio of wells containing exactly one cell to the total number of printed wells.
Only wells that actually contained a single cell were included in the subsequent culture observation. The microplates were incubated at 37 °C and 5% CO₂, with the culture medium refreshed every 2–3 days, and brightfield images of each well were acquired on days 1, 3, 5, 7, 14, and 20 using an inverted microscope. The single-cell survival yield was defined as the ratio of wells in which the deposited single cell had undergone division by day 2 to the total number of wells confirmed to contain a single cell after printing. Long-term clonal expansion was evaluated by monitoring the progressive growth of the deposited single cells into colonies over 20 days.
- Result
3.1 Working principle
The Soundpen™ CB developed in this work enables oil-free, label-free single-cell dispensing by integrating piezoelectric drop-on-demand droplet ejection with real-time, image-based morphological gating. As illustrated in Fig. 1, a cell suspension is loaded into the platform, where individual cells are identified in real time and acoustically ejected as aqueous droplets into downstream collection formats such as culture dishes or multi-well plates, supporting high-throughput and label-free single-cell handling without any oil phase or fluorescent labelling.
Figure 2 .Acoustic single-cell dispensing system and droplet characterization.
(a)Schematic of the single-cell printing system. Real-time AI-based image analysis monitors cells within the region of interest (ROI). Upon detection of a target single cell, a piezoelectric actuator is triggered to eject a droplet into the collection well. In contrast, droplets identified as containing no cell or multiple cells are diverted to the waste region, enabling on-demand and selective single-cell encapsulation. (b) High-speed imaging reveals the time-sequence droplet ejection process under acoustic actuation. A complete droplet is formed within 20 ms, demonstrating rapid response capability. (c) Each voltage condition corresponds to droplet formation from specific chip–cell combinations. Although optimal voltages exist for different experimental setups, the overall trend shows no clear linear correlation between voltage and droplet size. Bars represent mean ± SD. Color distinguishes devices. (d) (d) Printing accuracy of droplet ejection under varying voltages. Blue dots represent the normalized accuracy of droplet centering at each voltage level, with error bars indicating standard deviation. The red dashed curve shows the numerical fit based on an exponential saturation model. Representative droplet array images at each voltage are shown, illustrating the improvement in spatial precision with increasing voltage. (e) The distribution of single-cell, multi-cell, and empty (null-cell) droplets was evaluated across voltages ranging from 55 V to 75 V. As voltage increased, the proportion of single-cell droplets significantly improved from 88.5 ± 2.3% at 55 V to 98.2 ± 1.1% at 75 V. Correspondingly, the fractions of multi-cell and null-cell droplets decreased from 6.7 ± 1.8% and 4.8 ± 2.1% at 55 V to 1.1 ± 0.6% and 0.7 ± 0.5% at 75 V, respectively.
The acoustic droplet printer developed in this work integrates piezoelectric drop-on-demand droplet ejection with real-time, image-based morphological gating, enabling oil-free and label-free single-cell dispensing into open microwell plates. The overall workflow and principal components of the platform are illustrated in Fig. 2(a), including the dispenser chip, the dispensing signal input, the on-PC AI-based detection module, and the downstream collection of printed single cells and waste droplets.
Figure 3 Experimental validation of acoustic single-cell printing using CHO cells.
(a) A 384-well layout displaying droplet images acquired under a 10× objective. Each well was sequentially imaged to evaluate the cell occupancy within droplets. Among all droplets, 362 were identified as single-cell, 14 as multi-cell, and 8 as null-cell. Multi-cell droplets are marked with red boxes, and null-cell droplets with yellow boxes. Representative examples are highlighted and enlarged on the right: F4 (multi-cell, red box), J12 (null-cell, yellow box), and C21 (single-cell, black box). Single-cell droplets contain one clearly resolved cell, multi-cell droplets show clustered cells, and null-cell droplets lack visible cellular content. (b) The histogram shows the droplet volume distribution under 39 V using CHO cells. The mean droplet diameter was 150.2 pL with a coefficient of variation (CV) of 0.350%, indicating highly uniform droplet formation. The red curve represents a fitted normal distribution, and the histogram was normalized to probability. (c) Dual-layer donut charts illustrate the droplet composition and cell viability across four voltage conditions (48V, 57V, 62V, and 80V). The outer ring indicates the proportion of each droplet type: single-cell, multi-cell, and null-cell. The inner ring further divides viable and non-viable subtypes within single-cell and multi-cell droplets. Percentages are labeled on the respective segments for clarity. Notably, the proportion of single-cell droplets remains consistently high (>85%), with a predominance of viable cells in both single and multi-cell groups across all voltages. (d) Time-lapse microscopy images show the growth of a single CHO cell printed under optimized conditions. The cell maintained high viability and gradually proliferated into a visible colony over 20 days. By Day 7, clonal expansion was evident, and by Day 20, a dense cell cluster had formed. This confirms that the single-cell printing process preserves cellular activity and supports long-term survival and proliferation.
