Greater than 3-Log Reduction in Viable Coronavirus Aerosol Concentration in Ducted Ultraviolet-C (UV-C) Systems.

Control technologies to inactivate airborne viruses effectively are needed during the ongoing SARS-CoV-2 pandemic, and to guard against airborne transmitted diseases. We demonstrate that sealed UV-C flow reactors operating with fluences near 253 ± 1 nm of 13.9-49.6 mJ cm-2 efficiently inactivate coronaviruses in an aerosol. For measurements, porcine respiratory coronavirus (PRCV) was nebulized in a custom-built, 3.86 m wind tunnel housed in a biosafety level class II facility. The single pass log10 reduction of active coronavirus was in excess of 2.2 at a flow rate of 2439 L min-1 (13.9 mJ cm-2) and in excess of 3.7 (99.98% removal efficiency) at 684 L min-1 (49.6 mJ cm-2). Because virus titers resulting from sampling downstream of the UV-C reactor were below the limit of detection, the true log reduction is likely even higher than measured. Comparison of virus titration results to reverse transcriptase quantitative PCR and measurement of fluorescein concentrations (doped into the nebulized aerosol) reveals that the reduction in viable PRCV is primarily due to UV-C based inactivation, as opposed to physical collection of virus. The results confirm that UV-C flow reactors can efficiently inactivate coronaviruses through incorporation into HVAC ducts or recirculating air purifiers.

Introduction

SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2), the causative agent of COVID-19, may be spread by both direct aerosol routes (through direct inhalation of infectious particles[1]) and indirect aerosol routes (through deposition on surfaces). Coronaviruses can remain viable in an aerosol,[2,3] particularly at temperatures and relative humidities commonly encountered in commercial and residential buildings, and both viral RNA[4−6] and viable SARS-CoV-2[7] have been detected in the aerosol of hospital rooms in appreciable concentrations. This indicates the need for technologies that directly mitigate infectious virus present in an aerosol. Ultraviolet germicidal irradiation, that is, 100–280 nm wavelength UV–C light, has a more than 80 year history of application in preventing airborne disease transmission[8−16] with an emphasis on health care and school settings, where there is potential for a single infected individual to infect multiple individuals. UV–C photons directly damage nucleic acids, with a peak effectiveness near 260 nm.[17] Through comparative studies, UV–C light has been shown particularly effective in the inactivation of larger genome, single stranded RNA viruses.[18−21]

UV–C irradiation is hence promising toward the inactivation of coronaviruses.[22,23] A recent report[24] indeed confirms that UV–C light (207–222 nm) at a fluence of 1.7 mJ cm–2 leads to a 3-log reduction in human coronaviruses alpha HCoV-229E and beta HCoV-OC43. However, the method of UV–C implementation proposed was a low irradiative flux dosage applied over tens of minutes, since low doses are required in order to minimize human exposure to UV–C irradiation. An alternative to low UV intensities for longer durations is higher intensities with short duration via incorporation of UV–C systems directly into ventilation ducts[25] or into ducted, sealed air purification units (flow tube reactors). Along these lines, a number of groups have investigated the efficacy of flow tube reactors in inactivating aerosolized viruses.[16,23,26−29] Such sealed systems can be operated with much higher irradiation intensities as they minimize human exposure by design. However, more recent investigations of duct systems have focused on laboratory-scale flow rates of 0.2–4 L min–1[29] and 8.5–60 L min–1.[23,26−28] Flow rates need to be in excess of 500 L min–1 for air purification systems (even in small systems), and certainly need to exceed 1000 L min–1 to be of utility in ventilation systems (where multiplexing would still be needed). In 1964, Jensen[16] showed that a UV–C flow tube system operating at flow rates of 2830 L min–1 to 5660 L min–1 with fluences near 1.9 mJ cm–2 enabled inactivation of adenovirus, influenza A virus, coxsackievirus B-1, sindbis virus, and vaccinia virus in aerosols, with an inactivation efficiency in excess of 90% (log reduction of 1.0) and beyond 99.99% in some instances (log reduction >4). Jensen’s system, which was implemented in health care settings,[15] does not appear to have been previously tested with coronaviruses, and since its presentation, does not appear to have been replicated in the published literature. Nonetheless, the results shown with it suggest that high flow rate UV–C duct systems can be developed toward highly efficient inactivation of viruses, including SARS CoV-2. Motivated by the potential use of high flow rate duct UV–C systems toward SARS-CoV-2 inactivation, using porcine respiratory coronavirus (PRCV, an alpha coronavirus) as a surrogate for SARS-CoV-2, in the present study we examine the efficiency of a relatively compact, low pressure Hg lamp (monochromatic UV–C light) flow tube reactor tested at flow rates applicable to air purification and ventilation systems. In the subsequent sections, we describe the custom wind-tunnel designed to test the ducted UV–C system, measurement of the UV–C irradiation intensity distribution, system calibration, and inactivation studies with PRCV, with corresponding upstream and downstream sampling and measurement via virus titration, fluorescein dopant detection, and reverse transcription quantitative polymerase chain reaction (RT-qPCR). While we believe the results to be rather straightforward and anticipated based on prior successful implementation of UV–C irradiation in virus inactivation, explicit efforts are made in this manuscript to explain all measurement and data analysis methods in detail, as there is clearly increased importance in establishing proven technologies for airborne coronavirus removal and inactivation applicable to large volumes of air, or at flow rates higher than traditional bench scale experiments.

