Translational formulation of nanoparticle therapeutics from laboratory discovery to clinical scale

Three kinds of mixers were used in the current study (Fig. 1). The confined impinging jet mixer (CIJ) can be used in a batch, hand-held mode with syringes to feed the device, which produces NP formulations with sub-milligram active pharmaceutical ingredient (API) requirements [22]. The CIJ can also be driven by syringe pumps to make samples with larger volume of 200–300 mL [10]. The geometry and operation of the device have been previously reported [7]. Furthermore, two multi-inlet vortex mixers (MIVM-1.5L and MIVM-5L) were also used to generate NP formulations. The MIVM’s four-inlet geometry allows higher supersaturation during mixing than the CIJ and bypasses the secondary quenching step [23]; therefore the MIVM mixer has advantages for continuous and large scale production. Both mixer geometries produce NPs of the same size and stability, as will be shown below. The MIVM naming convention is based upon the approximate outlet flowrate, in liters per minute, at a mixer Reynolds number of 10⁵. While the MIVM-1.5L (Fig. 1b) can be used to produce any batch size by scaling production time, nanoparticle processing often involves other unit operations such as tangential flow filtration or spray drying. The mixer size should be matched to the flows and time scales of the other unit operations [9, 24]. Therefore, to avoid operating under conditions where the mixing and assembly regime has changed, a larger MIVM with a higher flow rate may be used. We designed the MIVM-5L to operate at a volumetric flow rate of 5 L/min at Re = 10⁵ and used a modified form of the design reported by Markwalter and Prud’homme [24]. We adopted a strategy that constrained several parameters within boundaries reported by Liu et al. as well as Markwalter and Prud’homme [24, 25]. The MIVM-1.5L and MIVM-5L mixers are geometrically similar with the vortex chamber of the 5L design being 2.5 times larger than the 1.5L design presented by Liu et al. [26]. A two-disk design was used to simplify machining and mixer assembly. The mixer was fabricated from stainless steel 316L with an electropolished surface and 20 RA finish.

To optimize the NP formulations, nanoparticles were first created via a CIJ. An organic stream of tetrahydrofuran (THF) with molecularly dissolved LMN and HPMCAS, was rapidly mixed against a deionized (DI) water stream into the mixing chamber of a CIJ in a 1:1 volume ratio [22]. The concentration in the organic stream was 7.5 mg/mL for LMN and 3.75 mg/mL for HPMCAS. With the CIJ, fluid was pressed manually from syringes at the same rate (~ 1 mL in 1 s), causing the two streams to merge into a mixing stream. The flow rate through the mixer was approximately 120 mL/min. The resulting mixed stream was collected in a quenching DI water bath to lower the final THF concentration to 10 vol%. Lyophilization was used to dry the CIJ samples.

In the MIVM, one organic stream containing 7.5 mg/mL LMN and 3.75 mg/mL HPMCAS-126 was mixed against three other water streams, with a volumetric flow rate of 1:9 (organic:water in total). The final organic solvent concentration as 10 vol%. Process development was carried out in the MIVM-1.5L using syringe pumps, which is convenient for samples from 20 to 300 mL. We then implemented Coriolis flow controllers (M14, mini CORI-FLOW, Bronkhorst, NL) to demonstrate a continuous process. The MIVM-5L was only operated with the flow controllers. The total flow rate was 160 and 550 mL/min for MIVM-1.5L and MIVM-5L, respectively. Based on the nanoparticle concentration, the mass production rate of MIVM-5L is 1 kg/day. Higher flow rates can further increase the mass production rates [24]. The MIVM-5L is designed to produce LMN NPs at 8 kg/day with Reynolds number of 10⁵. Spray drying was used to dry the MIVM samples.

Using a Zetasizer Nano-ZS (Malvern Instruments, Southboro, MA), NP diameter and polydispersity index (PDI) were determined, in triplicate, by dynamic light scattering (DLS) 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 International Organization for Standardization (ISO) standard document 13321. The calculations of PDI are defined in the ISO standard document 13321:1996 E.

Nanoparticle suspensions produced in either a CIJ or MIVM-1.5L were dropcast (~ 5 μL) onto a copper TEM grid (300 mesh carbon film, Electron Microscopy Sciences). Vapor-phase ruthenium staining was carried out by generating ruthenium tetroxide from ruthenium dioxide using sodium meta-periodate. The grids were placed in a sealed container with aqueous ruthenium solution until a cellulose sample indicated sufficient staining. Micrographs were obtained using a Philips CM-200 FEG-TEM at an accelerating voltage of 200 kV.

