key: cord-0734615-4boj0td7
authors: Hartigan, Devon R.; Adelfio, Miryam; Shutt, Molly E.; Jones, Stephanie M.; Patel, Shreya; Marchand, Joshua T.; McGuinness, Pamela D.; Buchholz, Bryan O.; Ghezzi, Chiara E.
title: In Vitro Nasal Tissue Model for the Validation of Nasopharyngeal and Midturbinate Swabs for SARS-CoV-2 Testing
date: 2022-03-29
journal: ACS Omega
DOI: 10.1021/acsomega.2c00587
sha: 777a80742b9d46e9cfd60c46ce2d5ffa9011dbbf
doc_id: 734615
cord_uid: 4boj0td7

[Image: see text] Large-scale population testing is a key tool to mitigate the spread of respiratory pathogens, such as the current COVID-19 pandemic, where swabs are used to collect samples in the upper airways (e.g., nasopharyngeal and midturbinate nasal cavities) for diagnostics. However, the high volume of supplies required to achieve large-scale population testing has posed unprecedented challenges for swab manufacturing and distribution, resulting in a global shortage that has heavily impacted testing capacity worldwide and prompted the development of new swabs suitable for large-scale production. Newly designed swabs require rigorous preclinical and clinical validation studies that are costly and time-consuming (i.e., months to years long); reducing the risks associated with swab validation is therefore paramount for their rapid deployment. To address these shortages, we developed a 3D-printed tissue model that mimics the nasopharyngeal and midturbinate nasal cavities, and we validated its use as a new tool to rapidly test swab performance. In addition to the nasal architecture, the tissue model mimics the soft nasal tissue with a silk-based sponge lining, and the physiological nasal fluid with asymptomatic and symptomatic viscosities of synthetic mucus. We performed several assays comparing standard flocked and injection-molded swabs. We quantified the swab pickup and release and determined the effect of viral load and mucus viscosity on swab efficacy by spiking the synthetic mucus with heat-inactivated SARS-CoV-2 virus. By molecular assay, we found that injected molded swabs performed similarly or superiorly in comparison to standard flocked swabs, and we underscored a viscosity-dependent difference in cycle threshold values between the asymptomatic and symptomatic mucuses for both swabs. To conclude, we developed an in vitro nasal tissue model that corroborated previous swab performance data from clinical studies; this model will provide to researchers a clinically relevant, reproducible, safe, and cost-effective validation tool for the rapid development of newly designed swabs.

The rapidly increasing demand for COVID-19 testing since the start of the 2020 pandemic has caused significant bottlenecks in testing capacity due to a global shortage of testing supplies, including specimen collection swabs. 1, 2 To help overcome the swab shortage, alternative swabs that could be mass produced at a relatively low cost (e.g., via injection-molded processing) have been recently developed and raced to the market. 3 Compared to a standard flocked, the injection-molded swabs are characterized by a nonabsorbent head and have demonstrated a more efficient release of viral RNA while absorbing less solution. 1,4,5 As a result, several prototypes of injection-molded swabs have been recently commercialized, including nasopharyngeal and midturbinate swabs like the IM2 and Rhinostic swabs. 3, 4 These new one-piece specimen collection swabs can be efficiently mass produced without multistep manufacturing methods and postprocessing. Vali-dation for swab prototypes typically requires preclinical testing before transitioning to clinical studies, which can take months to years. To streamline this initial preclinical validation process, there is a compelling need for the development of an in vitro experimental model that recapitulates physical and structural features of the human nasal cavities to bridge benchtop and clinical studies. An in vitro nasal model can be efficiently and safely used by stakeholders to perform preclinical evaluations, anticipating device design modifications, before confidently moving to a design-lock stage during clinical studies. 6, 7 There are currently no in vitro tissue models available for this purpose. Bench-top studies are performed by dipping in saline solutions without mimicking any of the physiological aspects of the nasal passage (i.e., architectural, mechanical, and physical structures) or of the actual swabbing procedure. 4, 8 Other alternative modes of validation for prototype swabs in the preclinical phase include swabbing the cheeks of participants and quantifying bacterial and cellular uptakes in comparison to the standard swabs or through clinical studies. 3, 9, 10 By solely relying on clinical studies, even for preclinical validation, the swabs meet a great deal of variability, thus there is a need to expand participant enrollment, with a significant increase in associated time and costs. Therefore, we hypothesize that the initial validation of swab prototypes on a simplified, reliable, and physiologically relevant in vitro nasal tissue model would provide more consistent and reproducible results, allowing investigators to assess swab performance in a time-and cost-efficient manner. 5, 10 Expanding the preclinical evaluation based on an in vitro tissue model will further support clinical studies to assess swab efficacy, streamlining the overall validation process.