Materials and Methods

UV–C Duct System

The UV–C duct system, manufactured by Novisphere LLC (Elk Grove Village, IL) is a 114.3 cm long, 30.5 cm diameter duct. Three low pressure UV–C Hg lamps, each of 60 W power were installed within the duct, aligned parallel to the direction of the flow. Upstream of the duct system we placed a MERV (minimum efficiency reporting value) 4 prefilter to remove any large particles (>10 μm) that may have been produced during tests. The lamps create an irradiated zone, which is nominally 29.8 cm in diameter and 72.4 cm in length. The ozone concentration in the unit operating at a flow rate of 2349 L min–1 was 10.85 ppb (examined with a model 49 C Ozone Analyzer, Thermal Environmental Instruments Inc.), compared to 11.23 ppb naturally in the laboratory where measurements were conducted. Irradiative flux distributions were measured in the axial center of the irradiated zone in the 200–850 nm range at five radial locations using an Ocean Optics USB 2000+ UV–vis spectrometer (White Bear Photonics, LLC, White Bear Lake, MN). Aside from the inlet and outlet, the duct system was sealed via silicone caulk and foam gasket.

Wind Tunnel Tests

The work conducted in this study followed the protocols approved by the University of Minnesota Institutional Biosafety Committee (IBC) protocol number 1808–36316H. To examine the efficacy of the UV–C duct unit in inactivating aerosolized coronaviruses, we constructed a 30.5 cm diameter, 3.86 m long wind tunnel unit housed within a biosafety level (BSL) class II facility at the University of Minnesota Veterinary Isolation Facility (Saint Paul, MN). A schematic diagram of the wind tunnel is depicted in Figure a, with an isometric CAD rendering shown in Figure b, and an image of the wind tunnel in the isolation facility shown in Figure c. The wind tunnel, made of galvanized steel, was operated as follows. Beginning at the upstream, air at the desired flow rate (controlled downstream) was actively pulled through a HEPA filter (Air Filters, Inc. model MPH24242MCUG2, Richmond, VA) into the wind tunnel. Close to the duct inlet in the present study we nebulized PRCV laden suspensions, under the conditions prescribed subsequently. The nebulizer employed was a large particle generating (LPG) air-jet nebulizer, used in recent studies of virus inactivation by nonthermal plasma systems.[30,31] Suspensions were fed into the nebulizer using a syringe pump (New Era Pump Systems, Inc., Farmingdale, NY). Nebulized suspensions were not actively dried in the wind tunnel since we found that the volumetric flow rate of air employed was sufficient to facilitate droplet drying prior to upstream and downstream measurements, with the relative humidity remaining in the 57–62% range throughout testing. The nebulizer was operated using an air compressor (Porter Cable, Jackson, TN) with a backing pressure of 138 kPa, leading to a carrier flow rate of 1.5 standard L min–1 injected into the wind tunnel. The carrier gas flow was bipolarly ionized using a corona discharge (created by a Simco-Ion model 91–6110, Alamedi, CA) in order to charge-neutralize the generated aerosol particles. While corona discharges commonly generate ozone, we found that the ozone concentration measured (with a Thermo Environmental Solutions model 49C) directly at the outlet of the ionizer was 184 ppb; this concentration was present within the 1.5 standard L min –1 flow and hence was significantly more dilute once introduced into the primary wind tunnel flow. Furthermore, in an independent testing (following a similar protocol to the present study), we found that this had no influence on the viability of aerosolized coronaviruses. Downstream 43.8 cm of the nebulizer, the aerosol passed around a custom-made mixing plate, which is a combined thin plate orifice