The central concept of our project is the design of a genetically modified probiotic that can act as a programmable therapeutic system for asthma. Our chassis,Lactobacillus plantarum, is generally recognized as a safe organism with natural anti-inflammatory properties and compatibility with the lung–gut axis. By equipping this probiotic with synthetic genetic circuits, we aimed to transform it into a smart therapeutic vehicle capable of producing, packaging, and delivering siRNA directly to the lung epithelium.
Compared to existing biologics and inhaled therapies, our project offers a localized, self-regulated,probiotic-based delivery system that minimizes systemic side effects, reduces treatment costs, and creates a sustainable therapeutic platform.
In order to demonstrate the feasibility and social value of the approach, and evaluate its translational potential, we validated our system through four complementary perspectives:
The KatA promoter is a well-characterized oxidative stress–responsive element in Gram-positive bacteria. Its regulation by the peroxide stress repressor PerR ensures tight repression under basal conditions and derepression upon exposure to hydrogen peroxide. As shown in Chen, Keramati, and Helmann katA–lacZ fusions, they demonstrated that KatA expression is strongly induced by H₂O₂, while perR mutants display constitutive promoter activity (1). Their findings also revealed that promoter regulation is modulated by metal ions (Fe²⁺, Mn²⁺), establishing KatA as a PerR-regulated oxidative stress–responsive element (2).
Prior studies and experimental data from the iGEM Team Technion-Israel show that KatA promoter can exhibit basal leakage under non-stressed conditions, which could compromise safety and specificity (3).
To address this, we integrated an additional repressor–operator system. Rep repressors have been successfully used to strengthen promoter control by binding to operator sequences and blocking transcription initiation. Meanwhile, the pH-sensitive promoter P170-CP25 is known to be activated under acidic conditions (pH ≤ 6.9), typical of inflamed lung microenvironments (4) for more details see Design page.
We placed a Rep operator downstream of the KatA promoter and constitutively expressed the Rep repressor from the p32 promoter. To add a second environmental checkpoint, we embedded a Lac operator in p32 and controlled LacR expression with P170-CP25. In healthy (neutral pH) airways, Rep is continuously produced and silences KatA. In inflamed (acidic) conditions, P170-CP25 induces LacR, which blocks Rep production and allows KatA to respond to H₂O₂ — creating a dual-gated AND logic system.
figure(2): Under acidic conditions (pH ≤ 6.9), LacR is activated via the P170-CP25 promoter, suppressing p32 and reducing Rep levels. This relief of repression enables KatA to activate CO-BERA expression, ensuring precise control in inflamed, low-pH environments.
Figure(3): showing the conditioned expression of CO-BERA.
As a result of our model we found that pKatA showed a high basal leakage reaching up to 40.08%.
Figure(4): shows pKatA expression activity in different concentrations of H₂O₂.
We developed a new conditioning device specific to our target asthma population. To demonstrate that our therapeutic circuit only activates under disease-specific conditions, we built a mathematical model of our dual-input AND gate Conditioning system. This gate required two asthma hallmarks simultaneously:
Low pH (6.9) → activates LacR, which represses Rep and removes the first block on the pKatA promoter that expresses CO-BERA.
High external H₂O₂ (>8 µM) → inactivates PerR and removes the second block on the pKatA promoter that expresses CO-BERA.
Only when both conditions are present together, CO-BERA transcription is permitted.
Figure(5): shows the result of the active condition, which leads to high expression of CO-BERA in the presence of 2 inputs, and the inactive condition that leads to suppression of CO-BERA expression to nearly zero.
These results validate that our AND gate successfully prevents accidental CO-BERA expression under healthy conditions, while ensuring robust activation in diseased airways, as mentioned in the Model page.
To validate our conditioning system, we will test the KatA–PerR system controlling CO-BERA expression. As designed, this promoter–repressor pair is expected to remain silent under basal conditions (low H₂O₂) and become active only under oxidative stress.
We plan to prepare two colonies of the engineered strain: one grown in media supplemented with H₂O₂ (induced condition) and the other in media without H₂O₂ (negative control). After culturing, total RNA will be isolated from both groups, converted into cDNA, and analyzed by RT-qPCR using CO-BERA–specific primers. The expected outcome is that CO-BERA transcription will be strongly upregulated in the H₂O₂-treated samples, while expression in the negative control will remain at minimal levels as shown in figure below. This will demonstrate both the responsiveness of the KatA promoter to oxidative stress and the low leakage under non-induced conditions, providing proof of concept for our conditioning system.