On the basis of our previously developed anterior nasal tissue model, here we describe the design and fabrication of an in vitro tissue model platform ( Figure 1 ) that aims to support preclinical validation of nasopharyngeal and midturbinate swabs, in an effort to significantly decrease swab validation time, allowing faster and more efficient distribution. 3, 5 The in vitro model is based on a three dimensionally (3D) printed nasal cavity to accurately mimic native tissue architecture, lined with a silk sponge to recapitulate the soft tissue structure, and does not require the use of cellular material. In addition, we varied viral load and mucus viscosities to better encompass the wide spectrum of clinical conditions and further investigated their effects on swab performance.

To further support the use of the in vitro nasal model, we developed several validation assays to assess the performance of nasopharyngeal and midturbinate injection-molded and standard flocked swabs. We proposed new assessments to streamline initial swab validation, including gravimetric analysis and release quantification of fluorescently labeled microparticles, that mimic cellular material to quantify swab pickup and release capabilities. In addition, we carried out a Quantitative Reverse Transcription Polymerase Chain Reaction (RT-qPCR) using spiked mucus samples to mimic clinical swabbing and compare in vitro performance of the different types of swabs. The proposed model is a novel approach to The human nasal cavity displays the entrance to the cavity, hard palate, septum, inferior turbinate, and nasopharynx. (B) The 3D model replicates the architecture and structure of the human nasal cavity, including the entrance to the cavity, hard palate, septum, inferior turbinate, and nasopharynx. The lateral and cross-sectional images are shown as well. (C) 4% w/v silk sponges line the cavity, saturated with an artificial nasal mucus that physiologically mimics the viscosity of nasal fluid. A nasopharyngeal swab is inserted all the way until it meets resistance, while a midturbinate swab is inserted halfway. Both swabs are twisted and held for 15 s before they are removed. The swabs can be placed in diagnostic assay solutions and be ready for post collection analyses (PCR, gravimetric, release etc.). A SEM micrograph displays the open pore structure of the soft tissue-like lining. Scale bar = 100 μm. Partially created with BioRender.com. support initial swab validation, as it accurately replicates the physiological components of the nasal cavity including architecture, structural elements, as well as the viscosities of physiological nasal fluids.

Experimental Swabs. Herein, we assessed the performance of injection molded swabs in comparison to the Clinical Laboratory Improvement Amendments (CLIA) use-approved class I exempt standard flocked swabs. Obecare sterile flocked nasopharyngeal (NP) swabs (Obecare, West Virginia) were used as the standard nasopharyngeal and midturbinate swabs (MT) in the experiments. The Obecare flocked-NP swabs are standard flocked swabs characterized by an adhesive coated surface and nylon fibers that are attached perpendicularly for maximum absorbance (Obecare, West Virginia). Injection molded-NP and injection molded-MT swabs were manufactured as a single element based on a biocompatible polymer injected into a mold of a swab and allowed to harden (Yukon Medical, Durham, NC; Figure 2 and Table S1 ).

3D Printed Nasal Tissue Model Preparation. To provide a benchtop validation system for experimental swabs, a 3D nasal tissue model was developed to mimic the human architecture and soft tissue properties ( Figure 1 ). The physiological architecture was recreated by replicating the nasal cavity, specifically the opening to the cavity, the inferior nasal concha (inferior turbinate), the septum, the hard palate, and the nasopharynx. The 3D-model design was previously generated by the Aerosol Research Lab at Carleton University in Ottawa, Canada. 11 The 3D model of the nasal cavities was then generated using acrylonitrile butadiene styrene (ABS, Gizmo Dorks LLC, Temple City, CA) filament with a fused deposition modeling 3D printer (ABS-P430, Stratasys, Eden Prairie, Minnesota).

