Quantifying human and environmental viral load relationships amidst mitigation strategies in a controlled chamber with participants having COVID- 19

Hooman Parhizkar Institute for Health in the Built Environment, University of Oregon https://orcid.org/0000-0003-0392-9459 Leslie Dietz Biology and the Built Environment Center, University of Oregon https://orcid.org/0000-0002-5623-1524 Andreas Olsen-Martinez Biology and the Built Environment Center, University of Oregon Patrick Horve Biology and the Built Environment Center, University of Oregon https://orcid.org/0000-0002-9318-9249 Liliana Barnatan Biology and the Built Environment Center, University of Oregon Dale Northcutt University of Oregon Kevin Van Den Wymelenberg (  kevinvdw@uoregon.edu ) kevinvdw@uoregon.edu


Introduction
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the causative agent of Coronavirus disease 2019 (COVID- 19), has resulted in 230,418,451 con rmed cases with more than 4,724,876 deaths globally, as of 24 September 2021 1 . There is substantial evidence that inhalation of aerosol particles containing viable SARS-CoV-2 virions is the primary route of human-to-human transmission [2][3][4][5][6][7] . Modeling of the impact of nonpharmaceutical interventions on the probability of COVID-19 infection and mortality rate through public health 8,9 and engineering perspectives 10 suggests that indoor congregation is the primary driver for COVID-19 disease transmission 11 . Therefore, better understanding and quantifying the relationship of human factors, design, and building operation practices on the abundance and dispersion of viral load in indoor spaces is necessary to combat disease transmission and provide environments for safe indoor congregation 12 .
Breathing and talking are some of the human expiratory activities that have been studied to determine how these activities are associated with concentrations of viral pathogens. These studies have contributed valuable information about the viral load of size fractionated aerosols 5,13,14 . In addition to human expiratory factors, indoor space design and engineering practices such as ventilation, ltration, and humidity control may in uence the abundance and infectious fraction of the environmental viral load, and therefore reduce inhalation dose [14][15][16][17][18][19][20] . However, these indoor environmental interventions need to be studied independently through controlled experiments to quantify their impacts, while minimizing confounding variables, especially with regard to aerosols that may contain SARS-CoV-2.
In this research, we sought to better understand viral abundance and dispersion associated with differing degrees of expiratory activity, ventilation, ltration, and humidi cation through controlled experiments in a quasi-eld setting. We measure viral RNA of SARS-CoV-2 using quantitative reverse-transcription polymerase chain reaction (qRT-PCR) techniques as a proxy of viral load in humans and environmental aerosols and surfaces. We studied 11 human participants that were diagnosed with COVID-19 in a controlled chamber measuring 4.3 m in length, 2.8 m in width, and 2.5 m in height (28.04 m 3 ). Our research protocol comprised a 3-day study for each participant in which human activity and environmental factors (ventilation rate, in-room ltration, humidity control) were studied as independent variables. In summary we found statistically signi cant: 1. positive relationships between viral load (RNA) found in human specimens and paired aerosol and surface samples at ~0 ACH and ambient conditions for sitting and standing trials (routine trials) as well as trials with high expiratory activities (coughing, speaking, and speaking loudly); 2. positive relationship between viral load in near eld aerosols captured during periods of higher expiratory activity and near eld particles of 0.3 µm -1 µm, 1 µm -2.5 µm, and 10 µm -25 µm in size, but no statistical signi cance for 2.5 µm -10 µm particles; 3. increased CO 2 concentrations and particulates in the range of 1-5 µm measured in the near eld as compared to the far eld for routine trials; 4. positive relationship between aerosol viral load in the far eld and the number of corresponding far eld particles detected in the range of 1-2.5 µm; 5. inverse relationships between viral load found in aerosols and degree of ventilation, as well as in-room ltration; . relationships between viral load and degree of relative humidity (RH); whereby higher RH is associated with lower viral load in aerosol samples and higher viral load in select surface samples, consistent with increased particle deposition on surfaces.

Results
A rapid deployment modular unit (RDM) was used as an environmentally controlled chamber ( Figure 1)  Trials were conducted in two different set-ups over three days. Trials with a S1 su x indicate Setup-1 where both air samplers were placed next to each other for short duration and higher expiratory tests ( Figure 1a).
During cough trials, participants were instructed to conduct 10 uncovered coughs into an area over the air samplers, particle counters (TSI AeroTrak 9306), and CO 2 (Onset HOBO MX1102A) sensors. During speak tests, participants were instructed to conduct continuous vocalization using a standardized CDC de ned passage 21 (Supplemental document, appendix A) for 5 minutes with normal and higher amplitude at their discretion, respectively 22 . A S2 su x indicates trials where participants conducted routine activities at a desk, including sitting and standing, sitting silently, sitting and participating in an online conference meeting, or were invited to walk on treadmill (physical activity day) (Figure 1b).