The KatA–PerR promoter system, supported by literature and experimental evidence, provides a robust oxidative stress–responsive switch. By integrating Rep repression and pH-sensitive P170 control, we engineered a dual-gated circuit that minimizes basal leakage and ensures activation only under inflammatory conditions. Together with modeling predictions, this multi-layered design confirms the potential to enable precise, condition-specific siRNA expression in inflamed airways.
RNA-binding proteins provide powerful tools to package different RNAs into vesicles, one of them is L7Ae, it is an extensively characterized archaeal ribosomal protein that binds with very high affinity and specificity to K-turn motifs in C/D box RNAs. Rozhdestvensky et al. (2003) validated this interaction by systematically analyzing the structural requirements for binding, showing that L7Ae specifically recognizes and stabilizes K-turn motifs in C/D box small nucleolar RNAs (snoRNAs) (7). This established a molecular framework for exploiting L7Ae–C/D box interactions in synthetic applications.
Building on this principle, iGEM 2024 HBMU-Taihe team demonstrated that incorporating multiple C/D box motifs into RNA transcripts greatly enhances their encapsulation into extracellular vesicles (8). Their experiments showed that two tandem C/D box repeats provided a strong balance between efficient RNA loading and transcript stability, confirming the utility of this strategy for vesicle-mediated RNA delivery (8).
Molecular dynamics simulations (CHARMM-GUI + GROMACS) predicted DUF4811–L7Ae fusions integrate stably into L. plantarum membranes.
The DUF4811-Linker-RNA binding protein (L7Ae) complex molecular dynamics simulation showed RMSD range within 1.5Å in 5ns, which is stable, and RMSF range within 0.1Å, which ensured complex inflexibility. This result proved that this complex is the most stable and rigid complex compared to the other transmembrane protein complexes.
figure(7): Graph shows the RMSD of one C/D box copy relative to L7Ae protein.
Figure(8): Graph shows the RMSD of two C/D box copies RNA relative to L7Ae protein.
To determine how many copies of C/D box would be suitable for our C/Dbox-L7Ae complex we made some in silico analysis and found that 1 copy has more in the docking score with L7Ae and lower ensemble diversity. We made a simulation for one copy of C/D box and two copies with L7Ae on CHARMM GUI and OPENMM to determine which C/D copies is more stable for our complex.The RMSD graph showed that one copy has lower RMSD and ends with plateau not rising as two copies of C/D box, which tells us that one copy is more promising. However we found that there were experimentals made by iGEM 2024 HBMU-Taihe team said that two C/D box copies give better results, so we ordered different versions of C/D box copies to determine which will give better results in experimental validation.
figure(9): Graph shows the RMSD of one C/D box copy relative to L7Ae protein.
Figure(10): Graph shows the RMSD of two C/D box copies RNA relative to L7Ae protein.
Leveraging these validated mechanisms, we fused DUF4811, a (bacterial transmembrane protein), with RNA binding protein (L7Ae) to anchor the RBP in bacterial membranes. This configuration ensures that C/D box–tagged RNA transcripts (CO-BERA) are selectively recruited and packaged into bacterial membrane vesicles (BMVs). We embedded two tandem C/D box repeats into CO-BERA, following the evidence from iGEM HBMU-Taihe that this arrangement yields robust enrichment of RNA cargo (8). By combining the mechanistic validation of L7Ae–C/D box binding with proven vesicle-loading efficacy, our system ensures selective and efficient packaging of therapeutic RNA scaffolds.
Figure(11): showing the loading system of CO-BERA.
After conducting molecular dynamics and simulation analyses, we will proceed with experimental validation by isolating BMVs from engineered bacteria as shown in figure below and comparing them with both positive and negative controls. Following the established ultracentrifugation and resuspension protocol, BMVs will be purified and analyzed for RNA content. The expected outcome is that CO-BERA will be successfully enriched within BMVs derived from the engineered strain, whereas minimal RNA will be detected in the negative control. This will demonstrate that the DUF4811–L7Ae fusion effectively directs selective packaging of therapeutic RNA scaffolds, thereby validating our loading system.