To mimic the soft tissue of the nasal cavities, aqueous silk sponges were prepared as previously reported. 12−14 Briefly, pure silk fibroin was extracted from Bombyx mori cocoons by degumming the fibers in a sodium carbonate solution (0.02 M; Sigma-Aldrich, St. Louis, Missouri) for 30 min to remove sericin. The degummed fibers were rinsed three times and dried overnight before the solubilization in 9.3 M lithium bromide (Sigma-Aldrich, St. Louis, Missouri) for 2 h at 60°C. The obtained solution was dialyzed for 3 days against DI water using standard grade regenerated cellulose dialysis tubing (3.5 kDa MWCO, Spectrum Laboratories Inc., Rancho Dominguez, California). The solution was then centrifuged to remove impurities. Subsequently, silk sponges were made according to the published protocol. 13 A total of 1.5 mL of a silk solution (4% w/v) was poured into a 24-well plate (VWR Scientific, Radnor, Pennsylvania) and frozen for two cycles of 24 h at −20°C and −80°C. The frozen plate was lyophilized for 3 days. The sponges were autoclaved at 121°C to induce the change in the secondary structure of the protein and induce water insolubility. Finally, sponges were cut into 0.5-mm-thick slices with an ad hoc sample cutter. The 3D printed model cavities were then lined with silk sponges with cyanoacrylate surgical glue (Henkel Loctite 4601 Medical Device Instant Adhesive Clear, Houston, Texas) to mimic the native soft tissue. The model was then rinsed with 70% v/v ethanol and ultrapure water to thoroughly remove any glue residue and biological material.

Synthetic Asymptomatic and Symptomatic Mucus Preparation and Characterization. Two nasal mucus conditions were designed to mimic the viscosity of asymptomatic and symptomatic nasal fluid conditions. 15 Poly(ethylene oxide) (PEO, Sigma-Aldrich, St. Louis, Missouri. MW 1 000 000) was used to replicate these conditions, as previously reported. 16 Upon preliminary investigations (data not reported here) and in previous literature, 16 PEO concentrations were chosen as 0.5% and 3.0% w/v for asymptomatic and symptomatic conditions, respectively. Viscosity analysis was performed by using a dynamic viscometer (Brookfield Viscometer-Massachusetts) to identify the physiological values of the nasal mucus. Briefly, PEO solutions (0.5% and 3% w/v) were incubated for 30 min at 37°C

. After the stabilization of the torque (equal or above 10%), 0.5 mL of PEO solution (N = 3 per condition) was loaded, while being maintained at 37°C, and the analysis was carried out between 0.1 and 100 s −1 .

The volume of PEO to fully saturate the nasal cavity was also considered. A symptomatic nasal cavity experiences rhinorrhea, nasal congestion, and excess mucus production. The 3% PEO volume mimicked these conditions; by almost oversaturating the model, there was evidence of PEO drainage. The asymptomatic case does not present the evidence of "runny nose" symptoms, with no drainage of PEO. 17 The volume to saturate the nasal cavity with asymptomatic and symptomatic ranged from 0.8 to 1 mL for both conditions; approximately 0.1 mL of each condition was added to the model after each swab to ensure continuous saturation.

Silk Sponge Morphological Characterization. The morphology and distribution of construct pores were Swab Pickup Quantification. To quantify swab uptake, the nasal in vitro tissue model was saturated with the artificial mucus matrix, and the following swabbing procedure was performed in accordance with CDC guidelines. 18 NP swabs were inserted into the nasal cavity until resistance was encountered, while the MT swab was inserted to the midway point, about 1.5-cm-deep. Both swabs were twisted around the surfaces five times, held in place for 15 s, and then removed. Each swab was then placed into phosphate buffer solution (1 × PBS) (VWR Scientific, Radnor, Pennsylvania) for further processing. In addition, we compared the swabbing workflow with the nasal in vitro tissue model (MODEL method) against the current gold-standard benchtop swab validation procedure, 1, 4, 8 which involves sequentially dipping swabs into tubes with relevant solutions (TUBE method).

The pickup swab quantification was performed by gravimetric analysis for injection molded-MT and injectionmolded-NP swabs in comparison with the commercially available flocked swab. The weight of each swab (N = 5) was recorded before and after the MODEL or the TUBE methods. Results were reported as mass uptake for three independent experiments.

