Tobramycin (C₁₈H₃₇N₅O_9, molecular weight 467.51 g/mol) was purchased from Hubei Widely Chemical Reagent Co. Ltd (China). Ethanol (HPLC grade, GENERAL-REAGENT^®) was obtained from Shanghai Titan Scientific Co. Ltd. (China). All other reagents used were of analytical grade. Enzyme-linked immunosorbent assay (ELISA) kits for IL-4, IL-6, and TNF-α were purchased from Hangzhou MultiSciences Biotech Co. Ltd. (China).

The operational parameters of the DPIs prepared by SFD are listed in Supplementary Table S1. Different amounts of tobramycin were dissolved in 50 mL of ultrapure water to obtain aqueous solutions of different concentrations (C _tob = 5, 15, 25, 35, and 45 mg/mL). The tobramycin aqueous solution was then transported with a hose (inner diameter = 5 mm) and sprayed into 1 L of liquid nitrogen using a laboratory spray dryer (Shanghai Ya Cheng Co. Ltd., YC-01) at different atomization pressures (P = 0.1, 0.15, 0.2, 0.25, and 0.3 MPa) and volumetric feeding rates (Q = 18, 21, 24, 27, and 30 mL/min). The sprayed product was then transferred to a lyophilizer (Foring Technology Development Co. Ltd., LGJ-22C) for freeze-drying.

The particle size distribution (PSD) was measured using laser diffraction (DLS) (HELOS/RODOS, SYMPATEC, Germany). Approximately 10 mg of the sample was loaded in a glass dispersion tube and dispersed at a pressure drop of 3.5 bar. An R5 lens with a measuring range of 0.45–875 µm was used for measurement. The values of D₁₀, D_50, and D₉₀ for the volumetric diameter were recorded. Span represents the width of the sample size distribution and was calculated with D₁₀, D_50, and D₉₀ using Equation 1 (Chua et al., 2023): S p a n = D 90 − D 10 D 50 (1)

Additionally, for porous or hollow particles, the aerodynamic diameter (D _a ) can be calculated using the particle tapped density ( ρ t ) and volume median diameter (VMD) of the particles using Equation 2 (Alhajj et al., 2021): D a = V M D ρ t (2)

Scanning electron microscopy (SEM, Hitachi S-4800, Hitachi High-Technologies Crop, Tokyo, Japan) was used to investigate the morphology of the samples. The samples were dispersed on conductive adhesive and imaged in the sample compartment at 5.0 kV for samples.

As an important factor for evaluating inhalation performance, the flowability of particles was investigated via the angle of repose test, in line with the standard protocol specified in the United States Pharmacopoeia (USP). A certain mass of powder often forms a cone when it is affected only by gravity. The steepest angle of descent relative to the horizontal plane is defined as the angle of repose (θ). The angle of repose was lower when the sample was non-cohesive and had good flowability. The angle of repose can be calculated using Equation 3 (Ghoroi et al., 2013): θ = arctan   2 H D (3) where H and D are the powder stacking height and circular stacking diameter, respectively.

In addition, many studies have used the Carr index for characterization and evaluation (Duhan et al., 2021). It is generally believed that the smaller the Carr index, the smaller the compressibility of the powder and the better its fluidity. The Carr index (CI) was calculated using Equation 4: C I   % = ρ t − ρ b ρ t × 100 % (4) where ρ_b and ρ_t are the bulk and tap densities of the sample, respectively.

The crystalline structures of the samples were tested using X-ray powder diffraction (SmartLab-SE, RIGAKU, Japan). Approximately 4 mg of each sample was tightly spread on a sample plate and subjected to Cu radiation at ambient temperature at a current of 15 mA and a voltage of 30 kV. The scanning range of the diffraction angle (2θ) was 5°≤2θ ≤ 60°.

Thermogravimetric analysis (TGA) was conducted to analyze the thermal stability of the DPIs samples and the relationship between the sample mass and temperature. Approximately, 3 mg of each sample was weighed and placed in an alumina crucible. The entire system was heated from 30 to 500°C at 10°C/min with a 20 mL/min nitrogen flow in a TGA device (TG 209 F3 Tarsus, NETZSCH, Germany).

