High‐Throughput Microfluidic‐Mediated Assembly of Layer‐By‐Layer Nanoparticles (2025)

2.1 Excess Polymer is Required for LbL Assembly Without Particle Aggregation during Batchwise Polyelectrolyte Adsorption

As a model system with relevance for targeted immunotherapy delivery to cancer cells, we studied anionic liposomes that were surface-conjugated with the potent cytokine IL-12 (IL-12 NPs), and subsequently coated by alternating layers of poly(L-arginine) (PLR) and poly(L-glutamate) (PLE). LbL assembly has been typically performed by alternating sequential incubations of the liposomal core particle with excess PLR and PLE in a low ionic strength buffer. Under these typical self-assembly conditions, each step of polyelectrolyte adsorption must be accompanied by extensive washing/purification to remove excess unbound polymer from the coated particles before the next round of adsorption.

To better understand and optimize LbL assembly, we first sought to define how the polymer:particle weight ratio varied during nanolayer deposition. Unlayered anionic IL-12 NPs were incubated with increasing PLR-to-liposome mass ratios, or“weight equivalents” (wt. eq.). Equivalently, purified PLR-coated cationic IL-12-NPs (PLR/NPs) were incubated with increasing PLE wt. eq relative to theliposome lipid mass. After polymer and NP mixing, we characterized the resulting unpurified polymer-NP assemblies' size and charge via dynamic light scattering (DLS) and electrophoretic mobility, respectively. Starting from very low polymer concentrations, increased polymer-to-NP wt. eq. led to increasing NP charge up to a charge neutralization point (isoelectric point), at which particles rapidly aggregated due to a lack of charge-charge repulsion (Figure 1a,b). Further increases in polymer-to-NP wt. eq. led to reductions in particle size as the overall NP charge inverted, and the zeta potential neared a plateau onset point (POP), where complete charge reversal was achieved. Given the known composition of the liposomes used here, we sought to estimate the charge neutralization and POP wt. eq. for PLR onto the IL-12 liposomes based on charge balance and found values of ≈0.05 and 0.1, respectively, which were similar to our experimental results (Figure1a, see Supplemental Information for derivation). Analogously, we further predicted charge neutralization and POP wt.eq. for PLE of ≈0.2 and ≈0.44 which were consistent with our empirical observations (Figure1b). However, increasing the polymer-to-NP wt. eq. beyond the zeta potential POP was required to form polymer-coated NPs with the lowest overall size and polydispersity index (PDI). Due to this reduced size and increased homogeneity, polymer:NP wt. equivalents beyond the POP are used for LbL-NP synthesis.^{[32, 40, 57-60]} These results suggested that excess polymers present during the layering prevent bridging of NPs by the polymer to avoid aggregate formation (Figure1c). Indeed, it has been previously shown that polymer deposition onto NP surfaces is a kinetically controlled reaction and is expected to be quantitative until the POP.^[61]

We next sought to assess whether incubation of NPs with excess polymer beyond the POP increased the amount of polyelectrolyte adsorbed to the NPs. Using conditions identified above that allowed polymer adsorption while maintaining a low polydispersity of the resulting NPs, bare IL-12 NPs or PLR/NPs were added to a solution of fluorescently-tagged PLR (0.3wt. eq.) or PLE (1 wt. eq.) under sonication to allow polymer adsorption.^{[32, 36]} The excess polymer was then removed via tangential flow filtration (TFF). This protocol of LbL deposition under sonication followed by TFF is our current best practice for the production of LbL-NPs.^[36] As expected, the production of LbL-NPs via the TFF protocol showed the characteristic inversion of surface charge and minor increases in overall size while maintaining overall low PDI as PLR and then PLE were adsorbed (Figure 2a). However, quantification of the amount of polymer bound to the NPs after purification demonstrated that only ≈0.1wt. eq. of PLR and ≈0.5-0.6wt. eq. of PLE were adsorbed (Figure2b). These results confirmed that incubation of NPs with polymer wt. eq. beyond the zeta potential POP (Figure1b,c) does not increase the amount of polymer incorporated into the LbL films.