Swab Release Quantification. To quantify swab release, we carried out two independent investigations. In order to efficiently assess cellular material uptake, we loaded the synthetic mucuses with 80% v/v fluorescently labeled microparticles (10 μm) to mimic cellular particulates into the artificial nasal solution. FITC-labeled microparticles (Sigma-Aldrich, St. Louis, MO), based on melamine resin, were homogeneously added to the 0.5% and 3% w/v PEO solution. The solution was then dispensed into the tissue model and allowed to saturate the silk sponges. The above-mentioned swabbing procedure was performed. Each swab (N = 5) was then removed and placed in a 1 mL volume of 1× PBS. Then, 100 μL aliquots were taken in triplicate and analyzed with a SpectraMax M2 plate reader at 490 nm excitation and 525 nm emission. A fluorescence signal was then reported as an expression of cellular-mimicking uptake.

To further assess swab uptake and the release of viral material, the nasal model was saturated with both nasal solutions spiked with heat-inactivated SARS-CoV-2 virus and strain USA-WA1/2020 (NR-52286, BEI Resources, ATCC, USA), and the swabbing procedure was performed, as described above. To investigate the effect of viral load on swab performance, the mucus was spiked with three different concentrations of inactivated virus (10 7 , 10 6 , and 10 5 copies/ mL). After the procedure, each swab was removed and placed into a tube with 350 μL of 1 × PBS. The vial with the swab was then vortexed for 30 s. Five microliters from each sample was then tested to quantify the detection of SARS-CoV-2. To evaluate the presence of SARS-CoV-2, we performed the CDC 2019-Novel Coronavirus (2019-nCoV) Real-Time RT-PCR Diagnostic Panel (https://www.fda.gov/media/134922/ download), per the manufacturer's instructions using the 2019-nCoV_N2 combined primer/probe mix with Quantabio Ultraplex One-Step RT-qPCR ToughMix. Amplification was performed following the manufacturer's instructions with a QuantStudio 5 Real-Time PCR System (Thermo Fisher Scientific, Waltham, MA, USA). The results for each swab (N = 5) were reported as cycle threshold (Ct) values.

Statistical Analysis. Statistical analysis was performed using a Student's t test (t test) and analysis of variance (ANOVA) single factor with a p value of <0.05 using Origin (Pro), version 2021b (OriginLab Corporation, Northampton, MA, USA). A t test was performed when comparing paired injection molded to standard flocked swabs in a gravimetric analysis, quantitative release, and RT-qPCR. ANOVA was performed to investigate the effect of swab type, mucus, and viral load.

Physical Characterization of Asymptomatic and Symptomatic Nasal Fluids. Viscosity analysis was performed to identify conditions for the symptomatic and asymptomatic physiological viscosities of the nasal mucus. The asymptomatic and symptomatic viscosities were mimicked by using PEO at 0.5% and 3% w/v, respectively. Both mucuses presented a shear thinning behavior, and the viscosity values were in the physiological range 7.61 ± 0.53 mPa·s for the asymptomatic and 2522 ± 243.3 mPa·s for the symptomatic (Figure 3) . Quantification of Swab Pickup and Release. The gravimetric analysis was conducted to understand mucus pickup, expressed as a difference in mass, between injectionmolded-NP and injection molded-MT and flocked swabs. The tissue-model analysis showed that the flocked-NP picked up 1.3 times more 3% w/v PEO than the injection-molded-NP swabs and picked up 4.1 times more 0.5% w/v PEO, while the flocked-MT swabs picked up 1.5 times more 3% w/v PEO than the injection molded-MT and 4.5 times more 0.5% w/v PEO. As a comparison method, we performed the same analysis by dipping the swabs into the same solution in a tube (TUBE method). The TUBE method showed a 2.9 times increase in pickup of the 3% w/v PEO from injection-molded-NP swabs in comparison to the MODEL collection method, while the flocked-NP swab picked up 2.4 times more. The flocked-NP swab picked up 5.2 times more 0.5% w/v PEO while the flocked-NP swab picked up 1.9 times more 0.5% w/v PEO. The gravimetric analysis concluded that all swabs picked up significantly more mucus in the TUBE method than the MODEL collection method, and that all flocked swabs, except for NP in the TUBE method in 3% w/v PEO, picked up significantly more mucus than the injection molded swabs across both methods ( Figure 4A) .

