Introduction

Immune modulation and desensitization is an area of significant current research that applies to clinical indications from oncology to food allergies. We present a new nanoparticle (NP) process and NP constructs to encapsulate proteins and address this broad range of therapeutic challenges. As an example, we apply these methods to generate polylactide-co-glycolide (PLG) NPs designed to serve as orally deliverable therapeutics for individuals with peanut allergies1. Currently, allergen immunotherapy (AIT) is the only approved clinical treatment for allergies in humans that can modulate the course of allergic disease (decrease in allergen-specific IgE, Th2 cytokines, and anaphylaxis). One major form of the immunotherapy delivered orally has been shown to modify the allergic response to food allergens by multiple administrations of increasing doses of “unmasked” sensitizing allergens, leading to Th1 and T regulatory (Treg) immune responses known to induce tolerance to the allergen2. However, because the allergens are not encapsulated, it also gives rise to undesirable adverse reactions that limit its therapeutic value, diminish its clinical use, and compromise its safety3. In initial clinical trials, a significant number of patients taking Palforzia®, an FDA-approved oral immunotherapy for peanut allergy, had an anaphylactic reaction requiring the administration of rescue epinephrin prior to reaching the maintenance dose4. In addition, the majority of participants undergoing such therapy regained reactivity upon discontinuation of the therapy.

Nanoparticle approaches to immune desensitization have shown significant promise5,6. First, encapsulation largely prevents premature engagement of the allergen with IgE-bearing effector cells, e.g. mast cells and basophils, to avoid anaphylaxis3. The NP construct enables uptake of the encapsulated antigen into antigen-presenting cells, which subsequently modulate T cell responses. In addition, the nanoparticle construct has the ability to serve as a co-delivery system to transport allergens in combination with selected adjuvants that serve to enhance and modulate immune responses. This allows a design strategy to skew the immune response away from Th2-driven allergic responses to Th1-driven and T-regulatory (Treg) responses that are known to achieve a successful AIT response. Finally, NPs can be designed for effective oral delivery, overcoming the poor immunogenicity (low bioavailability) of orally administered allergens, in part by the codelivery of adjuvant and the rapid uptake of NPs by dendritic cells (DCs)7.

Preclinical studies using DC targeting adjuvants such as Toll-like receptor (TLR) agonists delivered by NPs have been shown to lead to successful immune modulation8. For example, previous studies by Sampson et al.9 treating peanut-sensitized mice with peanut-encapsulated PLG nanoparticles and TLR agonists known to initiate and modulate immune responses demonstrated significant and long-lasting protection from anaphylaxis compared to unmasked peanuts. Other studies have also used NP constructs directed at peanut allergy2 and gluten sensitivity5 by intravenous injections. All of the above studies illustrate the power of nanoparticle delivery systems both to prevent anaphylaxis by encapsulating the allergen and in other cases to direct and optimize the induction of immune responses that inhibit allergic symptoms with the use of adjuvants. While these studies demonstrated proof of concept, the double emulsion process by which they were made is not efficiently scalable. Our process offers much better scalability and uses oral delivery for greater ease of treatment and safety.

Protein encapsulation for the purpose of immune modulation has been most widely investigated using biocompatible and biodegradable poly lactide-co-glycolide polymers (PLG), which are FDA-approved and have been successful in approved therapeutics such as the Lupron® depot10, which is a PLG microparticle. Current microparticles and NPs have almost universally been produced by the classic double emulsion process to encapsulate the protein antigen11. The double emulsion process is effective at trapping protein in larger microparticle constructs on the order of 10–40 microns, such as the Lupron depot. However, it has fundamental limitations when applied to nanoparticle-sized carriers. Those limitations are associated with the loss of protein during emulsification, which leads to poor encapsulation efficiency and low loading of protein into the final NPs. Currently, there are no FDA approved PLG submicron-sized nanoparticle drugs. In a recent study3, PLG polymer matrices were used to encapsulate peanut protein into 500–633 nm NPs; however, they only achieved protein loadings of 0.18 −1.0 wt% peanut protein, and only 3.5 wt% ovalbumin (OVA). These protein loadings are characteristic of the field11.

Translation of the concept of allergen desensitization with NPs has been impeded by the lack of a process to produce highly loaded PLG at scale. In this contribution, we present a new process that overcomes this translation limitation and enables the production of 100 g NP samples of 300 nm size with 5 wt% peanut protein loading. In addition, the process enables the incorporation of a DNA adjuvant into the protected protein core, and an activated E. coli lipid extract glyco-phospholipid adjuvant (ECPL) coating to stimulate antigen-presenting cell uptake in a mucosal application. This is important because adjuvants are a powerful tool not only to increase the immunogenicity of allergens but also to induce the desired type of immune response, such as allergen-specific Th1 and Treg responses, both needed to improve AIT. The TLR4 agonist E. coli phospholipids (ECPL) is known to increase DC activation and induce Th1-biased immune responses and the TLR9 agonist, E. coli DNA is known to activate DCs, Th1, and Tregs. Incorporating both is a key element of our PLG NP delivery system to successfully and synergistically suppress immune responses associated with allergy.

There are four innovations that enabled the successful production of these NPs. The first is realizing that the protein loss during classical PLG encapsulation occurs because the precipitation of the PLG matrix has been done by injection into an aqueous phase. Since the protein being encapsulated is soluble in the aqueous phase, there is a driving force for its escape from the precipitating PLG matrix. The smaller the NP that is being produced, the faster is that escape, since the diffusion escape time scales with the square of the particle size. Therefore, we adopted an organic antisolvent, isopropyl alcohol (IPA), which is an antisolvent for both the PLG and protein dissolved in DMSO. Consequently, there is no driving force for the loss of protein; this enables the loadings of peanut protein of 5%, which we obtain.

The second innovation is the incorporation of the ECPL coating on the NP. This coating is not applied from solution, but is rather “spread” or “smeared” on the NP surface by mechanical shear in a way that is reminiscent and motivated by the spreading of red blood cell membranes on the surfaces of NPs done by Zhang12 and Muzykantov13.