The infrared spectral chemical structures of the samples were analyzed using a Fourier transform infrared spectrometer (Nicolet iS5, Thermo Fisher China Technology, Shanghai, China). The test sample was placed on the “Attenuated Total Reflection” (ATR) detection component of the instrument and the probe was fixed. The scanning frequency was set at 32.

The thermal response profiles were obtained using differential scanning calorimetry (DSC) (DSC 214 Polyma, NETZSCH, Germany). The sample was weighed and placed in an aluminum crucible. The entire system was heated from 30°C to 280°C at 10 °C/min in a DSC device under a sustained 50 mL/min nitrogen purge.

According to the British Pharmacopoeia 2017, a next-generation impactor (Logan Instruments, New Jersey, USA) and the DPI inhaler (Carent, China) were used to evaluate the aerosol performance of the DPIs particles. The dispersion rate and duration were 60 L/min and 4 s, respectively. The weight method was used to determine the number of particles deposited at each stage by measuring the weight difference of the glass filter after particle deposition. Emission fraction (EF) is defined as the mass of powder leaving the inhaler relative to the total dose. Fine Particle Fraction (FPF) is defined as the recovered dose, which is the total mass of powders smaller than 5.0 µm. The mass median aerodynamic diameter (MMAD) is defined as the aerodynamic diameter, defining the aerodynamic diameter at which 50% of the particles collected from the first stage to the micropore collector have a larger mass and 50% are smaller. The MMAD and geometric standard deviation (GSD) were calculated by linear fitting of the cumulative mass and aerodynamic cut-off diameter on a logarithmic scale. The aforementioned in vitro aerodynamic parameters were collected for each sample and repeated six times.

The human alveolar basal epithelial A549 cell line (Wuxi Xinrun Biotechnology Co., Ltd.) was selected, and the cytotoxicity of the prepared product was evaluated using the MTT (3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide) method. Cell viability was determined by the reduction of MTT to formazan by mitochondria. The cells were cultured in RPMI 1640 medium supplemented with 10% Fetal Bovine Serum (FBS) in 5% CO₂ atmosphere, and the growth medium was changed every 2–3 days until the cells reached 100% confluence. Cell dissociation and harvesting was performed using trypsin. MTT assay was performed according to the manufacturer’s instructions. Briefly, 190 µL culture medium containing 5,000 cells was inoculated into each well of an experimental 96-well plate, and then incubated for 12 h to achieve 90%–100% confluence. Subsequently, 10 µL of testing products which were obtained by dissolving DPIs sample in PBS with the sample concentrations of 1, 0.5, 10⁻¹, 10⁻², 10⁻³, and 10⁻⁴ mg/mL were added to the corresponding wells and incubated. After 24-h incubation at 37°C and 5% CO₂ atmosphere, each well was treated with 10 µL of MTT (5 mg/mL) solution and further incubated for 4 h. After incubation, the culture medium was removed from the wells, and 150 µL of the termination solution, DMSO, was added to each well. After a final incubation period of 30 min, the absorbance of each well was measured at 490 nm. The cell viability (%) was calculated using Equation 5 (Adhikari et al., 2022): C e l l   v i a b i l i t y   % = M e a n   o f   A b s o r b a n c e : v a l u e   o f   t r e a t m e n t   g r o u p M e a n   o f   A b s o r b a n c e : v a l u e   o f   c o n t r o l × 100 % (5)

The antibacterial activity of the product prepared using the SFD methods against P. aeruginosa was evaluated by measuring the minimum inhibitory concentration (MIC) (Arauzo et al., 2021). The growing bacteria were diluted in phosphate-buffered saline (PBS) to a final concentration of approximately 10⁷ CFU/mL. They were then inoculated into the culture media prepared with tobramycin DPIs samples at different concentrations (0, 0.0625, 0.125, 0.250, 0.500, 1, 2, 4, 8, 16, 32, 64, 128, and 256 μg/mL). The bacteria were incubated on a shaking table at 37°C and 150 rpm for 24 h, and the optical density of the pathogenic bacteria in contact with the powder was measured at 600 nm (OD₆₀₀) using a microplate reader (Spectramax@i3x, Molecular Device, United States) to detect bacterial growth.

