The GDC-A drug product was prepared by spray-drying a liquid feedstock comprising a suspension of micronized GDC-A particles in an aqueous phase containing dispersed submicron PFOB-in-water emulsion droplets (Weers and Tarara, 2014; Weers et al., 2019). The emulsion droplets were prepared by first forming a dispersion of multilamellar vesicles (MLV) of DSPC and CaCl₂ (2/1 mol/mol ratio) in hot water (T > 65°C) above the hydrated gel-to-liquid crystal phase transition (T_m) of the phospholipid. The phospholipid was dispersed with a high-shear mixer (UltraTurrax T-25, (IKA, Wilmington, NC, USA). PFOB was added to this vesicle dispersion while mixing to form a coarse PFOB-in-water emulsion. This emulsion was then homogenized using a high-pressure homogenizer (EmulsiFlex C5, Avestin, Inc., Ottawa, Canada) for ∼10 min at 5,000 psi, ∼10 min at 10,000 psi, ending with a single discrete pass at 17,000 psi. The resulting fine emulsion has droplet diameters between ∼200 and 500 nm (Weers and Tarara, 2014). The size of the emulsion droplets and as a result the size of the pores in the small porous particle post spray-drying can be controlled by varying the ratio of PFOB/DSPC. Droplets that are starved in DSPC will have a larger droplet size in order to maintain the area/molecule for DSPC at the PFOB/water interface (Weers and Tarara, 2014).

During formulation development, different approaches were used to incorporate GDC-A into the feedstock. Forming stable suspensions of hydrophobic drugs like GDC-A in water can be challenging, as they often wet poorly, preferring to agglomerate at the air-water interface (Figure 1A) or adhere to surfaces in the manufacturing equipment.

Owing to its low surface tension at the air/water interface (∼20 mN/m), DSPC is a good dispersant for hydrophobic drugs (Kabalnov et al., 1995). The challenge with DSPC is that it is poorly soluble in water (∼10^–10 mol/L) (Marsh, 1990). When present in vesicles or emulsion droplets in a continuous water phase, molecular diffusion of DSPC molecules to a drug particle’s surface is exceedingly slow (Kabalnov et al., 1995). Hence, effective mixing of the phospholipid and drug is key to enabling preparation of a suitable suspension. This is not trivial. Fortunately, hydrophobic materials (e.g., leucine, phospholipids, cyclosporine, corticosteroids), are easily wetted and form stable suspensions in PFOB (Weers and Tarara, 2014; Weers et al., 2019; Miller et al., 2021). Hence, wetting the drug in PFOB (either in the coarse emulsion, or prior to emulsion formation in a solid-in-oil-in-water dispersion) reduces the risk that a poor suspension may be produced, especially at larger scales. In the present study, three methods for incorporation of the drug into the feedstock were studied. or the process designated as PSph A, the drug is dispersed in the fine emulsion. In PSph B, the drug is added to the lipid dispersion, while in PSph C the drug is dispersed in the PFOB in a solid-in-oil-in-water dispersion prior to fine emulsion formation. All three methods were able to effectively disperse the drug. The resulting complex dispersion comprising suspended drug particles in the oil-in-water emulsion was passed through a high-pressure homogenizer for a single discrete pass at 17,000 psi. The final aqueous feedstock comprising the suspended GDC-A particles and sub-micron emulsion droplets had an appearance that resembled milk, with no visible flocs of drug (Figure 1B).

Spray drying was conducted with a Büchi B-191 mini spray dryer (Flawil, Switzerland). For the purposes of this work, the atomization and drying processes were held constant, using parameters consistent with previous work on suspension-based PulmoSphere™ formulations (Weers et al., 2019). The spray-drying conditions utilized were an inlet temperature of 105°C, an outlet temperature of 60–65°C, an aspirator setting of 100%, an atomizer gas pressure of 70 psi, and a liquid feed rate of 5.0 ml/min. A custom-built (Adams and Chittenden, Berkeley, CA) glass cyclone was used. The aqueous feedstock was atomized with a twin fluid atomizer to produce 8–10 μm liquid droplets containing suspended drug particles and approximately one thousand emulsion droplets (Figure 2) (Ivey and Vehring, 2010; Weers et al., 2019). To prevent particle sedimentation having a negative impact on batch uniformity, the liquid feedstock was stirred during the spray drying process. During the initial moments of the drying process, the lower boiling process aid (water) begins to evaporate. This leads to a decrease in size of the atomized droplets, concentration of the slowly diffusing drug particles and emulsion droplets at the air-water interface, and the formation of a hollow core at the center of the droplet (Vehring, 2008; Ivey and Vehring, 2010; Weers and Tarara, 2014; Weers et al., 2019). Eventually, the second process aid (i.e., the oil phase in the dispersed emulsion droplets) evaporates, leaving behind pores where the sub-micron emulsion droplets were originally found. The process results in production of crystalline drug particles embedded in a nanostructured porous shell of phospholipid (Ivey and Vehring, 2010; Weers et al., 2019).