The core of the printer is a disposable dispenser chip, which consists of an upper sample chamber connected to a tapered channel narrowing toward the nozzle outlet. This geometry promotes smooth convergence of the cell suspension flow and facilitates reproducible droplet pinch-off. Upon applying a tuned dispensing signal to the chip, the liquid meniscus at the nozzle first protrudes (t ≈ 2 ms), elongates into a ligament (t ≈ 6–10 ms), undergoes Rayleigh-type necking (t ≈ 12 ms), and finally detaches as a discrete spherical droplet by t ≈ 16 ms (Fig. 2b). The full cycle of necking, elongation, and pinch-off is confined within a single actuation period, and no satellite droplets were observed under the optimized driving waveform, ensuring that only one well-defined droplet is released per trigger event, which is a prerequisite for accurate single-cell deposition.
A region of interest (ROI) is defined directly at the nozzle outlet and continuously monitored by an on-PC AI-based detection module. For each droplet passing through the ROI, a multi-parameter morphological gating algorithm simultaneously evaluates three descriptors—size, circularity, and solidity—and classifies the droplet into one of three categories: NULL (empty droplet), Single (single-cell droplet), and Multi (multi-cell droplet). Only droplets classified as Single are deposited into the target microwell plate as printed cells, whereas NULL and Multi droplets are diverted into the waste reservoir. This closed-loop design ensures that each well is seeded with exactly one verified single cell, and the process repeats automatically until the entire plate is populated.
3.2 Monodispersity characterization
To establish the operational envelope of the acoustic droplet printer, droplet generation was systematically characterized on 22 independently fabricated dispensing chips (Device 1–Device 22), each operated at its own working voltage (Fig. 2c). The optimal working voltages of the 22 chips were distributed between approximately 19.5 V and 55 V, and the corresponding droplet volumes fell in the range of approximately 85–270 pL. The error bars associated with each chip were consistently small, indicating good intra-chip reproducibility in droplet volume.
Notably, no clear monotonic correlation was observed between the working voltage and the droplet volume across devices—chips operated at similar voltages could produce markedly different droplet volumes. This indicates that each chip exhibits a characteristic response jointly determined by its nozzle geometry, fluidic resistance, and the viscosity of the loaded suspension, and therefore possesses its own optimal driving voltage. Consequently, a pre-calibration step is required for every new chip prior to any printing experiment, in order to identify the working point at which it stably generates droplets of the desired volume. In the present work, this pre-calibration was performed automatically by a built-in volumetric feedback loop, typically completed within several minutes for each new chip, which also mitigates chip-to-chip variation arising from microfabrication tolerances.
3.3 Printing accuracy and single-cell purity
The spatial precision of droplet deposition was evaluated over a driving voltage range of 56 V to 70 V (Fig. 2d). At 56 V, the normalized printing accuracy was only about 0.30, with microscopy images revealing clearly scattered deposition patterns. As the driving voltage increased, the accuracy rose rapidly and reached approximately 0.9 at around 68–70 V, at which point the printed droplet arrays appeared tightly clustered on the substrate. The voltage–accuracy relationship was well described by an exponential saturation model:
With , and .This trend can be interpreted from a kinetic-energy perspective: at low driving voltages, the ejection velocity is small and the droplet trajectory is easily perturbed by residual nozzle oscillations and asymmetries in meniscus pinch-off; as the driving amplitude increases, the kinetic energy imparted to the droplet rises markedly, and the droplet flies ballistically toward its target along a more stable trajectory. The saturation of the accuracy at approximately 0.9 rather than 1.0 reflects irreducible error sources such as nozzle-to-substrate stand-off variation. Based on these results, a driving voltage of approximately 70 V was adopted in the subsequent printing experiments to balance printing accuracy and operational stability.