C57BL/6 mice (Hangzhou Ziyuan Laboratory Animal Technology Co., Ltd.), aged 6–8 weeks were used in this study. They were placed and fed under a 12-hour light/12-hour dark cycle. All animal and cell experiments were performed in accordance with the guidelines of the Pharmaceutical Animal Experimental Center of China Pharmaceutical University with the approval number SYXK(SU)-2023–0019.

Mice were divided into healthy groups (six mice), and treated group (six mice), which received tobramycin powders prepared by SFD method 30 mg/kg. The survival rates of each group of mice as well as physical examinations of the skin, eyes, respiratory system, and behavior were observed and recorded every day. The mice were enthanized after a 14-day therapy period. During necropsy, the viscera and body were carefully and thoroughly examined, dissected, gathered, and weighed. In gross necropsies of mice, we meticulously analyzed the body exterior, orifices, and abdominal and thoracic cavities. Dehydrated hearts, livers, spleens, lungs, and kidneys were fixed with 4% paraformaldehyde fixative, embedded in paraffin, sectioned using a microtome at a thickness of 5 μm, mounted on glass slides, stained with hematoxylin and eosin, and prepared for histopathological analysis under a light microscope.

Animals were randomly grouped into five: NC (PBS-treated, negative control, n = 12); BLM (Bleomycin-treated, shortened to BLM-treated, n = 12); BLM + PA (BLM-and-P. aeruginosa-treated, n = 12); ITP (BLM-and-P. aeruginosa-treated and inhaled sample A4, n = 12); and IV (BLM-and-P. aeruginosa-treated and intravenous injection administration of tobramycin, n = 12).

Specifically, BLM, BLM + PA, ITP, and IV groups of animals were treated with a single intratracheal injection of bleomycin (3 mg/kg; Shanghai Yuanye Biotechnology Co., Ltd.) to induce pulmonary fibrosis (Kang et al., 2022). The NC group received an equal amount of PBS. Seventeen days after BLM injection, a bacterial suspension of 50 μL 1 × 10⁹ CFU/mL P. aeruginosa was injected into the BLM + PA, ITP, and IV groups. After 1 day of inoculation with PA, group ITP was treated using a dry powder insufflator (DP-M, TOW Intelligent Technology Co., Ltd, China) for intratracheal administration of DPIs sample A4 prepared by SFD method for three consecutive days, and the dosage of administration was 5 mg/kg/dose twice daily. Group IV was administered intravenously for three consecutive days at a dose of 5 mg/kg/dose twice a day. All mice were weighed daily for 25 days. The dosing amount for each group was referenced from the dosing amount of tobramycin in traditional dosage forms used in clinical applications, and was determined according to the conversion of body surface area between humans and experimental animals.

Generally, larger DPIs particles (>5 µm) are easily deposited in the mouth or wall of the upper respiratory tract. In contrast, smaller particles (<0.5 µm) can be expelled from the lungs during exhalation because of their smaller volume and lower inertia tract (Hebbink et al., 2022). Particles ranging from 0.5 to 2 µm are considered to be effectively deposited in deeper pulmonary (Huang, 2023). In this study, the SFD method was applied to prepare tobramycin DPIs microparticles with sizes ranging from 1 to 5 µm. The D₅₀ and D_a values of the DPIs samples were measured using DLS and are shown in Supplementary Table S2. The DPIs prepared by the SFD method exhibited smaller particle sizes and narrower distributions than API. To investigate the effects of operation parameters, DPIs samples were prepared with different Tob concentration (Sample A1-A5; C _tob = 5, 15, 25, 35 and 45 mg/mL), atomization pressure (Sample A6-A10; P = 0.1, 0.15, 0.2, 0.25 and 0.3 MPa), and volumetric feeding rate (Sample A11-A15; Q = 18, 21, 24, 27 and 30 mL/min) when applying SFD method. According to Figure 1a, significant decline of D₅₀ (from 10.81 ± 0.70 μm to 3.52 ± 0.04 μm) and D_a (from 5.07 ± 0.10 μm to 1.06 ± 0.04 μm) can be obtained when increasing C _tob from 5 mg/mL to 35 mg/mL. However, further increasing C _tob to higher value (45 mg/mL), the value of D₅₀ and D_a improved to 8.85 ± 0.32 μm and 4.12 ± 0.82 μm, respectively.