with an opening diameter of 15.24 cm, followed by a thin plate disc downstream of the orifice, with a diameter of 7.62 cm. 2.4 duct diameters (73.2 cm) downstream of the first mixing plate, an upstream sampling probe with a diameter of 3.47 cm was placed at the center of the duct and oriented to directly sample the flow with an Andersen cascade impactor (equivalent to a BGI Inc. nonviable Andersen cascade-impactor) operated with stages 0, 5, 6, and the total filter installed. The impactor was operated via a vacuum pump set at a flow rate of 90 standard L min–1 where the 50% cut sizes of stages 0 and 5 are 5.2 and 0.22 μm, respectively. While the stage 6 cut size of the Andersen impactor is not known at this flow rate,[32] we estimated that we collected all particles above 200 nm efficiently, and in monitoring the virus titer on a filter downstream of the impactor (as noted subsequently), we never recovered any viable virus, suggesting that all virus-carrying particles were sampled by the impactor stages utilized. Andersen impactors were selected as the samplers to maximize the sampling flow rate (in comparison to more efficient, but lower flow rate samplers[33,34]) while maintaining the ability to collect particles near the size of single virions (which commonly available liquid impingers cannot[35,36]). Velocity traverse measurements at the sampling location in the 580–2350 L min–1 range (using a TSI Inc. model 9545 VelociCalc, Shoreview, MN) where the flow rate was turbulent, revealed a radial coefficient of variation below 0.11 for the velocity. For the flow rate range examined in this study, the sampling probe was not isokinetic; the sampling flow velocity was higher than the mean velocity in the tunnel. However, as the majority of virus-laden particles were below 5 μm in diameter (Supporting Information (SI)) and results are presented as downstream/upstream sampling ratios, we do not believe nonisokinetic sampling had a major influence on the presented results. For the mixing plate and sampler configuration utilized at the sampling location, using an optical particle sizer (TSI Inc. model 3330), we found that the coefficient of variation on the particle concentration was less than 0.15 across the diameter of duct for particles 5 μm and smaller in diameter.

Figure 1

(a). A schematic diagram of the test system used in the UV–C duct system evaluation. The sampling flow rate shown of 90 L min–1 is the default setting. (b). An isometric view of the test system with dimensions labeled. (c). Photograph of the system within the biosafety level class II facility.

The UV–C duct unit was connected in the wind tunnel after the upstream sampler. Subsequent to the UV–C test unit, the aerosol passed through a second, identical mixing plate and to a downstream sampling probe (86.8 cm from the mixing plate), of equal size to the upstream sampler also connected to an Andersen impactor operated under identical conditions to the upstream impactor. Correlation testing of particle concentrations using the optical particle sizer revealed downstream/upstream particle concentration ratios between 0.99 and 1.05 for particles in the 300 nm to 4.75 μm range across the flow rate range examined in the absence of the test UV–C duct. Following the downstream sampler, the flow passed through a 6.27 cm thin plate orifice, across which the static pressure was monitored, and which was calibrated (using a dry gas meter) to yield the wind tunnel flow rate. Following the orifice meter, the flow passed through a second HEPA filter and to vacuum blower controlled by a variable transformer. Target volumetric flow rates in the wind tunnel were achieved by adjusting the pump voltage to attain a desired static pressure drop across the orifice meter and then accounting for the downstream sampler flow.

