Formulation and in vitro characterization of PLGA/PLGA-PEG nanoparticles loaded with murine granulocyte-macrophage colony-stimulating factor.

Abstract

Granulocyte-Macrophage Colony-Stimulating Factor (GM-CSF) has demonstrated notable clinical activity in cancer immunotherapy, but is limited by systemic toxicities, poor bioavailability, rapid clearance, and instability in vivo. Nanoparticles (NPs) may overcome these limitations and provide a mechanism for passive targeting of tumors. This study aimed to develop GM-CSF-loaded PLGA/PLGA-PEG NPs and evaluate them in vitro as a potential candidate for in vivo administration. NPs were created by a phase-separation technique that did not require toxic/protein-denaturing solvents or harsh agitation techniques and encapsulated GM-CSF in a more stable precipitated form. NP sizes were within 200 nm for enhanced permeability and retention (EPR) effect with negative zeta potentials, spherical morphology, and high entrapment efficiencies. The optimal formulation was identified by sustained release of approximately 70% of loaded GM-CSF over 24 hours, alongside an average size of 143 ± 35 nm and entrapment efficiency of 84 ± 5%. These NPs were successfully freeze-dried in 5% (w/v) hydroxypropyl-β-cyclodextrin for long-term storage and further characterized. Bioactivity of released GM-CSF was determined by observing GM-CSF receptor activation on murine monocytes and remained fully in-tact. NPs were not cytotoxic to murine bone marrow-derived macrophages (BMDMs) at concentrations up to 1 mg/mL as determined by MTT and Trypan Blue exclusion assays. Lastly, NP components generated no significant transcription of inflammation-regulating genes from BMDMs compared to IFNγ+LPS “M1” controls. This report lays the preliminary groundwork to validate in vivo studies with GM-CSF-loaded PLGA/PEG-PLGA NPs for tumor immunomodulation. Overall, these data suggest that in vivo delivery will be well tolerated.

INTRODUCTION

Granulocyte-macrophage colony-stimulating factor (GM-CSF) is a hematopoietic cytokine that drives the mobilization, differentiation, maturation, and function of granulocyte and/or macrophage progenitor cells (1-3). Recombinant human GM-CSF (Leukine^®, Sargramostim) is FDA-approved for treatment of chemotherapy-induced neutropenia, mobilization of hematopoietic progenitor cells into the peripheral blood prior to leukapheresis, and myeloid reconstitution following bone marrow transplantation (4). GM-CSF also plays a critical role in the immunomodulation of macrophages and dendritic cells by increasing their activation, recruitment, function, and antigen presentation capabilities to bridge the gap between innate and adaptive immunity (5-7). This has since led to investigations using GM-CSF to augment anti-cancer immunotherapy (8-10). GM-CSF has demonstrated promising clinical benefits as an adjuvant to cancer cell vaccines in a variety of solid tumors (3, 8), and has synergized with anti-PD-1 or anti-CTLA-4 checkpoint blockade (11-14). Despite tumor-specific effects, GM-CSF is often administered systemically by subcutaneous or intravenous injection. This requires high doses delivered into the circulation for adequate amounts to reach the tumor site, which concomitantly increases the likelihood of off-target effects associated with GM-CSF therapy and limits therapeutic potential. Systemic delivery of GM-CSF is associated with pericardial effusion, sequestration of granulocytes in pulmonary circulation, transient supraventricular arrhythmia, renal and/or hepatic dysfunction, and other disruptive side-effects that may lessen or ultimately lead to a stoppage in treatment (4). GM-CSF delivery is further complicated by the cytokine’s poor bioavailability, rapid clearance, and instability during in vivo delivery (15).

Nanoparticles (NPs) are known for their ability to protect therapeutic cargo in circulation. Importantly, NPs may also selectively accumulate and release therapeutic payload at the tumor site due to leaky tumor vasculature and poor lymphatic drainage in a phenomenon known as the enhanced permeability and retention (EPR) effect (16). NPs present a particularly attractive tumor-targeting delivery modality for GM-CSF that could direct therapeutic benefit to the tumor site and minimize off-target effects. NPs 50-200 nm in size can reach maximum accumulation at the tumor site by the EPR effect in as little as 4-6 hours after administration (17, 18). Therefore, sub-200 nm NPs with sustained release properties over approximately 24 hours may be optimal for delivery of GM-CSF to tumors.

NPs composed of FDA-approved poly(lactic-co-glycolic acid)(PLGA) are known for their controlled release properties, as well as their exceptional biodegradability and biocompatibility. Nevertheless, the entrapment of hydrophilic, structure-sensitive proteins within hydrophobic PLGA matrices poses a significant challenge in nanoformulations. Standard formulation techniques for protein-loaded PLGA NPs include the double emulsion (water/oil/water) solvent evaporation method. However, this method often utilizes protein-damaging agitation methods such as sonication (19) as well as toxic organic PLGA solvents and amphiphilic surfactants such as dichloromethane (DCM) and polyvinyl alcohol (PVA) that can denature proteins and are associated with residual in vivo toxicities (20, 21). Further, proteins in solution are subject to denaturation during the emulsification process via adsorption at the oil/water interface, as the conformational flexibility of proteins in solution may permit the externalization of hydrophobic pockets to face the oil interface resulting in denaturation (22). Methods that avoid the use of toxic solvents and surfactants and provide the least amount of shear stress on the protein cargo are optimal for the nano-encapsulation of proteins. Inclusion of PLGA-PEG copolymer can stabilize the formulation in the absence of surfactant, improve encapsulation of hydrophilic proteins, and create stealth properties. Recently, M. Haji Mansor et al. reported an attractive phase separation technique for the formation of protein-loaded PLGA/PLGA-PEG NPs. This novel technique utilized non-toxic solvents glycofurol and isosorbide and did not require surfactant. Protein was also precipitated prior to encapsulation, which limited molecular motions and further ensured protein stability during the encapsulation process (23).

In this report, PLGA/PLGA-PEG NPs loaded with GM-CSF were prepared and characterized using the previously reported phase separation method (23). The purpose of this work was to generate a formulation suitable in size for EPR effect with sustained release of functional protein and exceptional biocompatibility to rationalize future in vivo studies of NP-mediated delivery of GM-CSF to tumors.