Third, unexpectedly, the NPs are formed in the IPA solution without the need for a stabilizer. In previous PLG microparticle studies, a polymeric stabilizer has had to be added to the antisolvent stream to prevent NP aggregation during precipitation or processing. In our process, the NPs are self-stabilizing and have extraordinarily narrow polydispersities with polydispersity index (PDI) metrics of around 0.1 compared to the 0.3–0.8 values typically seen for NPs produced through double emulsion processes3,11.

Fourth, and most unexpected, was that uniform NPs are produced by a globally heterogeneous precipitation process. The universal understanding of how to control particle size and uniformity during precipitation is to conduct the precipitation under uniform supersaturation to drive uniform precipitation kinetics14. This is accomplished either by using intense mixing to homogenize the solvent/antisolvent streams15,16, or to use microfluidics to reduce dimensions so that the diffusion time scales are short and solvent uniformity is rapidly achieved17. We have found that an intentionally heterogeneous, layered two-fluid process with mild mixing produced the desired NPs. We have termed the precipitation a “Tequila sunrise” precipitation, since the dye layered visualization experiments (Fig. 1) are reminiscent of that drink. Quite notably, this layered system allows scale up from laboratory scale to production scale involving 190 L batches that enable the production of NPs containing 5 g of protein that would provide 2500 doses for testing in a clinical trial. This intentionally heterogeneous precipitation means that particle formation occurs continuously under a time-dependent concentration driving force, but still produces uniform NPs.

Fig. 1: Protein encapsulating nanoparticle (NP).
figure 1

a Core comprises protein and DNA in PLG matrix. Shell comprises deoxycholate and E. coli total lipid extract (ECPL). b Schematic of process steps for NP precipitation and post processing. c Visualization of Tequila sunrise layered fluid system for PLG NP precipitation. The stir bar in (1) is 5 mm as a scale calibration. In this demonstration experiment the IPA antisolvent phase (1) contains green fluorescein dye. The dense DMSO layer (2) containing PLG and DNA with red Eosin Y dye is layered under the antisolvent layer. The low-speed impeller is placed at the interface of the two fluids then stirred (3–5). d a representative particle size distribution from precipitation is shown, with an average NP size of 305 nm and polydispersity index (PDI) of 0.1. e SEM of representative NP dispersion, with 1 micron scale bar. f PLG NPs trigger activation of mouse and human TLR2,4 and 9 as determined in TLR-transfected HEK293 reporter cells through measurement of NF-κB activity. Error bars are ± standard deviation of triplicate replicates.

Full size image

In this study, we first present the development of the Tequila Sunrise precipitation process, and then the post-processing by tangential flow ultrafiltration (TFF), and stabilization with cryoprotectants so that frozen samples can be stored, thawed, and administered. We describe the application of these techniques to generate NPs encapsulating either protein from a crude peanut extract or chicken ovalbumin (OVA). NPs from the second protein demonstrate the “platform” capability of the process. We then present biological assays to demonstrate TLR activation by NPs carrying selected adjuvants, a substantial reduction in the propensity of nano encapsulated peanut antigen to cause peanut allergen-induced basophil activation using basophils from peanut-allergic patients, and OVA-specific T cell responses induced in vitro and in vivo through oral administration of OVA-encapsulated nanoparticles to OVA-responsive mice.

Results and Discussion

Design of PLG nanoparticles for peanut allergy desensitization via sublingual/buccal oral administration

The design of PLG constructs takes into consideration the best options to develop NPs for successful treatment by AIT, based on previous findings18,19.

Figure 1 provides a schematic representation of the particle’s components and a block diagram of the steps in the process. The PLG matrix encapsulates both the peanut protein allergens as well as sheared E. coli DNA as an adjuvant to activate TLR 9. Coating the nanoparticle is a mixture of ECPL (as E. coli total lipid extract) and sodium deoxycholate, which provides colloidal stability and activates TLRs (TLR4) on dendritic cells of the immune system.

Nanoparticle (NP) preparation using the Tequila Sunrise process

Briefly, (see Methods for detailed description) the peanut protein and DNA in aqueous buffer were added to DMSO containing the PLG polymer. The solution was concentrated by rotary evaporation. The dense DMSO phase was carefully layered below the less dense IPA antisolvent phase. In Fig. 1 green fluorescein dye was added to the IPA and red Eosin Y dye was added to the DMSO to visualize this layering. An impeller placed at the interface between the two phases slowly mixed the two initially clear layers. Turbidity, as NPs are produced, is observed at the interface as mixing proceeds until the entire volume comprises the turbid NP dispersion in a predominantly IPA antisolvent phase. No additional stabilizer is introduced during the precipitation, and extremely narrow size distributions (Fig. 1, Table 1) and reproducibility are obtained: a polydispersity (PDI) of 0.1 is measured for the distribution, and triplicate precipitation runs yielded a standard deviation of the measured size of only 2 nm (Table 1). A three-fold volume of aqueous deoxycholate solution was added to increase the polarity of the solvent and to coat the NPs with the negatively charged deoxycholate. The deoxycholate introduction increased the electrostatic repulsions between NPs to reduce aggregation during TFF concentration and produced the negative surface charge (ζ = − 40 mV (Table 1)) that upregulates immune cell uptake. TFF was conducted on 190 L volumes of dispersion to remove IPA and concentrate the dispersion. NP size increased to 330 nm, but with excellent reproducibility (11 nm standard deviation between three runs). The E. coli extracted phospholipids oil is not aqueous-soluble, so the ECPL was transferred to the NP surface by high shear homogenization. Trehalose at 100 mg/mL was added to the dispersion as a cryoprotectant to enable freeze-thawing stability. The 305 nm NPs increased in size by only 15 nm when thawed and gently shaken.

Table 1 Nanoparticle sizes at various steps in the process
Full size table

From 8.57 g of peanut protein introduced into the process, the final concentrated dispersion contained 5.1 g of protein with 4.9 g encapsulated in PLG and 0.2 g as soluble protein. The overall encapsulation efficiency is 57%. The precipitation into the IPA antisolvent minimized protein escape during initial NP formation, and the strong electrostatic stabilization of the NPs minimized losses of NPs to the filter surfaces during TFF. These two factors contributed to the high overall efficiency of the process. BCA assay determined that the amount of unencapsulated (free) protein in the final dispersion aqueous phase is less than 4%. This low level of free protein ensures increased NP safety.