3.4 Single-cell encapsulation purity characterization
Building on the printing accuracy characterization, the single-cell encapsulation purity was further evaluated across driving voltages ranging from 55 V to 75 V (Fig. 2e). At each voltage, the dispensed droplets were classified by direct microscopic inspection as single-cell, multi-cell, or null-cell, and their respective proportions were quantified.
The results showed a clear dependence of single-cell purity on the driving voltage. At 55 V, single-cell droplets accounted for approximately 89% of all dispensed events, while multi-cell and null-cell droplets represented approximately 6% and 7%, respectively. As the voltage increased, the single-cell fraction rose progressively, reaching approximately 97% at 70 V, at which point the multi-cell and null-cell fractions dropped to their minimum values of approximately 2% and 3%, respectively. Further increasing the voltage to 75 V produced no significant change in any of these metrics.
This trend is consistent with the improvement in printing accuracy: at higher driving voltages, the droplet kinetic energy increases and the flight trajectory become more stable, enabling the droplet to land accurately at the target position. Meanwhile, the higher ejection velocity shortens the latency window between detection and ejection, reducing the occurrence of multi-cell and null-cell droplets. Taking the printing accuracy (Section 3.3) and the single-cell encapsulation purity together, 70 V was selected as the representative operating voltage for all subsequent CHO cell printing experiments, at which both metrics approach their respective plateaus, balancing dispensing precision with operational stability.
3.5 CHO cell printing and volume distribution
Building on the preceding characterizations, a CHO-K1 cell suspension was printed into a standard 384-well flat-bottom microplate under the optimized operating conditions, and brightfield images of all 384 wells were acquired sequentially after printing (Fig. 3a). Among the 384 printed wells, 362 contained single-cell droplets (e.g., C21, black box), 14 contained multi-cell droplets (e.g., F4, red box), and 8 contained null-cell droplets (e.g., J12, yellow box), corresponding to plate-wide fractions of 94.3%, 3.6%, and 2.1%, respectively.
It is worth noting that the plate-wide single-cell fraction (94.3%) is slightly lower than that measured under comparable voltage conditions in Fig. 2e. This difference can be attributed to two factors. First, full-plate printing involves hundreds of consecutive ejections and a substantially longer operation time, during which the cell suspension may experience sedimentation or local concentration drift within the sample chamber. Second, the plate-wide statistics are based on a much larger sample size and therefore more faithfully reflect the long-run average performance of the printer rather than a peak condition captured in a shorter voltage-sweep experiment. Accordingly, the full-plate purity, though lower than the peak observed during voltage scanning, represents a more realistic estimate of the steady-state performance under routine use.
In parallel, the droplet volume distribution was measured under loaded CHO suspension conditions (Fig. 3b). At a driving voltage of 39 V, the measured droplet volumes followed a narrow normal distribution, with a mean value of 150.2 pL and a coefficient of variation (CV) of only 0.350%, indicating that the printer could stably generate highly monodisperse picolitre-scale droplets even when loaded with a cell suspension. This CV is more than one order of magnitude smaller than that reported for inkjet-based dispensers operating in comparable volume regimes.
In the context of single-cell encapsulation, the stability of droplet volume affects not only geometric reproducibility but also the statistical behavior of cell loading. At a fixed cell concentration, the droplet volume CV propagates through the Poisson-loading process into the well-to-well distribution of cell counts, so tighter volume uniformity translates into more predictable and reproducible single-cell occupancy across the entire microplate. This feature makes the printer particularly well suited to applications that demand high well-to-well consistency, such as large-scale clonal screening and single-cell quantitative analysis.
3.6 Cell viability
Beyond dispensing purity, the impact of the printing process on cell viability and proliferative capacity is equally critical for the practical value of the platform. To this end, the droplet composition and cell viability after printing were comprehensively analyzed under four representative driving voltages of 48 V, 57 V, 62 V, and 80 V (Fig. 3c). In each dual-layer donut chart, the outer ring shows the proportions of single-cell, multi-cell, and null-cell droplets, while the inner ring further subdivides the single-cell and multi-cell populations into viable and non-viable subtypes. The results show that the single-cell droplet fraction remained above 85% under all four voltage conditions; more importantly, viable cells constituted the overwhelming majority within both the single-cell and multi-cell populations. Even at the highest voltage of 80 V, the viable-cell fraction within single-cell droplets remained around 90%, indicating that the acoustic driving amplitudes tested here did not cause significant damage to cell membrane integrity.
