Parallel DLD microfluidics for chloroplast isolation and sorting

Abstract

Chloroplasts are characteristic organelles of plant cells, essential for photosynthesis and various other metabolic processes, including amino acid, lipid, and hormone biosynthesis. Beyond their classical functions, chloroplasts have emerged as promising targets in biotechnology, particularly in therapeutic applications and biofuel production. However, their isolation remains technically challenging due to the limitations of conventional methods, which typically require complex protocols, specialized equipment, and trained personnel. Here, we present a microfluidic-based platform that enables size-based chloroplast separation using deterministic lateral displacement (DLD). Our device integrates four parallel DLD arrays, each with a distinct critical diameter (CD). This configuration enables bandpass filtering and allows the simultaneous isolation of chloroplasts of various sizes within a single device. Shared inlets and uniform flow conditions across all arrays enhance reproducibility compared to conventional techniques. Unlike traditional sucrose density gradients, which lack precise size-based separation, our system achieves separation efficiencies of 50–85% for chloroplasts ranging from 3 to 8 μm, with recovered fractions having purities of 17–66%. This platform provides a rapid, automated, and scalable solution for chloroplast isolation, with significant potential applications in plant research, biotechnology, and synthetic biology.

---

Introduction

Chloroplasts, the characteristic organelles of plant cells, exhibit diverse morphologies, from ovoid to discoid shapes, with dimensions varying between 1 to 10 μm depending on the species. Enclosed by a double membrane and containing their own extranuclear DNA, chloroplasts possess a remarkable degree of functional semi-autonomy.¹ Their primary function is photosynthesis, the process by which light energy is converted into chemical energy, sustaining life on Earth. In addition, chloroplasts participate in a range of essential metabolic pathways, such as nitrate and sulfate assimilation, as well as the biosynthesis of essential compounds, including chlorophyll, amino acids, fatty acids, and carotenoids.^2,3 These functions are fundamental for plant growth, development, and survival. Beyond their central metabolic roles, chloroplasts also play key functions in plant responses to biotic and abiotic stress. They are active sites of reactive oxygen species (ROS) generation and contribute to cellular signaling and the synthesis of defense-related compounds that enable plants to cope with environmental conditions and pathogen attack.^4,5 Recent advances have expanded the potential applications of chloroplasts beyond their classical roles. For instance, algal chloroplasts have been explored as biofactories to produce anticancer immunotoxins.^6,7

While chloroplasts typically range in size from 1 to 10 μm, mature and fully functional chloroplasts in mesophyll cells of higher plants generally measure between 5 and 8 μm in diameter.^8,9 In contrast, chloroplasts within the 2–5 μm range often correspond to immature or differentiating plastids, such as those found in developing leaves, meristematic tissues, or in cells undergoing plastid-type transitions (e.g., proplastid to chloroplast, or etioplast to chloroplast).^10,11 These smaller chloroplasts are of particular interest in studies investigating plastid biogenesis, development, regulation, and stress-induced plastid remodeling.^12,13 However, conventional isolation techniques, such as sucrose density gradients, lack the precision required for the selective enrichment of these subpopulations, frequently yielding heterogeneous fractions that complicate downstream analyses. Consequently, effective methods for size-based separation of chloroplasts have remained elusive—a critical limitation for studying their developmental stages and the associated gene expression.

This growing scientific interest in chloroplast function and utility has driven the development of methods aimed at isolating chloroplasts while maintaining their structural integrity.¹⁴ These methods are essential for detailed biochemical, molecular, and physiological analyses, supporting efforts to elucidate chloroplast roles in plant biology, improve biofuel production efficiency, and enhance carbon sequestration strategies.^15–17 Chloroplasts extraction typically begins with the mechanical disruption of leaf tissue in a buffer solution containing osmotic stabilizers and essential cofactors.¹⁴ This is followed by controlled mechanical homogenization to release intact chloroplasts from the cells, filtration to remove larger debris and differential centrifugation to enrich the sample with chloroplasts, Fig. 1a. Final purification is commonly performed using density gradient centrifugation, often with Percoll or sucrose to separate organelles based on buoyant density. While effective, these conventional methods are time-consuming, require multiple centrifugation and washing steps, and specialized equipment and must be performed by trained personnel, which limits their accessibility and scalability,^18–20Fig. 1b.

Fig. 1 Chloroplast isolation: conventional vs microfluidic approaches. (a) Schematic of the multi-step conventional extraction process from spinach leaves, outlining key steps: cutting leaves, enzymatic treatment, homogenization, centrifugation, and washing steps. (b) Schematic depicting sucrose density gradient centrifugation for conventional chloroplast purification, followed by centrifugation and successive washing steps. (c) Schematic of the streamlined microfluidic isolation method, showing direct sample injection through a single inlet and simultaneous collection of four distinct sorted populations into separate vials.

Over the past decade, microfluidics has attracted increasing interest in the field of plant biotechnology, owing to its diverse applications, particularly in the development of “plants-on-a-chip” platforms.²¹ These systems have advanced the study of root development, microorganism interactions, and real-time monitoring of plant responses to biotic and abiotic stress under controlled conditions with high precision and reproducibility.^22–25 Despite the advancements achieved, critical areas remain to be explored, such as the isolation of plant cells and organelles in microfluidic devices. To date, no microfluidic devices have been specifically developed for chloroplast isolation, highlighting the need for novel strategies tailored to the unique properties of plant organelles. To address this challenge, we present a microfluidic device that applies DLD principles to achieve size-based separation of chloroplasts, Fig. 1c. Furthermore, our approach offers significant advantages, including a substantial reduction in processing time—from hours to minutes—by eliminating labor-intensive steps such as centrifugation and repeated washing. The device performs the entire isolation process on-chip, allowing for hands-free separation and passive collection at the designated outlets once connected and running. This semi-automated operation enhances reproducibility and scalability, making the system particularly well-suited for high-throughput applications, including gene expression studies.