Analysis of TLR activity of nanoparticles

Activation of TLR-mediated signaling leads to the stimulation of the NF-κB/AP-1 transcription pathway, which can be measured using a reporter gene construct whose expression is responsive to NF-κB/AP-1. For each of the TLRs examined, Fig.1f presents the fold induction of reporter expression observed at the dilution of peanut PLG NPs that produced maximal induction (1:100 for TLR2 and TLR4, 1:10 for TLR9). These results demonstrate that the bacterial DNA and ECPL that were incorporated within the peanut PLG NPs are bioavailable and able to activate their target TLRs in the cell-based bioassay. It should be noted that each TLR responds to a different pathogen-associated molecular pattern, for which they manifest distinct affinities. In addition, the availability of nanoparticle-associated adjuvants to interact with TLRs is influenced by the geometry of the nanoparticles. Since DNA is encapsulated inside of the nanoparticles, it is perhaps not surprising that a higher concentration of nanoparticles is required to maximally activate the DNA-responsive TLR9, as compared to the nanoparticle concentrations that are required to maximally activate TLR4 and 2, which respond to lipids that are accessible at the surfaces of the nanoparticles. Thus, the PLG NPs that we have developed will deliver their peanut protein payload in association with bioactive bacterial adjuvants that allow these NPs to be tuned so as to produce desired immunological responses.

Design of PLG nanoparticles encapsulating ovalbumin (OVA)

NPs encapsulating OVA were prepared to demonstrate that the Tequila Sunrise process was a platform that could be applied to proteins more widely. These experiments were run at 1.18 g OVA scale. The Tequila sunrise precipitation was identical to that performed for the peanut protein NPs. However, there were two differences in the post-processing steps. The first was that the deoxycholate was not required to stabilize the NPs. We believe that this is attributable to the greater hydrophilicity of the OVA protein relative to peanut proteins. This is consistent with the fact that albumin, another hydrophilic protein, is by itself sufficient to adequately stabilize nanoparticles in the commercial paclitaxel-albumin nanoparticle drug Abraxane20,21 without the need for additional components to add electrostatic repulsions. The second difference is that the ECPL was added immediately after nanoprecipitation rather than after an intermediate deoxycholate addition.

Immunological characterization of PLG nanoparticles

To test the NP constructs, peanut PLG and OVA PLG NPs were prepared to analyze their safety and efficacy of immune modulation. Since oral delivery has been a challenge due both to poor bioavailability and low immunogenic response, it was important to demonstrate that the NP constructs can provide the necessary signals to induce the desired immune responses when used orally.

Below we demonstrate that encapsulating peanut extract within PLG NPs and exposing basophils from allergic individuals to these peanut PLG NPs leads to substantially less response compared to unencapsulated peanut extract, thus improving the safety of peanut delivery to allergic individuals. Finally, we demonstrate that the peanut PLG NPs activate T cells in vitro and in vivo, and thereby generate immune responses known to inhibit allergic responses.

Demonstration that basophils from peanut-allergic patients are substantially less responsive to encapsulated peanut proteins

To determine if PLG NPs carrying encapsulated peanut extract and administered orally would increase the safety of the delivery system compared to peanut powder, human basophils prepared from peanut-allergic individuals were exposed to the peanut PLG NPs, unencapsulated peanut protein or an empty vehicle control (without peanut extract and adjuvant). Basophil activation was evaluated by determining the percentage of basophils expressing CD63 following stimulation with various concentrations of peanut, which correlates directly with histamine release22. As seen in Fig.2a, human basophils from a peanut-allergic patient showed substantially more CD63+ basophils at a lower dose of unencapsulated peanut extract, as compared to that seen when CD63+ basophils were exposed to the same quantity of peanut extract encapsulated in peanut PLG NPs. In addition, the analysis of 7 individuals, plotted in Fig.2c at a concentration of peanut extract needed to elicit a half-maximal basophil response, showed that the dose of peanut PLG NPs required to induce basophil activation is considerably greater (~10-fold) than the dose required by unencapsulated peanut extract.

Fig. 2: The activation of basophils from peanut allergic individuals.
figure 2

a Representative dose-response curve of basophil activation obtained with one of the study subject samples. Basophil activation is shown as percentage of CD63+ basophils in response to increasing concentration of the peanut extract protein (PNE). The blue dotted line is unmasked peanut extract (PNE) compared with encapsulated peanut nanoparticles (PN + PNE) (green) and empty nanoparticles (Empty NP) (dotted green). Empty NP are composed of PLG materials without PNE and adjuvants. The concentration of the empty nanoparticles used for the stimulation was calculated based on the ratio of PNE to the nanoparticle’s material in the NP + PNE formulation so each point contains the same mass of PLG with or without PNE. “0” on the X-axis corresponds to the negative (no stimulant) control condition. b The percentage of the CD63+ basophils when stimulated with fMLP (green bar) or anti-human IgE (red bar) as positive controls for basophil activation. c EC50 values (Y-axis, logarithmic scale) obtained with the 7 study samples stimulated with PNE (blue bars) or with PN + PNE (green bars).

Full size image

Significantly, these data demonstrate that human basophils were approximately 10 times less responsive to the peanut PLG NPs than to the unencapsulated peanut extract.

These data provide ex vivo evidence that encapsulating the peanut in NPs substantially reduces the level of IgE-mediated basophil degranulation, which suggests that encapsulation would increase the safety of oral immunotherapies, including AIT.

Demonstration that the new PLG nanoparticle process produces nanoparticles that activate OVA-specific T cell responses in vitro

To test the ability of T cells to respond to encapsulated antigen, NPs encapsulating OVA were prepared as described, with the PLG matrix encapsulating both OVA proteins as well as sheared E. coli DNA as an adjuvant to activate TLR9. The ECPL coating was applied to activate TLR2 and TLR4 on dendritic cells.