Lyophilization was carried out using a benchtop VirTis Advantage (Gardiner, NY) with appropriate cryoprotectants (HPMC E3). In our previous study with clofazimine [16, 17], HPMC E3, a water-soluble HPMC polymer, was used for HPMCAS NPs. The HPMC E3 serves as a cryoprotectant and prevents aggregation between the HPMCAS NPs during freezing and drying. 1 mL NP suspension were mixed with 0.1 mL cryoprotectant solutions to reach a 1:1 mass ratio of NP:cryoprotectant. The mixtures were then flash frozen by fast immersion in a dry ice/acetone cooling bath (− 78 °C) for 1 min with mild agitation. The frozen samples were then immediately transferred to the lyophilizer with shelf temperature at − 20 °C under vacuum (< 1 × 10⁻³ bar). After 2 days, dried powders were removed, sealed, and stored at − 20 °C. Lyophilization was only used for NP suspension generated by CIJ as the baseline for dissolution test.

A mini spray-drier B-290 (BÜCHI Corporation, New Castle, DE), equipped with a two-fluid nozzle, was used for drying the NP suspension in an open mode. After FNP, the NP suspension was mixed with the excipient, HPMC E3, at a mass ratio of 1:1. The suspension was then fed by a peristaltic pump into the spray-drier. The spray nozzle consisted of a tip and a cap with diameter of 0.7 and 1.5 mm, respectively, and the drier was equipped with a high-performance cyclone provided by BÜCHI. Compressed nitrogen at 480 kPa was used to atomize the liquid phase into droplets, and the flow rate was controlled by a rotameter. The inlet temperature, outlet temperature, drying gas flow rate, liquid feed rate and the gas flow rate of the aspirator were shown in Table 1. Spray dried powders were collected in scintillation vials, sealed, and stored at a vacuum desiccator and room temperature (20 °C) before use.

PXRD was performed using a Bruker D8 Advance Twin diffractometer equipped with Ag Kα radiation (λ = 0.56 Å) and LYNXEYE-XE detector. In each test, approximately 10 mg of powder was loaded into a polyimide capillary with an inner dimeter of 1 mm. Then the tube was mounted on a capillary stage, which rotated at a speed of 60 rpm during operation. Signals were collected between values of 3°–20° (2θ, corresponding to a Cu Kα 2θ value of ~ 8°–58°) with a step size of 0.025° (0.070° for Cu Kα radiation) and a count rate of 5 s/step. All PXRD results are presented with 2θ value corresponding to a Cu Kα radiation.

DSC experiments were performed with a TA Instrument Q200 (New Castle, DE) with hermetically sealed aluminum pans. Dried samples (5–10 mg) were equilibrated at 20 °C under dry N₂ atmosphere (50 mL/min), and then heated from 20 to 200 °C at a heating rate of 5 °C/min. The scan was analyzed by TA Instruments Universal Analysis 2000 software.

FaSSGF, FaSSIF and FeSSIF buffers were prepared following the manufacturer’s instructions. Triplicate experiments were performed for each sample, and free LMN powder was used as a control. For release under gastric conditions, dried powders were first resuspended in water and then diluted with pre-warmed FaSSGF (37 °C) to achieve a drug concentration of 50 μg/mL. The suspensions were then incubated at 37 °C (NesLab RTE-111 bath circulator, Thermo Fisher Scientific, Waltham, MA) for 30 min without agitation to mimic physiological gastric conditions and transit time in the stomach [27]. Since Brownian motion kept the small particles well dispersed, the effect of gastric mixing was not considered. Aliquots were taken at 5, 10, 20, and 30 min, which was centrifuged at 21,000g for 10 min to pellet NPs. For release under intestinal conditions, the solutions after the FaSSGF protocol were diluted 10× with 1.1× FaSSIF (pH = 6.5) or FeSSIF (pH = 5.8) with a final LMN concentration lower than its solubility limit in both buffers. Aliquots were taken at 30, 60, 120, 240, and 360 min, and were centrifuged at 21,000g for 10 min. Centrifugation provides complete separation of the nanoparticles from the supernatant, as confirmed by the lack of DLS signal in the supernatant after centrifugation. All supernatants were then removed, frozen, and lyophilized for later tests, and the sampling time points were defined as the incubation time from the assay start to the sampling.

High performance liquid chromatography (HPLC) was used to analyze the supernatants from the dissolution tests with a Gemini C18 column (particle size 5 μm, pore size 110 Å). The dried powder from the supernatants was resuspended in a mixture of acetonitrile (ACN) and THF (90/10, v/v), and then further sonicated to dissolve LMN. To pellet the insoluble bile salts from the buffers, each aliquot was centrifuged at 21,000g for 3 min. The supernatant was then filtered through a GE Healthcare Life Sciences Whatman™ 0.1 µm syringe filter. An isocratic mobile phase of ACN:water (60/40, v/v, both with 0.05 vol% trifluoroacetic acid) at 45 °C was applied to detect LMN with a flow rate of 1 mL/min. The LMN peak at 347 nm eluted at 6.8 min. The standard curve linearity was verified from 25 to 0.5 μg/mL with an r² value of at least 0.999 (Additional file 1: Figure S1).