Near and far eld aerosol samples and paired human specimens
To quantify the relationship between viral loads (RNA copies) in human nasal and aerosol samples, we paired the outcome of each aerosol sample collected with its corresponding shallow nasal sample for both near and far AerosolSense samplers during trials when participants were sitting or standing for one hour at ~0 ACH under typical ambient conditions without environmental interventions (routine trials). Figure 2a shows the relationship between nasal viral load and near eld and far eld aerosol viral load for all routine trials. Note that negative samples are de ned with a value of 40 C T .
The coe cients associated with signi cant regression models presented in Figure 2a indicate that an increase in viral load equivalent to -1 C T in human nasal samples is associated with increased near eld viral load of -0.32639 C T (R 2 = 0.2276, P = 0.001092) and increased far eld viral load of -0.4014 C T (R 2 = 0.4026, P = 1.721e-06). The difference of means between the aerosol C T value of near eld and far eld aerosol samples was 1.0583 C T , whereas far eld samples represent lower viral load, however the paired t-test differentiating near eld and far eld samples was not signi cant (P = 0.05955) (Figure 2b, note that black solid horizontal line represents median in all box plots). Therefore, we also report the signi cant coe cient for all nasal and aerosol samples in routine trials which indicates that an increase in viral load equivalent to -1 C T in nasal samples is associated with an increase in room aerosol viral load of -0.36216 C T (R 2 = 0.3119, P = 1.675e-08, Supplemental gure 1). Based upon qRT-PCR theory, a -1 C T difference is approximately equivalent to double the viral load 23 . To our knowledge this is the rst reported relationship between environmental aerosol viral load and human viral load in a controlled environment (28,040 L 3 room, ~0 ACH, one-hour trials, single COVID-19 positive individual).
In addition to these viral dispersion characteristics, among all routine trials, we found a statistically signi cant difference between the mean CO 2 concentration recorded at near eld and far eld, whereas CO 2 concentrations of near eld were 80 PPM higher than in the far eld (P = 0.0004009) ( Figure  We explored the relationships between aerosol viral load, particle counts, and CO 2 concentration for all routine trials. We did not nd any signi cant correlation between near eld aerosol viral load and the corresponding number of near eld particles for any size bin for routine trials. As shown in Figure 2e, we identi ed a signi cant relationship between aerosol viral load and far eld particle counts within the size bin 1-2.5 µm. The signi cant coe cient in Figure 2e indicates that an increase in far eld aerosol viral load equivalent to -1 C T is associated with ~27 more particles in the range of 1-2.5 µm (R 2 = 0.1112, P = 0.04313) in the far eld.
We report a statistically signi cant positive correlation between the average far eld CO 2  High-touch surfaces, settling plates, and paired human specimens Human specimens were compared to paired samples collected from the participants' phone (screen), computer (adjacent to keyboard), and chair (described as high-touch surfaces), and from near eld settling plates (on participant's desk) and far eld plates (adjacent to far eld air sampler). Figure 3a illustrates the signi cant linear regressions for the viral load (RNA) on each high-touch surface relative to paired nasal samples. Figure 3b illustrates the signi cant linear regressions for viral load in settling plates (near and far) relative to paired nasal samples. There are no signi cant differences between the viral loads found in near eld and far eld setting plates, nor are there signi cant differences between any of the high-touch surfaces (Supplemental gures 4 & 5). Figure 3c illustrates the signi cant regressions for all sampling types relative to human nasal samples within a single gure and indicate that high-touch surfaces and aerosol samples have higher viral loads than settling plate surfaces.
High expiratory activity, particles, and aerosol viral load We nd a signi cant correlation between aerosol viral load associated with high expiratory activities and paired nasal samples whereas an increase in viral load equivalent to -1 C T in human nasal samples is associated with increased immediate eld (<1m, Figure 1a) Interestingly, the 0.3 µm -1 µm size bin indicates the highest correlation coe cient between immediate eld particle counts and immediate eld aerosol viral load. While the relationship between the particles of 1 µm -2.5 µm and immediate eld viral load is signi cant, there is no signi cant relationship found for 2.5 µm -3 µm, 3 µm -5 µm and 5 µm -10 µm.
Among high expiratory trials, we observed an increase in immediate eld viral load equivalent to -1 C T to be associated with an increase of ~1000 particles of the size 0.3 µm -1 µm, and an increase in ~100 particles of the size 1 µm -2.5 µm, and ~ one particle of the size 10 µm -25 µm in the immediate eld. It is important to stress that these results are relevant to immediate eld particulates dominated by bioaerosols.
Our ndings for immediate eld trials support previous research in which SARS-CoV-2 RNA was identi ed in ne particles 5 . While we did not nd any statistically signi cant relationship between aerosol viral load and particle counts of 5 µm -25 µm during routine trials in the near eld (1.2m) or the far eld (3.5 m), during immediate eld (<1m) high expiratory trials we identi ed a signi cant relationship for large particles (10 µm -25 µm) and immediate eld aerosol viral load; we hypothesize that may be due to immediate eld respiratory droplets prevalent in high expiratory activities 11,13,24 .
The impact of ventilation and ltration on aerosol and surface viral load Indoor air exchange rate, measured in Air Changes per Hour (ACH), has previously been demonstrated to reduce indoor particulates and therefore hypothesized to reduce the concentration of viral aerosols, corresponding inhalation dose, and consequently the probability of indoor occupants acquiring infection [25][26][27] . Few studies have measured the relationship between ventilation, ltration and aerosol viral load 28 .
Therefore, we investigated the impact of alternate air exchange rates, using 100% outside air (OSA) and ltration levels during removal mechanism trials. As shown in Table 1, the removal mechanism day began with a baseline ~0 ACH trial, followed by four 100% OSA ventilation trials (two at ~9 ACH and two at ~3 -4.5 ACH) provided by an exhaust fan ( tted with HEPA lter for infection control). Thereafter, a single trial with two in-room HEPA lters (without OSA) was conducted. All removal mechanism trials and the ~0 ACH control trials were conducted for a duration of one hour. We found a signi cant difference between control trials and all removal mechanism trials (P = 0.029, Figure 5a). In Figure 5a we show a signi cant difference between control trials and paired removal mechanism trials, while in Figure 5b we show a signi cant correlation for all control trials at ~0 ACH and all ventilation trials with 100% OA organized by mean CO 2 concentration. Trials with less than ~4.5 ACH (including ~0 ACH trials) were associated with signi cantly higher aerosol viral loads in the near eld when compared with trials greater than ~9 ACH, with a mean difference of -3.6 C T (P = 0.037, unpaired t-test, Figure 5c). Even though the mean difference of aerosol viral load in the far eld for trials with less than ~4.5 ACH (including ~0 ACH trials) was higher than trials with greater than ~9 ACH, we did not observe a statistically signi cant difference for far eld aerosol viral load (P = 0.085, unpaired t-test, Figure   5c). When examining total room aerosol viral load (near eld and far eld together), we report that trials with less than ~4.5 ACH (including ~0 ACH trials) were associated with statistically higher viral load than trials with greater than ~9 ACH, with a mean difference of -3.2 C T (P = 0.01153, unpaired t-test, supplemental gure 9). Our research provides further evidence that improved ventilation is bene cial for both near eld and far eld aerosol viral load. Given these relationships within this room (Figure 5b), ventilation trials indicate that an increase in ~128 PPM of CO 2 concentration corresponds with an increase in aerosol viral load equivalent to -1 C T , thus, approximately a doubling of the viral load. Moreover, ltration trials indicate that there is a signi cant difference between trials with only in-room HEPA ltration (~1000 m 3 /hr) and paired control trials at ~0 ACH, whereas HEPA trials have lower viral load equivalent to 3.240741 C T (P = 0.029), thus, approximately an order of magnitude reduction (Figure 5d).
Our results provide evidence that increased air exchange (~9 ACH with 100% OSA) or in-room HEPA ltration (~1000 m 3 /hr) yields reduced aerosol viral load, and reason therefore suggests these measures are likely to reduce inhalation dose and the probability of infection in indoor spaces. We found no statistical difference between aerosols captured during control trials with ~0ACH and those with ~3 -4.5 ACH; however, this may be related to limitations in sample size. Among three types of high-touch surfaces collected in this study, increased ACH was associated with lower viral load on participant's computers, with a mean difference of 4.033908 C T (P = 0.002323) whereas phone and chair samples showed no signi cant difference with air exchange rate (Supplemental gure 10).