When C _tob was ranging from 5 mg/mL to 35 mg/mL, an increase in the feeding solution’s viscosity was observed, which consequently led to higher surface tension. Therefore, smaller droplets and ice crystals were formed during the spray- and freeze-drying processes, which would eventually translate into smaller particles. However, when the concentration was further increased to 45 mg/mL, the particle size was primarily determined by intermolecular forces. Tobramycin is a water-soluble aminoglycoside antibiotic, and its hydroxyl and amino groups can form intermolecular hydrogen bonds during freeze-drying. This enhances the van der Waals forces between particles, leading to particle aggregation due to mutual attraction. Thus, the increased concentration contributed to the strengthening of powder interactions, resulting in larger particles (Farinha et al., 2023; Mueannoom, 2012).

Figure 1b shows that the nebulizer pressure slightly affects the DPIs particle size. The D₅₀ and D_a particle size decreased by 40.14% (from 5.88 ± 0.34 μm to 3.52 ± 0.04 μm) and 60.00% (from 2.65 ± 0.51 μm to 1.06 ± 0.04 μm) in our experiment when the nebulizer pressure was tripled from 0.1 MPa to 0.3 MPa. Elevating the nebulizer pressure from 0.1 MPa to 0.3 MPa tended to result in a decrease of particle size. First, a heightened nebulizer pressure typically corresponded to a more vigorous atomization airflow, which led to a higher dispersion of liquid droplets and mitigation of inter-droplet interaction. Second, as the nebulizer pressure increased, the solution was sprayed into minute droplets with augmented surface area and surface energy, which resulted in the formation of finer particle (Farinha et al., 2023; Schappo et al., 2021).

The relationship between the volumetric feeding rate (Q) and particle size is shown in Figure 1c. A clear relationship exists in which a higher volumetric feeding rate leads to smaller D₅₀ and D_a sizes. When the volumetric feeding rate increased from 18 mL/min to 21 mL/min, the particle size reduced from 10.31 ± 0.30 μm (D₅₀) and 4.13 ± 1.03 μm (D_a) to 6.53 ± 0.04 μm (D₅₀) and 2.52 ± 0.19 μm (D_a). Further increasing the volumetric feeding rate from 24 mL/min to 30 mL/min results in particle size reduction by 46.09% (D₅₀, from 6.53 ± 0.04 μm to 3.52 ± 0.04 μm) and 57.94% (D_a, from 2.52 ± 0.19 μm to 1.06 ± 0.04 μm), respectively. The heightened flow velocity effectively enhanced the atomization efficiency of the spray nozzle, facilitating the dispersion of liquid droplets into finer mist, which increased the surface area of the droplets. Thus, more droplets could quickly contact liquid nitrogen and lose water when dispersed. Simultaneously, the atomized droplets collided with the internal airflow, and the elevated feed flow velocity may have influenced the airflow within the system, altering the suspended time and dispersion extent of particles in the air. This reduced the aggregation of particles, which helped the droplets to form smaller particles. After preliminary screening and optimization, A4 was used in further research to evaluate their fluidity and aerodynamic performance.

To explore the connection between Tob concentration (C _tob ), atomization pressure (P), and volumetric feeding rate (Q), and pinpoint the key factor influencing the particle size of D_a, we applied a machine learning approach, specifically a decision tree regressor, to construct a model for the research problem at hand. In this model, C _tob , P, and Q acted as features (independent variables), whereas D_a served as the target (dependent variable). The model demonstrated a performance metric with a squared error of 1.23, computed as the mean across 10 simulations with fine-tuned hyperparameters. To ensure a well-rounded model and prevent overfitting, we set the minimum depth of the decision tree to 4. The resulting decision tree is visually represented in Figure 2a, where the nodes represent the experimental conditions and illustrate the distribution of samples across these conditions. Moreover, our investigation delved into the significance of each factor, as shown in Figure 2b, demonstrating that C _tob played a primary role, while Q assumed secondary importance. This observation was confirmed by decision tree analysis, where C _tob was identified as the root node, underscoring its pivotal role in influencing the overall outcomes. Through the evaluation of process parameters, it is demonstrated that in the subsequent more refined process optimization, the upper limit of the increased C _tob applicable to the SFD method is 35 mg/mL. Additionally, the Q is a factor worthy of further exploration. Given the flow rate limitations of the experimental equipment, it is feasible to further increase the volumetric feeding rate (>30 mL/min) by improving the equipment, so as to explore a more optimal formulation.