Materials and methods

Materials

Phosphate buffered saline (PBS, 14190144), 1,4-ditiotreitol (DTT, D0632-5G), 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES, H3375-250G), Pluronic F-127 (P2443), phenylmethanesulfonyl fluoride (PMSF, P7626-5G), bovine serum albumin (BSA, A7030-10G), propylene glycol methyl ether acetate (PGMEA, 484431), isopropanol (IPA, V000139), chlorotrimethylsilane (386529), and polystyrene particles (sizes: 2 μm, 80177; and 4 μm, 81494) were obtained from Sigma-Aldrich, USA. Sorbitol (S1876-500), ascorbic acid (A92902-100G), magnesium chloride (MgCl₂, M8266-100G), and 4-morpholinepropanesulfonic acid (MOPS, M1254-25G) were acquired from Sigma life science, Germany. Red-stained aqueous fluorescent particles (size: 3 μm, R0300) and Fluoro-Max, green-dyed aqueous fluorescent particles (size: 5 μm, G0500) were purchased from Thermo Fisher Scientific, USA. Sodium chloride (NaCl, CAS# 7647-14-5) was obtained from J. T. Baker, USA. Ethylenediaminetetraacetic acid (EDTA, 28030) was obtained from Norgen Biotec Corp, Canada. Polyvinylpyrrolidone (PVP-40, CAS# 9003-39-8) was purchased from Biobasic, USA.

Buffers composition

The following buffers were prepared and used during chloroplast isolation and handling procedures. For the isolation buffer: 100 mM HEPES (pH 8), 300 mM sorbitol, 5 μM ascorbic acid, 1 M NaCl, 5 mM MgCl₂, 5 mM EDTA, 2 mM DTT, 1% (w/v) polyvinylpyrrolidone (PVP-40), and 0.05% (w/v) bovine serum albumin (BSA). For the washing buffer: 75 mM HEPES (pH 7.75), 200 mM sorbitol, 500 mM NaCl, 5 mM MgCl₂, 1% PVP-40, 0.05% BSA, and 2 mM DTT. For the visualization buffer: 20 mM MOPS, 1 mM phenylmethylsulfonyl fluoride (PMSF), 50 mM EDTA, 0.05% (w/v) BSA. All buffers were prepared using ultrapure water, filtered through a 0.22 μm membrane, and stored at 4 °C until use.

Microfluidic device fabrication

The microfluidic device was designed using AutoCAD (version 2020, Autodesk) and fabricated via standard soft lithography techniques. Master mold preparation began with 4 inch silicon wafers, which were cleaned with IPA and treated with oxygen plasma for 5 min (Zepto, Diener Electronic). A negative photoresist (GM 1060, Gersteltec) was spin-coated at 1000 rpm for 40 s, achieving an approximate thickness of 15 μm. The coated wafers were rested for 1 h to ensure uniformity. A soft bake was performed on a hot plate, starting at 50 °C and increasing by 5 °C every 2 min until reaching 65 °C, and maintained for 10 min. The temperature was then ramped up to 95 °C at the same rate and held for 30 min. Patterning was performed using a micropattern generator (μPG-101, Heidelberg Instruments) with a laser output power of 30 mW. Post-exposure baking followed the same temperature profile as the soft bake. Wafers were developed in PGMEA to remove unexposed resist and rinsed with IPA. A hard bake at 135 °C in a convection oven was performed for 2 h to complete the master mold. To facilitate demolding, the wafers were treated with chlorotrimethylsilane for 30 min under an extraction hood. Microfluidic devices were fabricated using PDMS (Sylgard 184, Dow Corning) mixed at a 10 : 1 ratio of oligomer to curing agent. The mixture was degassed and cured for 1 h at 80 °C. After curing, inlets and outlets were punched, and both the device and glass slides underwent oxygen plasma treatment for final assembly.

DLD pressure drop simulations

Numerical analysis was performed in COMSOL using the laminar flow module to determine the pressure drop along each DLD array of the device. The 2D model of each of the four arrays was imported separately from AutoCAD 2023 as a CAD file and analyzed individually. The properties of the liquid used in the simulation were those of pure water at standard conditions: a density of 998 kg m⁻³ and a dynamic viscosity of 0.001 Pa s. The mesh generated was composed of tetrahedral elements with a calibrated size, ranging from a maximum of 38.7 μm to a minimum of 1.72 μm, and a maximum element growth rate of 1.15. Simulations of the four arrays were conducted with a channel depth of 15 μm at a constant flow rate of 50 μL min⁻¹. Results were analyzed and plotted in Prism 6 (GraphPad).

Chloroplast extraction from Spinacia oleracea L. leaves

Spinach plants (Spinacia oleracea L.) were obtained from a local food supplier. The leaves were stored at 4 °C in complete darkness for 24–48 h to reduce the starch content in the plastids. For chloroplast extraction, 15–30 g of fresh spinach leaves were cut into small segments of approximately 1 cm and submerged in 50 mL of isolation buffer. The prepared sample was transferred to a pre-cooled Büchner flask, which was placed in a cooler. A vacuum was applied for approximately 10 min to facilitate infiltration. The leaves were then homogenized using an immersion blender (Oster®) in short pulses (5–7 pulses) to minimize chloroplast damage. The resulting homogenate was filtered through a 40 μm pore-size mesh, and the filtrate was centrifuged at 200 × g for 15 min at 4 °C. The supernatant was transferred to a 50 mL tube and centrifuged again at 2000 × g for 15 min at 4 °C to pellet the chloroplasts. The supernatant was discarded, and the pellet was carefully resuspended using a micropipette in the washing buffer. The washing process consisted of centrifugation at 2000 × g for 10 min at 4 °C and was repeated twice more to ensure removal of impurities. Finally, the purified chloroplasts were resuspended in 5 mL of separation buffer for subsequent use.