MATERIALS AND METHODS

Materials

Recombinant murine (rm)GM-CSF (Cat# 415-ML/CF), colony-stimulating factor-1 (CSF-1), and IFNγ were purchased from R&D Systems, as well as GM-CSF DuoSet ELISA (Cat# DY415), and ELISA Ancillary Reagent Kit 2 (DY008). PLGA-COOH (Poly(D,L-lactide-co-glycolide) lactide:glycolide 75:25, Mw 4,000-15,000 (Cat# 719919), monomethoxy-PEG (5 kDa, Cat# 303143), DL-lactide, DL-glycolide, tin(II) 2-ethylhexanoate, benzyl alcohol, 2-hydroxypropyl)-β-cyclodextrin (HPBCD), Tween^® 20, bovine serum albumin (BSA), and all other chemicals and solvents were purchased from Sigma Aldrich and used as received without any further purification/modification unless otherwise indicated. Low endotoxin certified RPMI 1640 was purchased from Corning (Ref# 10-041-CV), low endotoxin (certified less than 0.06 EU/mL) fetal bovine serum (FBS) was purchased from Atlanta Biologicals (Cat# S11150H), and polymyxin B sulfate (endotoxin inhibitor) was purchased from CalBiochem (Cat# 5291). Brilliant Violet 510 anti-Ly6C antibody (clone HK1.4, Cat#128033) and FITC anti-CD11b antibody (clone M1/70, Cat#101206) was purchased from BioLegend, and anti-STAT5 pY694 antibody (clone 47, Cat #612567) was purchased from BD Biosciences. CyQUANT™ 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) Cell Proliferation Assay Kit was purchased from ThermoFisher. Trypan blue dye was purchased from Invitrogen. Lipopolysaccharide (E. coli) was purchased from Enzo Life Sciences. The RNeasy^® Mini Kit was obtained from Qiagen (Ref# 74106) and the Superscript™ III First-Strand Synthesis SuperMix Kit from Invitrogen. SYBR green was purchased from Applied Biosystems. Validated primers for inflammatory genes were purchased from Integrated DNA Technologies (IDT).

Synthesis of PLGA-COOBn polymer and PLGA-PEG co-polymer

PLGA-COOBn and PLGA-PEG were synthesized by ring-opening polymerization (23). Briefly, for PLGA-COOBn, DL-lactide and DL-glycolide were mixed at a molar ratio of 3:1 with benzyl alcohol as the initiator. The mixture was then heated to 140 °C under argon atmosphere until melted, followed by the addition of 0.04% (w/w) tin(II) 2-ethylhexanoate. The solution was then stirred at 180 °C for 3 hours to allow for polymerization, then allowed to cool. The polymer was dissolved in DCM followed by precipitation with heptane and the supernatant discarded. Next, the polymer was washed two times with heptane and one time with diethyl ether. Finally, the precipitate was dried under vacuum (~10⁻³ bar) for 24 hours. For PLGA-PEG, the same procedure was utilized except for the use of monomethoxy PEG 5k as the initiator rather than benzyl alcohol, and the polymer was precipitated with diethyl ether pre-chilled to −20 °C and then washed with heptane.

Polymer characterization

Proton nuclear magnetic resonance (¹H-NMR, 400 MHz) spectra were recorded at room temperature (Jeol JNM-ECZ spectrometer) to characterize the polymer/co-polymer and to estimate the Mn (number averaged molecular weight). Chemical shifts are reported in parts per million (ppm), patterns are designated as singlet (s) and multiplet (m), and the spectra were calibrated with respect to the residual hydrogen signal from the solvent used to dissolve the polymer; DMSO-d₆ (2.50 ppm) for PLGA-COOBn and CDCl₃ (7.26 ppm) for PLGA-PEG.

Preparation of empty and GM-CSF loaded nanoparticles

First, to enhance protein stability during encapsulation, GM-CSF was precipitated using a method adapted from Giteau et. al (24). Briefly, lyophilized rmGM-CSF as provided by the manufacturer was resuspended in MilliQ water containing 0.544 M NaCl to a concentration of 4 mg/mL. 25 μL protein solution containing 100 μg GM-CSF was then added to 975 μL glycofurol and left on ice for 30 minutes to precipitate. To test functionality of precipitates (as described in later sections), the sample was subject to centrifugation at 8,500xg for 30 min at 4 ºC to obtain a pelleted precipitate that was then dissolved in PBS.

NPs loaded with GM-CSF were prepared using an adapted phase separation method (23). PLGA-COOBn, PLGA-PEG, and PLGA-COOH were dissolved at 12%, 9%, or 6% w/v in dimethyl isosorbide (DMI) and mixed in varying ratios (see Table I and II) to a total volume of 300 μL in a 5 mL vial containing a stir bar (10 mm x 6 mm). Then, 100 μL protein precipitate suspension in glycofurol containing 10 μg GM-CSF was added under magnetic stirring at 1300 RPM for 30 seconds. For synthesis of empty NPs, the 100 μl of protein precipitates were replaced with 100 μl of glycofurol alone. Phase separation was first initiated under continual stirring by the dropwise addition of 100 μL aqueous buffer (0.05M sodium citrate-HCl buffer solution, pH = 5.9). This pH was chosen to match the isoelectric point (pi) of GM-CSF (5.84) in order to maximize protein loading by minimizing aqueous protein solubility during the encapsulation process. After 1 minute, 500 μL additional aqueous buffer was added dropwise every 30 seconds for a total of 4 times (4 x 500 μL). The suspension was then left to stir for 1 minute before dilution with 30 mL of MilliQ water and then left to rest for 1 hour to allow residual solvents to diffuse from the newly formed NPs. All steps above were carried out at room temperature. The suspension was then centrifuged at 10,000xg for 30 minutes at 4°C and supernatant carefully decanted. The particle pellet was resuspended with 30 mL of MilliQ water and centrifugation repeated. The supernatant was again decanted, and the final pellet (36 mg for Formulation 1-4, 27 mg for Formulation 5 and 18 mg for Formulation 6) was resuspended in aqueous solution to achieve a final particle suspension of desired concentration. The theoretical drug loadings (DL) for GM-CSF calculated using Eq (1) was 0.03% for Formulations 1-4, 0.04% for Formulation 5, and 0.05% for Formulation 6. These values are similar to other published formulations (23).

Table I.

Physiochemical characteristics of GM-CSF-loaded PLGA/PLGA-PEG NP formulations in which each polymer was implemented at a concentration of 12% w/v in DMI.