Relative humidity and aerosol viral load
Relative humidity is hypothesized to impact aerosol pathogens and disease transmission in three ways; (1) improved human immune response 26 (2) reduced viability in aerosols at relative humidity between 40-60% 11,15 , and (3) increased particle deposition 29 . The structure and behavior of aerosol pathogens, speci cally particle size, settling rate, and diffusion, are each affected by relative humidity 29,30 . In this study, we aimed to measure environmental viral load at different relative humidity conditions. Two dehumidi ers and two humidi ers were used to regulate relative humidity to low and high levels during the "relative humidity" trials. All relative humidity trials were conducted for 1-hour. Each participant's relative humidity day started with a 1-hour control trial with ~0 ACH and relative humidity at ambient conditions, followed by two 1hour dehumidi cation trials and two 1-hour humidi cation trials. Room aerosol C T values were paired with mean relative humidity values (ranging from 20-70%) recorded for each trial.
Relative humidity trials indicate that an increase of ~11.85% in relative humidity corresponds with a decrease in aerosol viral load equivalent to 1 C T (p = 0.008), thus, approximately a 50% reduction in aerosol viral load, as shown in (Figure 6a). Similarly, an increase of ~10.02% in relative humidity corresponds with an increase in surface (chair, computer, phone) viral load equivalent to -1 C T (p = 0.01) as shown in Figure 6c, consistent with increased particle deposition. Figure 6b shows the signi cant decrease in aerosol viral load equivalent to 3.28908 C T (paired t-test, P = 0.0002643) for humidi cation trials as compared to dehumidi cation trials.
Conversely, Figure 6d shows the signi cant increase in computer surface viral load equivalent to -2.873077 C T (paired t-test, P = 0.01593) for humidi cation trials as compared to dehumidi cation trials. This is one of the rst studies that investigated the role of relative humidity on viral RNA in aerosols and surfaces in a realistic setting. Our results suggest that increased relative humidity corresponds with decreased viral load in aerosols and increased viral load on select indoor surfaces, consistent with an increased rate of particle deposition. Since several studies have demonstrated that there is a substantially higher risk for aerosol mediated transmission than fomite mediated transmission 31 , active humidity control (including humidi cation, or reduced dehumidi cation) could be implemented to reduce aerosol mediated COVID-19 transmission risk reduction in indoor spaces. Of course, humidi cation controls must be properly maintained and managed to avoid condensation and mold propagation.