Bone marrow-derived dendritic cells (BMDC) were pulsed with OVA PLG NPs or unencapsulated OVA for 2 h, washed, and then cultured with OVA-specific CD8+ and CD4+ T cells derived from T cell receptor transgenic mice (OT1 and OT2 respectively). As can be seen in Fig. 3b–g (and summarized in a), both CD4+ and CD8+ T cells responded by proliferation to the OVA PLG NPs at all doses tested (dark green bars).

Fig. 3: Comparison of T cell activation (proliferation and cytokine production) induced by bone marrow-derived dendritic cells (BMDC) pulsed with OVA PLG NPs and unmasked unencapsulated OVA in vitro.
figure 3

Bone marrow-derived FLT3L dendritic cells (BMDCs) were pulsed with OVA PLG NPs including incorporated adjuvant (NP + OVA+adjuvant), unencapsulated OVA alone (OVA), and no antigen control (Ctr) at various doses of antigen for 1 h. The adjuvant includes both ECPL that coats the NPs and sheared DNA embedded within the NPs. CD4+ (OT2) and CD8+ (OT1) T cells were derived from spleens of OT 2 and OT1 T cell receptor transgenic mice and were labelled with CFSE. The BMDCs and T cells were cocultured for 3 days, and the proliferative response of the CD4+ (OT2) and CD8+ (OT1) T cells was analyzed 3 days later (CFSE% T cells). b–g Bar graphs show CD4+ and CD8+ T cell proliferation in response to BMDCs, pulsed with (100, 10, 1 µg/mL OVA or no OVA). a 3D Summary of CD4+ and CD8+ T cell proliferation shown in b–g. Data are representative of three independent experiments. i–l, CD4+ T cell cytokines production (IFNγ, IL-10, IL17, IL4). m, n CD8+ T cell cytokines production (IFNγ, granzyme B (Gzb)). h 3D summary of CD4+ and CD8+ T cell cytokine production. BMDC were pulsed with NP + OVA+adjuvant, NP+adjuvant, unencapsulated OVA along with adjuvants (OVA&adjuvant), and unencapsulated OVA alone (OVA). Data are representative of three independent experiments. P-values were calculated by an unpaired t test. Error bars are ± standard deviation of triplicate repeats.

Full size image

In addition, the OVA PLG NPs induced a more robust T cell response at lower doses of OVA (10 and 1 μg/mL) than did unencapsulated OVA, which showed little response at those two doses. The presence of the OVA protein itself was required for the NP-induced proliferative response since the empty NPs prepared with only the adjuvants did not drive T cell proliferation (Fig. S1a, b in supplementary information).

These findings also show Fig.3i–l (and summarized in h) that a substantial number of OVA specific CD4+ T cells (77%) express IFN-γ when exposed to OVA NPs (dark green bars). To a much lesser extent they express IL4, IL17, and IL10 (less than 6%). Unencapsulated OVA or unencapsulated OVA administered together with adjuvant induces fewer than 17% cytokine-expressing cells (red bars). Upon stimulation, CD8+ T cells show IFN-γ/granzyme expression (Fig. 3m,n) that is greater than that seen with unencapsulated OVA administered together with adjuvant, and with NP with adjuvant only.

These data demonstrate that OVA PLG nanoparticle-pulsed dendritic cells induce skewed Th1, inducing primarily IFN-γ producing CD4+ T cells. IFN-γ was produced by both CD4+ and CD8+ T cells, which serves as a significant deterrent of allergy-associated Th2 immunity.

Demonstration that the new PLG nanoparticles activate antigen-specific T cell proliferative responses in vivo

To analyze the ability of OVA PLG NPs and NPs without OVA to activate T cell proliferative responses in vivo, isolated OVA-specific CD4+ (OT2) and CD8+ (OT1) T cells (labeled with CFSE) were adoptively transferred to WT mice. As seen in Fig. 4a, b, 20–60% of T cells responded to the OVA-encapsulated NPs in vivo. This response is higher (ranging from 2.5 to 60-fold) than what is seen with NPs without OVA.

Fig. 4: The proliferative response of CD4 (OT2) and CD8 (OT1) T cells to OVA nanoparticles in vivo.
figure 4

Isolated CD8+ (OT1) and CD4+ (OT2) T cells were adoptively transferred into WT mice. 18 h later the mice were immunized (oral gavage) with NPs incorporating OVA and adjuvant (NPs+OVA+adjuvant) or NPs incorporating adjuvant only (NPs+adjuvant). b The proliferation of T cells from mesenteric lymph nodes was measured at day 3 in 4 independent experiments. a 3D summary of the 4 experiments shown in b. P-values were calculated by an unpaired t test. Data are representative of three independent experiments. Error bars are ± standard deviation of triplicate repeats.

Full size image

These data indicate that oral administration of OVA NPs is capable of activating CD4+ and CD8+ T cells to a substantially greater extent in vivo than NPs without OVA. Taken together, the NP platform provides an approach to AIT that is both practical from the manufacturing perspective and promising with regard to its potential efficacy and safety.

Conclusions

We have presented a new scalable platform process to prepare poly(lactide-co-glycolide) PLG nanoparticles (NPs) that encapsulate proteins. NPs encapsulating peanut protein were prepared for allergy immunmodulation therapy (AIT), and NPs encapsulating OVA were prepared as a demonstration of the broad applicability of the process. Peanut-encapsulating NPs of 300 nm size with narrow size distributions contain 5 wt% protein were prepared at 100 g NP scale. The initial precipitation and post-processing steps enable the preparation of complex NPs that contain DNA and ECPL that drive IFNγ producing CD4+ and CD8+ T cells, and highly negative surface charges established by deoxycholate coating to maximize immune cell uptake. The process, termed the “Tequila Sunrise”, involves a non-homogeneous, layered precipitation that is reproducible at scale--three replicate runs showed a reproducibility of NP size of 11 nm. It is a departure from the conventional water-in oil-in water (W/O/W) emulsion precipitation processes for NPs, which only achieve low loading efficiency and low protein loading. In contrast, for the entire precipitation and post-processing 59% of the peanut protein charged to the process is captured as NP product after precipitation and post processing. The 100 g run would represent 2500 doses of peanut protein for AIT administration.