Porcine Respiratory Coronavirus Challenges

Porcine respiratory coronavirus (ATCC VR-2384) was grown and titrated in Sus scrofa testis (ST) cells. The ST cells were grown in Dulbecco’s Modified Eagle Medium (DMEM, Life Technologies, Grand Island, NY) supplemented with 10% fetal bovine serum (FBS, Life Technologies, Grand Island, NY), and 1× antibiotic-antimycotic (Life Technologies, Grand Island, NY). PRCV was propagated in ST monolayers between 80 and 90% confluency. The monolayers were washed with phosphate-buffered saline and infected with PRCV. The infected cells were maintained in DMEM supplemented with 2% FBS and 1× antibiotic-antimycotic and incubated at 37 °C with 5% CO2 until cytopathic effects (CPE) appeared, within 2–3 days. The cultures were frozen at −80 °C, thawed once, centrifuged at 3000g for 10 min, titrated, aliquoted into 21 mL volumes, and frozen at −80 °C until used in the experiment.

PRCV suspensions were loaded into the syringe pump and nebulized at a liquid feed rate of 3 mL min–1. The size distribution function of nebulized particles, in the absence of virus but with all other suspension components, was measured using a scanning mobility particle sizer (TSI model 3034) as well as a TSI aerodynamic particle sizer (TSI model 3321). Size distribution functions by number and by volume are plotted in the SI, revealing a mode diameter by number below 100 nm, and near 3 μm by volume (see SI Figure S1). As shown in prior studies,[35,37] the presence of viruses has a negligible influence on nebulized particle size distributions, which is governed by the produced droplet diameter and the volume fraction of nonvolatile material (of which viruses constitute a negligible fraction). Fluorescein was added to the PRCV suspensions to a concentration of 300 μg mL–1. The fluorescein was used as a physical tracer to determine the physical penetration of particles through the system. Nebulization and sampling were carried out in triplicate, both with and without the UV–C duct system installed. We operated at system flow rates of 684, 1674, and 2439 standard L min–1 with the UV–C duct system installed (corresponding to UV–C exposure times of 5.1, 1.9, and 1.3 s, respectively), and 684, 1505, and 2409 standard L min–1 without the UC-V system (correlation tests). Each replicate test consisted of 30 min of nebulization with samples collected in both the upstream and downstream samplers, with two exceptions. For the 684 L min–1 tests (both with and without the UV–C duct system), and the 2409 L min–1 correlation test (without the UV–C system), each replicate consisted of 20 min of nebulization. Furthermore, for the 2409 L min–1 correlation test, we used stages 0, 5, 6, and 7 of the Andersen impactor, which has 50% cut sizes of 6.5, 0.55, and 0.26 μm for stages 0, 5, and 6, respectively, and a lower unspecified cut size for stage 7, at a sampling flow rate of 60 L min–1. In total, measurement condition required 180–270 mL of ∼107 TCID50 mL–1 PRCV suspension; with 1.35 L of 107 TCID50 mL–1 PRCV suspension required for study completion. All upstream and downstream samples for a given flow rate were collected using the same virus suspension in the nebulizer, enabling results to be pooled as they were collected under identical conditions.

The wind tunnel was first checked daily to ensure air was entrained into the unit from the upstream HEPA filter with the vacuum blower operating, such that in the event of an accidental leak, air would penetrate into the unit and not allow any virus aerosol to penetrate into the room. For each test, the compressed air pump was turned on first, which was regulated at 138 kPa and the dispersion air flow applied to the LPG nebulizer was monitored by a flow meter at 1.5 L min –1. The vacuum blower was turned on after initializing the nebulizer and set to the desired system flow rate, minus the flow rate of the downstream sampling pump. Sampling pumps were then turned on. Samples from Andersen impactor plates were collected using a cell scraper and 3 mL per stage of the collection liquid, DMEM supplemented with 2% FBS and 1× antibiotic-antimycotic (100 units mL–1 of penicillin, 100 μg mL–1 of streptomycin, and 0.25 μg mL–1 of Gibco Amphotericin B). The final stage filter of each impactor was also eluted with 15 mL of the collection liquid. Viral particle recovery from the final stage filter was increased by adjusting the pH of the collection liquid to pH 9 using 1 M NaOH solution.[38] After vortexing at 3,000 rpm for 2 min, the filter fibers were separated by centrifugation at 3000g for 20 min, then the pH of the decanted supernatant containing the eluted PRCV was readjusted to pH 7 with 5 M HCl solution. After sampling, all liquid samples were divided into four aliquots of ∼0.5 mL in 1.7 mL sterile microcentrifuge tubes. Two hundred μL of eluted sample from each impactor stage for upstream and downstream collections were pooled separately. All collected individual and pooled samples were stored at −80 °C until testing. Pooled Andersen stages and final stage filter samples were used to measure upstream and downstream concentrations of (1) viable viruses, carried out by virus titration, (2) PRCV nucleic acid, carried out by RT-qPCR, and (3) fluorescein, carried out by fluorimetry. However, as final filter samples did not yield detectable viable concentrations of PRCV, we do not utilize these results in discussing UV–C duct performance.