The pulmonary delivery and deposition efficiency of DPIs rely not only on the particle size but also on their morphology, density, fluidity, and other properties (Huang, 2023). Therefore, according to the above results and discussion, samples A4 was chosen for further evaluation of their fluidity and aerodynamic performance. The yield, density, angle of repose, and Carr index of A4 were investigated. The yield of A4 was 63.21 %± 2.11%, and the angle of repose, the Carr index value of sample A4 were 36° and 37.54%, respectively. Overall, sample A4 shows slightly good flow characteristics and powder fluidization behavior, which play a crucial role in the aerosol performance of the DPIs (Liu, 2024).

SEM images of the tobramycin API and samples A4 were obtained to characterize the particle size, microstructure, and morphology. As shown in Figure 3a, raw tobramycin exhibited a larger average particle size of 25.26 µm (D₅₀) with an irregular shape and broader size distribution. In contrast, sample A4 showed the smallest D₅₀ size (3.52 µm), narrower size distribution, and reticular spherical morphology with a loose and porous structure (Figure 3b). The formation mechanism can be explained as follows: The droplets generated in the spray process were frozen instantaneously to form spherical ice crystals after contact with liquid nitrogen. Subsequently, water was removed through ice sublimation during the freeze-drying process, resulting in the formation of a fine porous structure. Previous studies have demonstrated that this morphology is more suitable for lung drug delivery because the porous structure can improve the specific surface area of particles, leading to an increased contact area between the solvent and particles. Moreover, its low density results in a small aerodynamic diameter, which can make DPIs delivery deeper into the lungs (Ye et al., 2017).

The ATR-FTIR, XRPD, DSC, and TGA results of the API and samples A4 are shown in Figure 4. The results of the ATR-FTIR spectrum (Figure 4a) showed that the absorption and intensity of the drug substance and tobramycin granule sample was consistent with the API, indicating that the SFD processes did not change the original structure. Specifically, the spectrum exhibited bands at 3,400–3,200 cm⁻¹ attributable to N–H or O–H stretching, and bands at 3,000–2,800 cm⁻¹ attributable to C–H stretching. The spectrum simultaneously showed a band at 1,550–1,600 cm⁻¹ due to N–H bending, a typical band around 1,460 cm⁻¹ caused by CH₂ scissoring, bands at 1,300–1,380 cm⁻¹ caused by O–H in-plane bending vibration, and a band around 1,019 cm⁻¹ caused by C–N or C–O stretching.

The crystallinity of the API and sample A4 were analyzed using XRPD, and the results are shown in Figure 4b. It is obvious to see the perfect crystalline of tobramycin API. The XRPD pattern of tobramycin API has obvious peak intensities. In contrast, there were no peaks in the XRPD spectrum of sample A4, indicating that the SFD produced an amorphous powder, which might be caused by the rapid arrangement of tobramycin molecules in the process of SFD. Moreover, it has been demonstrated that drugs in an amorphous state generally show enhanced solubility. Thus, DPIs in an amorphous state could be released rapidly after deposition in the lung, leading to a relatively higher drug concentration and improved treatment effect (Guo et al., 2023).

The DSC curves of the Tobramycin API and samples A4 are shown in Figure 4C. Compared with previous studies (Fernández, 2018; Rosasco, 2018), the DSC curve of Tobramycin API exhibited two endothermic peaks at ∼170 and ∼230°C and one exothermic peak at ∼210°C. These three thermal events in Figure 4c are marked as 1-3, corresponding to the first endothermic peak 1, which can be ascribed to the melting of the metastable form of tobramycin, and exothermic peak 2, which was due to the stable anhydrous form of crystallization before melting. Corresponding to the melting behavior, the DSC curve showed a sharp endothermic peak three at ∼230°C. In summary, tobramycin API showed two different crystal phases at ∼170 and ∼230°C, with a recrystallization peak between the two points and degradation starting at the second melting point peak. In contrast, the thermogram of sample A4, which was prepared by the SFD method, had no endothermic peak at approximately 230°C, which is consistent with the results of similar technology and processes in other studies (Miller et al., 2015).