The release quantification of FITC-labeled microparticles was performed to efficiently mimic cellular uptake and subsequently correlate with cellular material release via RT-qPCR analysis. The injection-molded-NP swabs released 2.6 times more microparticles than the flocked swabs in 3% w/v PEO, while in 0.5% w/v PEO the injection-molded-NP swabs released 3.2 times more microparticles. Overall, the injectionmolded-NP and injection-molded-MT swabs released statistically significantly more microparticles in both asymptomatic and symptomatic conditions in comparison to flocked swabs ( Figure 4B) .

Quantification of Swab Performance. Bench-top validation of swab performance was performed on both injection molded and flocked swabs with the nasal in vitro tissue model saturated with synthetic mucus, under symptomatic and asymptomatic conditions, spiked with different loads of SARS-CoV-2 heat-inactivated virus, in an effort to encompass clinical variability. The Ct values for all of the swabs were compared across the three virus concentrations for both mucus viscosities. Our analysis showed, as expected, that as the concentration of the virus increased there was also a decrease in Ct values. In fact, there was a 4.48 Ct decrease in the injection-molded-NP swabs in 3% w/v PEO from 10 7 to 10 5 copies/mL, and a 4.04 Ct shift in the flocked-NP swabs under the same conditions. Injection-molded-NP swabs with 10 7 copies/mL in 0.5% w/v PEO, instead, showed a 5.33 higher Ct value than the 3% w/v PEO Ct. Furthermore, all injection molded and flocked swabs loaded with 10 5 copies/ mL of virus in 0.5% w/v PEO showed a statistical difference between the paired swab groups in favor of injection molded swabs. For 10 6 copies/mL, the NP and MT swabs in 0.5% w/v PEO were statistically different, but in favor of the flocked swabs. With 10 7 copies/mL, only the injection-molded and flocked-MT swabs were statistically different in 0.5% w/v PEO. For the symptomatic mucus (3% w/v PEO), the MT swabs were statistically different only when they were loaded with 10 6 or 10 5 copies/mL, indicating that the injection-molded-MT swabs perform better at a lower virus concentration compared to the flocked swabs ( Figure 5 ).

The increasing number of COVID-19 positive cases in the United States led companies to develop new swabs to overcome pandemic associated testing bottlenecks. However, new swabs need to undergo extensive validations prior to reaching the market. To support initial preclinical validation, we developed an in vitro 3D-printed nasal tissue model that recapitulates key features of the nasal cavity (i.e., architecture, soft tissue, and mucus viscosity). Alternative strategies aimed to create tissue models that mimicked the paranasal sinuses and skull and were flexible enough to guide surgeons in their preoperative practice simulations. 19 Although useful, those models were meant to be a valuable tool for surgeons and not for testing and research purposes. For this reason, we have developed a model that could efficiently and safely support the research and development stage of medical devices to streamline the validation of new swab prototypes and increase their clinical relevancy before clinical studies.