Viable PRCV from pooled Andersen stages were titrated in ST cells on the same day as collection, using the 50% tissue culture infectious dose (TCID50) method. Seven serial 10-fold dilutions (10–1 to 10–7) of each sample were prepared in DMEM, supplemented with 2% FBS and 1X antibiotic-antimycotic. Four replicate wells of 1-day-old ST monolayers in 96-wells cell culture plates were infected with 100 μL aliquots of each sample dilution. Cytopathic effects were examined under inverted microscope after 5 days of incubation at 37 °C with 5% CO2. Viral titers were calculated using the Karber method[39] and expressed as TCID50 mL–1.

Viral RNA was extracted from 50 μL of sample using MagMAX-96 Viral RNA Isolation kit (Applied Biosystems by ThermoFisher Scientific, Lithuania) according to the manufacturer’s instructions on a semiautomatic MagMAX Express-96 Deepwell Magnetic Particle Processor (Applied Biosystems by Thermo Fisher Scientific, U.S.). The RNA was eluted in 50 μL of elution buffer and stored at −80 °C until used for viral genome quantification. For the RT-qPCR, we used in-house developed PCR primer set and probe shown in SI Table S1. The RT-qPCR primers were designed to target a conserved 112 bp region, corresponding to the region between nucleotides 84 and 195 of PRCV S gene.[40] The primers and probe were manufactured by Integrated DNA Technologies (IDT Inc., IA). The reactions were performed using AgPath-ID One-Step RT-PCR kit (Applied Biosystems by Thermo Fisher Scientific, Austin, TX). The reaction mixture (25 μL) consisted of 5 μL of template RNA, 12.5 μL of 2X RT-PCR buffer, 1 μL 25X RT-PCR Enzyme Mix, 0.50 μL of 10 μM forward primer (200 nM final concentration), 0.50 μL of 10 μM reverse primer solution (200 nM final concentration), 0.30 μL of 10 μM probe (120 nM final concentration), and 5.20 μL of nuclease-free water. The RT-qPCR was performed in the Applied BioSystems 7500 Fast Real-Time PCR thermocycler system using the following conditions: 45 °C for 10 min, 95 °C for 15 min followed by 45 cycles of 95 °C for 15 s and 60 °C for 45 s. In each run of RT-qPCR, standard curve samples and no template control were used as positive and negative controls, respectively. The PRCV standard/calibration curve was constructed for absolute quantification of viral genome copy number, in which we used serial 10-fold dilutions of a 509 bp RT-PCR purified amplicon of PRCV S gene (including the 112 bp target sequence of the RT-qPCR prime/probe set). The 509 bp PRCV S gene fragment was produced by RT-PCR reaction using an in-house developed primer set shown in SI Table S1. RT-qPCR results were expressed as cycle threshold (Ct) values. The Ct values were used along with the standard curve to calculate the absolute genome copy number of PRCV, expressed as genome copies mL–1.

Fluorimetry was carried out from samples of pooled Andersen impactor plates using a microplate reader (model Synergy H1, BioTek, Winoosk, VT). Samples with a volume of 100 μL each were loaded into 96-well black plates (Thermo Fisher Scientific, Roskilde, Denmark), and sealed with transparent film on the top to avoid evaporation. The fluorescein intensity was measured at excitation and emission wavelengths of 485 and 515 nm respectively, with a gain of 30.