Formulation Number	Proportion (% v/v)			Size (nm) a	Zeta potential (mV) b	Entrapment efficiency (%) c
	PLGA-COOH	PLGA-COOBn	PLGA-PEG
1	33	33	33	198 ± 56	−6.9 ± 0.7	91 ± 6
2	17	50	33	169 ± 42	−5.4 ± 0.4	95 ± 4
3	0	67	33	179 ± 44	−6.0 ± 0.7	96 ± 3
4	0	50	50	152 ± 40	−7.0 ± 0.6	88 ± 6

Size (nm, nanometers), zeta potential (mV, millivolts), and entrapment efficiency values are listed as average ± SD of n=3.

^a=NPs were diluted in ultrapure MilliQ water to a suitable concentration.

^b=NPs were suspended at a concentration of 100 μg/mL in 0.01M NaCl at pH = 7.

^c=For 10 μg total GM-CSF per batch.

Table II.

Average size, average zeta potential, and entrapment efficiency of GM-CSF-loaded NPs composed of 50% PLGA-PEG and 50% PLGA-COOBn with differing concentrations of polymers in dimethyl isosorbide.

Formulation Number	Proportion (% v/v)			Polymer concentration (% w/v)	Size (nm) a	Zeta potential (mV) b	Entrapment efficiency (%) c
	PLGA-COOH	PLGA-COOBn	PLGA-PEG
5	0	50	50	9	143 ± 35	−7.8 ± 1.7	84 ± 5
6	0	50	50	6	176 ± 42	−8.0 ± 1.2	92 ± 2

Size (nm, nanometers), zeta potential (mV, millivolts), and entrapment efficiency values are listed as average ± SD of n=3.

^a=NPs were diluted in ultrapure MilliQ® water to a suitable concentration.

^b=NPs were suspended at a concentration of 100 μg/mL in 0.01M NaCl at pH = 7.

^c=For 10 μg total GM-CSF per batch.

$$\text{DL}(\%)=\frac{\text{Initial mass of protein used as a starting material}}{TotalmassofPLGAandPLGA-PEG}x100$$ (1)

Size distribution and zeta potential (ZP)

NP suspensions diluted in MilliQ water were analyzed immediately after preparation or immediately upon reconstitution of freeze-dried NPs. Average size and size distribution of NPs were determined by Nanoparticle Tracking Analysis (NTA) (NanoSight NS3000, Malvern Panalytical) using a Blue (488 nm) laser beam. NPs were diluted in MilliQ water to optimal concentrations, and sizes were analyzed in triplicate for each run. ZP measurements were conducted by Electrophoretic Light Scattering (ZetaSizer, Malvern Panalytical) in which NPs were placed in a cuvette at 100 μg/mL concentration in 0.01M NaCl with pH titrated to 7 using 0.01M NaOH or 0.01M HCl immediately prior to measurement. Samples were measured at room temperature and analyzed in triplicate with each run representing 5-10 individual measurements.

Freeze-drying of nanoparticles

The final NP pellet was resuspended in 20 mL of MilliQ water with 5% (w/v) 2-hydroxypropyl-β-cyclodextrin (HPBCD) as a cryoprotectant and freeze-dried (Christ alpha 4-2 LD plus freeze-dryer). NPs were characterized by size and ZP both before and after freeze-drying to determine stability throughout the freeze-drying process, which was determined by comparing size and ZP readings before and after freeze-drying. The ratio of final to initial size (S_f/S_i) and final and initial standard deviations (SD_f/SD_i) was used to determine NP stability, along with changes in ZP.

Scanning Electron Microscopy

1 μL drop of NPs diluted in ultrapure MilliQ water was placed onto a silicon wafer and left to dry in a sterile hood at room temperature overnight. The samples were placed in a Denton Benchtop Desk V Sputter for application of a thin coat of gold-palladium and imaged with a Hitachi S-4700 Scanning Electron Microscope running at 5 KV.

Quantification of GM-CSF

Recombinant murine GM-CSF concentration was quantified using an enzyme-linked immunosorbent assay (ELISA) according to the manufacturer’s instructions (R&D Systems). This assay was implemented due to the high specificity for GM-CSF to allow for accurate detection and quantification. The concentration of kit standards used to generate a calibration curve for GM-CSF range from 7.8-500 pg/mL GM-CSF, which defines the limits of detection of the ELISA. Because we loaded 10 μg GM-CSF total into the NPs, our samples for entrapment and release studies required significant dilution to fall in the range of detection. Samples were diluted appropriately in Reagent Diluent (R&D Systems, DY995). The plates were measured for absorbance at both 450 nm and 540 nm with a BioTek plate reader. Samples were assessed in duplicate, and final optical density (OD) measures were taken by subtracting 540 nm values (the plate) from 450 nm OD values.

Protein entrapment efficiency

The entrapment efficiency (EE) of GM-CSF was determined by the quantity of unentrapped protein measured by ELISA assay in relation to the initial mass of protein used as starting material. Briefly, 1 mL supernatant was collected following each centrifugation of newly formed loaded NPs (as described above) and empty NPs as a control. Samples were then diluted in reagent diluent for quantification by recombinant murine GM-CSF ELISA as described in the previous section. The calculated protein masses from both supernatants were added together to represent the total unentrapped protein fraction and the percentage of entrapment efficiency was calculated using Eq. (2).

$$\text{EE}(\%)=\frac{\text{Initial mass of protein}-\text{mass of unentrapped protein}}{Initialmassofprotein}x100$$ (2)

In vitro protein release studies

The final NP pellets (36 mg for Formulation 1-4, 27 mg for Formulation 5, and 18 mg for Formulation 6, containing 8.4-9.6 μg of GMCSF, see %EE Table I and II) were resuspended in 6 mL of 0.05M Tris-HCl buffer (pH=7.4) + 0.9% (w/v) NaCl and 0.1% (w/v) BSA as a protein stabilizer, alongside their respective empty NPs as a control. The release studies were performed with 2 mL of solution in 2 mL Eppendorf tubes incubated at 37 ºC in a shaker (300 rpm). At pre-determined time intervals, 50 μL of sample was removed, diluted with 100 μL water, and subjected to centrifugation at 8,500xg for 30 min at 4 ºC. The supernatant, which contained the released protein at the respective time point, was then carefully collected and further diluted for protein quantification by ELISA (see above section). The release studies were performed in triplicate.