Our initial anterior nasal passage model consisted of silicone tubing lined with a cellulose sponge with sizes only compatible with the human nostril. 5, 18, 20 This model was subsequently modified with the proposed 3D-printed tissue model, to replicate the entire structure of the midturbinate and nasopharyngeal walls of the nasal cavity. The degree of precision achieved in the reproduction of the nasal cavity in our 3D model was accomplished by averaging the computed tomography (CT) scans of 30 healthy patients' nasal cavities provided by the Aerosol Research Lab at Carleton University in Ottawa, Canada. 11 The model was then lined with a silk sponge, as a replacement for the cellulose sponge to better mimic soft tissue mechanical properties, and with synthetic mucus fluids to resemble both symptomatic and asymptomatic fluid viscosities. Silk protein was chosen because of its structural and mechanical properties and inertness. Silk is a versatile, biocompatible, and biodegradable material with tunable mechanical properties and is extensively used in tissue engineering for mimicking soft and high-strength human tissues. 7,12,21−23 The compressive modulus for nasal tissue falls between 0.44 and 0.97 MPa, and 4% w/v lyophilized silk sponges are within a comparable range. 12, 24 The cellulose sponge from the original model has a greater pore size and density than the silk sponges and requires more PEO to become fully saturated. In addition, silk processing can be easily tuned to vary the compressive moduli and porosity of soft tissues, 12 if needed. In our model, silk was used in a sponge format to replicate the soft architecture of the nasal tissue with a controlled pore size. Lastly, to mimic the nasal mucus, we utilized PEO, a hydrophilic polymer with physical and mechanical properties that can be tuned based on its molecular weight. 25 Due to its viscoelastic properties, PEO at different concentrations creates a range of viscous solutions that can be used to mimic physiological mucus. 26 Another advantage to using PEO is its compatibility with biomolecular assays; in fact, PEO does not interfere with RT-qPCR amplification at low viscosities compared to other viscous body fluids. 27 To match the viscosity of the nasal fluid in asymptomatic and symptomatic conditions, we tested several PEO concentrations finding that 0.5% and 3% w/v were compatible with the physiological range of nasal fluid viscosities. 15 In general, human mucus viscoelasticity is characterized by a shear thinning behavior with a viscosity range between 10 and 10 6 mPa·s. 15 Furthermore, low and high mucus viscosities have been associated with asymptomatic and symptomatic nasal mucus viscosities (∼13 and 1400 mPa·s) in artificial mucus compositions. 28 Our synthetic mucus formulation confirmed the shear thinning behavior 16 and matched the viscosity range for both nasal fluid conditions (Figure 3 ). Due to the variation in viscosity between the two solutions and limitation of the instruments (i.e., spindle size, torque %, and fixed revolutions per minute (RPM) range values), different shear rates were tested for the two conditions, and fewer points were analyzed for the lower viscosity (0.5%), albeit sufficient to characterize the artificial nasal matrix behavior. Further analysis will be conducted in the future with more sensitive instruments to obtain a more precise shear−viscosity curve. To support our model as a suitable tool for swab validation, we developed and performed several assays to assess swab pickup and release efficiencies and then evaluate data agreement against the available literature. In addition, an in vitro tissue model would provide more controlled experimental conditions in comparison to clinical studies that have shown greater variability, arising from differences in sampling methods, nasal cavity structure, nasal fluid viscosity, and other conditions that vary from patient to patient and season to season. 3, 4 Moreover, the disadvantages of relying on clinical trials for initial swab validation are also the bureaucratic aspects (i.e., recruitment, regulatory requirements, and cost). Thus, the tissue model would be a great tool to support initial research and development explorations with clinical relevancy for swab design and optimization. Data from clinical trials showed that injection molded swabs perform similarly to flocked swabs, and those results are comparable with our previous findings. 3 −5 In fact, the IM2 injection molded swabs had an agreement of 96% with the FLOQ standard flocked swab during clinical studies, as also supported by our findings with the in vitro tissue model. 3 Injection molded swabs pick up significantly less viral material than standard flocked swabs, but the difference is offset by greater release ability of the injected molded swabs, mainly due to the hydrophobic and nonabsorbent head compared to the high retention flocked swabs. Taken together these results explain why injection molded swabs perform better or the same as flocked swabs when detecting SARS-CoV-2 virus in RT-qPCR.

We initially quantified swab pickup in our model via gravimetric analysis and subsequently quantified viral release via RT-qPCR. Current methodologies simulate the specimen collection by dipping the swab into a spiked COVID-19 negative nasal fluid or water and estimating the swab pickup by measuring, pre and post, its weight or volume (TUBE method). However, such analyses can lead to misleading results when analyzing the data, 1,4 because they replicate neither the nasal architecture and physiological fluids nor the actual collection procedure; furthermore, when comparing swab typologies, data in the literature reported that flocked swabs dipped in water pick up 10.7 times more water than the injection molded swab due to their difference in geometry, material, and device fabrication. 4 Another discrepancy in the TUBE method as a workflow to correctly assess pickup and release when using contrived samples is the viscosity of the solution used. For example, a PurFlock Ultra flocked swab picked up 6.3 × 10 4 copies of viral material from a tube method 1 , while a flocked swab picked up 1.6 × 10 4 copies from our model with 0.5% w/v PEO. Our analysis in fact showed that the standard NP picked up 1.9 times more mucus (0.5% w/v PEO) from the TUBE method than the MODEL method. This suggests that the dipping TUBE model allows for a much greater absorption of liquid from the swab, and therefore more viral material, in comparison to a standard swabbing procedure, introducing artifacts in the data collection ( Figure 4) . Not only did all swabs pick up more mucus in the TUBE method compared to the MODEL method but the flocked swabs also picked up more or the same amount of mucus compared to the injection molded swabs in both collection methods. This has also been shown in other studies, where injection molded swabs pick up significantly less mucus with a more efficient release than flocked swabs. 4, 5 The MODEL method incorporates porous silk sponges completely saturated with PEO, similarly to how the human nasal cavity is saturated with fluid. 29 The TUBE method, in comparison, recreates a dipping procedure with a large excess of fluid. This contributes to the large offset in uptake registered for the flocked swabs that have a greater absorption capacity in comparison to solid head swabs. This supports the importance in replicating the native architecture and tissue saturation in the validation process.