Chloroplast separation using a sucrose gradient

Five sucrose solutions with concentrations of 2 M, 1.75 M, 1.5 M, 1.25 M, and 1 M were prepared. In 25 mL Falcon tubes, 5 mL of each solution was carefully layered in descending order of concentration, starting with the most concentrated solution at the bottom and ending with the least concentrated at the top, creating a density gradient. On top of the sucrose gradient, 5 mL of chloroplasts extracted from spinach leaves, dissolved in separation buffer, were added. The tubes containing the samples were centrifuged using a swing-bucket rotor at 3500 × g for 10 min at 4 °C to achieve phase separation. After centrifugation, the distinct phases were observed, and 2–4 mL from each phase was collected into separate tubes. To each tube, 5 mL of washing buffer was added, and the samples were centrifuged at 3500 × g for 15 min at 4 °C. This washing procedure was repeated twice to ensure the removal of impurities. Finally, the samples were resuspended in 5 mL of visualization buffer. Each phase was loaded into a Neubauer chamber (BR717810, Sigma-Aldrich) and observed under an inverted microscope (Axio Observer A1, Carl Zeiss) equipped with a color camera (Zeiss Axiocam 208 color). Bright-field micrographs were captured to quantify the chloroplasts present in each phase. The separation related steps required a total of 3.5 hours.

Device characterization

Microfluidic device performance was characterized by microbead separation. To prevent particle adhesion, the surfaces of the microfluidic channels were passivated by filling them with a 1% (w/v) Pluronic F-127 solution in PBS and incubating the channels for 10 minutes before starting the experiment. The outlets were then sealed to ensure the removal of air bubbles and to maintain continuous flow during the assay. A microbead mixture was prepared consisting of polystyrene microbeads with diameters of 2 μm, 3 μm, 4 μm, and 5 μm, in 500 μL of PBS to achieve a final concentration of 6 × 10⁶ microbeads per mL. The pretreated device was loaded with PBS at the buffer inlet. Then, the microbead mixture was injected, and the pressure in the tubing was adjusted until the microbeads were properly focused. After 30 min, the separated samples were collected from the relevant outlets using pipette tips. The recovered samples were resuspended in 100 μL of PBS for further analysis. Finally, micrographs of the collected samples were taken using an inverted fluorescence microscope (Axio Observer A1, Carl Zeiss) equipped with a monochromatic camera (Axiocam 506 Mono, Carl Zeiss).

Separation of chloroplasts using a microfluidic device

The microfluidic channels were treated with a 1% (w/v) Pluronic F-127 solution prepared in PBS. The channels were filled with the solution and incubated for 15 min before starting the experiments. Next, the outlets were closed to eliminate air bubbles and ensure continuous flow during the assays. In the pretreated device, PBS was injected at a pressure of 160 mbar using a pressure controller (MFCA-EZ, Fluigent) at the buffer inlet. The chloroplast sample was then introduced at a pressure of 80 mbar, with a concentration of 8.3 × 10⁶ chloroplasts per mL, and the flow was maintained for 30 min. The device outlets were connected to Tygon tubing (AAD04103, Saint-Gobain PPl) that directed the separated chloroplast samples into Eppendorf tubes, which were cooled to 0 °C for preservation. The isolated samples were recovered from the corresponding outlets after the separation process was completed. Finally, the recovered samples were loaded into a Neubauer chamber (BR717810, Sigma-Aldrich) and observed under a microscope (Axio Observer A1, Carl Zeiss) equipped with a color camera (Zeiss Axiocam 208 color). Bright-field micrographs were acquired to quantify the chloroplasts present in the samples. Separation efficiency was defined as the ratio of target-sized chloroplasts recovered at a specific outlet to the total number of chloroplasts of that size collected across all outlets. Purity was calculated as the proportion of target-sized chloroplasts relative to the total chloroplasts present in a specific outlet. These metrics were applied to both the chloroplast separation assays and the characterization experiments performed with polystyrene microbeads. Total recovery was calculated as the percentage of chloroplasts collected from all outlets relative to the total number introduced at the device inlet. Additionally, the absolute concentration of recovered chloroplasts (expressed as chloroplasts per μL) was determined using the Neubauer chamber, according to the following formula:

Image analysis and size estimation

Micrographs of both chloroplasts and polystyrene microbeads were processed using ImageJ software (NIH, USA) to quantify their morphological dimensions. First, the spatial scale was calibrated using the Set Scale function by associating a known distance of 50 μm (corresponding to the size of the Neubauer chamber grid) with the number of pixels, thereby establishing a uniform spatial resolution for all measurements. The images were converted to 8-bit format to ensure compatibility with thresholding processes. A manual threshold was then applied to appropriately highlight the particles of interest. The thresholded image was converted into a binary mask using the convert to mask function, followed by morphological operations (close, invert, and erode) to improve segmentation, close discontinuities, invert contrast, and reduce edge noise.

Particle detection was performed using the analyze particles function, setting a minimum area threshold of 2 μm² to exclude small artifacts. For each detected particle, the projected area (in square micrometers) was recorded, and all detections were automatically labeled and overlaid on the original image for visual verification. The equivalent circular diameter was calculated from the projected area using the equation: where A is the area of the particle. The equivalent diameter provides a standardized scalar value that enables consistent comparison of particle sizes regardless of shape complexity, and is widely applied in morphometric analyses of heterogeneous or asymmetric particles.²⁶

This image analysis protocol was applied uniformly to both microbeads and chloroplasts to ensure methodological consistency in size characterization. To facilitate comparative analysis, the resulting diameter values were grouped into discrete size intervals using a standard rounding function. Values with a first decimal place of five or greater were rounded up to the nearest whole number; otherwise, they were rounded down. This allowed for practical comparison with the nominal bead diameters provided by the manufacturer and enabled consistent size classification of chloroplasts, despite their natural morphological variability. The protocol was applied to all experimental conditions, including samples collected from the outlets of the microfluidic device and those processed by conventional sucrose gradient centrifugation. The final size distribution data were visualized using Prism 6 (GraphPad).