The TGA curves of the tobramycin API and samples A4 are shown in Figure 4d. The mass loss before 200°C can be attributed to the loss of moisture and volatile components in the particles, dehydration reactions, and release of water molecules. The deviation of the two curves was precisely the result of changes in the crystal shape. In addition, tobramycin is known to be hygroscopic in nature (Patere, 2018). As sample A4, which is in an amorphous state, is more easily soluble, the decrease in sample A4 was greater than that of API. This result was in line with expectations. However, when the temperature rose to approximately 285 °C, the curves decreased dramatically, which corresponded to the mass loss resulting from the collapse of the tobramycin structure. From the perspective of thermal stability, the remaining weights after dehydration at 285 °C were close to 90%, similar to that of tobramycin API. Therefore, samples A4 exhibited good thermal stability.

The in vitro aerosol performance of tobramycin API and DPIs samples (A4) were tested by NGI, as shown in Figure 5. Aerosol performance and particle characteristics are listed in Table 1. The interstage distribution map shows that sample A4 was mainly deposited in stages 4–7, making these particles very suitable for delivering the drug to the lungs. The EF range of API was 97.91%, which was much higher than that of A4 (89.02%). This difference can be attributed to the fact that the particles prepared by the SFD method were loose and porous. Meanwhile, the phenomenon of electrostatic adsorption can be observed during the deposition experiment in vitro (Kaialy, 2016). It might stem from the fact that the edges or tips of the pores in porous particles are prone to form charge accumulation, leading to an increase in the local charge density.

Although API have excellent fluidity, and they are easier to inhale and deposit in vivo and in vitro, which is equivalent to a lower dose loss when inhaled tobramycin DPIs (Sun et al., 2020). On the other hand, the FPF of samples A4 was 83.31%, which were significantly higher than those of the APIs (18.65%). Meanwhile, the MMAD of samples A4 was1.31 μm, which is much higher than the APIs valued 12.51 μm, indicating that the size of inhalable particles is best transported to the respiratory bronchioles and alveoli with small deviations through sedimentation and diffusion (a target for the treatment of Pseudomonas aeruginosa infection CF by inhaling tobramycin). In addition, the GSD value indicates the polydispersity of the particle-size distribution. The GSD value of A4 was lower than API, implying that it was suitable for pulmonary administration (Huang, 2023). Therefore, this study paves the way for the effective treatment of excipient-free tobramycin DPIs through the lung pathway in future CF.

To determine the influence of storage stability, A4 was stored in commercial packaging boxes which were airtight and light-proof for 28 days under two conditions: 25°C and 60% relative humidity (RH) and 40°C and 75% RH. After testing the storage stability, the aerodynamic performance parameters of all samples were measured, as shown in Table 2 and Figure 6, and compared with the values of fresh DPIs samples to evaluate the stability of particle storage. As the temperature and humidity of the storage environment increased, the EF range of the stored samples slightly increased owing to the increase in density after moisture absorption, overcoming electrostatic adsorption and reducing particle gaps, resulting in a slight improvement in flowability, although there was no significant difference under various conditions. Moreover, because amorphous powder is more hygroscopic than crystalline powder, increasing the humidity results in particle agglomeration after the absorption of water molecules, and the higher the humidity, the more obvious the agglomeration phenomenon. Previous studies have reported that crystallization/recrystallization of an amorphous form is a complex phenomenon governed by many factors, including intermolecular interactions such as hydrogen bonding and/or π–π interactions, and ionic interactions between two molecules can form in some amorphous particles (Lu, 2019; Müller, 2015). A higher water content may lead to particle growth owing to an increase in interparticular forces (e.g., capillary forces). This led to a significant decrease in FPF (p < 0.05) and a significant increase (p < 0.05) in MMAD of samples stored at 40°C and 75% RH for 28 days compared with the fresh sample. Based on the 28-days storage stability results, A4 shows relatively superior results. However, in order to ensure the long-term stability of A4, it is recommended to use materials with excellent moisture-proof performance for the primary packaging to prevent moisture absorption. In addition, desiccant packets can be considered for addition in the secondary packaging to further control the internal humidity environment.