Assessment of GM-CSF bioactivity

The bioactivity of precipitated and released GM-CSF was determined by the ability of the protein to activate membrane-bound GM-CSF receptors on murine bone marrow monocytes. Phosphorylated signal transducer and activator of transcription-5 (pSTAT5) is indicative of GM-CSF receptor complex activation (GM-CSFR; common beta and alpha subunits) as a result of binding biologically active GM-CSF (25). We performed intracellular staining for pSTAT5 on monocytes following treatment with GM-CSF to test its bioactivity. Both GM-CSF precipitates and released protein collected from NPs at pre-determined time points were tested in comparison to non-precipitated/non-packaged recombinant GM-CSF control directly as supplied by the manufacturer. To isolate bone marrow monocytes, the hindlimbs from 8-to-20-week-old C57Bl/6J mice (either male or female) were flushed with low endotoxin-certified RPMI 1640 using a 26 1/2 gage needle and cultured in RPMI 1640 supplemented with 10% low endotoxin FBS and 10 μg/mL polymyxin B sulfate at 37 °C with 20% O₂ and 5% CO₂. Cells were seeded at 200,000 cells per 100 μL in 96-well V-bottom plates followed by immunostaining with Brilliant Violet 510 anti-Ly6C antibody and FITC anti-CD11b for 15 minutes at 37 °C. Next, the cells were treated with 100 ng/mL GM-CSF that was either precipitated in glycofurol and re-dissolved in PBS, or collected from in vitro release, followed by incubation for an additional 15 minutes at 37 °C. 100 ng/mL recombinant protein as supplied from the manufacturer at was also prepared as a positive control for functional GM-CSF for each experiment. The cells were then fixed in complete medium (RPMI + 10% FBS) with a final concentration of 2% paraformaldehyde at 37 °C for 10 minutes, followed by permeabilization in ice cold 90% methanol + 10% PBS (pH=7.2-7.4) at −20 °C for at least 30 minutes. Finally, intracellular staining was performed with an anti-pSTAT5 antibody for 30 minutes at room temperature in the dark before being subjected to flow cytometric analyses on a BD LSRFortessa. Data were normalized to percent of pSTAT5 detected from the rmGM-CF control within each individual experiment for adequate comparison of groups between experiments. Flow cytometry plots were analyzed using BD FCS Express software.

In vitro cytotoxicity assays

For a thorough assessment of NP cytotoxicity, murine bone marrow-derived macrophages (BMDM) were utilized in a Trypan blue exclusion assay to assess changes in cell membrane integrity, as well as an MTT assay to quantify changes in cell viability in response to NPs.

For the Trypan blue assay, bone marrow cells were isolated as described above from male or female WT C57BL/6J mice and cultured in 48-well plates in macrophage differentiation medium (low endotoxin RPMI 1640 supplemented with 10% low endotoxin FBS, 10 μg/mL polymyxin B sulfate, and 20 ng/mL CSF-1) for 5 days. Media was supplemented with 5 ng/mL additional CSF-1 every other day, followed by a media change on the 5^th day to remove undifferentiated cells and achieve the final adherent BMDM culture. BMDM were incubated with 0.01, 0.1, or 1 mg/mL empty NPs diluted in macrophage differentiation medium for a dose-escalation cytotoxicity study, or with 5% DMSO as a control for membrane disruption of cells and Trypan blue positivity. After 24 hours, the cells were imaged with an inverted light microscope (Leica) and total cells versus Trypan blue positive cells were counted in each well. Samples were analyzed in triplicate, with 3-4 fields of view counted per well and averaged to obtain final values of percent Trypan blue negative cells .

For the MTT assay, isolated bone marrow cells were seeded in 96-well flat-bottom culture plates at 500,000 cells per well in 200 μL differentiation medium (described above), which we found to be an optimal initial seeding density for the MTT signal from the final differentiated BMDM culture achieved 5 days later. Resultant BMDMs were then treated with the same conditions as described above for the Trypan blue assay. After 24 hours of incubation at 37 ºC and 5% CO₂, the old medium was removed and replaced with 100 μL fresh differentiation medium. 10 μL of 12 mM MTT stock solution was then added to the cells, alongside a negative control in which 10 μL MTT stock solution was added to 100 μL of culture medium alone, and cells were incubated at 37 ºC for 4 hours. Next, 100 μL sodium dodecyl sulfate (SDS)-HCl solution was added to each well by thorough mixing followed by another 4-hour incubation at 37 ºC. Lastly, samples were mixed well with a multichannel pipette before reading absorbance at both 570 nm and 630 nm and subtracting the 630 OD values to obtain sample OD values. OD values of the medium plus MTT reagents alone were used as a blank and subtracted from sample values to obtain final OD values for each sample. Percent cell viability of each sample was calculated by taking the OD value for NP-treated cells divided by the OD value for control cells and multiplying times 100. Samples were assessed in triplicate and ODs were averaged to calculate a final percentage value for each.

In vitro macrophage immunogenicity study

Murine BMDM were generated as described above in 6-well culture plates, incubated with 1 mg/mL NPs for 24 hours, and analyzed for transcription of Il-12, Tnfa, Il-6, Nos2, and Arg1 mRNA by Quantitative Polymerase Chain Reaction (qPCR). Macrophages were starved of CSF-1 for at least 12 hours and polymyxin B was removed from the media, followed by either no treatment or incubation with 1 mg/mL NPs for 24 hours. Untreated M0 (quiescent) macrophages and M1 (stimulated/inflammatory) macrophages were used as controls alongside the NP-treated cells. The M1 controls were generated by treatment with 20 ng/mL rmIFNγ at the same time of NP treatment, followed by addition of 500 ng/mL LPS to the M1-designated wells 6 hours later and incubated overnight. BMDM were then processed for total RNA isolation with RNeasy^® Mini Kit and complementary DNA (cDNA) was then synthesized from 250 μg RNA with Superscript™ III First-Strand Synthesis SuperMix Kit. qPCR was then performed with SYBR Green Master Mix in relation to Rpl4 housekeeping RNAs using the comparative threshold method (26).

Statistical analyses

One-way ANOVA and a threshold p-value of 0.05 was employed to determine significant differences between groups. For Figure 6, a mixed-effects model was used to compare groups between M1 and NP adjusting for M0 (by ratio) in the presence of repeated measures from different genes within each mouse. This approach can handle fixed and random effect model parameters, as well as nested or unbalanced designs with repeated measures from the same mice. Models included the fixed effects of groups (M1 versus NP), genes, and gene-group interaction, with random effects of genes within mice to account for the correlation among repeated measurements on the same mice and among genes within the same mice, respectively. Specific comparisons between M1 and NP groups were evaluated using statistical contrasts of M1 and NP groups over each gene based on the fitted mixed-effect model.