Results and discussion

Device design and simulation insights

Conventional isolation techniques, such as sucrose gradient centrifugation, lack the precision necessary to selectively enrich distinct chloroplast populations with high purity and reproducibility within the 2 to 5 μm size range. This limitation often yields heterogeneous fractions, significantly complicating downstream analyses that require the recovery of specific chloroplast subsets. The ability to isolate these defined populations with greater fidelity would provide a valuable tool for exploring fundamental aspects of plastid biology, including development, functional specialization, and metabolic diversity under both physiological and experimental conditions. Recognizing this inherent challenge in current methodologies, we therefore aimed to develop a microfluidic device capable of sorting this defined population of chloroplasts for more targeted investigations.

Among the separation methods developed through microfluidic technology, deterministic lateral displacement (DLD) stands out for its ability to separate particles or cells based on size, shape, and deformability. DLD has been successfully applied to a wide range of biological samples, including leukocytes,²⁷ mammalian cells,²⁸ parasites,²⁹ minicells,³⁰ and circulating tumor cells.³¹ Thus, our device employs passive separation with DLD and, importantly, integrates multiple DLD arrays on a single chip. This allows for simultaneous sorting of particles by size, as each array can be designed with a specific geometry to target different size ranges within the same device. In our implementation, the microfluidic device incorporates four DLD arrays arranged in parallel (Fig. 2a), with each array specifically configured to separate particles within a 1 μm size window, covering the range of 2 to 5 μm. To facilitate the simultaneous operation of the four arrays, the inputs of the arrays are connected to common inlets: one for the sample and one for the focusing buffer. The device features individual outlets to collect the separated particles of different sizes, with three outlets for each array, resulting in a total of 12 outlets (3 outlets by 4 arrays). This design allows for easy and simultaneous particle separation across different particle sizes.

Fig. 2 Design of a parallel DLD device for size-based chloroplast separation. (a) Top view photograph of the microfluidic device, highlighting the arrangement of four parallel DLD sorting arrays. Arrows show the direction of the flow. Inlets are in the middle of the device and the outlets are found on the right side. (b) Particle trajectories predicted by theoretical calculations, demonstrating the expected behavior of >3 μm (i), <3 μm (ii), and <2 μm (iii) particles within an example of the four DLD arrays, engineered to separate 3 μm particles. Posts are drawn to scale; microbeads are not. (i) Initial separation of >3 μm particles is shown at the chip entrance. (ii) Further downstream, particles <3 μm are separated. (iii) >3 μm and <2 μm particles exit through outlets O2A and O2B, respectively. (c) Table outlining the critical geometric parameters governing DLD separation, specific to each of the four arrays. (d) Detailed geometric specifications of the pillar arrangements in the four DLD arrays, including S1 and S2 section configurations. Scale bar: 40 μm (unless specified).

Each of the four arrays functions as a bandpass filter, designed to separate particles that fall within a specific size range of 1 μm. The filter effectively removes particles larger than critical diameter 1 (CD1) and displaces particles larger than the critical diameter 2 (CD2), thereby increasing efficiency and purity of the separation. As shown in Fig. 2a, arrays 1, 2, 3, and 4 were specifically designed to sort particles with theoretically target diameters within the following ranges: >1 to ≤2 μm, >2 to ≤3 μm, >3 to ≤4 μm, >4 to ≤5 μm, respectively. However, we anticipate a significant deviation from these theoretically designed ranges. This is due to the inherent approximations in the underlying theory, manufacturing inaccuracies, and other potential influencing factors such as particle concentration and surface effects. Fig. 2c presents the geometric parameters used for each DLD array, and Fig. 2d illustrates them schematically. Further information on the estimation of the critical diameter and how it relates to the system parameters can be found in the Supplementary Information. Each DLD array, designed for a specific target diameter, comprises two distinct sections with different pillar arrangements. In the first section (S1 in Fig. 2b), the pillars follow a geometric configuration extending from the inlet to the array's midpoint. In the second section (S2 in Fig. 2b), the pillar arrangement transitions to a different configuration from the midpoint to the outlets. The S1 sections selectively displace particles larger than the critical diameter 1 (CD1), directing them toward the waste outlet (outlet OxA). In our example in Fig. 2b, all particles larger than 3 μm are displaced and exit through O2A. Following this, in the S2 section, particles with a size greater than the critical diameter (CD2) undergo displacement toward the target outlet (OxC). Conversely, particles smaller than CD2 continue to flow in a zigzag mode through the pillar arrays without exhibiting a net lateral displacement, thus proceeding directly to the waste outlet (OxB). Referring again to Fig. 2b, particles approximately 3 μm in size are displaced to outlet O2C, while all the particles continue in zigzag flow to outlet O2B. As a result, the device is capable of efficiently sorting particles based on size within a compact and integrated platform, making it suitable for our applications that require precise particle fractionation of chloroplasts.