The nanoparticles are capable of producing immunological responses induced by their encapsulated proteins in cell-based ex vivo assays as well as in vivo, following oral delivery in mice. This feature suggests the intriguing possibility that this nanoparticle production strategy can serve as a platform for the development of a wide range of protein or RNA-based oral vaccines. We also show that the cargo encapsulated in these nanoparticles is protected from exposure to mast cells and basophils, thus reducing the likelihood that their administration to highly allergic patients would initiate severe anaphylactic responses. Thus, these nanoparticles have the potential to be especially useful as agents for inducing allergic desensitization. In summary, we have developed a method that supports the production at scale of nanoparticles whose size, surface properties, and content can be adjusted to meet the specifications required by a variety of immunomodulatory applications.

Methods

Peanut Protein NP Preparation

Protein carrier phase preparation

This detailed description is for the large-scale peanut protein encapsulation process, which produced 1.7 L of concentrated peanut protein NPs (at 3 mg/mL peanut protein) at the final scale. However, initial laboratory scale studies produced NPs at 40 mL scale with the same NP properties. The OVA processing is similar. The matrix PLG (50:50 poly(DL-lactide-co-glycolide) with a 0.26–0.54 dL/g inherent viscosity from Lactel (B6010-1, Birmingham, AL) was dissolved in DMSO at a concentration of 100 mg of peanut protein per gram of PLG. The pH is adjusted to 9 with 0.1 N NaOH, and DNA, (Scientific Protein Laboratories, Waunakee, MI) is added at 4 mg/100 mg peanut protein. The peanut protein is defatted, lyophilized peanut extract (Stallergenes Greer; Lenoir, NC) composed of 91% peanut protein by dry weight, and this material was used as-is without further purification. Peanut and DNA solution (493 mL) at 1.7 wt% peanut is added to 8.976 kg of DMSO containing 78.2 g PLG. A clear solution is obtained. However, the resulting solution is too dilute in PLG to precipitate and form properly sized NPs in IPA. Therefore, the PLG and peanut concentrations were raised to 27.4 mg/mL PLG and 2.7 mg/mL protein by rotary evaporation at 2 mTorr and 75 C. In this concentration step, 65% of the solution mass was removed. The resulting solution appeared clear, but quite likely contains nucleating or phase separating protein and PLG by virtue of depletion flocculation, which is known to occur in protein and polymer solutions23. This possibility is supported by the fact that direct dissolution of the PLG, protein, and DNA at the final concentrations could not be achieved, even with moderate heating.

Nanoparticle precipitation

The 3645 ml of antisolvent IPA mixed with 405 mL of DMSO was introduced into a 10 L vessel. The 900 mL of concentrated protein PLG solution, which has a density of ~1.1 g/cm3, was carefully layered under the IPA/DMSO solution. A four-bladed stirring paddle with a diameter of 114 mm was placed at the interface in the 217 mm diameter vessel. The impeller blade was placed at the interface with 67% of the blade in the DMSO layer. The overhead stirrer was set to 80 rpms to provide gentle stirring for 2 min, at which time the entire vessel was turbid indicating the layering had been eliminated. At that point, the stirring speed was increased to 150 rpm for 1 min. Our goal had been 200–350 nm NPs, which is expected to optimize NP uptake by the cells involved in antigen processing and presentation. With pure IPA in the antisolvent phase, we found that the NP size was smaller than this goal. By adding 10% DMSO to the IPA antisolvent phase, we decreased the antisolvent quality, which slowed the precipitation and resulted in the desired NP size. The sizes for each step of the process are given in Table 1.

The particle size distributions are exceedingly narrow and reproducible, as shown by using a Zetasizer Nano-ZS (Malvern Instruments, Southboro, MA) at 25 °C with a detection angle of 173°. DLS data were processed with Malvern’s software using a cumulant model for distribution analysis. The cumulant analysis is defined in the International Organization for Standardization (ISO) standard document 13321. The calculations of PDI are defined in the ISO standard document 13321:1996 E. PDI’s of 0.1 or less are considered monodisperse. For three separate experimental runs using the Tequilla sunrise precipitation process, the z-average size varied by only 11 nm (standard deviation, Table 2, for replicate runs) and the PDIs were 0.09, 0.10 and 0.09.

Table 2 Sizes of three replicate lots of PLG particles prepared by nanoprecipitation from the multiphase system at a 2.9 g scale with respect to encapsulated protein
Full size table

Coating with deoxycholate

While these particles were stable as formed, during the subsequent concentration by tangential flow ultrafiltration (TFF), they aggregated. To enhance NP repulsions, electrostatic stabilization was incorporated by coating them with the negatively charged amphiphilic bile salt—deoxycholate. The efficacy of this approach had been demonstrated previously with emulsion-based processing of NPs24. Sodium deoxycholate was added at 0.11 mg/mL in a 10 mM pH 8.2 ammonium bicarbonate (AmBic) buffer at a volume ratio of approximately 3:1 aqueous buffer to IPA dispersion: 4.95 L of the IPA + DMSO and protein NP product was homogenized with 16.2 L of AmBic + deoxycholate buffer to produce a final batch of 21.15 L. The addition was performed using a high shear mixer (Silverson, L5M-A), mounted with an in-line mixing assembly operated at 10,000 rpm for 6 min. The deoxycholate concentration was calculated to create a 2 nm thick uniform coating on the surfaces of the NPs. In addition to providing steric stabilization of the NPs during processing, Hughes et al.3 have shown that highly negatively charged NPs enhance immune cell uptake. Three runs of precipitation were combined to yield 63.45 L.