Results and Discussion

Ultraviolet Irradiative Flux Distribution and Fluence

Measured irradiative flux distributions (units of mW cm–2 nm–1) are shown in Figure , along with the relative locations where measurements were made along the cross section of the UV–C duct. Characteristic of low pressure Hg lamps, the flux distributions are monochromatic with a peak at 252.7 nm ±1 nm, close to the expected peak for low pressure Hg lamps. Zoomed-in insets in the 245–260 nm region of the irradiative flux distributions are also shown in Figure . Integration of the distribution over the highlighted band in Figure yields the total flux in this wavelength range; the average of these yields a flux of 11.17 mW cm–2 for the tested unit. With approximate dimensions of 72.4 cm length and 29.85 cm diameter for the UV irradiated region, the flow residence time under UV irradiation is 4.44, 1.81, and 1.25 s for operating flow rates for 684 L min–1, 1674 L min–1, and 2439 L min–1, respectively. Consequently, the fluences (products of residence time and irradiative flux) are 49.63 mJ cm–2, 20.28 mJ cm–2, and 13.92 mJ cm–2 for these three operating flow rates. These fluences are one[29] to two[16,26] orders of magnitude higher than utilized in previously studied duct systems; even before testing, this suggests that (1) the current system will be extremely efficacious in virus inactivation and (2) the system could be operated at even higher flow rates and remain effective. We remark, however, that increasing the flow rate complicates sampling and measurement, as we are not able to sample the entirety of the flow upstream or downstream for measurements and increasing the system flow rate dilutes the viable virus concentration.

Figure 2

Measured irradiative flux distributions at different points within the UV–C duct. Each measurement location is noted in the plot. fwhm: full width at half-maximum (in nm) for the main peak in each flux distribution. Total flux denotes the area under the curve for the highlighted region.

Virus Removal and Inactivation

PRCV, an alpha coronavirus, was chosen as a surrogate for other coronaviruses because it is not known to infect humans and leads largely to asymptomatic infection in swine, it requires BSL class II facilities to work with in experiments (and not BSL class III facilities as SARS-CoV-2 does), and we were successful in propagating the virus to high titers (>107 TCID50 mL–1) in the volumes necessary for the presented experiments (180–270 mL per test). Table summarizes the results from all triplicate experiments, including fluorimetry results, virus titers, and RT-qPCR measurements for impactor plate samples collected upstream and downstream of the UV–C duct. Penetration tests correspond to instances with the UV–C duct in place and operating, while correlation tests refer to instances where the UV–C duct was removed. The latter tests were performed to evaluate systematic losses in the duct system itself. None of the three correlation measurements performed reveal losses in either particles (fluorescein) or viruses (titer and RT-qPCR) from upstream to downstream in the absence of the duct; in fact, for several tests, the average downstream concentrations were higher than the upstream concentrations but within the expected variation rates. In analyzing results in the presence of the duct, we therefore do not make any direct corrections for correlation testing results, and instead opt to display raw results simply to make clear the results of all testing. Importantly, virus titer results, expressed as TCID50 mL–1, have a lower limit of detection of 102 mL–1. For measurements below this threshold (events where virus was not detected), as is common practice, we assigned a value of 3.16 × 101 TCID50 mL–1 for the virus titer, which is a half log increment below the limit of detection and is an upper limit on the virus titer for the samples in question. Evident in Table , with the UV–C duct in place, downstream virus titers (but not RT-qPCR results) were always below the limit of detection, hence in analyzing results for removal efficiency subsequently, we report lower limits on the removal efficiency, and the true efficiency is always higher than the reported number.

Table 1

Summary of the Porcine Respiratory Coronavirus Titers, RT-qPCR Signal, and Averaged Fluorescein Signal for All Tests

	Fluorescein penetration tests [a.u.]
	684 L·min–1		1674 L·min–1		2439 L·min–1
Replicate	Upstream	Downstream	Upstream	Downstream	Upstream	Downstream
1	1.32 × 103	1.28 × 103	9.39 × 102	3.32 × 102	7.30 × 102	2.82 × 102
2	1.72 × 103	7.28 × 102	1.20 × 103	3.77 × 102	8.50 × 102	3.96 × 102
3	1.27 × 103	1.27 × 103	4.51 × 102	3.87 × 102	6.76 × 102	3.05 × 102
Average	1.44 × 103	1.09 × 103	8.62 × 102	3.65 × 102	7.52 × 102	3.28 × 102

Values of 3.16 × 101 correspond to virus titer levels below the limit of detection, with this value utilized as the upper limit of the possible virus titer.