Based on these results, we theorized that the assembly of LbL-NPs at the POP could allow us to omit purification steps, as no excess polymers should be present in the solution. To better understand the limitations of performing LbL-NP assembly at the POP polymer-to-NP wt. eq., we compared particles generated with POP wt. eq. of PLR or PLE to the standard layering with excess polymers under bath sonication followed by TFF. While we could generate reasonably sized LbL-NPs with POP wt. eq. of PLR in small-scale test batches (<50µL and <1mgmL⁻¹ lipids), a noticeable increase in NP size was observed when layering with a larger batch size of NPs (≈5mg) was attempted (Figure2c). Omitting sonication during the layering led to a further increase in the resulting NP size compared to the standard approach. Similarly, PLE layering with POP wt. eq. (0.5wt.eq. PLE-to-lipids) on purified PLR-NPs resulted in NP aggregation during layering (Figure2d). Thus, large-scale LbL assembly in static solution even in the presence of sonication could only successfully be performed under conditions of excess polymer, requiring subsequent purification steps.

2.2 Microfluidics-Enabled Mixing Generates Homogeneous LbL-NPs Without Intermediate Purification Steps

As polymer wt. eq. beyond POP only increases the number of free polymer chains in the solution without affecting the LbL coating composition, we theorized that mixing limitations during the standard bath sonication protocol lead to the observed NP aggregate formation. To overcome these limitations and manufacture large-scale LbL-NPs without the need for purification, we explored using microfluidic (MCF) channels for the mixing of polymers at their POP wt. eq. with NPs. We used a commercial MCF cartridge with a bifurcating mixer.^{[48, 50]} This MCF chip platform has been developed for lab- and clinical-scale cGMP manufacturing of NPs, making it an ideal candidate for clinical-scale manufacturing of LbL-NPs.^[48] For mixing, solutions of liposomes and polyelectrolytes were each introduced into one of two entry flow ports (Figure 3a); the two solution streams converged at the mixer and the mixed sample was collected from the outlet port.

We first evaluated the effect of flow rate through the channels on the adsorption of PLR to anionic IL-12 liposomes, combining PLR and liposomes at the POP. The target was to achieve NPs similar in size to that of the standard optimized bath sonication protocol. Increasing the flow rate led to reduced PLR-NP Z-avg size and PDI (Figure3b) while maintaining charge conversion (Figure3c), demonstrating improved sample homogeneity without impacting polymer adsorption. Importantly, at flow rates of >3 mLmin⁻¹, PLR-NPs were prepared with the same size and PDI achieved by the standard protocol expected for layered, non-aggregated particles. Higher flow rates than 10mL were unachievable due to flow rate limitations of the MCF chip. Nonetheless, the MCF mixing enabled large-scale polymer deposition at the POP wt. eq.