In all four arrays, in which chambers have the same surface area, the pillar diameter remains constant, with variations occurring only in the geometric arrangement. As a result, the total number of pillars differs: the arrays with critical diameters of 5, 4, 3, and 2 μm contain 21 271, 27 392, 31 593, and 33 422 pillars, respectively. This variation in the number of pillars influences the fluidic properties, as a higher number of pillars increases flow resistance, resulting in an imbalance in flow rates across the arrays, as they are all connected to the same inlet. This flow imbalance could result in uneven flow distribution across the arrays, potentially compromising the device separation efficiency. To address this issue, we conducted COMSOL simulations to analyze and then compensate in the design for the pressure drop caused by fluidic resistance in each array.

Fig. 3a presents the simulation results as a color map, illustrating that the pressure drop increases with the number of pillars. Fig. 3b illustrates the pressure drop along each array, reaching 0.68 psi, 1.1 psi, 1.48 psi, and 1.67 psi (46.9 mbar, 75.8 mbar, 102.0 mbar, and 115.1 mbar) for pillar arrays of 5 μm, 4 μm, 3 μm, and 2 μm, respectively. To compensate for these, the outlet channel widths were adjusted to 25, 34, 40, and 60 μm for the corresponding pillar arrays while maintaining a constant channel length of 3.25 mm. Fig. 3c shows the effect of this adjustment by comparing flow rates before and after compensation. The results indicate that the four arrays achieve similar flow rates, effectively eliminating discrepancies for consistent performance. Simulations also estimate the device's operating pressures, showing that the buffer requires twice the sample pressure to generate an enveloping buffer flow, as shown in Fig. 3d. This ensures the sample stream remains focused along the pillar array.

Fig. 3 Computational modeling of the parallel DLD device for optimized chloroplast separation. (a) Simulated pressure distribution within the four parallel sorting arrays. Scale bar: 200 μm. (b) Longitudinal pressure drop profiles along each array, from inlet to outlet. (c) Flow rate profiles at the array inlets, comparing pre- and post-optimization results from computational simulations. (d) Simulated flow streamlines at the array input, illustrating the controlled introduction of sample and buffer flows. Scale bar: 100 μm.

Experimental validation with microbeads

Before isolating chloroplasts in the microfluidic device, we characterized its separation capability by evaluating the efficiency and purity of polystyrene microbead fractions. We used microbeads with diameters ranging from 2 to 5 μm to simulate the behavior of different chloroplast populations. For the experiment, we injected a particle mixture (6 × 10⁶ microbeads per mL) into the device and adjusted the sample inlet and sheath flow pressures to 172 mBar and 344 mBar, respectively. These pressure values were optimized through preliminary tests, where a sheath-to-sample flow ratio of 2 : 1 was found to enable optimal hydrodynamic focusing at the entrance of the DLD arrays (Fig. S1†). After 30 minutes of operation—the time required to collect a 20 μL sample volume suitable for quantification using a Neubauer chamber—we collected the separated microbeads at the designated outlets using pipette tips, Fig. 4a. This ensured efficient sample collection without affecting channel resistance. Subsequently, we acquired images of each sample collected at the outlets to quantify the particle distribution across the different outlets, Fig. 4b. The size distribution of the microbeads used in these experiments—obtained from image analysis and equivalent diameter calculation—is presented, along with the corresponding standard deviations, in ESI† Table S1. Fluorescence imaging performed during the experiment confirmed proper device operation. As shown in Fig. 4c, the 5 μm (green) and 3 μm (red) microbeads followed distinct trajectories, demonstrating effective separation: 3 μm beads migrated toward the upper-right region and exited through outlet O3, while 5 μm beads were directed to the waste outlets.

Fig. 4 Characterization of the parallel DLD device for microbeads separation. (a) Photograph of the DLD device with dye, highlighting the sample and buffer inlets, and pipette tips connected at the outlets for collecting separated microbeads. Scale bar: 5 mm (b) representative bright-field and fluorescence micrographs of the device outlets. Scale bar: 50 μm. (c) Fluorescence micrographs visualizing the separation of 5 μm (green) and 3 μm (red) microbeads by their trajectories. Scale bar: 200 μm. (d) Bar graphs illustrating microbead separation efficiency at device outlets for different microbead sizes (n = 3, error bars: standard deviation). (e) Bar graphs showing sample purity at device outlets for different microbead sizes (n = 3, error bars: standard deviation).

We calculated separation efficiency as the fraction of target particles recovered relative to the total collected at all outlets. The device achieved optimal separation performance, with efficiencies of 82% (CV: 10%) and 89% (CV: 12%) for the 5 μm and 4 μm microbeads in outlets O1 and O2, respectively. For the 3 μm and 2 μm particles, efficiencies reached 89% (CV: 11%) and 88% (CV: 7.4%) in outlets O3 and O4, respectively. However, outlet O4 showed a higher degree of cross-contamination, containing particles of three different sizes. In contrast, the other outlets typically exhibited contamination from only one or two non-target particle sizes, Fig. 4d. Regarding purity, we measured the proportion of target particles in each specific outlet. We obtained 96% (CV: 0.6%) in O1 for the 5 μm microbeads and 82.4% (CV: 1.5%) in O2 for the 4 μm microbeads. For the 3 μm and 2 μm particles, purity reached 94.6% (CV: 4.8%) in O3 and 80.8% (CV: 24.5%) in O4, as shown in Fig. 4e.

These results demonstrate that the microfluidic device effectively and precisely separates polystyrene microbeads of different sizes, validating its ability to fractionate particles based on diameter. The separation efficiency was high for all microbeads, with values exceeding 80%, a commonly accepted threshold for reliable classification in particle per cell separation studies,³² ensuring correct sorting at the designated outlets (O1–O4). Furthermore, purity reached satisfactory levels, particularly for the 5 μm and 3 μm particles, with values up to 96%.