Using the average concentration values from triplicate measurements, in Figure we plot the log reduction in particles and viruses based on fluorimetry, virus titer and RT-qPCR at the three test flow rates. Log reduction (LR) is calculated based on the equation:[41]where Cup refers to the average upstream concentration and Cdown the average downstream concentration. Log reduction is directly linked to removal efficiency (RE) via the equation: RE = (1–10–LR) × 100%. The error bars in Figure represent one standard deviation on the log reduction, estimated using the root-sum-square (RSS) method as shown in the SI. Log reduction based upon fluorescein measurements is representative of the mass-based log removal of particles by collection (neglecting photobleaching). RT-qPCR based log reduction is representative of combined physical virus removal, and damage to the RNA sequence exploited in the PCR assay. Conversely, virus titer-based log reduction reflects the combined physical removal and UV–C induced inactivation of viruses. Evidenced in Figure , for the averages from fluorescein penetration tests, log reductions are below 0.5. We attribute this to gravitational settling and inertial impaction of supermicrometer particles as they move through the system; the housing holding the UV–C lamps forces the flow to bend at multiple locations inside the duct system, providing the opportunity for collection via settling and impaction. RT-qPCR results reveal slightly higher log reductions in the 0.75–1.0 range, suggesting that in addition to physical collection, there is some degree of RNA modification via UV–C absorption, leading to a reduction in RNA concentrations inferred from RT-qPCR assays. However, as reported by Myatt et al.[28] for rhinovirus, we do not find substantial (i.e., orders of magnitude) reductions in RNA concentrations because of passage through the UV–C systems. However, it is important to note that RT-qPCR detection methods that use primers and probes that amplify and detect longer RNA fragments with pyrimidine-rich regions may capture a wider area of UV–C damage which in turn could provide stronger correlation with cell culture assays.[42] Both fluorescein tracer and RT-qPCR results can be contrasted with virus titer determination; we find log reductions in excess of 2.20 at the highest flow rate (99.4% efficient), and in excess of 3.39 (99.96% efficient) and 3.73 (99.98% efficient) for the lower two flow rates (with the words “in excess” noted because the downstream titers were below the limit of detection). With the high fluences in the duct, we suspect the true log reductions and removal efficiencies are significantly higher. The lower limits provided here are based upon the amount of PRCV we are able to aerosolize and ultimately sample; we remark that determining the true inactivation efficiency of highly efficient control technologies for viruses at elevated flow rates is challenging because of practical limitations in virus propagation and aerosolization. Despite specific efforts made in this study to maximize virus titer, starting suspension volume, aerosolization time and rate, and sampling flow rates, our wind tunnel system faces limitations in determining true virus log reductions in excess of 4.0. This points to the need to develop high flow rate, highly efficient virus sampling techniques, which will also undoubtedly find application in better quantifying viable virus aerosol concentrations.[7]

Figure 3

Log reduction based upon fluorescence, virus titers (lower limits), and RT-qPCR detected RNA at test flow rates of 684 L min–1, 1674 L min–1, and 2439 L min–1. Corresponding removal efficiencies (RE) are also displayed. Error bars represent one estimated standard deviation of the log reduction.

We therefore confirm that UV–C flow tube units can be an integral part of control technologies designed to mitigate airborne coronavirus transmission. The tested duct unit, though similar to units described and applied previously,[15,16] is distinct in design from more common UV–C HVAC units, where UV–C lights are typically placed near filtration systems and designed to inactivate infectious particles collected by the filter media. That said, we do believe such UV–C flow tube reactors would be best incorporated into systems integrating multiple control technologies, such as fibrous filters (for efficient particle removal, after the UV–C unit), activated carbon filters (for VOC removal upstream, to mitigate gas phase reactions catalyzed by the UV–C unit), or electrostatic precipitators (also for particle collection and possibly virus inactivation[41]).