Tangential Flow Ultrafiltration for solvent removal and concentration

Tangential flow ultrafiltration (TFF) was used to remove antisolvent, to remove unencapsulated protein, and to concentrate the final dispersion. This was accomplished in two TFF steps. The first step used a modified poly(ethersulfone) membrane (Repligen, hollow fiber, mPES) with a 750 kDa pore size. The unit had a surface area of 41,000 cm2 and was operated at a shear rate of 19,000 s−1 with a transmembrane pressure of 30 psi. These conditions were chosen to minimize NP deposition on the membrane surface during TFF. The sample was diluted 3X to 190 L buffer to dilute the IPA concentration, and then concentrated by 14X, followed by diafiltration at constant volume with 20X volume of 10 mM phosphate buffer, 20 mM NaCl and 0.08 mg/mL deoxycholate. TFF took 2 h. The added deoxycholate maintains the charge and stability of the NPs since labile deoxycholate passes through the membrane along with the unencapsulated peanut protein and DNA. The holdup volume in the large TFF unit was too large to enable concentration to the desired final concentration of peanut protein of 2 mg/ml. Therefore, a second, smaller-scale TFF unit with the same mPES membrane (8,500 cm2 area, and 16,400 s−1 shear rate) was used for the final 3.6X wash and concentration. The final solution comprised 10 mM phosphate buffer at pH 9 and 0.08 mg/mL deoxycholate. The total volume was reduced from 13.5 L to 1.7 L. Various NP dispersions were tested, under USP <467>, for residual DMSO and isopropanol solvents after TFF. All tested dispersions contained less than the 5,000ppm limit of each solvent, with DMSO under the 1,250ppm and isopropanol under the 50ppm detection limits of the assay.

Coating with E. coli total phospholipid (ECPL) extract

The lipid extract of E. coli from Avanti Polar Lipids (# 100500, Alabaster AL) consists mainly of phosphatidylethanolamine, phosphatidylglycerol, and cardiolipin as well as an additional 18% by mass of the polar components of E.coli cell membrane. The ECPL extract is insoluble in aqueous buffer. Therefore, its deposition on the NP surface involves mechanical shear of the lipid oil onto the interface. This is accomplished by high shear homogenization with a Silverson L5M-A multifunctional Lab Mixer mounted with a standard head and high shear screen, operated at 10,000 rpm for 4 min. The mechanical deposition results in some aggregation, which is minimized by continued shear. Shearing decreased the NP size from 330 nm to 305 nm. Spinning down the NPs and measuring the lipid concentration in the supernatant showed that 56% of the lipid was associated with the NP surface.

Freeze Thaw Stability

One application of these nanoparticles is to serve as an orally deliverable agent with which to suppress allergic reactivity in patients with peanut allergy. For this purpose, the final dosage form is a packet containing 1 mL of liquid dispersion. This would be shipped and stored frozen, and then thawed and administered. After an evaluation of a number of cryoprotectants, it was found that 100 mg/mL trehalose provided stability during freeze thawing. NPs 305 nm before freeze thaw were 320 nm after storage for 7 days at −80 °C.

OVA protein NP preparation

Protein carrier phase preparation

Ovalbumin (1.18 g) (Sigma-Aldrich) was dissolved in 27.41 g of DI water with magnetic stirring. To this was added 4.63 mL of a 10 mg/mL solution of DNA (Scientific Protein Laboratories, Waunakee, MI) in water. The pH was adjusted to 8.9 with 1 N NaOH (110 µL) and it was further diluted with an additional 27.41 g of water giving a 19.8 mg/mL and 0.777 mg/mL solution of ovalbumin and DNA respectively in pH 8.9 water.

A solution of 11.58 g of PLG in 1.33 kg of DMSO was prepared with stirring for 20 min until the PLG was dissolved. To this DMSO solution, the DNA and protein solution was slowly added. The solution was stirred for an additional 20 min and then transferred to a 3 L round-bottomed flask. Rotary evaporation at 2 Torr in a 75 °C water bath concentrated to solution to a final mass of 465 g.

Nanoparticle precipitation

The nanoprecipitation of the OVA particles was performed by mixing the DMSO solution of PLG, protein, and DNA with an antisolvent solution of DMSO and IPA. A 4 L beaker was charged with 1.273 kg of IPA and 198 g of DMSO. Using a glass tube connected to a peristaltic pump, 440 g of the DMSO solution of protein, DNA, and PLG was pumped gently under the IPA/DMSO layer such that two discrete layers were formed. An overhead stirrer with a stainless-steel stirring shaft and cross-shaped impellor was inserted such that ¾ of the impellor was in the bottom DMSO solution and was set to stir at 50 rpm for 2.5 min. The stirring rate was increased to 150 rpm and continued until a white homogeneous solution was obtained. By DLS, the particles were found to be 286 nm with a PDI of 0.15.

Coating with E. coli total phospholipid (ECPL) extract

The dispersion was homogenized with a solution of ECPL extract (Avanti Polar Lipids, # 100500, Alabaster AL) comprising 1.44 g in 7.2 kg of water, prepared with overhead stirring using a homogenizer (Silverson L5M-A, East Longmeadow, MA). Mixing involved pumping the E.coli extract solution into the nanoprecipitation mixture at 1.2 L/min while recirculating through the homogenizer at 10,000 RPM. Once all the E.coli extract was added the solution was homogenized for an additional 2 min. At the end of the homogenization, the DLS size was 225 nm with a PDI of 0.11.

Tangential Flow Ultrafiltration (TFF)

The dispersion was diluted with 18.8 L of 10 mM AmBic. Two TFF loops were run. The first TFF loop (Repligen, Waltham MA), hollow fiber, mPES, 750 kDa, 8500 cm2) was preconditioned with 10 mM AmBic. At a shear rate of 1.4k s−1 the solution was concentrated to 2000 mL and then diafiltered against 40 L of a buffer consisting of 10 mM potassium phosphate, 60 mM NaCl, and 0.1 mg/mL ECPL. After the 20-fold wash, 1200 mL of the resulting solution (69% of the total) was transferred to a second TFF loop (Repligen, hollow fiber, mPES, 750 kDa, 115 cm2), which had been preconditioned with a buffer solution of 10 mM phosphate and 0.1 mg/mL ECPL. The particle solution was first concentrated to 280 mL, washed 3.5-fold with the same buffer solution at 15k s−1 shear, and then concentrated to a final volume of 153 mL. Table 3 shows the final NP size during the process steps. The uniformity of the NP size is retained at each step in the process.