Our loading system successfully integrates the DUF4811–L7Ae fusion with tandem C/D box motifs to achieve selective enrichment of therapeutic RNA in BMVs. Supported by literature, modeling predictions, and planned wet-lab validation, this system demonstrates a reliable strategy for packaging RNA scaffolds into bacterial vesicles, ensuring both specificity and efficiency.
siRNAs are probably one of the most studied silencing mechanisms for mammalian cells. More specifically, once one strand of the short RNA duplex is loaded into the RNA-induced silencing complex (RISC), this single guide strand causes sequence-specific cleavage of complementary messenger RNAs to inhibit protein translation. Davidson and McCray cemented this principle by reviewing and synthesizing results from several preclinical studies, showing that synthetic and vector-delivered siRNAs can efficiently knock down gene expression in vivo, thus emphasizing better designs to reduce off-target effects and maintain therapeutic efficacy (5).
For asthma, thymic stromal lymphopoietin (TSLP) has emerged as a validated therapeutic target. TSLP drives type 2 immune responses and airway inflammation by activating dendritic cells. Corren et al. provided clinical validation of TSLP targeting: in a randomized controlled trial, they showed that the anti-TSLP monoclonal antibody tezepelumab significantly reduced asthma exacerbations and improved lung function in patients with uncontrolled asthma (6).
Building on this validated mechanism, we implemented KatA to control the expression of CO-BERA, our dual-siRNA therapeutic scaffold. Under healthy conditions (low H₂O₂), PerR repression minimizes basal transcription, reducing off-target effects. In inflamed lung tissue, where H₂O₂ levels are elevated, PerR is inactivated and KatA is derepressed. This condition-specific activation allows KatA to drive expression of CO-BERA, enabling targeted therapeutic activation only during oxidative stress associated with lung inflammation.
Figure(13): showing the mechanism of TSLP knockdwon .
After validating that our AND gate ensures disease-specific expression of CO-BERA and an effective Loading system, we searched for whether the CO-BERA construct could truly be processed inside host cells into functional 2 siRNAs capable of knockdown of TSLP mRNA.
The dicer enzyme will bind to CO-BERA and then break it to 2 guide siRNAs, the guide siRNA will bind to Argonaute protein to form RISC complex to degrade the TSLP mRNA. We have made molecular dynamics simulation to detect the stability of that complex which showed this RMSD graph that shows the stability of the argonaute protein with RMSD range around 2-2.7 Å after an initial equilibration phase indicating the stability of the protein, while the guide siRNA shows higher RMSD values (4-5.5 Å) which is expected, as RNA is generally more flexible than protein and as referred to (Department of Biotechnology, School of Biosciences and Technology, Vellore Institute of Technology, Vellore, 632014, Tamil Nadu, India) (18) which have done simulation similar to ours of siRNA with argonaute, indicates that the average RMSD of the simulation was 0.568 nm (5.68 Å) as shown in figure below, indicating a quasi-stable state with moderate conformational rearrangements. And that the more dynamic nature of the AGO-siRNA complex may be biologically relevant, as previous studies have shown that structural flexibility and RNA strain contribute to efficient guide strand binding and RISC loading. Sciabola et al. (2013) (19).
figure(14): Graph shows the RMSD of guide siRNA-argonaute complex
To validate this, we built a kinetic model of RNA processing by dicer and compared the outcomes of ASO versus BERA (single siRNA) versus CO-BERA (dual siRNA).
Figure(15): shows the difference between natural vs ASO vs BERA vs CO-BERA knockdown efficiency.
In this model, we validated every step from CO-BERA endocytosis to TSLP mRNA knockdown and its final effect on TSLP protein concentration.
This specifically validates our concept through the following:
Successful generation of siRNA: CO-BERA consistently produced two siRNAs for every transcript, confirming that the processing by Dicer was correct.
More efficient than BERA and ASO: This nearly doubled the amount of siRNA that was produced compared to BERA, thereby causing stronger knockdown of TSLP mRNA from 82.5% to 99.7% efficiency.
Validation through PCR-based analysis of endogenous TSLP knockdown
We will first confirm CO-BERA expression and siRNA loading. Next, the basal level of TSLP mRNA in human cell lines will be assessed before treatment. Engineered bacteria will then be co-cultured with human cells for 6–24 hours, during which host media is expected to trigger the release of bacterial membrane vesicles carrying CO-BERA. Following treatment, total RNA will be isolated from human cells, converted into cDNA, and analyzed by RT-qPCR to measure silencing efficiency. In parallel, Western blot analysis will be performed to confirm the reduction of TSLP protein expression. Collectively, these experiments are expected to validate that CO-BERA–delivered siRNAs successfully mediate RNAi silencing in target cells, thereby demonstrating the therapeutic potential of our system.
Validation through GFP reporter assay
As part of an additional validation strategy, we will clone the TSLP and GFP expression cassette into a Level-1 TU1-lacZ plasmid and transfect it into human cell lines.