Based on these promising findings, we devised a two-stage microfluidic mixing process designed to allow two rounds of polyelectrolyte adsorption at POP wt. eq., in order to generate LbL-NPs without purification steps. IL-12-liposomes and PLR were combined in a first-stage MCF mixer followed by a 30min incubation step to ensure polymer adsorption and perform quality tests. The output of the first stage and a PLE solution were used as the input in a second-stage mixer (Figure3d). When we performed POP wt. eq. PLR deposition onto IL-12-NPs at increased scales (>1 mL and >5 mg) using MCF, we could readily achieve the target PLR-NP sizes of the standard bath sonication production protocol across multiple batches (Figure3e). Similarly, depositing PLE at its POP wt. eq. onto MCF-generated PLR-IL-12 NPs could yield PLR/PLE IL-12 NPs at equal or smaller sizes than the standard bath sonication protocol with reproducibility across batches (Figure3f). To validate that all polymers were adsorbed on NPs using the MCF protocol, we used fluorescently tagged PLE polymers to assemble PLE/PLR IL-12 NPs using both the standard TFF-based protocol and the new MCF LbL-NPs. We separated out any free polymer in the LbL-NPs using centrifugal filters and, as expected from the POP of ≈ 0.5wt. eq. for PLE, only ≈50% of PLE was bound to TFF-LbL-NPs prior to purification due to the excess polymer used during assembly with 1wt. eq. of PLE. On the other hand, both purified TFF-LbL-NPs and MCF LbL-NPs had >95% of PLE bound to NPs (Figure3g), confirming the lack of free polymer. Analysis of sample morphology via negative stain transmission electron microscopy (TEM) confirmed that both TFF-based and the MCF-based PLE/PLR-IL-12 NPs had a polymer film on the NP based on the altered staining around the liposome (Figure3h). Further, we did not find signs of significant polyplex formation in either LbL-NP formulation on the TEM micrographs (Figure S1, Supporting Information)

To compare the reproducibility between LbL assembly on IL-12 liposomes via either the standard TFF-based approach or MCF, we compiled the data from independent batches of TFF-LbL (n = 10) and MCF-LbL (n = 8) over the course of over six months. As expected, both methods yielded overall similar NPs with Z-avg 110–150nm, zeta potential of ≈−50mV, and PDI ≤ 0.2 (Figure S2; Table S1, Supporting Information). However, MCF-LbL NPs yielded smaller size and PDI with reduced standard deviations, demonstrating increased sample homogeneity and reduction in amount of partially aggregated LbL-NPs.

2.3 MCF LbL-NPs Maintain Desired Particle Properties In Vitro and Maintain IL-12-LbL-NP Efficacy In Vivo in a Metastatic Ovarian Cancer Mouse Model

PLE/PLR nanolayers assembled on IL-12 liposomes have two functions: First, the PLE layer promotes the targeting of the particles to ovarian cancer cells, and second, prevention of NP endocytosis following binding to the cancer cells, leading to high retention of the LbL-NPs on the cancer cell membrane.^{[8, 55-57]} As the process of polymer adsorption is a kinetically controlled assembly dependent on mixing, concentration, shear stress, and ionic strength, it was critical to ensure that MCF-LbL NPs maintained the desired properties of TFF-LbL-NPs.^[62-69]

To determine whether PLE/PLR-IL-12-NPs generated via MCF assembly had the expected cell-targeting properties, we dosed a murine metastatic ovarian cancer cell line, OV2944-HM1 (HM-1) with fluorescent NPs prepared using both the traditional bath sonication with TFF process and MCF assembly. We then quantified NP association with the cells after 4 or 24 h. As expected, MCF LbL-NPs and TFF-based LbL-NPs had significantly increased HM-1 cell association compared to unlayered (UL) NPs (Figure 4a). Moreover, MCF-LbL-NPs had equal or better HM-1 association compared to TFF LbL NPs, confirming that MCF layering maintained and potentially improved the desired high ovarian cancer cell affinity. However, the most critical characteristic of PLE/PLR LbL-NPs is its cell-membrane retention which enables high intratumoral extracellular residence time of the cytokine payload,IL-12.^[57] Thus, we evaluated the subcellular localization of NPs 24 h after dosing via confocal microscopy. While UL particles were all endocytosed (Figure4b), both TFF-LbL NPs (Figure4c) and MCF-LbL NPs (Figure4d) presented with NPs on the cancer cell membrane, confirming that PLE/PLR-IL-12-NPs retained their cell surface retention properties when prepared via MCF mixing.