Table 3 Sizes of ovalbumin NPs during processing steps
Full size table

Physicochemical Characterization of the particles

Nanoparticles are characterized by size, zeta potential (as shown in Table 1), protein loading, DNA loading, and ECPL coating.

Protein concentration in NPs

A BCA assay was used to measure protein loading in the NPs and the ratio of encapsulated to unencapsulated protein. In brief, the NP dispersion was centrifuged (21.1 krcf for 20 min) to pellet the NPs. The supernatant was removed by filtering through a 100 nm filter and assayed by BCA.

To measure the protein encapsulated in the particles, a separate sample of the dispersion was digested in 0.1 M NaOH overnight to digest the PLG and release the protein. The solution becomes clear as the scattering from the PLG particles is lost since the ester bonds are hydrolyzed. The final solution was again measured by BCA against a standard of protein that had also been treated with the same concentration of base. The encapsulated protein is then determined by subtracting the free protein from the total protein.

For the peanut NPs, 59% of the initial protein was found in the concentrated NP dispersion at the end of TFF 2, for a total of 5.1 g of the initial 8.57 g of peanut protein. Protein encapsulation efficiency, defined as the mass of protein encapsulated in the NPs after processing divided by the mass of protein initially added to the precipitation, is 57% based on 4.9 g of encapsulated protein measured in the final dispersion. The total free protein (unencapsulated in the NP) in the dispersion was measured as 0.2 g for a total of 5.1 g in the final sample. We have employed a calculation of the ratio of free to encapsulated protein to define a metric we termed ‘safety factor’. The definition is based on administering a peanut protein dose using our NP dispersion that would be equivalent to the amount of peanut protein that would be given for current oral protein desensitization therapy (OIT). The NPs have a safety factor greater than 25; since, less than 4% of the peanut protein is free in solution. In contrast, with oral or sublingual approaches to peanut desensitization, 100% of the administered protein is immediately available. The 5.1 g of protein corresponds to approximately 2500 doses if administered at the peanut protein levels in the current OIT.

For the OVA particles, which is our second example protein, the final yield is 38%, which is higher than obtained by traditional PLG NP processes. The lower yield when compared to peanut protein is attributed to the lesser hydrophobicity of the ovalbumin relative to peanut protein. The less hydrophobic protein does not precipitate quite as easily with the PLG polymer in the initial precipitation.

PCR quantification of bacterial DNA

The presence of the DNA in the particles was quantified using PCR. Samples of the PLGA NPs were aliquoted at 50uL each as “Total” and “Pellet”. The tube labeled as “Pellet” was centrifuged at 10,000 rpm for 5 min, the supernatant was taken out to a separate tube labeled as “Sup”, the leftover pellet (nanoparticles) was then resuspended in 250uL of DMSO. The “Total” and “Sup” samples were dissolved in 9 volumes of DMSO. The assay was carried out in a 20 µL reaction containing 0.8 nM of E.coli primers specific to the ybbW gene, 0.2 nM of the E.coli probe (FAM), SprinTaq Master mix (containing 0.5 mM of dNTPs, hot-start Taq DNA polymerase and other buffer components), and template DNA (described below).

A standard curve was generated with 0.3 ng/uL to 0.003 ng/uL (10-fold step down) of the known E.coli DNA control (30 ng/uL stock). “Total”, “Pellet” and “Sup” samples from each NFD sample were then serial diluted and quantified against the known E.coli control DNA standard curve.

The cycling conditions for the reaction were the following: Hot-start 95 °C for 2 min, then followed by 45 cycles of (denaturation 95 °C for 3 sec and annealing 60 °C for 30 sec.

Quantification of the ECPL coating

The presence of the phospholipids was determined using a phospholipid assay from Abcam (AB241005, Cambridge MA). The sample was diluted 1:10 with 5% solution of Triton X by mixing 100 µL of sample with 900 µL of 5% Triton X solution in a 1.5 mL Eppendorf tube. The sample was mixed using an end-over-end rotator for 30 min after which the nanoparticles were separated by centrifuging at 21.1 krcf for 20 min. The supernatant was then run on the phosphatidylethanolamine Abcam assay. The assay being able to detect only PE, and the total phospholipid extract being a mixture containing 57% PE by weight, it was necessary to divide the obtained value by 0.57 to determine total phospholipid amounts.

Analysis of TLR activity of nanoparticles

TLR activation studies were conducted by InvivoGen, San Diego, CA. In the TLR activity assays, HEK293 cells are transfected to express a TLR of interest along with a cDNA encoding a NF-κB/AP-1-inducible reporter gene that encodes a secreted embryonic alkaline phosphatase reporter gene. Test materials are added to the medium bathing the cells, and colorimetric measurement of alkaline phosphatase enzymatic activity in the medium 16–24 h after the addition of the test material provides a measure of the quantity of bioavailable TLR-activating substances associated with the test material. Prior to being tested in the TLR-activation assay, the peanut PLG NPs were subjected to centrifugation (14.5 krcf for 30 min) on a sucrose step gradient (130% top layer, 200% bottom layer) to separate the NPs from any unincorporated ECPL and DNA to ensure that that measured TLR activation is produced exclusively by NP-associated TLR ligands. Assays were performed in triplicate. Positive controls were performed by incubating cells expressing a TLR of interest with a standardized preparation that contains an activating ligand of that TLR. For TLR2, the positive control was heat-killed Listeria monocytogenes (108 cells/mL), for TLR4 the positive control was E. coli K12 LPS (100 ng/mL), and for TLR9 the positive control was CpG ODN 2006 (10 μg/mL). TLR-expressing cells that were not incubated with peanut PLG NPs were used to establish the baseline signal in the assay. Cells that did not express exogenous TLR were also incubated with peanut PLG NPs, which demonstrated that the peanut PLG NPs did not produce non-specific TLR-independent activation of the NF-κB/AP-1 reporter.