Evaluation of the microfluidic device for chloroplast separation

To assess the efficiency of the microfluidic device in chloroplast separation, experiments were conducted to characterize its performance in terms of efficiency and purity compared to conventional methods. Spinach (S. oleracea) leaves were used as the source material, as they are widely recognized as a model system in chloroplast research due to their high chloroplast content, ease of tissue handling, and well-characterized plastid biology. These qualities make spinach an ideal plant for developing and benchmarking chloroplast isolation methods. The size distribution of chloroplasts extracted from spinach leaves was first determined by bright-field microscopy and image analysis of individual organelles. This analysis generated a histogram identifying the predominant size range prior to separation in the microfluidic device. To ensure consistency in the quantification of particle size, we employed the equivalent circular diameter (D_eq.), defined as the diameter of a circle having the same projected area as the chloroplast. Given that chloroplasts exhibit ovoid or discoid shapes, this parameter allows for size characterization independent of particle elongation or orientation. D_eq. values were computed based on the segmented area obtained through image analysis and provide a robust single-value descriptor for comparing particles of irregular shape. Subsequently, the extracted chloroplasts were separated using the microfluidic device. To preserve chloroplast integrity and minimize metabolic activity or degradation, both the extraction process and the microfluidic separation were performed at a controlled temperature of 4 °C—a condition commonly used in organelle handling protocols to maintain structural and functional stability. The collection efficiency was analyzed as a function of size.

First, we characterized the chloroplasts extracted from spinach leaves, determining that their sizes ranged from 3 μm to 8 μm (Fig. S2†). This size range is consistent with the presence of predominantly mature, fully developed chloroplasts in mesophyll cells, which are abundant in mature spinach leaves. In contrast, smaller plastids in the 2–5 μm range are more commonly associated with young or meristematic tissues, or with early stages of plastid differentiation,⁹ which are not prevalent in fully expanded spinach leaves. We then injected the chloroplast sample into the microfluidic device, setting inlet pressures of 180 mBar for the sheath flow and 60 mBar for the sample, corresponding to a sheath-to-sample flow ratio of 3 : 1. This configuration, selected based on preliminary tests (Fig. S1†), enabled stable and efficient hydrodynamic focusing. In addition, a moderate pressure regime was chosen to minimize the risk of potential mechanical damage, based on the hypothesis that chloroplasts, due to their fragile structure, could be susceptible to flow-induced deformation under high shear stress conditions. Although no direct evidence of such damage in chloroplasts has been reported, similar effects have been observed in organelles with comparable characteristics, such as mitochondria, when exposed to intense flow. Therefore, this strategy was considered appropriate to preserve the structural and functional integrity of chloroplasts during the separation process. After 30 minutes of operation, we collected the chloroplasts at the designated outlets and acquired images of each sample to quantify their distribution (Fig. 5a). The results demonstrated an effective size-based separation of the chloroplasts (Fig. 5b). Larger chloroplasts (8 μm) were predominantly collected at outlet O1, where approximately 85% (CV: 17%) of the particles were recovered. Additionally, 7 μm chloroplasts, the next largest size, accounted for 60% (CV: 7.6%) of the particles collected at this outlet. At outlet O2, 6 μm chloroplasts were the most abundant, comprising approximately 50% (CV: 13.6%) of the particles of that size. Similarly, 5 μm chloroplasts were primarily collected at outlet O3, reaching 55% (CV: 22.8%). In contrast, smaller chloroplasts were concentrated at outlet O4, with 75% (CV: 11%) of the 3 μm chloroplasts and 61% (CV: 9%) of the 4 μm chloroplasts collected. These results confirm that the microfluidic device enables a progressive size-based separation of chloroplasts, supporting its applicability in organelle fractionation studies (Fig. 5b).

Fig. 5 Comparative chloroplast isolation from Spinacia oleracea L.: microfluidics vs. sucrose gradient methods (a) bright-field micrographs of samples collected at the inlet and outlet O1, including a magnified view of an isolated chloroplast. Scale bars: 50 μm and 8 μm. DLD device performance: chloroplast separation efficiency (b) and sample purity (c) at outlets 1–4 (n = 3, error bars: standard deviation). (d) Sucrose gradient-based separation: schematic representation of the gradient phases and corresponding photograph of the separated phases. Sucrose gradient method: chloroplast separation efficiency (e) and purity (f) from phases 1–4 (n = 3, error bars: standard deviation).

In addition to the normalized data, we estimated the absolute concentrations of recovered chloroplasts (chloroplasts per μL) and calculated the overall recovery percentage. The mean recovery value was 17.9% ± 1.3%, which, although it reflects some inherent challenges in the current design, also highlights clear opportunities for optimization—both in device geometry and in operational parameters (ESI† Table S2). Although the device was initially designed to target chloroplasts within the 2–5 μm size range—validated using polystyrene beads—it effectively fractionated real chloroplasts across a broader range of approximately 3–8 μm. We believe this extension of the separation range can be attributed to two main factors: the deformability and morphology of chloroplasts. In terms of deformability, chloroplasts exhibit a Young's modulus on the order of 26 kPa, whereas the polystyrene particles used for characterization have a modulus of approximately 3.5 MPa.³³ This represents a two-order-of-magnitude difference, indicating that chloroplasts are significantly more deformable. Under flow conditions, this property may reduce their effective hydrodynamic size, causing them to behave like smaller particles during separation. A similar phenomenon has been reported in the separation of other organelles, such as mitochondria.^34,35 Like chloroplasts, mitochondria possess a double membrane and are structurally soft. In such cases, separation efficiency has been shown to decrease, as mitochondria fail to reach equilibrium positions as rigid particles do.³⁴

Furthermore, the geometry of chloroplasts—typically ovoid or discoid—may influence their orientation and trajectory within the microfluidic channel. Previous studies have shown that human erythrocytes, having similar, non-spherical aspect ratios as chloroplasts, align along their major axis (∼8 μm) in channels approximately 9 μm in height but reorient along their thickness (∼2 μm) in channels around 3.5 μm high, significantly impacting their migration behavior.³⁶ By analogy, similar shape-dependent reorientation may occur with chloroplasts. Taken together, these observations provide a plausible explanation for the broader separation range observed and underscore the importance of considering both mechanical and geometric properties when designing microfluidic systems for organelle separation.

