Microfluidic vortex focusing for high throughput synthesis of size-tunable liposomes

Liposomes are widely used nanocarriers for delivering therapeutic agents across oncology, infectious disease, immune modulation, and vaccines. Because pharmacokinetics, biodistribution, cell uptake, and toxicity are strongly size-dependent, precise control of liposome size with low polydispersity is essential. Current batch methods (e.g., solvent injection, detergent removal, membrane/high-pressure extrusion, sonication/freeze–thaw) offer high throughput but limited control and high polydispersity, and scale transitions are difficult. Continuous-flow microfluidics improves control during lipid self-assembly via hydrodynamic flow focusing or rapid micromixing, but scaling throughput while maintaining tight size control remains challenging. This study addresses the need for a high-throughput, size-tunable, low-variance liposome synthesis technology by introducing microfluidic vortex focusing (MVF), which combines hydrodynamic focusing with vortical mixing in an axisymmetric flow cell to accelerate solvent exchange and control the kinetics of vesicle formation.

Batch techniques such as solvent injection, detergent removal, membrane extrusion, and high-pressure extrusion can achieve large-scale production but often yield broad size distributions and require multiple handling steps. Hydrodynamic flow focusing in microfluidics narrows the lipid stream to reduce diffusion length scales, enabling smaller, less polydisperse vesicles by tuning the buffer:lipid flow rate ratio (FRR). Chaotic advection micromixers (e.g., herringbone, baffles, toroidal/twisted channels) enhance mixing at Re≈80–100 but typically produce higher size variance and limited size tuning. Vortex mixers that inject both streams tangentially can improve mixing rates over ethanol injection, yet reported polydispersity indices (PDI) commonly exceed 0.2 due to spatial variations in the mixing zone. Attempts to scale microfluidic platforms (modified focusing and micromixer designs) often sacrifice size control due to larger geometries needed for higher flow rates. Hydrocyclone-based devices have been miniaturized for particle separations; leveraging a similar geometry with modified flow paths provides a basis for MVF to achieve rapid mixing with spatially constrained focusing for improved size control at high throughput.

Device design and operation: An axisymmetric hydrocyclone-inspired chamber was used. Aqueous buffer enters tangentially into an annular region to generate a rotating flow that proceeds into a conical mixing chamber. Lipid solution in ethanol is injected axially through a central inlet within the annular region and emerges into the center of the conical chamber sheathed by the rotating buffer, transferring rotational momentum to the lipid stream. The annular inner surface is tapered to minimize discontinuities and promote focusing; mixed flow exits via a lower axial outlet.

Device fabrication: Devices were fabricated by high-resolution SLA-DLP 3D printing (Perfactory 4 DLP-SLA; 25 µm z-step; 75 mm objective; 1920×1200 pixels). Overall dimensions: chamber diameter 1.5 mm, length 8.4 mm; vortex formation gap 300 µm; inlet/outlet channel diameters 300 µm. Lipid injection channel tapered to a tip thickness of 150 µm (minimum reliably formed), nozzle length ≤1 mm to avoid warping/clogging. Print orientation aligned device axis perpendicular to stage to form the tapered annulus reliably and reduce asymmetry. Exposure intensity set to 97% of nominal to balance resolution and mechanical stability. Post-processing included IPA rinse/flush, air flush, and secondary UV curing (Otoflash, 500 flashes). Internal structures were assessed via micro-CT. Surface roughness in the chamber: Ra ≈1.32 µm (axial) and 0.45 µm (radial). Material: acrylic-based photopolymer resin (HTM140) exhibiting ethanol compatibility and sufficient rigidity (tensile strength ~56 MPa); post-curing essential to prevent solvent-induced failures.

Numerical modeling: COMSOL Multiphysics simulations modeled co-flow of water and ethanol with the MVF geometry; for comparison, a hydrodynamic focusing configuration was simulated by injecting buffer axially (no tangential inlet). Stationary solver with second-order discretization; incompressible fluids and no-slip boundary conditions. Total flow rates (TFR) from 10–80 mL/min and FRR (water:ethanol) from 10–50 were evaluated. Ethanol mole fraction threshold of ~0.5 was used to approximate onset of vesicle formation based on prior DMPC aggregation studies. Simulations assessed streamlines, ethanol concentration profiles, and axial evolution of peak ethanol fraction.

Liposome synthesis: Lipid films were prepared from DMPC:cholesterol:DCP or DMPC:cholesterol:PEG2k-PE (5:4:1 molar ratio) by dissolving in chloroform, drying under nitrogen, and vacuum desiccation (4 h). Films were solvated in dehydrated ethanol (molecular sieves ~20 wt% for ≥24 h) to prepare stock solutions (typically 20 mM), then diluted as needed. For PEGylated studies, 10 mol% PEG2k-PE was used. Lipid ethanol solutions and 1× PBS buffer were filtered (0.22 µm) and loaded into syringes connected to lipid and buffer inlets, respectively. The device was submerged in a 40 °C water bath, primed with PBS, and operated at prescribed TFR and FRR values. Unless otherwise specified, TLC values were 10 or 20 mM for performance studies; for high-throughput trials, TLC up to 30 mM (DCP-based) and 60 mM (PEGylated) were used. Operating ranges: FRR from 10 to 100; TFR from 10 to 80 mL/min.

Characterization: Dynamic light scattering (Zetasizer Nano ZS) at 25 °C measured size distributions and PDI; samples filtered (0.22 µm) prior to measurement. Cryo-TEM (JEM 2100 LaB6) provided direct size distribution imaging for representative samples. Long-term stability was assessed by storing selected formulations at 4 °C for 99 days and re-measuring size distributions.