Basophil Activation Test

Whole blood samples were collected in heparinized tubes from peanut allergic patients recruited under the FARE protocol, which was reviewed and approved by the Mount Sinai Program of the Protection of Human Subjects (IRB). The recruited subjects had a physician-confirmed history of an allergic reaction to peanut and a serum peanut-specific IgE antibody level that exceeds the 95% predictive value for clinical reactivity, i.e. >15 kUA/L25. For basophil activation 50 μL of whole blood was incubated (in the presence of IL3 at 1 ng/mL) with 10-fold serial dilutions of the peanut extract (0.001 to 10 µg/mL), or the nanoparticle-encapsulated peanut extract (0.001 to 10 µg/mL), or the empty nanoparticles formulation at concentrations calculated based on the ratio of peanut extract protein to the nanoparticles material in the nanoparticle-encapsulated peanut extract formulation, with no stimulant, with polyclonal rabbit anti-human IgE antibody (1 μg/mL, Bethyl Laboratories, Montgomery, TX), or with formyl-methionyl-leucylphenylalanine (fMLP, 1 μM, Sigma-Aldrich, St. Louis, MO). Antibody cocktail of anti-human CD63-BV421, CD123-PE-Cy5, HLADR-PE-Cy7, CD3-AF488, CD41a-APC, CD14-AF700, CD19-APC-H7 (all by BD Bioscience, San Jose, CA) and CD203c-PE (Beckman Coulter, Indianapolis, IN) for cell staining was also included in the stimulation mixture of final volume of 125 µL. Flow cytometry analysis was performed within 96 h using Cytoflex S flow cytometer (4 lasers, 13 colors, RUO, Beckman Coulter). Flow cytometry data were analyzed using Kaluza Analysis Software, v. 2.1 (Beckman Coulter). Basophil population was gated as CD123+, HLADR−, (CD3, CD19, CD14 –all negative), CD41aall. Basophil activation phenotype was expressed as percentage of CD63+ positive basophil.

Mouse strains

C57BL/6 WT mice and congenic C57BL/6-Ly5.1 [B6.SJL-PtprcaPepcb/BoyCrl] WT mice were purchased from Charles River Laboratories (Wilmington, MA). OT1 [C57BL/6 Tg(TcraTcrb)1100Mjb/J] and OT2 [B6.Cg-Tg(TcraTcrb)425Cbn/J] mice were purchased from Jackson Laboratories (Bar Harbor, ME). OT1 and OT2 mice were crossed onto the CD45.1 mice. We used age- and sex-matched (male and female) mice that were between 6 and 16 weeks of age in all experiments. All protocols used in this study were approved by the Institutional Animal Care and Use Committee at the Yale University School of Medicine (IACUC # 11398).

BMDC coculture

Bone marrow derived FLT3L-DC were generated as described26. In brief, bone marrow (BM) cells were prepared by flushing femurs and tibias with RPMI media using a 27-gauge needle followed by RBC lysis and filtering through a sterile 70 mm cell strainer. BM cells were cultured for 10 days at 8 × 106 cells/ 4 mL in RPMI 1640 culture medium containing 200 ng/ml murine FLT3L (Peprotech, 250-31 L).

Flow cytometry

Preparation of single-cell suspensions was blocked using a Fc receptor-blocking solution TruStain FcX™ (BioLegend, 101320) for 10 min at 4 °C prior to immunostaining. Subsequently, the cells were stained with corresponding antibodies for 30 min at 4 °C. Then, cells were washed to remove excess antibodies and resuspended in FACS buffer. Stained Samples were run on a Beckman CytoFLEX flow cytometer and then analyzed using FlowJo software.

The following antibodies were used for staining different cell subsets (from BioLegend): Anti-TCRb (H57-597), Anti-B220 (RA3-6B2), Anti-CD4 (RM4-5), Anti-CD8 (53-6.7), Anti-CD45.2 (104), Anti-CD45.1 (A20), Anti-IL-4 (11B11), Anti-IL-17a (TC11-18H10.1), Anti-IL-17a (TC11-18H10.1), Anti-IL-10(JES5-16E3), Anti-Granzyme B (GB11), Anti-IFN- γ (XMG1.2). LIVE/DEAD™ Fixable Aqua Dead Cell Stain Kit (Thermofisher, L34965) was used for gating out dead cell.

T cell proliferation and cytokine assay

For analysis of in vitro T cell proliferation, OVA-specific CD4+ or CD8+ T cells were prepared from spleen of OT2 or OT1 TCR transgenic mice using the EasySepTM CD4+ or CD8+ T Cell Isolation Kit (StemCell Technologies, 19852 or 19853) according to the manufacturer’s instructions. The CD4+ T or CD8+ T cells were labelled with CFSE (Thermofisher, C34554). FLT3L DC was pulsed with OVA or different nanoparticle preparations as the indicated condition for 1 h. Then DC and T cells were cocultured at 1:10 ratio and OVA-specific CD4+ T and CD8+ T cell proliferation were analyzed 3 days later. For cytokine stimulation, cocultured cells were harvested at day 3, and single-cell suspensions were prepared and restimulated with 1X eBioscience Cell Stimulation Cocktail (Thermo Fisher Scientific, 00-4970-93) for 5 h with 10 mg/ml Brefeldin A (Biolegend, 420601) for the last 4 h. The cells were surface-stained and fixed, and then permeabilized for intracellular cytokine staining with BD Cytofix/Cytoperm™ Fixation/Permeabilization Kit (BD biosciences, 554714). For analysis of in vivo T cell proliferation, OVA-specific CD4+ or CD8+ T cells were prepared and labeled with CSFE as above. Two million cells were transferred intravenously into recipient mice. 18 h later, the mice were treated with various NP preparations via oral gavage. For the OVA NP, 1.5 mg OVA was given and for the NP without OVA the concentration was based on the ratio of peanut extract protein to the nanoparticle material in the nanoparticle-encapsulated peanut extract formulation. T cell proliferation in mesenteric lymph nodes were analyzed 3 days after immunization.

Statistical analysis

All statistical analysis was performed using GraphPad Prism software. Data were analyzed with a two-tailed unpaired t-test with Prism software. Statistical significance is defined as *P < 0.05, **P < 0.01, ***P < 0.001 and ***P < 0.0001.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.