To further evaluate the long-term proliferative capacity of the printed cells, individual CHO cells deposited into 384-well plates under the optimized conditions were tracked by brightfield microscopy for up to 20 days (Fig. 3d). On day 1, only a single cell was visible in each well; by days 3–5, the cells had begun to adhere and divide, and by day 7 early clonal colonies were clearly discernible. Continued culture to days 14–20 yielded progressively denser cell clusters, ultimately forming visually confluent colonies. These results confirm that the acoustic droplet printing process preserves not only the immediate viability of the dispensed cells but also their long-term proliferative and clone-forming capacity, directly supporting downstream applications such as clonal expansion, monoclonal antibody screening, and other biomedical workflows that require clonally derived samples.
- Discussion
In this work, an acoustic droplet printer integrating piezoelectric drop-on-demand droplet ejection with real-time, image-based morphological gating was developed and applied for oil-free, label-free single-cell isolation and downstream clonal expansion. By combining a hydrophilically treated glass–polymer dispensing chip with a multi-parameter morphological gating algorithm, the printer generated monodisperse picolitre droplets with a coefficient of variation of 0.350%, achieved single-cell encapsulation purities of up to 98.2% under optimized driving voltages and 94.3% averaged across a full 384-well microplate, preserved cell viability across the operating voltage range, and supported long-term clonal expansion of dispensed single CHO cells into near-confluent colonies over 20 days of culture. The overall performance validates the proposed architecture as a practical and scalable tool for single-cell dispensing in biomedical research and biopharmaceutical development.
Compared with conventional oil-based droplet microfluidic systems, the present printer generates free-flying aqueous droplets that are deposited directly into open microwell plates. This aqueous-phase operation eliminates the need for fluorinated carrier oils, surfactants, and post-dispensing de-emulsification steps, and thus naturally supports downstream workflows such as direct cell culture, in-well staining, and ELISA-based secretion assays without any intermediate transfer of cellular samples. Compared with fluorescence-activated sorting platforms, the multi-parameter morphological gating used here is fully label-free and operates only on bright-field images acquired at the nozzle outlet. This relaxes the requirement for fluorescent labelling of target cells, avoids the associated sample preparation steps and potential perturbations to cell physiology, and enables the platform to process unmodified suspensions directly from routine cell culture. Apart from these advantages, the platform operates at ambient temperature on a compact footprint and is compatible with any standard 384-well microplate without further modification, making it suitable for deployment in routine biomedical laboratories with limited specialized infrastructure.
The multi-parameter morphological gating algorithm constitutes one of the key design features of the platform. By requiring the simultaneous satisfaction of three independent morphological descriptors, including cell size within the P10–P90 range, circularity ≥ 0.85, and solidity ≥ 0.95, the algorithm effectively rejects three common failure modes that confound single-parameter gating schemes: (i) cellular debris, which typically passes a size filter but fails the circularity criterion; (ii) small cell doublets, which may satisfy the circularity criterion but fail the size or solidity criterion; and (iii) out-of-focus background particles, which fail the solidity criterion. In addition, the pre-profiling procedure adaptively tailors the three gating windows to each new cell type, eliminating the need for manual parameter tuning when the sample changes. This adaptability is particularly useful for biomedical laboratories handling heterogeneous sample types, such as CHO, hybridoma, HEK293, and primary B cells, in which cell size and morphology vary appreciably.
The characterization results further revealed that the driving voltage modulates two distinct device-level metrics, i.e., the spatial printing accuracy and the single-cell encapsulation purity, through related but not identical mechanisms. The spatial accuracy improves with voltage primarily because the increased droplet kinetic energy reduces the relative contribution of trajectory perturbations, which is a purely fluid-dynamical effect. The single-cell purity, in contrast, improves because the sharper pinch-off at higher voltages generates tighter droplet volume distributions and shortens the detection-to-ejection latency window, both of which reduce the rate of cell mis-classification by coupling fluid dynamics to the operation of the gating algorithm. Notably, the two trends saturate at slightly different voltages, with the accuracy plateauing near 70 V, while the purity continues to improve marginally up to 75 V. At higher driving amplitudes such as 80 V, the viable single-cell fraction drops by approximately 7 percentage points, with a concurrent increase in the fraction of null droplets, indicating that overly aggressive driving begins to compromise both cell viability and droplet ejection stability. Taken together, these observations indicate that 70 V offers a favourable trade-off among spatial accuracy, single-cell purity, and preserved cell viability, and the practical working voltage may be further tuned for specific applications, for example, slightly higher values for clonal cell-line development where purity is critical, or slightly lower values for more fragile primary cells where shear minimization is preferred.