Co-transfection with our designed siRNAs will enable us to assess TSLP knockdown at both the mRNA and protein levels by monitoring the decrease in green fluorescence as shown in figure below. Functional assays, including qPCR, will be employed to quantify silencing efficiency and confirm that the siRNA candidates effectively reduce TSLP expression in vitro.
By combining rationally designed siRNAs with the validated asthma target TSLP, our system demonstrates a powerful therapeutic framework. Both validation strategies—PCR-based analysis of endogenous knockdown and GFP reporter assays are expected to confirm efficient and specific silencing at the mRNA and protein levels. Together with computational predictions, these approaches provide strong evidence that CO-BERA–delivered siRNAs can act as an effective and targeted RNAi therapy.
Wild-type Listeriolysin O (LLO) is a cholesterol-dependent cytolysin secreted by Listeria monocytogenes to enable bacterial escape from endosomes. However, with its lack of control in pore formation, it becomes highly cytotoxic. Bavdek et al. (2007) demonstrated that LLO is tightly regulated by cholesterol and acidic pH, confining its function essentially to endosomes (9). A point mutation, L461T, is interesting in that it shifts the profile of the activity: the mutant retains its ability to form pores but with approximately 100-fold less virulence and is activated near neutral pH (10). Therapeutic use of LLO-L461T is safer and more productive. Validation studies confirmed that this mutant preserves efficient membrane disruption while greatly reducing host cell pathogenesis, enabling endosomal escape without compromising cell viability (10).
We employ LLO-L461T to facilitate cytosolic release of CO-BERA RNAs. By coupling efficient pore formation with reduced virulence activity, this mutant provides a reliable means of endosomal escape. Its earlier activation not only improves safety but also safeguards CO-BERA from lysosomal degradation, ensuring greater stability and functional availability of the therapeutic RNA inside target cells.
To validate the mutated Listeriolysin O, we made a molecular dynamics simulation to ensure the activity of pore formation within the phagosomal membrane in the pH level of the early endosome which is 6.5.
The simulation was made for 5 ns, we extracted the RMSD and the RMSF graphs, which showed the conformational changes that happens from pre forming to forming. However, this simulation was not enough for validating the Listeriolysin O pore formation activity as it needs a simulation for nearly 1000ns-1500ns to be validated.
By integrating the LLO-L461T mutant, our system harnesses efficient endosomal escape while minimizing cytotoxicity. The mutation preserves pore-forming ability under near-neutral pH but with drastically reduced virulence, ensuring safe cytosolic delivery of CO-BERA. This clinically inspired strategy enhances RNA stability and therapeutic availability, providing a balanced solution between efficacy and biosafety.
Toxin–antitoxin systems are widely used in synthetic biology to maintain plasmid stability and serve as kill switches. The natural pair PemK/PemI has a long history of application: earlier studies (e.g. Tsuchimoto et al.) showed that PemK alone arrests growth, while PemI rescues this effect, and loss of the plasmid triggers cell death in plasmid-free segregants (11). More recently, PemI/PemK has been validated in non-lab contexts: addition of pemI/pemK to broad host-range vectors in Xylella fastidiosa and E. coli achieved nearly complete plasmid retention in the absence of antibiotic selection (12). Also, comparative studies show that combining TA systems with plasmid partitioning mechanisms enhances stability, especially for low copy number plasmids (13).
Our model shows that the phoB promoter and RNA 2U thermosensor act only under lung conditions. This ensures the circuit becomes active at low phosphate and body temperature, preventing unintended activation elsewhere.
The model demonstrates that transcription and antitoxin translation reach a steady state inside the lung. In contrast, outside conditions shift the balance, where degradation dominates and toxins begin to accumulate.
1. Toxin–Antitoxin Balance
Equilibrium between toxin and antitoxin is maintained only within the lung environment. Once bacteria escape this niche, the balance collapses, allowing free toxin to accumulate and triggering the self-killing mechanism.
2. Bacterial Population Outcome
Survival is restricted to lung-specific conditions. In every other environment, toxin activity overwhelms the system, ensuring bacterial death and strengthening biosafety.
This figure shows PemI/K kill switch component changes over time equal to 10 hours and the deactivation of the circuit at time (5h) that leads to decreased complex and antitoxin concentration and toxin accumulation.
This figure shows mRNA change over time before and after deactivation of the circuit.
This figure shows free toxin concentration over time.
Together, these results highlight how the PemIK model provides a strong safety mechanism: it preserves bacterial function only in the lung and prevents survival anywhere else.