For aerosol performance testing, the spray-dried GDC-A bulk powder was hand-filled into size 3, inhalation grade hydroxypropylmethylcellulose (HPMC) capsules (Quali-V^®, Qualicaps, Whitsett, NC). The fill mass varied depending on the nature of the formulation. Powder consolidation within the capsule was achieved by gently tapping the capsule on a hard surface to enable increases in fill mass.

Determination of GDC-A content and purity in formulated bulk powder and aerosol test samples were determined by reverse phase high performance liquid chromatography (RP-HPLC) with detection at 220 nm. Samples were analyzed with an Agilent 1260 Infinity II HPLC system (Wilmington, DE, United States). Separation was achieved with Agilent InfinityLab Poroshell 120 EC-C18, 3.0 × 150 mm, 2.7 µm column (P/N 693,975–302) using a gradient method (solvent A = water with 0.025% trifluoroacetic acid (TFA); solvent B = acetonitrile with 0.025% TFA. GDC-A was quantified using a single-point calibration with a reference standard. Samples were extracted using water:acetonitrile 1:1 (v/v). The analyte retention time was 9.5 min, and the total run time was 22 min.

Primary particle size distributions were determined via laser diffraction (Sympatec GmbH, Clausthal-Zellerfeld, Germany). The Sympatec H3296 unit was equipped with an R2 lens, an ASPIROS micro dosing unit, and a RODOS/M dry powder-dispersing unit. Approximately 2–5 mg powder was filled into tubes and fed at 5 mm/s into a RODOS operated with 4 bar dispersion pressure and 65 mbar vacuum. Powders were introduced at an optical concentration of approximately 1–5% and data was collected over a measurement duration up to 15 s. Particle size distributions were calculated by the instrument software using the Fraünhofer model. Reported values represent the mean and standard deviation of three measurements.

Tapped densities were determined using a cylindrical cavity of known volume (0.593 cm³). Powder was filled into this sample holder using a microspatula. The sample cell was then gently tapped on a countertop. As the sample volume decreased, more powder was added to the cell. The tapping and addition of powder steps were repeated until the cavity was filled, and the powder bed no longer consolidated with further tapping. The tapped density is defined as the mass of this tapped bed of powder divided by the volume of the cavity.

X-ray powder diffraction (XRPD) patterns of selected powders were measured with a Rigaku Miniflex Model 600 diffraction system equipped with a D/Tex solid state “strip” detector. Each sample was prepared by packing bulk powder into a sample holder. Samples were scanned from 3° 2θ to 35° 2θ at a scan rate of 1° 2θ/min, using a Cu radiation source with a wavelength of 1.54 Å, operated at 40 kV and 15 mA.

The aerosol performance of spray-dried formulations of GDC-A were assessed with a variant of the RS01 dry powder inhaler, DPI (Plastiape, S.p.A., Osnago, Italy). The RS01 DPI is a portable, capsule-based DPI that is available with different resistances to airflow through the device. The variant utilized modifies the air inlet in the base assembly to increase device resistance to ∼0.12 cm H₂O^0.5 L⁻¹ min.

Aerodynamic particle size distributions (APSD) were determined with a Next Generation Impactor (NGI) equipped with a USP induction port. No pre-separator is needed for engineered particles in the absence of carrier particles. The impactor stages were coated with a solution comprising 50% v/v glycerol, 5% v/v Tween 20, and 45% v/v water (1 ml of solution for stages 2–7 and 2 ml for stage 1 and MOC). Tests were conducted in accordance with USP 〈601〉 Aerosols ‘Aerodynamic Size Distribution, Apparatus six for Dry Powder Inhalers’ and Ph. Eur. 2.9.18 ‘Preparations for Inhalation; Aerodynamic Assessment of Fine Particles; Apparatus E’. APSD measurements were conducted at a volume of 4 L and a pressure drop of 4 kPa. This corresponds to a volumetric flow rate of about 54 L/min. Delivered doses (DD) are reported for NGI testing by summing the mass of drug in the throat and on the various stages in the impactor. GDC-A was extracted and quantitated using the RP-HPLC detailed above.

Delivered dose measurements were conducted with a dose uniformity sampling apparatus (DUSA), with quantitation of drug content using the RP-HPLC detailed above. A total of ten measurements were made for delivered dose uniformity (DDU) testing, and the mean DD and standard deviation were calculated.

The physicochemical properties of spray-dried formulations of GDC-A are detailed in Table 1. Unless otherwise noted, the formulations within Table 1 were prepared at 10% v/v PFOB.