Figure 6. Effects of Formulation 5 (empty NPs) on macrophage inflammatory gene transcription.

Each data point equals the average ± SD of n=3 or more.

Study approval and animals

Adult wild type C57BL/6J mice (8 weeks old) were purchased from The Jackson Laboratory, 000664. All mice utilized in this study were 8-to-20-week-old males and females. Animal treatment and euthanasia were carried out in strict accordance with protocols approved by the Institutional Animal Care and Use Committee at West Virginia University and in compliance with the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) International.

RESULTS

Characterization of PLGA-COOBn and PLGA-PEG polymers

¹H-NMR spectrum of synthesized PLGA-COOBn (Figure S1A) revealed the presence of a benzyl end group at 7.30-7.40 ppm, glycolide units at 4.80-5.00 ppm, and lactide units at 1.46 ppm and 5.10-5.30 ppm. Using the benzyl signals as reference for integration (5H), a Mn of 3.8 kDa was calculated, and the molar ratio of lactide:glycolide was determined to be 75:25. ¹H-NMR spectrum of synthesized PLGA-PEG (Figure S1B) revealed the presence of ethylene glycol units at 3.62 ppm, glycolide units at 4.60-5.00 ppm, and lactide units at 1.50-1.60 ppm and 5.10-5.30 ppm. Using the PEG signals as reference for integration (456H), polymer Mn of 30.2 kDa was calculated and the molar ratio of lactide:glycolide was determined to be 75:25.

Formation and characterization of NPs created from PLGA polymers at 12% (w/v) concentration

GM-CSF-loaded NPs were prepared using a method modified from M. Haji Mansor and colleagues (23). The physiochemical characteristics of NPs created from PLGA polymers at an initial concentration of 12% (w/v) (Formulations 1-4) are outlined in Table I. All formulations yielded notably high entrapment efficiency of GM-CSF, which ranged from 88 to 96 percent of the 10 μg GM-CSF initially loaded into each formulation. An increase in entrapment efficiency was observed as the proportion of PLGA-COOBn increased from 33% to 67% w/v in Formulations 1 to 3, while a small decrease in entrapment was observed for higher PLGA-PEG content (50% w/v) in Formulation 4. All NPs generated sizes below 200 nm and negative ZPs. Formulation 4, composed of 50% w/v PLGA-PEG, had the smallest average size of 152 ± 40 nm and homogeneous size distribution (Figure 1A). The SEM image obtained for Formulation 4 (Figure 1B) shows spherical NPs slightly below 200 nanometers in size, consistent with the value determined using Nanoparticle Tracking Analysis. All other formulations showed similar sizes in the range of 150-200 nm with uniform size distributions (Table I and Figure S2), suggesting these formulations are candidates for passive targeting of select tissues with leaky blood vessels, such as tumors, via the enhanced permeability and retention (EPR) effect.

Figure 1. Size distribution and morphology of Formulation 4.

A) Size distribution was determined by Nanoparticle Tracking Analysis and is representative of n=3 batches. B) SEM image.

Next, all formulations in Table I were monitored for release of GM-CSF in vitro to determine if the desired criteria of sustained release over a 24-hour period could be met. A quantity of GM-CSF was released from all NP formulations in the form of an initial “burst release” within one hour, which ranged from ~20 to 50 percent of entrapped protein (Figure 2 and Table S1). A small increase in protein release was observed in Formulations 3 and 4 between 1 to 2 hours; however, no further protein was released from any formulation over the remaining 24 hours incubation (<10%). All 4 formulations lacked sustained release properties over 24 hours, therefore prolonged release time points were not collected. Surprisingly, these data show that changing polymer composition had minimal effects on GM-CSF release patterns in our study, although a few trends were noted. Specifically, Formulations 1 and 2, which both contained polar carboxylic acid end-capped PLGA, released the least amount of protein as both barely achieved 30%. More GM-CSF was released from Formulations 3 and 4, which both lacked PLGA-COOH. Formulation 4, which contained a higher proportion of PLGA-PEG (50% w/v) than the other 3 NP Formulations, generated the highest percent release of GM-CSF on average. Of note, released GM-CSF from Formulations 1 and 4 were tested for functionality and were found to have bioactivity preserved (Figure S3). These data suggest that although Formulation 4 showed the smallest size and uniformness, which could result in higher extravasation and superior accumulation in tissue with leaky vasculature, further optimization of formulation parameters was necessary to achieve sustained in vitro release.

Figure 2. In vitro protein release study of Formulations 1-4.

Each data point represents average ± SD of n=3 independent release experiments. 100 percent release corresponds to 1.5 μg/mL of GM-CSF. Data points were collected at 1, 2, 18, and 24 hours. Data points are slightly offset for figure clarity.

Formation and characterization of NPs created from PLGA polymers at 6% and 9% (w/v) polymer concentrations

Several studies report that decreasing the concentration of polymer incorporated into the formulation improves in vitro release of the encapsulated drug (27-29). Therefore, we thought to improve the release profile of Formulation 4 by decreasing polymer concentration during the formulation process. NPs were created with the composition of Formulation 4, but with a polymer concentration decreased from 12% (w/v) to 9% (w/v) (Formulation 5), and further decreased to 6% (w/v) (Formulation 6) (Table II). Using lower polymer concentrations did not substantially alter the size or ZP as particles still had average size under 200 nm and similar negative ZP. In fact, Formulation 5 had a highly uniform size distribution (Figure 3A) with an average size of 143 ± 35 nm, which is slightly smaller than Formulation 4, making Formulation 5 even more attractive for delivery by passive targeting. Interestingly, Formulation 6, created with even lower polymer concentration (6% w/v), had an average size of 176 ± 42 nm, slightly larger than Formulation 4 (Figure 3B). Entrapment efficiencies of NPs also remained high for both Formulations 5 and 6.

Figure 3. Characterization of Formulations 5 and 6.

Size distribution of A) Formulation 5 (9% w/v polymers) and B) Formulation 6 (6% w/v polymers) was measured by Nanoparticle Tracking Analysis and is representative of n=3 batches per formulation. C) In vitro protein release of Formulations 5 and 6. Each data point represents average ± SD of n=3 independent release experiments. 100 percent release corresponds to 1.5 μg/mL of GM-CSF.