The combination of label-free operation, oil-free dispensing, full-plate single-cell purity above 94%, and preserved clonogenicity renders the platform well suited for a range of biomedical workflows. In monoclonal antibody discovery, the dispensed B cells or hybridoma cells can be cultured directly in assay-compatible plates and subsequently screened for antibody secretion by ELISA, without the need for intermediate transfer steps that often reduce recovery and perturb cell state. In CHO cell-line development for biologics manufacturing, the platform provides an image-based record of monoclonality at the moment of deposition, which directly addresses a key regulatory requirement and may substantially compress the timeline of clonality confirmation compared with conventional limiting-dilution cloning. In rare-cell recovery scenarios, such as the isolation of circulating tumour cells (CTCs) from peripheral blood or the retrieval of patient-derived cells for personalized therapy development, the morphological gating strategy is naturally compatible with the deposition of individual cells into traceable positions, which facilitates downstream molecular or functional analysis. Apart from these cell-isolation-centred applications, the platform can also be used to dispense cells into wells pre-loaded with siRNA, sgRNA libraries, or candidate compounds, enabling single-cell-level functional readouts that complement transcriptomic screening.
Several limitations of the present study should be acknowledged. First, only CHO cells were evaluated in this work. Although CHO cells are industrially relevant and widely used in biopharmaceutical research, further validation with more mechanically fragile cell types, including primary lymphocytes, induced pluripotent stem cells, and patient-derived tumour cells, is still needed before extrapolating the performance to those contexts. Second, although the full-plate single-cell purity of 94.3% reported here exceeds typical values reported for earlier image-guided dispensers, it is not yet 100%. The residual multi-cell and null events most likely originate from out-of-plane cell stacking within the tapered microchannel, in which the channel depth permits more than one cell to overlap along the optical axis, producing false single-cell calls from a two-dimensional image-based detection scheme. This upper-bound limitation could in principle be mitigated by adopting a shallower microchannel, a confocal-style detection module, or a stereoscopic imaging configuration. Third, the throughput of the present system is limited by the AI inference latency together with the mechanical traversal time between wells, reaching approximately one well per second, which is considerably lower than FACS but remains adequate for most plate-based applications. Fourth, the AI detection module has been evaluated mainly on relatively clean cell suspensions, and its performance on high-debris samples, such as those obtained from enzymatic tissue dissociation, warrants further characterization. Finally, the current study does not include a direct side-by-side benchmarking against FACS or other commercial single-cell printers, and such comparisons will be valuable for more rigorously positioning the platform in future work.
Several extensions are particularly promising as future directions. First, integrating a fluorescence detection channel into the existing optical module would enable phenotypic pre-selection before deposition, for example, selecting antibody-secreting cells stained with a fluorescent antigen, thereby combining the gating specificity of FACS with the gentleness and oil-free nature of the present platform. Second, parallelizing the dispensing chip with multiple nozzles operating simultaneously could push the effective throughput to the level of tens to hundreds of wells per second, narrowing the gap with FACS for screening-scale applications. Third, direct integration with downstream readout platforms, such as single-cell RT-qPCR, on-chip ELISA, or imaging cytometry, would establish a closed-loop pipeline from cell suspension to assay readout, which would be particularly valuable for industrial antibody discovery and personalized cell therapy development.
- Conclusion
In this work, a compact acoustic droplet printer integrating piezoelectric drop-on-demand droplet ejection with real-time, image-based morphological gating was developed and applied for oil-free, label-free single-cell isolation and downstream clonal expansion. By combining a hydrophilically treated glass–polymer dispensing chip with a multi-parameter gating algorithm that evaluates cell size, circularity, and solidity simultaneously, the printer generated monodisperse picolitre droplets with a coefficient of variation of 0.350%, achieved single-cell encapsulation purities of up to 98.2% under optimized driving voltages and 94.3% averaged across a full 384-well microplate, preserved CHO cell viability across the operating voltage range, and supported long-term clonal expansion of dispensed single cells into near-confluent colonies over 20 days. Apart from the demonstrated performance, the platform offers distinct advantages in operational simplicity, label-free detection, and downstream compatibility, all while maintaining a compact footprint suitable for routine biomedical laboratories. This work provides a practical and accessible tool for high-purity single-cell isolation, with immediate applications in monoclonal antibody discovery, CHO cell-line development for biologics manufacturing, rare-cell recovery, and single-cell functional screening. Future work will extend the validation to diverse mammalian cell types, incorporate a fluorescence detection channel for phenotypic pre-selection, and integrate the platform with downstream analytical workflows such as ELISA and single-cell RT-qPCR to enable end-to-end single-cell pipelines for biomedical research and therapeutic development.