Suspension-based PulmoSphere formulations are often enriched in drug content relative to their theoretical content, presumably due to selective loss (out the exhaust of the spray dryer) of small unilamellar vesicles (post-homogenization) that do not contain drug. For GDC-A, the degree of enrichment was between 3.0 and 11.4%, decreasing with increases in drug content. Studies with other suspension based PulmoSphere formulations suggest that the degree of enrichment for a given formulation and process is highly reproducible between batches.

The suspension-based PulmoSphere manufacturing process for GDC-A, including high-shear mixing, high-pressure homogenization and spray drying resulted in purity and related substance profiles that were identical to the neat drug. Chemical stability of the formulated GDC-A drug product was assessed by placing both neat drug substance and batch R19017 (70% API, PSph C) at 40 °C for 1 month in an open dish configuration. The purity of both the neat drug substance and batch R19017 remained at ∼100.0% (normalized peak area), and all peaks remained below the limit of quantitation of 0.5 ng. Moreover, no changes in drug content were observed for batch R19017.

XRPD results indicate that the suspension-based PulmoSphere manufacturing process preserves the crystalline nature (Form A) of GDC-A in the drug product (Figure 3). Each diffraction peak observed in the drug substance is also present in the spray-dried powder. In this example, the diffraction peaks of the spray-dried powder are intense due to the high drug loading (70% w/w). The additional peak observed at 21.2°2θ is due to the DSPC excipient (Sun et al., 1996; Miller et al., 2015). This corresponds to a d-spacing of ∼4.2 Å. This length scale is consistent with the chain-chain spacing in hydrated gel-phase DSPC. The high crystallinity of the formulated drug product is maintained over 26 months following room temperature storage in a dry box as evidenced by the overlapping XRPD pattern with the t = 0 sample.

Scanning electron microscopy (SEM) images of selected GDC-A formulations are detailed in Figure 4. As the drug loading is increased, there is a transition from particles that are PulmoSphere-like with a porous morphology (42% w/w GDC-A) to particles that have an appearance that is more consistent with the micronized API (70% w/w GDC-A).

The shift in particle morphology with increases in drug content is also reflected by linear decreases in the size and increases in tapped density of the GDC-A powders (Table 1; Figure 5). As the theoretical drug loading was increased from 42 to 70% w/w, the size of the spray-dried particles was markedly decreased from 2.9 to 1.9 μm for the X₅₀, and from 4.9 to 3.6 μm for the X₉₀. The tapped density of the powders was increased from 0.05 g/cm³ to 0.13 g/cm³. Hence, the less PulmoSphere excipient that is present, the smaller the particles and the greater the tapped density. As the drug loading increases, the particles increasingly resemble the neat drug crystals (as given by the data for the 100% API). At a 70% drug loading, the X₅₀ remains about 0.4 μm larger than that of the neat drug, however. The increases in tapped density and decreases in primary particle size distribution at higher drug loadings both favor higher TLDs for GDC-A, provided that the particles can be adequately dispersed with a portable dry powder inhaler.

The aerosol performance was assessed with a Next Generation Impactor (4 kPa pressure drop with the RS01 DPI) for batches with varying drug content and fill mass (Table 2). A representative aerodynamic particle size distribution is presented for batch 19,015–1 in Figure 6. The impactor DD measurements were generally greater than 80% of the nominal dose. The MMAD of the powders decreased with increased drug loading, presumably driven by the decrease in primary particle size. For the 70% w/w GDC-A formulations (10% v/v PFOB), the MMAD values were in the range from 2.4 to 2.7 μm, with FPF_<5µm values of 82–87% of the DD. This equates to an FPD_<5µm of up to 25 mg from a size three capsule. This exceeded the target FPF_<5µm of 5–20 mg. If comparable aerosol performance were observed in a multi-dose dry powder inhaler (MD-DPI) with a receptacle volume of 50 μL, an FPD_<5µm of ∼4 mg could be achieved. This exceeds by an order of magnitude the highest FPD_<5µm observed with MD-DPIs for the delivery of potent bronchodilators and inhaled corticosteroids for the treatment of asthma and chronic obstructive pulmonary disease.

In a separate experiment conducted at Genentech, the aerosol performance of neat GDC-A was assessed. The mean DD was just 48.0%. For the fraction of particles exiting the device, the MMAD was low (2.1 μm) and the FPF_{<5 μm} expressed as a percentage of the DD was comparable to the PulmoSphere formulations (79.2%). Nonetheless, the low DD leads to a much lower FPF_<5μm when expressed as a percentage of the nominal dose (i.e., 38.0 vs 71.0% R19017-3).