Formulations 5 and 6 were then monitored for in vitro release of GM-CSF to determine the impact of lesser polymers on protein release (Figure 3C and Table S1). The average total percent of released GM-CSF was improved from approximately 40% in Formulation 4 to 70% in Formulation 5. For Formulation 5, an initial “burst” was still observed in the first hour in which 40% of entrapped GM-CSF was released. However, the release further improved over time for a maximum peak release of approximately 70% after 18 hours, and no further increase through 72 hours. Released protein from Formulation 5 was also tested for bioactivity on bone marrow monocytes and found to remain fully functional throughout the 72-hour study (Figure 4). Importantly, GM-CSF also remained biologically active following precipitation in glycofurol, centrifugation, and resolubilization in PBS (pH=7.2-7.4), which was used as an additional control for bioactivity to demonstrate that the precipitation step prior to entrapment and release did not compromise protein function. Further decreasing polymer to 6% (w/v) did not improve the release profile observed in Formulation 5, as Formulation 6 showed a higher initial burst of 55% GM-CSF released in the first hour, which increased only slightly over 24 hours for a total release of 60%. Based on these data, it was determined that NPs made of 50% PLGA-PEG and 50% PLGA-COOBn at 9% w/v concentrations (Formulation 5) represented our optimal formulation to move forward based on the observation of sustained release of functional protein over 24 hours in conjunction with small, uniform size and a negative ZP value. These data suggest that the concentration of polymer incorporated into PLGA/PLGA-PEG NPs is a key formulation parameter that impacts both the quantity and pattern of GM-CSF release from polymeric PLGA matrices while having minimal effects on other essential NP characteristics for passive targeting (size and ZP).

Figure 4. Bioactivity of precipitated and released GM-CSF from Formulation 5.

A) Representative flow cytometry plot used to assess bioactivity of GM-CSF by quantification of intracellular staining of phosphorylated STAT5 in murine bone marrow monocytes (Cd11b⁺Ly6C^hi) treated with GM-CSF. B) Quantification of percent phosphorylated STAT5 observed throughout processing and release (Formulation 5) normalized to recombinant GM-CSF control from each individual experiment. Each data point equals the average of n=3 or more ± S.D. ** = p ≤ 0.01.

Freeze-drying of nanoparticles

Because PLGA-based NPs stored in solution may undergo aggregation and fusion, and/or degrade over time by hydrolysis of esters to cause NP breakdown and cargo leakage, Formulation 5 was freeze-dried for long term storage. 5% HPBCD was chosen as the cryoprotectant similar to (23), as this study and several others report that HPBCD has superior cryoprotectant capabilities compared to other traditionally used saccharides such as mannitol (23, 30, 31). Freeze-drying allows for long-term storage of dehydrated NPs but can also cause NP degradation and/or modification of physicochemical properties (32). Therefore, stability of NPs throughout the freeze-drying process was determined by measuring their size and ZP both before and after freeze-drying (Table III). Freeze-dried NPs formed a fine powder that readily dissolved with gentle agitation consistent to other reports. The ratio of final size to initial size (S_f/S_i) value was calculated to be 1.06, indicating less than 10% deviation in the average size and, therefore, stability maintained during freeze-drying. Further, the differences in standard deviation for the size (SD_f/SD_i) was determined to be 1.02, which indicated a distribution of size maintained after freeze-drying. NPs also demonstrated minimal change in ZP to further support that NPs remained stable throughout freeze-drying. Lastly, both the in vitro release pattern and bioactivity of GM-CSF from the reconstituted freeze-dried NPs were similar to what was observed prior to freeze-drying (Figure S4).

Table III.

Characterization of Formulation 5 before and after freeze-drying (F.D.) in 5% (w/v) HPBCD.

Size (nm) a				Zeta potential (mV) b
Before F.D.	After F.D.	Sf/Si	SDf/SDi	Before F.D.	After F.D.
140 ± 36	149 ± 37	1.06	1.02	−9.4 ± 0.5	−9.6 ± 0.3

Size (nm, nanometers) and zeta potential (mV, millivolts) values are from one batch for a true representation of changes from final to initial measurement and are representative of similar data repeated with n=3 total batches.

^a=NPs were diluted in ultrapure MilliQ water to a suitable concentration.

^b=NPs were suspended at a concentration of 100 μg/mL in 0.01M NaCl solution to pH=7.

Characteristics of empty vs. GM-CSF-loaded nanoparticles

Next, in order to investigate the biocompatibility of NPs on primary murine macrophages, empty particles without GM-CSF were synthesized. This was because the effects of NPs alone are of interest and GM-CSF cargo will confound experimental results as GM-CSF itself induces a pro-inflammatory profile from macrophages (33, 34). Empty NPs of Formulation 5 were synthesized and characterized (Table IV). As expected, empty NPs did not have substantial differences in size or ZP when compared to the loaded formulation. This can easily be explained by the small protein payload these NPs carry. Therefore, the use of empty NPs for in vitro cytotoxicity and immunogenicity assays was justified.

Table IV.

Characterization of Formulation 5, both empty and GM-CSF-loaded.

Formulation	Size (nm) a	Zeta potential (mV) b
Formulation 5 - Empty	136 ± 29	−8.9 ± 0.5
Formulation 5 - Loaded	143 ± 35	−7.8 ± 1.7

Average size (nm, nanometers) and zeta potential (mV, millivolts) values listed as average ± SD of n=3.

^a=NPs were diluted in ultrapure MilliQ water to a suitable concentration.

^b=NPs were suspended at a concentration of 100 μg/mL in 0.01M NaCl solution to pH=7.

In vitro cytotoxicity and immunogenicity studies

In order to predict biocompatibility in vivo, empty NPs from Formulation 5 were incubated with murine BMDM and assessed for viability and inflammatory gene transcription changes. BMDM were the cell of choice for two main reasons: 1) primary cells provide more translational results than traditional experiments using cell lines, and 2) NPs frequently encounter mononuclear phagocytic cells such as monocytes and macrophages in vivo and these cells are primarily responsible for their clearance, therefore it is imperative to understand the potential cytotoxic and/or immunogenic effects NPs may exert on macrophages encountered in target tissue or during clearance to predict their biocompatibility. BMDM treated with empty NPs with concentrations as high as 1 mg/mL generated a slight but significant decrease in cell viability vs 0.1 and 0.01 mg/mL NPs, however viability remained above 80% in comparison to untreated cells via MTT assay (Figure 5A). As per ISO 10933-5, percentages of cell viability above 80% are considered non-cytotoxic (35). These results were further corroborated by analysis of cell membrane integrity by the Trypan blue exclusion assay in which the percentage of Trypan blue negative cells were 99% after culture with NPs (Figure 5B) in similar conditions. Both experiments utilized 5% DMSO as a control of cell death, which reduced cell viability to 0.9% and yielded no detectable Trypan blue negative cells.