We integrate the PemI/PemK system into our system vector to ensure stable inheritance of the CO-BERA construct without continuous antibiotic pressure. Given the recent evidence that PemI/PemK can be highly effective in non-lab environments (X. fastidiosa), and that combining TA with partitioning yields robust maintenance even for low copy plasmids, we anticipate high levels of retention over time. This stability supports consistent therapeutic output and reduces the risk of plasmid loss in vivo.
Our system incorporates multilayered biosafety as shown in figure below, ensuring strict containment of engineered strains.
Dry powder inhalers (DPIs) are established as a leading delivery format for asthma because they combine long-term formulation stability, breath-activated dosing that improves patient adherence, and efficient lung deposition when particle aerodynamic diameters are tuned to ~1–5 µm (14). Reviews of DPI device design and particle engineering summarize advances that improve aerosol performance and dose reproducibility (14).
Clinical guidance and large observational studies also support DPI use for maintenance respiratory therapy: DPIs perform at least comparably to pMDIs or nebulizers for maintenance treatment in appropriately selected patients and have advantages for portability and environmental impact; however, inspiratory flow dependence is an important practical consideration, especially in children or severe disease (15).
Our aerodynamic deposition model validated that our inhaled probiotic can deliver a safe and effective therapeutic dose to the lung. Using experimentally grounded parameters (potency, aerosol viability, and regional deposition fraction), the deterministic calculation predicted 5×10⁶ CFU deposited per puff as shown in figure below.
This distribution plot provides confidence intervals for bacterial lung deposition, validating our deterministic prediction of 5 × 10⁶ CFU (0.5 × 10⁷) as the final delivered dose. Such value is well within the expected variability range.
In addition, our model showed that the inhaled probiotic dose falls within the therapeutic window, hence, assuring its safety and efficacy. Simulation has shown that the delivered bacteria maintain stability under normal conditions and also fare well against changes in powder potency and inhalation force. The model is based on known aerosol science phenomena and further checked against pharmaceutical expert opinion, adding a plausibility check to real-world feasibility. Sensitivity analysis found particle size and aerosolization survival to be a big concern; hence, they direct our formulation and device design strategy.
Stakeholder Validation: Dr/ Suzan, who is a representative of the European pharmaceutical company and biotech engineers, confirmed that DPIs are the most clinically viable administration method for our probiotic-based therapeutic.
Our system can be realistically delivered using an inhaler format already familiar to asthma patients.
Freeze-drying (lyophilization) guarantees better maintenance of viability than spray-drying, especially for those probiotic strains that are sensitive to heat, shear, or osmotic stress (16). Comparative studies have always reported a higher survival rate and better functional stability in lyophilized preparations than in spray-dried ones (16). Stabilizing excipients such as trehalose, lactose, and skim milk powder serve as stabilizers during lyophilization by taking the place of water molecules in cell membranes and proteins so as to prevent denaturation and structural collapse (17).
We utilize freeze-drying as the stabilization strategy for our engineered probiotic chassis carrying the therapeutic systems. By incorporating trehalose and lactose into the lyophilization matrix, we aim to maximize cell viability during storage and reconstitution, ensuring consistent delivery performance in dry powder formats. This approach combines industrially scalable preservation with the robustness needed for clinical translation.
Our freeze-drying model validated that this method achieves significantly higher bacterial survival compared to spray-drying, confirming it as the superior processing strategy for maintaining probiotic potency. Simulations consistently supported freeze-drying as the more reliable option under varying process conditions, with advantages extending to storage stability and long-term viability.
This bar chart demonstrates freeze-drying's superior survival rate (72.7%) versus spray-drying (59.5%) for L. plantarum processing in our validations. The 13.2% advantage supports our technology selection decision based on quantitative modeling rather than assumptions in our validations.
Dr/ Ahmed Abdelsabour, who is the manager of the manufacturing process in EIPCO Pharmaceutical Factory, highlighted freeze-drying as the most scalable and regulatory-accepted approach for the manufacturing of live bacterial therapeutics.
By grounding the model in literature data and bioprocess principles, and aligning it with manufacturing feasibility and regulatory standards, this proof of concept ensures that our technology choice was made based on rigorous quantitative analysis rather than assumptions that show higher survival with lower process risk.