Having validated that LbL-NP assembly via MCF maintained the desired cancer cell association properties, we next wanted to validate that the loaded IL-12 maintained a similar bioactivity to that of TFF-based IL-12 NPs. We dosed HEK-Blue IL-12 reporter cell lines with either free IL-12 or IL-12 bound to PLE/PLR-IL-12-NPs generated via the standard TFF-based protocol or MCF. The IL-12 on MCF-LbL maintained its activity like that of the TFF-LbL (Figure 5a,b). We then sought to evaluate the particle performance in vivo in mice bearing peritoneally disseminated metastatic HM-1 tumors. We used the same dosing paradigm shown previously to extend survival in this mouse model with IL-12-loaded PLE/PLR-LbL-NPs (Figure5c).^[57] Both TFF-based and MCF-based LbL-NPs were found to have better control of tumor growth based on whole-mouse tumor bioluminescence readings (Figure5d), significantly extending survival of mice compared to either free IL-12 or UL IL-12 NPs (Figure5e).

2.4 MCF LbL-NPs can be Readily Assembled with Varied Polymer Chemistries and NP Core Sizes for Screening of LbL-NPs Interactions

The LbL platform is a modular approach for tuning NP surface chemistries with a wide range of polymer chemistries showing promising results in preclinical models.^{[8, 70]} To validate that MCF assembly of LbL films was amenable to other polymer chemistries, we performed titration experiments of IL-12 loaded PLR-NPs with varying wt. eq. of hyaluronic acid (HA), poly-L-aspartic acid (PLD), or polyacrylic acid (PAA). Each polymer chemistry showed a distinct POP wt. eq. which followed the expected trend of mass charge density (PAA>PLD>HA) for each of the polymers tested (Figure 6a). Importantly, we could employ these POP wt. eq. for each of these varying polymer chemistries to generate monodisperse LbL-NPs using MCF mixing (Figure6b,c).

In addition to varying chemistries, LbL assembly may be performed on differing NP cores.^{[18, 32]} However, little has been explored on the effect of NP size on LbL-NP properties. Thus, we sought to layer carboxy-modified latex (CML) beads with varying core sizes with PLR and PLE. As expected, smaller CML particles required larger wt. eq. to reach the POP given their higher surface area to mass ratios (Figure6d). However, no discernable trend for the plateau zeta potential values could be determined likely due to the effect of NP size and roughness on the apparent zeta potential.^{[71, 72]} Nonetheless, we were able to readily assemble PLR/PLE coated CML beads of varying sizes (Figure6e). TEM confirmed that particles maintained their homogeneity (Figure S3, Supporting Information). When evaluated for their binding toward HM-1 cells via flow cytometry, all LbL-NPs showed an increase in median fluorescence intensity (MFI) compared to UL NPs (Figure6f). Interestingly, we found that increasing UL NP size increased NP association whereas smaller LbL-NPs showed higher association. We theorized that this effect on small NPs was due to their increased surface area which should allow for higher LbL-NP film interaction with cancer cells. Indeed, when we assessed the correlation between the log-fold change in LbL-NP MFI relative to UL-NP MFI and the available NP surface area, we found that the data showed a clear significant correlation (Figure6g).

The experiments above carried out layering in deionized water, but LbL assembly is also sometimes carried out in the presence of salts to create LbL films with thicker layers.^{[32, 73]}We thus tested whether LbL-NPs could be generated in the presence of salt-containing buffers. Here we employed anionic liposomes devoid of IL-12 with a composition optimized for siRNA loading and delivery from LbL-NPs.^[32] We first examined varying PLR:liposome mass ratios to find the POP of PLR required for LbL assembly in 25mM HEPES and 20mM sodium chloride (NaCl), a buffer composition previously found optimal for loading of siRNA into LbL films.^[32] ≈ 0.15wt. eq. of PLR was required to reach the POP (Figure S4a,b, Supporting Information). We next compared the size and zeta potential of LbL-NP synthesized from small-scale tests under bath sonication to increased bath sonication and TFF to that of MCF LbL-NPs. Similar to IL-12-LbL-NPs, we found that assembly with MCF enabled more homogenous particle formulations (Figure S4c–e, Supporting Information).