• Simulations showed that MVF accelerates solvent exchange versus hydrodynamic focusing: peak ethanol mole fraction fell to ~0.5 within ~0.5 mm (~0.75 ms) downstream of the junction in MVF, compared to ~3.1 mm (~2.50 ms) for hydrodynamic focusing at TFR 60 mL/min and FRR 1:30, indicating much shorter growth times for bilayer fragments and enabling smaller vesicles.
• Device fabrication via SLA-DLP achieved functional MVF structures with ~50% yield; typical failure mode was partial/full lipid channel clogging. Chamber roughness Ra ≈1.32 µm (axial), 0.45 µm (radial). Resin exhibited robust ethanol compatibility.
• Effect of FRR (TFR fixed at 60 mL/min; TLC 10 mM; DMPC:cholesterol:DCP 5:4:1): Increasing FRR from 10 to 100 reduced mean diameter, reaching a minimum of ~61 nm at FRR=100. PDI remained low and nearly constant across FRR with an average ~0.04. Results were highly repeatable (N≥6 per condition).
• Effect of TFR (FRR=50): Increasing TFR decreased both mean diameter and PDI—distinct from conventional hydrodynamic focusing and chaotic micromixing where size/PDI are typically TFR-insensitive. Enhanced rotational velocity at higher buffer flow rates likely augments mixing.
• Effect of lipid concentration: Increasing TLC from 10 mM to 20 mM produced small increases in mean diameter and PDI.
• PEGylation (10 mol% PEG2k-PE; FRR=50; TLC 10 mM): PEG-containing formulations showed substantially smaller vesicles at all TFR, with up to ~60% size reduction at TFR=80 mL/min, yielding minimum mean diameter ~27 nm. PDI averaged ~0.14 and was relatively invariant with TFR.
• Stability: Selected formulations stored 99 days at 4 °C showed no detectable changes in mean size or PDI.
• Throughput: Operating near the laminar limit (TFR up to 80 mL/min) with elevated lipid concentrations achieved mass production rates of ~7.2 g/h for DMPC:cholesterol:DCP (30 mM) and >20 g/h for DMPC:cholesterol:PEG2k-PE (60 mM) at FRR=10. Throughput exceeded prior high-aspect-ratio hydrodynamic focusing by >50× and was nearly an order of magnitude higher than emerging chaotic advection platforms.
• Flow regime: Reynolds number approached the laminar limit around TFR ~60 mL/min and entered transitional flow at ~80 mL/min; higher TFR risks turbulence, degraded focusing, and increased PDI.

MVF integrates hydrodynamic focusing with vortical mixing to confine and rapidly mix lipid and buffer streams, drastically shortening solvent exchange times relative to conventional hydrodynamic focusing. The accelerated approach to the critical solvent polarity threshold limits bilayer fragment growth time, producing smaller vesicles with narrow distributions. The ability of TFR to modulate mixing intensity via rotational flow offers an additional control parameter absent in standard flow focusing or chaotic mixers, explaining observed decreases in both size and PDI with increasing TFR at fixed FRR. Low PDI (~0.04) across broad FRR in non-PEG formulations demonstrates improved uniformity relative to chaotic mixers (often PDI>0.2). PEG-lipid inclusion further reduces size, potentially by slowing intermediate structure growth due to lower diffusivity, albeit with higher PDIs (~0.14). Achieved mass production rates (>20 g/h) place MVF within pilot/manufacturing throughput ranges while maintaining size control, addressing a key barrier to translation. The technology’s additive manufacturing enables reproducibility and accessibility without specialized microfabrication. Collectively, results substantiate MVF as a platform that overcomes the scale–control trade-off inherent to existing methods.

The study introduces microfluidic vortex focusing as a continuous-flow method that combines hydrodynamic focusing and vortical mixing to produce size-tunable, low-polydispersity liposomes at high throughput. MVF reliably generated vesicles as small as ~61 nm (non-PEG) with PDI ~0.04, and as small as ~27 nm with PEG-lipids (PDI ~0.14). Size and PDI decreased with increasing FRR and, uniquely, with increasing TFR. The platform achieved mass production rates up to >20 g/h while maintaining tight size control and long-term stability. Fabrication by SLA-DLP offers cost-effective, reproducible devices. Future work should quantify drug encapsulation efficiencies for hydrophilic and hydrophobic agents under MVF conditions, optimize downstream concentration and buffer exchange strategies for high-FRR operation, extend to diverse lipid chemistries and payloads, and adapt MVF principles to other nanoparticle systems (e.g., solid lipid or inorganic nanoparticles).

• High FRR operation, often necessary for minimal sizes, yields dilute suspensions requiring downstream concentration and buffer exchange (e.g., ultrafiltration), adding process complexity.
• Drug encapsulation efficiency under MVF, for both hydrophilic (aqueous phase) and hydrophobic (lipid phase) cargos, was not determined and may vary with FRR and mixing kinetics.
• Throughput increases are constrained by the laminar-to-transitional flow boundary (~60–80 mL/min); exceeding this can destabilize focusing and increase PDI.
• Device fabrication yield was ~50% due to potential nozzle/channel clogging and print defects; process refinements may be needed for manufacturing robustness.
• Results focus on specific lipid systems (e.g., DMPC:cholesterol:DCP and PEG2k-PE). Generalizability to unsaturated lipids or other formulations may require additional validation.
• While simulations used a solvent mole fraction threshold (x_ethanol≈0.5) to infer onset of vesicle formation, this approximation may vary across lipid chemistries and conditions.