Figure 5. Effects of Formulation 5 (empty NPs) on cell viability.

NPs were tested for cytotoxic effects by A) MTT assay, in comparison to 5% DMSO controls, and in relation to untreated (UTX) cells, and B) Trypan blue exclusion assay in comparison to UTX and 5% DMSO controls. Each data point equals the average of n=3 ± S.D. ** = p ≤ 0.01, * = p ≤ 0.05.

In addition, it was determined that BMDM cultured with 1 mg/mL NPs for 24 hours elicited significantly less transcription of the inflammation-regulating genes Nos2, Il6, Il12, Tnfa, and Arg1 than macrophages stimulated by rmIFNγ and LPS for polarization to M1 inflammatory macrophages (Figure 6). Therefore, it was concluded that empty NPs of this formulation did not initiate a significant inflammatory response from primary macrophages, in vitro. Overall, these data suggest that NPs will generate negligible immunogenic or cytotoxic effects on monocytes/macrophages during in vivo encounters/clearance and is indicative of in vivo biocompatibility in future applications.

DISCUSSION

The immunostimulatory properties of GM-CSF present an attractive therapeutic for immunomodulation of cancer and other diseases. NPs serve as a biodegradable and biocompatible delivery system with controlled release and/or targeted delivery properties that can limit toxicities of systemic free GM-CSF. GM-CSF (both human and murine) has been incorporated previously into various micro- and nano-delivery systems. Petit et al. encapsulated both murine and human GM-CSF within PLGA/PLA microspheres optimized to achieve long-term circulation and cumulative extended release (36). Hill et al. employed liposomes loaded with GM-CSF and IL-12 for direct intratumor delivery (37). Other studies include the immobilization of GM-CSF on silica particles (38), encapsulation GM-CSF in microsphere-sized liposomes (39, 40), co-encapsulation of GM-CSF with other cytokines and vaccine adjuvants in a liposomal influenza vaccine (41), and encapsulation of GM-CSF inside injectable hydrogels (42). However, none of these formulations were designed for passive targeting of tumors, which could benefit patients with both accessible and non-accessible tumors and minimize toxicity. Therefore, this study aimed to develop GM-CSF-loaded nanometer-sized particles using FDA-approved PLGA and evaluate these NPs in vitro as a potential candidate for in vivo delivery to tumors.

Standard formulation techniques for protein encapsulation into NP matrices include the use of toxic halogenated solvents (43-45) listed as Class 2 by the International Conference on Harmonization (ICH), labeling them as significant threats to patient safety (46). Less toxic solvents like acetone, ethyl acetate, and DMSO (47-49) may be used but are still recognized by the ICH as hazardous to humans. Swed et al. reported a novel non-toxic and non-denaturing phase separation technique that incorporated the biocompatible and non-volatile solvents glycofurol and DMI to create nanometer-sized PLGA particles loaded with various proteins. (50). These solvents have been shown to be highly biocompatible (51, 52). M. Haji Mansor et al. utilized a similar method to create protein-loaded NPs consisting of varying ratios of PEG-PLGA copolymer, PLGA-COOR, and/or PLGA-COOBn polymer (23). Mixture with hydrophobic PLGA permits manipulation of PLGA end groups to influence NP physiochemical characteristics and more protein-dependent parameters such as entrapment and release profiles. Our report looked to further optimize this formulation for packaging and sustained release of GM-CSF.

Since encapsulation of a protein in solution can lead to its denaturation, GM-CSF was first precipitated in glycofurol which allows for effective dehydration of protein molecules for precipitation. Glycofurol has been used to form small homogenous precipitates of proteins suitable for NP encapsulation, which is easily reversible, and does not alter protein bioactivity (23, 24, 53-55). Here, the bioactivity of GM-CSF precipitates was fully preserved after redissolution (Figure 4).

GM-CSF-loaded NP Formulations 1-4 were prepared to explore the effect of polymer composition on NP physiochemical characteristics. NTA was utilized to track size distribution and has been shown to be advantageous over Dynamic Light Scattering (DLS) techniques as NTA detects multiple populations not always shown by DLS (56, 57). All formulations generated were sufficient in size to justify use for in vivo passive targeting by the EPR effect (50-200 nm). NPs comprised of 50% PEG-PLGA and 50% PLGA-COOBn (Formulation 4) resulted the smallest size and more narrow size deviation. Other studies have shown decreases in size and size distribution for PEGylated PLGA vs PLGA alone (58, 59). We did not explore combinations without PLGA-PEG because PEG plays a stabilizing role in the absence of surfactants, and because NPs did not form without PLGA-PEG in previous studies using this formulation method (23). Furthermore, the amphiphilic PEG-PLGA can improve stability at the oil/water interface during NP formation and enhance encapsulation efficiency of hydrophilic proteins into hydrophobic PLGA matrices while preventing protein adsorption/denaturation during entrapment (58, 60). All formulations had notably high entrapment efficiencies, and the ratio of PLGA-COOBn/PLGA-COOH had negligible influence on entrapment efficiency. It is expected that the ionized groups on solubilized proteins might contribute to electrostatic interactions with charged polymer groups such as COOH (deprotonated at pH=5.8), which may impede or promote their entrapment. Instead, minimizing aqueous solubility of GM-CSF by precipitation and encapsulation at pH near its pi could increase propensity of the protein to undergo hydrophobic interactions with PLGA matrices and in-turn maximize encapsulation efficiency (61). Other studies detail the importance of aqueous phase pH on protein encapsulation with maximal encapsulation achieved at pH near pI (23, 62, 63).