Delivered dose uniformity (DDU) of spray-dried GDC-A (Batch R19017) was determined at a 4 kPa pressure drop (N = 10). The mean delivered dose was 86.7% with a standard deviation of 4.4%; which suggests it would meet the FDA MDI/DPI guidance requirement for DDU.

Formulation of GDC-A as a crystalline solid using the suspension PulmoSphere process is expected to provide numerous benefits relative to formulation as an amorphous solid using conventional solution-based spray drying processes. These include improvements in physical and chemical stability, improvements in lung targeting due to the ability to sustain drug levels in the lungs, and minimization of potential tolerability issues associated with high dose delivery. These features are discussed in more detail below.

Son et al. categorized inhaled dry powders in terms of their TLD (Son et al., 2021). TLD values less than 1 mg were defined as low doses, TLD values between 1 and 20 mg were defined as moderate doses, those between 20 and 100 mg as high doses, and TLD values >100 mg were defined as ultrahigh doses. The GDC-A drug product with its target TLD of 5–20 mg falls within the moderate dose category. The results presented for the suspension PulmoSphere formulation herein demonstrate the ability of the suspension PulmoSphere powder to meet the target dose (based on using the FPD_<5μm as a metric for TLD).

Son et al. also introduced a new metric for assessing high dose delivery termed the ‘product density’ (Son et al., 2021). The product density ( ρ p r o d u c t ) is simply the T L D achieved with a portable DPI divided by the volume of the receptable ( V r ) that the powder is contained within. The product density can be subdivided into three terms that capture the essence of key design features of a high dose product (Eq. (4)). ρ p r o d u c t = T L D V r =   ( m p o w d e r V r )   ( m d r u g m p o w d e r )   ( T L D m d r u g ) (4)

The three terms comprise: 1) the mass of powder ( m p o w d e r ) that can be filled into the receptacle volume (‘packing density’); 2) the mass of drug ( m d r u g ) divided by the mass of powder (‘drug loading’), and; 3) the fraction of the drug that is delivered into the lungs (‘aerosol performance’). These terms enable comparison of the GDC-A PulmoSphere formulations with other marketed products comprising spray-dried porous particles.

For GDC-A batch R19015-2, a powder fill mass of 51.25 mg was achieved in a size three capsule (V_r = 0.30 ml), for a packing density of 170.8 mg/ml. This is greater than the packing density (131.1 mg/ml) achieved with the solution-based PulmoSphere formulation in the TOBI Podhaler drug product, and significantly greater than is achieved with the large porous particle formulation in Inbrija (52.6 mg/ml) (Geller et al., 2011; Son et al., 2021). It should be noted that the GDC-A powder was hand-filled versus the commercial products which are filled on a drum filler. With that said, our experience is that we are able to achieve even greater fill masses with a drum filler due to compression of the powder into pucks during filling. Hence, one would expect the improvements in packing density might be even greater. The drug loading in this formulation was 0.72 mg/mg, and the FPD_<5μm was 0.68 mg/mg, resulting in a product density of 83.8 mg/ml. This is about 1.8-fold greater than is observed in TOBI Podhaler (47.1 mg/ml), and nearly 4-fold greater than Inbrija (22.1 mg/ml) (Son et al., 2021).

The achievable TLD from a single receptacle can be increased significantly by using larger sized capsules (assuming equivalent aerosol performance). For the product density of 83.8 mg/ml observed, the FPD_<5μm increases to 31.0, 41.9, 57.0, and 79.6 mg in size 2 (V_r = 0.37 ml), size 1 (V_r = 0.50 ml), size 0 (V_r = 0.68 ml), and size 00 capsules (V_r = 0.95 ml), respectively. Perhaps of equal importance, an FPD_<5μm of 4.2 mg can be achieved in a 50 μL blister, enabling moderate doses to be delivered in a multi-dose DPI. The potential fill mass in a size 00 capsule is ∼162 mg. This would require an inhaled volume of about 3.6 L to empty the powder contents, or between two and four inhalations depending on the patient population (Son et al., 2021).

This study did not explore drug loadings greater than 70% w/w. It remains to be seen whether the drug loading in suspension PulmoSphere powders can be increased further, and when the influence of the phospholipid shell on powder properties is lost. Buttini et al. demonstrated packing densities as high as 176.5 mg/ml and a product density of 104.4 mg/ml in spray-dried tobramycin base coated with sodium stearate (Buttini et al., 2018). They were able to deliver a TLD of ∼70 mg from a size 0 capsule. Pilcer et al. also achieved excellent aerosol performance with lipid-coated crystals of tobramycin base (Pilcer et al., 2009).

Hence, the formulation of crystalline drugs via suspension-based spray drying should be considered as a viable formulation alternative for the delivery of high doses of poorly soluble hydrophobic drugs, such as GDC-A.