The system constitutes a therapeutic framework composed of conditional gene activation, selective RNA loading, siRNA-based TSLP silencing, and LLO-L461T-mediated endosomal escape. In addition, the plasmid-payload constructs employ toxin-antitoxin containment and freeze-drying to ensure biosafety and stability, deliverable via a dry powder inhaler. These systems jointly ensure that our system can be developed into safe, effective, and clinically feasible probiotics for asthma.
1-Chen L, Keramati L, Helmann JD. Coordinate regulation of Bacillus subtilis peroxide stress genes by hydrogen peroxide and metal ions. Proc Natl Acad Sci U S A. 1995;92(19):8190-4.
2-Lee JW, Helmann JD. Functional specialization within the Fur family of metalloregulators. Biometals. 2007;20(3–4):485-99.
3-iGEM Team Technion-Israel (2019). Basic Part. iGEM 2019. Available from: https://2019.igem.org/Team:Technion-Israel/Basic_Part.
4-Madsen SM, Arnau J, Vrang A, Givskov M, Israelsen H. Molecular characterization of the pH-inducible and growth phase–dependent promoter P170 of Lactococcus lactis. Mol Microbiol. 1999;32(1):75-87. doi:10.1046/j.1365-2958.1999.01326.x.
5-Davidson BL, McCray PB Jr. Current prospects for RNA interference-based therapies. Nat Rev Genet. 2011;12(5):329-40.
6-Corren J, Parnes JR, Wang L, Mo M, Roseti SL, Griffiths JM, et al. Tezepelumab in adults with uncontrolled asthma. N Engl J Med. 2017;377(10):936-46.
7-Rozhdestvensky TS, Tang TH, Tchirkova IV, Brosius J, Bachellerie JP, Hüttenhofer A. Binding of L7Ae protein to the K-turn of archaeal snoRNAs: a shared RNA binding motif for C/D and H/ACA box snoRNAs in Archaea. Nucleic Acids Res. 2003;31(1):234–42.
8-iGEM HBMU-Taihe 2024. GPC3 CDS–2×C/D Box RNA packaging construct. iGEM Registry Part: BBa_K5073025. Available from: https://parts.igem.org/Part:BBa_K5073025.
9-Bavdek A, Gekara NO, Priselac D, Gutierrez Aguirre I, Darji A, Chakraborty T, et al. Sterol and pH interdependence in the binding, oligomerization, and pore formation of Listeriolysin O. Biochemistry. 2007;46(2):442–53.
10-Glomski IJ, Gedde MM, Tsang AW, Swanson JA, Portnoy DA. The Listeria monocytogenes hemolysin has an acidic pH optimum to compartmentalize activity and prevent damage to infected host cells. J Cell Biol. 2002;156(6):1029–38. Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC308949/
11-Tsuchimoto S, Ohtsubo H, Ohtsubo E. Two genes, pemK and pemI, responsible for stable maintenance of mini-R100 plasmid. J Bacteriol. 1988;170(4):1461–6.
12-Lee MW, Rogers EE, Stenger DC. Xylella fastidiosa plasmid-encoded PemK toxin is an endoribonuclease. Phytopathology. 2012;102(1):32–40.
13-Stenger DC, Burbank L, Lee MW. Plasmid vectors for Xylella fastidiosa utilizing a toxin–antitoxin system for stability in the absence of antibiotic selection. Phytopathology. 2016;106(4):384–92.
14-Islam N, Gladki E. Dry powder inhalers (DPIs)—a review of device reliability and innovation. Int J Pharm. 2008;360(1–2):1–11.
15-Laube BL. The expanding role of aerosols in systemic drug delivery, gene therapy, and vaccination. Respir Care. 2005;50(9):1161–76./p>
16-Morgan CA, Herman N, White PA, Vesey G. Preservation of micro-organisms by drying; a review. J Microbiol Methods. 2006;66(2):183–93.
17-Carvalho AS, Silva J, Ho P, Teixeira P, Malcata FX, Gibbs P. Relevant factors for the preparation of freeze-dried lactic acid bacteria. Int Dairy J. 2004;14(10):835–47.
18-riram, S., Palanichamy, C., Subash, P.T. et al. Molecular dynamics simulations based siRNA design against GPR10 reveals stable RNAi therapeutics for hormone-dependent uterine fibroids. Sci Rep 15, 31708 (2025). https://doi.org/10.1038/s41598-025-16936-z
19-Sciabola, S., Cao, Q., Orozco, M., Faustino, I. & Stanton, R. V. Improved nucleic acid descriptors for SiRNA efficacy prediction. Nucleic Acids Res. 41 (3), 1383–1394. https://doi.org/10.1093/nar/gks1191 (2012).