On the other hand, a release quantification analysis needs to be performed in order to obtain reliable and consistent molecular data for the virus detection. This type of analysis has been done in the past with different microorganisms, where contrived biological samples were pipetted onto swabs, eluted in a buffer, and then processed for RT-qPCR. 30 This method, however, does not actually replicate the specimen collection and release process that, again, are dependent on the swab pickup features as well as the anatomical structure and geometry of the patient cavities. Our data, in fact, showed that the injection molded NP swabs released more microparticles compared to the standard swabs which can be attributed to their geometry and structural properties. In fact, flocked swabs do not release as many microparticles as the injection molded swabs due to their nylon fibers, which are meant for maximum absorbance and high retention. On the contrary, the hydrophobic plastic nature of the injection molded swabs allows them to release most of the sample they collect 4 (Figure 4 ). The 10 μm in diameter fluorescently labeled microparticles are an accurate representation of viral nasal epithelial cells that a swab might pick up during testing in the nasal cavity. 31 This supports the use of these particles for release quantification of each swab.

As a final test, we concluded the validation of our model by performing RT-qPCR. We initially tested the limit of detection of the model by spiking the synthetic nasal mucus with 10 5 , 10 6 , and 10 7 copies/mL of heat inactivated SARS-CoV-2 virus. In general, clinical studies typically compare Ct values to a control since the viral load is unknown during the specimen collection and diagnostic process. Our analysis demonstrated that when the model is spiked with decreasing concentration of COVID-19 virus, the Ct values were acceptable among all of the conditions except for 10 5 in 0.5% w/v (38 Ct); however, according to the World Health Organization (WHO), a Ct value between 37 and 40 is at the acceptance limit. 32 The same behavior was evident also when the virus was spiked at 10 7 copies/mL in the 0.5% w/v PEO in which there was a difference of five cycles for injection-molded-NP and four cycles for flocked-NP, compared to 3% w/v PEO, even if the values are in the acceptance range (Ct ≤ 37). Those differences must be caused by the lower viscosity of the mucus, which causes a major dispersion of the virus in the sponge and, consequently, a lower pickup or release by the swab (Figures 5 and S1 ). Due to the greater viscosity of the symptomatic mucus and drag experienced by the particles, the movement of the viral material is severely limited compared to a less viscous solution. This reduces the absorption of viral particles into the porous silk sponge and allows the swab to pick up more PEO and viral material. 33 In addition, injection molded swabs confirmed comparable performance to flocked swabs for higher viral loads, while they outperform flocked swabs in all symptomatic conditions. This supports the importance of replicating the physical properties of the native tissue in the validation process.

We demonstrated that physiologically relevant tissue models can serve as powerful tools for medical device validation, significantly reducing time and finances required for the preclinical stage. In vitro testing eliminates the excess cost and time spent on adjusting prototypes during clinical investigations. This model provides a more effective and accurate way to validate swabs on the bench, as the current swab preclinical validation method does not consider the physiological architecture of the nasal cavity as well as mucus characteristics. The physiologically accurate architecture, soft tissue properties, and fluid viscosity should be taken into consideration for future in vitro model design for medical device validation, which could potentially go beyond swab validation. As validation techniques mimic the human body more and more, major design refinements of medical device prototypes could be streamlined in the preclinical phase.