In terms of purity—defined as the proportion of target chloroplasts in each specific outlet—the device achieved a maximum average purity of 66% for 4 μm chloroplasts (CV: 6.7%). For the other sizes, the purity values were 24.5% (CV: 4.5%) for 3 μm chloroplasts, 44% (CV: 8.1%) for 5 μm, 39.6% (CV: 8.4%) for 6 μm, 44.4% (CV: 13%) for 7 μm, and 17.6% (CV: 11.8%) for 8 μm (Fig. 5c). When comparing the separation achieved with our microfluidic device to that obtained using the conventional sucrose gradient-based method, significant differences were observed (Fig. 5d). The conventional method exhibited a broad distribution of chloroplast sizes across the gradient phases, indicating a dispersion in the fraction of sizes recovered in each phase. In phase 1, chloroplasts of various sizes were identified, with concentrations of 30.5% for 8 μm, 32.4% for 7 μm, 26.8% for 6 μm, 14.7% for 5 μm, and 5.9% for 4 μm. Similarly, phase 2 showed an overlap of sizes, with values of 31.9% for 8 μm, 35.6% for 7 μm, 31% for 6 μm, 22.6% for 5 μm, and 13.1% for 4 μm (Fig. 5e). In phase 3, the size distribution changed, with a higher proportion of 4 μm chloroplasts (52%), followed by 3 μm (31.4%), 5 μm (42.2%), 6 μm (25%), 7 μm (17%), and 8 μm (20.8%). Finally, in phase 4, smaller chloroplasts predominated, with concentrations of 35% for 3 μm, 28.8% for 4 μm, 20.3% for 5 μm, 17% for 6 μm, 14.8% for 7 μm, and 16.6% for 8 μm (Fig. 5e). These results suggest that the density gradient-based method enables a partial separation of chloroplasts by size, although with considerable overlap between the different fractions, which could compromise the recovery of homogeneous populations.

In terms of purity, the samples recovered in the different phases of the density gradient method reached a maximum average purity of 45.8% for 5 μm chloroplasts. For the other sizes, the purity values were 43.5% for 4 μm, 33.1% for 6 μm, 10.3% for 7 μm, 8.4% for 8 μm, and 1.7% for 3 μm (Fig. 5f). Additionally, a high coefficient of variation (CV > 50%) was observed, indicating low reproducibility of the density gradient separation method. This variability can be attributed to the inherent complexity of the procedure, which involves multiple centrifugation steps, successive washes, and strict control of critical variables such as temperature and gradient concentration. The requirement for precise handling at each stage of the process, along with the reliance on operator expertise, increases the risk of variations in the results, thereby compromising the reproducibility of the method. The density gradient method proved to be both complex and time-consuming. The multistep protocol required one to two full working days of a skilled technician. Steps involved in separation, including gradient formation and sample collection, were particularly intensive and required 3.5 hours to complete. This duration is substantially longer compared to the time required for the DLD method.

Conclusions

The traditional separation of chloroplasts using sucrose gradients presents significant limitations in terms of specificity, reproducibility, and operational complexity. In this study, we proposed a microfluidic platform based on deterministic lateral displacement (DLD) as an innovative and efficient alternative for the size-based isolation of chloroplasts. This strategy allowed us to overcome methodological challenges associated with conventional approaches, providing a more precise and accessible tool for studying plant cell organelles.

Our experiments demonstrated that the microfluidic system successfully separated chloroplasts into four defined size ranges (3 to 8 μm), achieving separation efficiencies ranging from 50% to 85%. These values outperformed those obtained through traditional sucrose gradient methods, which ranged between 31% and 52%. Moreover, an improvement in the purity of the recovered fractions was observed, with levels ranging from 17% to 66%, compared to the 2% to 46% range achieved with conventional techniques. Separation time was significantly shortened from several hours for conventional gradient methods to just a few tens of minutes for the DLD approach. These results show that the proposed microfluidic approach represents a substantial advancement in size-based chloroplast isolation. Our DLD-based microfluidic device significantly reduced the overlap between populations of different sizes, thereby improving the precision in the recovery of specific fractions. This enhanced resolution is particularly valuable for downstream applications requiring high-purity organelle samples. By decreasing population heterogeneity, the platform enables more consistent and reliable analyses in plant research. Taken together, these features position DLD-based microfluidics as a powerful, scalable, and robust strategy for organelle separation, opening new possibilities for precise and reproducible studies in plant cell biology and biotechnology.

Data availability

All data necessary to reproduce the findings of this study are included within the article and its ESI.†

Conflicts of interest

The authors declare no conflict of interest. JLGC is an employee of F. Hoffmann-La Roche AG.

Acknowledgements

This work was supported by Mexico's Secretariat of Science, Humanities, Technology and Innovation (SECIHTI) through grants no. 313088, CB-286368, CF-102963. VML (CVU: 375624) was awarded a postdoctoral scholarship, and OGCP (CVU: 741461) received a research assistantship, both funded by SECIHTI.