Reference
(1) Du, L.; Liu, H.; Zhou, J. Picoliter droplet array based on bioinspired microholes for in situ single-cell analysis. Microsystems & Nanoengineering 2020, 6 (1), 33. DOI: 10.1038/s41378-020-0138-2.
(2) Utharala, R.; Grab, A.; Vafaizadeh, V.; Peschke, N.; Ballinger, M.; Turei, D.; Tuechler, N.; Ma, W.; Ivanova, O.; Ortiz, A. G.; et al. A microfluidic Braille valve platform for on-demand production, combinatorial screening and sorting of chemically distinct droplets. Nat Protoc 2022, 17 (12), 2920-2965. DOI: 10.1038/s41596-022-00740-4
(3) Zhang, P.; Liu, C.; Modavi, C.; Abate, A.; Chen, H. Printhead on a chip: empowering droplet-based bioprinting with microfluidics. Trends in Biotechnology 2024, 42 (3), 353-368.
(4) Tiller, T.; Meffre, E.; Yurasov, S.; Tsuiji, M.; Nussenzweig, M. C.; Wardemann, H. Efficient generation of monoclonal antibodies from single human B cells by single cell RT-PCR and expression vector cloning. Journal of Immunological Methods 2008, 329 (1), 112-124.
(5) Bendall, S. C.; Nolan, G. P. From single cells to deep phenotypes in cancer. Nature Biotechnology 2012, 30 (7), 639-647.
(6) Irish, J. M.; Hovland, R.; Krutzik, P. O.; Perez, O. D.; Bruserud, Ø.; Gjertsen, B. T.; Nolan, G. P. Single Cell Profiling of Potentiated Phospho-Protein Networks in Cancer Cells. Cell 2004, 118 (2), 217-228.
(7) Thompson, A. M.; Paguirigan, A. L.; Kreutz, J. E.; Radich, J. P.; Chiu, D. T. Microfluidics for single-cell genetic analysis. Lab on a Chip 2014, 14 (17), 3135-3142,
(8) Crouch, E. E.; Doetsch, F. FACS isolation of endothelial cells and pericytes from mouse brain microregions. Nature Protocols 2018, 13 (4), 738-751
(9) Leibacher, I.; Schoendube, J.; Dual, J.; Zengerle, R.; Koltay, P. Enhanced single-cell printing by acoustophoretic cell focusing. Biomicrofluidics 2015, 9 (2), 024109.
(10) Gross, A.; Schöndube, J.; Niekrawitz, S.; Streule, W.; Riegger, L.; Zengerle, R.; Koltay, P. Single-cell printer: automated, on demand, and label free. J Lab Autom 2013, 18 (6), 504-518.
(11) Lulevich, V.; Shih, Y. P.; Lo, S. H.; Liu, G. Y. Cell tracing dyes significantly change single cell mechanics. J Phys Chem B 2009, 113 (18), 6511-6519
(12) Riba, J.; Gleichmann, T.; Zimmermann, S.; Zengerle, R.; Koltay, P. Label-free isolation and deposition of single bacterial cells from heterogeneous samples for clonal culturing. Scientific Reports 2016, 6 (1), 32837.
(13) Ammar, A. E. S. Single-cell cloning and its approaches. Frontiers in Cell and Developmental Biology 2025, Volume 13 – 2025, Review.
(14) Weinguny, M.; Klanert, G.; Eisenhut, P.; Jonsson, A.; Ivansson, D.; Lövgren, A.; Borth, N. Directed evolution approach to enhance efficiency and speed of outgrowth during single cell subcloning of Chinese Hamster Ovary cells. Computational and Structural Biotechnology Journal 2020, 18, 1320-1329.
(15) Visal, T. H.; den Hollander, P.; Cristofanilli, M.; Mani, S. A. Circulating tumour cells in the -omics era: how far are we from achieving the ‘singularity’? British Journal of Cancer 2022, 127 (2), 173-184.