A global shortage of collection specimen swabs has been among the several bottlenecks in COVID-19 testing, during the 2020 pandemic. Several companies have created new injection molded swabs that can be mass produced quickly and cost efficiently. To validate these swabs, we have developed an in vitro tissue model of the human nasal cavity. This model accurately mimics the architecture and structure of the cavity and is lined with silk sponges to resemble the nasal soft tissue. An artificial mucus was also developed from PEO to replicate two different physiological conditions, asymptomatic and symptomatic nasal fluid viscosities. This model was used to validate a new injection molded swab and to provide comparable RT-qPCR results to a standard flocked swab, showing the importance of replicating physical and structural features of the native tissue as part of the validation process. Not only does this model provide more reliable results in comparison to standard dipping preclinical validation methods, but the accessibility of fused deposition modeling printers and the relatively low cost of PEO, silk, and ABS would provide a time-and cost-effective tool for medical device validation. 

Comparison analysis of different swabs and transport mediums suitable for SARS-CoV-2 testing following shortages

Nasal swab sampling for SARS-CoV-2: a convenient alternative in times of nasopharyngeal swab shortage

Clinical Diagnostic Study of a Novel Injection Molded Swab for SARS-Cov-2 Testing

Accessioning and automation compatible anterior nares swab design

Preclinical Validation of a Novel Injection-Molded Swab for the Molecular Assay Detection of SARS-CoV-2

Multilayered dense collagen-silk fibroin hybrid: a platform for mesenchymal stem cell differentiation towards chondrogenic and osteogenic lineages

Coculture of dorsal root ganglion neurons and differentiated human corneal stromal stem cells on silk-based scaffolds

The influence of a swab type on the results of point-of-care tests

Comparison of nasopharyngeal nylon flocked swabs with universal transport medium and rayon-bud swabs with a sponge reservoir of viral transport medium in the diagnosis of paediatric influenza

Open Development and Clinical Validation of Multiple 3D-Printed Nasopharyngeal Collection Swabs: Rapid Resolution of a Critical COVID-19 Testing Bottleneck

Creation of a standardized geometry of the human nasal cavity

Lyophilized silk sponges: a versatile biomaterial platform for soft tissue engineering

Materials fabrication from Bombyx mori silk fibroin

Acellular bi-layer silk fibroin scaffolds support functional tissue regeneration in a rat model of onlay esophagoplasty

Micro-and macrorheology of mucus

Rheological characterization of poly (ethylene oxide) solutions of different molecular weights

Upper respiratory viral load in asymptomatic individuals and mildly symptomatic patients with SARS-CoV-2 infection

Interim Guidelines for Collecting, Handling, and Testing Clinical Specimens for COVID-19

3D Rapid Prototyping for Otolaryngology-Head and Neck Surgery: Applications in Image-Guidance, Surgical Simulation and Patient-Specific Modeling

Review of Computer-Aided Sinus Surgery

Silk as a biomaterial

Functionalized 3D-printed silk-hydroxyapatite scaffolds for enhanced bone regeneration with innervation and vascularization

Assembly and Application of a 3D Human Corneal Tissue Model: A Step-by-Step Guide

Biomechanical characterisation of the human nasal cartilages; implications for tissue engineering

Applications of poly (ethylene oxide) in controlled release tablet systems: a review

Method of Providing a Synthetic Mucus in vivo

Detection of respiratory viruses in sputum from adults by use of automated multiplex PCR

Development of a physiologically relevant dripping analytical method using simulated nasal mucus for nasal spray formulation analysis

A numerical simulation of air flow in the human respiratory system for various environmental conditions

Swab sample transfer for point-of-care diagnostics: characterization of swab types and manual agitation methods

Culture and comparison of human bronchial and nasal epithelial cells in vitro

Visualization of particle/liquid movements in high viscous fluids during end-over-end rotation

The authors declare no competing financial interest.

This study was funded by the NIH RADx-Tech program under 3U54HL143541-02S1 and 3U54HL143541-02S2. The views expressed in this manuscript are those of the authors and do not necessarily represent the views of the National Institute of Biomedical Imaging and Bioengineering; the National Heart, Lung, and Blood Institute; the National Institutes of Health; or the U.S. Department of Health and Human Services. We would like to thank M2D2 for 3D printing access and Professor F. Omenetto and graduate student N. A. Ostrovsky-Snider (Tufts University) for helping us with the viscosity measurements. We would also like to thank the Aerosol Research Lab at Carleton University in Ottawa, Canada for the .stl file of the nasal cavity model.