Formulations containing PLGA-COOH (Formulations 1 and 2) released less GM-CSF in vitro than formulations which lacked this polymer (Formulations 3 and 4). Poor protein release from PLGA with unencapped carboxylic acid terminal groups has been reported previously (53, 64, 65). Overall, release patterns in Formulations 1-4 consisted of initial “burst” at 1-2 hours followed by no further increases in released protein up to 24 hours. This “burst” is a common challenge with protein-loaded PEGylated PLGA NPs and usually occurs when more protein associates near the hydrophilic external PEG layer rather than the hydrophobic PLGA core (66). Furthermore, the incomplete protein release observed in this study is a common limitation of PLGA NPs (67), which may be due to adsorption of protein at the surface of PLGA (68-70). Of Formulations 1-4, the highest cumulative GM-CSF release was observed from 50% PLGA-PEG and 50% PLGA-COOBn NPs (Formulation 4). In addition to its favorable properties in promoting stability and heightened protein encapsulation, PEG is also associated with improved protein release from polymeric NPs, as it may create a swollen matrix upon rapid hydration (71).

Studies show that decreasing the concentration of polymer incorporated into the formulation can improve sustained release (27-29). Decreasing polymer concentration from 12% (w/v) to 9% (w/v) in Formulation 5 resulted in sustained release of ~70% of total entrapped GM-CSF over 24 hours. Because maximum accumulation of 50-200 nm NPs within tumors by passive targeting can occur in as little as 4-6 hours after administration (17, 18), the release profile of Formulation 5 was considered optimal. These NPs also yielded a size even smaller than Formulation 4, which could enhance their tumor-targeting capacity. Unexpectedly, further decreasing polymer concentration to 6% (w/v) in Formulation 6 resulted in NPs size similar to Formulation 4. Although decreasing polymer concentration is also associated with a small decrease in entrapment efficiency, the amount of protein entrapped remained high for both Formulations 5 and 6. However, NPs never reached 100% release over 72 hours, even for Formulation 5, reinforcing incomplete protein released as a limitation to PLGA NPs. GM-CSF was preserved in its native form and function at all stages of release in Formulation 5, which is important to 1) ensure therapeutic effectiveness, and 2) prevent release of immunogenic denatured protein aggregates (72). Ultimately, Formulation 5 was selected as the optimal formulation based on its small size, in vitro sustained release of functional GM-CSF, and high PEG composition.

PEG is used to create “stealth” NPs by forming a hydrating layer that sterically interferes with surface protein adsorption and opsonization of NPs, enhancing circulation time, in vivo. (66, 73, 74). PEGylation is beneficial for stability and clearance, but especially because substantial cellular uptake of GM-CSF-loaded NPs is undesirable due to the cell-exterior location of GM-CSF receptor complexes. Released GM-CSF must bind to the cell surface receptors to initiate internal signaling and drive therapeutic effects. Our optimal formulation consisted of 50% PEG-PLGA with 5 kDa PEG intended to decrease cell internalization. Liposomes coated with 5 kDa PEG had significantly prolonged circulation time and reduced phagocytic uptake by the reticuloendothelial system than those coated with 750 Da (75). Another study showed that the amount of protein adsorbed onto the surface of PLA-PEG NPs significantly decreased as PEG molecular weight increased to 5 kDa, while 10, 15 or 20 kDa PEG yielded no further decrease in adsorption (76). Further investigations to understand in vivo cell uptake and biodistribution of Formulation 5 are underway in our laboratory.

All formulations possessed slightly negative (under −10 mV) ZPs. Slightly anionic NPs experience even longer half-lives than more anionic NPs (77), which allow for maximum circulation and tumor accumulation. PEGylation of PLGA generally results in a less negative ZP while increasing COOH results in a more negative value. Surprisingly, NP ZP seem unaffected by the polymer composition. This could be explained by a coating of the NP by PEG chains with carboxyl groups buried inside the NP. The polymer concentration did not have a significant effect on the ZPs, indicating similar polymer arrangements (58). As expected, the loading of the NP with GM-CSF did not result in changes in size and/or ZP due to small overall amount of protein in relation to polymer mass.

Monocytes/macrophages are sentinels of the immune system. In addition to being our target cell of interest, they will encounter NPs in both circulation and tissues. Therefore, it is essential to understand how empty NP components influence monocyte/macrophage viability and inflammatory response. Formulation 5 (empty NPs) had negligible effects on viability (MTT) and membrane integrity (Trypan blue) of murine BMDM at high concentrations of up to 1 mg/mL. These data speak for the overall biocompatibility of PLGA-COOBn and PLGA-PEG. High biocompatibility could also be attributed to lack of harmful solvents and surfactants in our formulation, as commonly used residual stabilizers like PVA and poloxamer 188 can directly contribute to NP toxicity at high concentrations (78-80). While basic studies focus primarily on cytotoxicity, assessing potential effects of empty NPs on the inflammatory response of primary immune cells is often overlooked. This is a key step to predict in vivo tolerance as some nanomaterials have been implicated in inflammatory and allergic responses (81). It is especially crucial to avoid these responses from NPs when they carry immune-modulating therapies. Standard assays to predict immunogenic responses include nitric oxide production in RAW264.7 cells (80) or NF-κB activation in human THP-1 cells in response to NPs (82). Instead, we cultured primary BMDM with empty NPs to observe potential changes in transcription of inflammation-regulating genes that are indicative of classically-activated M1 macrophage polarization. NP-treated cells were compared to 1) M0 unstimulated “quiescent” BMDM, and 2) classically-activated M1 BMDM by IFNγ priming and subsequent LPS activation. BMDM initiated significantly less transcription of Nos2 (iNOS), Il-6, Il-12, Tnfα, and Arg1 mRNA in response to 1 mg/mL NPs over 24 hours compared to M1 activated cells. This suggests that NPs to not initiate an inflammatory response from macrophages after exposure to empty NPs up to 1 mg/mL, in vitro.

CONCLUSION

Optimized GM-CSF-loaded NPs achieved 85 ± 5 % entrapment efficiency of GM-CSF and sustained release over a 24-hour period while remaining sub-200 nm in size to take advantage of the EPR effect. Precipitated and released GM-CSF was found to remain fully bioactive. NPs were conveniently freeze-dried for long-term storage. Empty NPs were found to be highly biocompatible as reflected by no cytotoxic effects on murine BMDM. Immunogenically, empty NPs did not induce an inflammatory transcriptional profile from BMDM, suggesting that NPs themselves will not exert unwanted immunomodulatory effects in vivo. These studies support the continued study of targeted and controlled release of GM-CSF with PLGA/PLGA-PEG NPs for tumor immunomodulation. Overall, we can predict that in vivo delivery of our NPs will be well-tolerated.