References

1. J. Dobrogojski, M. Adamiec and R. Luciński, Acta Physiol. Plant., 2020, 42, 98.
2. P. E. Jensen and D. Leister, F1000Prime Rep., 2014, 6, 40.
3. E. Lopez-Juez and K. A. Pyke, Int. J. Dev. Biol., 2005, 49, 557–577.
4. Y. Song, L. Feng, M. A. M. Alyafei, A. Jaleel and M. Ren, Int. J. Mol. Sci., 2021, 22, 13464.
5. G. R. Littlejohn, S. Breen, N. Smirnoff and M. Grant, New Phytol., 2021, 229, 3088–3107.
6. M. Akram, M. A. Khan, N. Ahmed, R. Bhatti, R. Pervaiz, K. Malik, S. Tahir, R. Abbas, F. Ashraf and Q. Ali, AMB Express, 2023, 13, 23.
7. M. Tran, C. Van, D. J. Barrera, P. L. Pettersson, C. D. Peinado, J. Bui and S. P. Mayfield, Proc. Natl. Acad. Sci. U. S. A., 2013, 110, E15–E22.
8. K. A. Pyke, Plant Cell, 1999, 11, 549–556.
9. J. V. Possingham and M. E. Lawrence, Int. Rev. Cytol., 1983, 84, 1–56.
10. W. Sakamoto, S. Miyagishima and P. Jarvis, Arabidopsis Book, 2008, vol. 6, p. e0110.
11. M. Liebers, B. Grübler, F. Chevalier, S. Lerbs-Mache, L. Merendino, R. Blanvillain and T. Pfannschmidt, Front. Plant Sci., 2017, 8, 23.
12. K. X. Chan, S. Y. Phua, P. Crisp, R. McQuinn and B. J. Pogson, Annu. Rev. Plant Biol., 2016, 67, 25–53.
13. B. J. Pogson, N. S. Woo, B. Förster and I. D. Small, Trends Plant Sci., 2008, 13, 602–609.
14. Md. S. Islam, G. L. Buttelmann, K. Chekhovskiy, T. Kwon and M. C. Saha, Plants, 2019, 8, 606.
15. P. Nawkarkar, S. Chugh, S. Sharma, M. Jain, S. Kajla and S. Kumar, Curr. Genomics, 2020, 21, 610–623.
16. H. Daniell, S. Jin, X. Zhu, M. A. Gitzendanner, D. E. Soltis and P. S. Soltis, Plant Biotechnol. J., 2021, 19, 430–447.
17. S. Li, X. Li and S.-H. Ho, Sep. Purif. Technol., 2022, 291, 120951.
18. M. A. Block and C. Albrieux, in Methods in Molecular Biology, Humana Press Inc., 2018, vol. 1829, pp. 123–135.
19. L. F. Núñez-Becerril, S. Hernández-Castellano, V. Limones-Briones and C. De-la-Peña, in Methods in Molecular Biology, Humana Press Inc., 2024, vol. 2827, pp. 377–383.
20. O. Bhattacharya, I. Ortiz and L. L. Walling, Plant Methods, 2020, 16, 131.
21. H. Jiang, Z. Xu, M. R. Aluru and L. Dong, Lab Chip, 2014, 14, 1281–1293.
22. H. Massalha, E. Korenblum, S. Malitsky, O. H. Shapiro and A. Aharoni, Proc. Natl. Acad. Sci. U. S. A., 2017, 114, 4549–4554.
23. M. Fendrych, M. Akhmanova, J. Merrin, M. Glanc, S. Hagihara, K. Takahashi, N. Uchida, K. U. Torii and J. Friml, Nat. Plants, 2018, 4, 453–459.
24. Z. Khan, H. Karamahmutoğlu, M. Elitaş, M. Yüce and H. Budak, Plants, 2019, 8, 14.
25. M.-F. Noirot-Gros, S. V. Shinde, C. Akins, J. L. Johnson, S. Zerbs, R. Wilton, K. M. Kemner, P. Noirot and G. Babnigg, Front. Plant Sci., 2020, 11, 408.
26. R. E. Cavicchi, M. J. Carrier, J. B. Cohen, S. Boger, C. B. Montgomery, Z. Hu and D. C. Ripple, J. Pharm. Sci., 2015, 104, 971–987.
27. O. G. Chavez-Pineda, R. Rodriguez-Moncayo, A. M. Gonzalez-Suarez, P. E. Guevara-Pantoja, J. L. Maravillas-Montero and J. L. Garcia-Cordero, Lab Chip, 2024, 24, 2575–2589.
28. K. Matsuura, S. Hashioka and K. Takata, Anal. Sci., 2024, 40, 1801–1807.
29. S. H. Holm, J. P. Beech, M. P. Barrett and J. O. Tegenfeldt, Anal. Methods, 2016, 8, 3291–3300.
30. A. Sherbaz, B. M. K. Konak, P. Pezeshkpour, B. Di Ventura and B. E. Rapp, Micromachines, 2022, 13, 365.
31. H. Okano, T. Konishi, T. Suzuki, T. Suzuki, S. Ariyasu, S. Aoki, R. Abe and M. Hayase, Biomed. Microdevices, 2015, 17, 59.
32. M. A. Mansor, M. A. Jamrus, C. K. Lok, M. R. Ahmad, M. Petrů and S. S. R. Koloor, Sens. Actuators Rep., 2025, 9, 100277.
33. T. Yamada, H. Arakawa, T. Okajima, T. Shimada and A. Ikai, Ultramicroscopy, 2002, 91, 261–268.
34. D. Kang, S. Oh, P. Reschiglian and M. H. Moon, Analyst, 2008, 133, 505.
35. C. Lee, Y. Chen, P. Wang, D. C. Wallace and P. J. Burke, Anal. Chem., 2022, 94, 6930–6938.
36. S. H. Holm, J. P. Beech, M. P. Barrett and J. O. Tegenfeldt, Anal. Methods, 2016, 8, 3291–3300.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5lc00348b

‡ These authors contributed equally.

§ Current address: Institute of Human Biology (IHB), Roche Pharma Research and Early Development, Roche Innovation Center Basel, Basel, Switzerland.

---