Limitations
All participants were given the opportunity to opt out of the study at any time, thus two subjects only completed the rst day of study. There were some modest inconsistencies between trial durations in order to accommodate participants' needs. Not all participants walked on the treadmill, and some walked at different speeds or for different durations. While this was an extensive study design, conducted over three days per participants, the total number of unique participants (n=11), and limited age range (18-24 years of age) of participants, presents some limitations to generalizability. RNA samples were not assessed for viability.

Method
Institutional Approval and Data Availability Biological protocols were reviewed and approved by Advarra Institutional Biosafety Committee (IBC) (Protocol #PROTO202000132). Advarra IBC is an authorized external IBC for the University of Oregon and is registered with the National Institute of Health (NIH). Human participant protocols were reviewed and approved by the University of Oregon Institutional Review Board (IRB) (Protocol #12292020).

Participant Recruitment
University of Oregon COVID-19 protocols require individuals living in the university residence halls to spend their isolation period at an off-campus quarantine dormitory room for 14 days. Individuals positive for COVID-19 were identi ed through the University of Oregon Monitoring and Assessment Program (MAP). Following transfer to the isolation dormitory, individuals were recruited into the program to conduct a 3-day study at the RDM which was located in the dormitory parking lot. All participants volunteered to conduct different activities involved with this study with no penalty associated with leaving the research at any time.

RDM layout
The interior volume of the RDM was 28,080 L. Interior temperature was maintained at 22ºC +/-4 ºC with three portable electric resistance heaters. Relative humidity was adjusted using two portable humidi ers and two dehumidi ers, respectively. Outdoor ACH of ~3 -4.5 and ~9 were provided through a HEPA ltered (CleanShield HEPA 550, ALORAIR) exhaust air removal from the RDM with make-up air via in ltration and an operable window that opened during maximum ventilation trials. Filtration was provided with two in-room HEPA lters with combined Clean Air Delivery Rate (CADR) of ~1000 (600 Cubic Feet per Minute) .
Temperature, RH, and CO 2 were monitored and recorded using multiple data loggers (Onset HOBO MX1102A).

Sample Collection
Samples were collected 8-12 times throughout a day as follows: Day 1-Physical activity day (~0 ACH, room condition): plates (Millipore Sigma #P5731-500EA), and active air samplers (ThermoFisher #2900AA). Environmental swabs were collected from the participant's cell phone, computer, chair, and exhaust inlet. For walking on treadmill trials, samples from treadmill handrail, front rail, and bottom were collected. Flocked nylon swabs pre-moistened with DNA/RNA Shield (Zymo Research, Catalog #R1100) were used to swab the sampling location in a zig-zag 'S' pattern for 15-20 seconds and then returned to a labeled 5ml tube containing 1 ml of DNA/RNA Shield. Settling particulates were captured using both components (base and lid) of standard Petri dishes. Following the sampling period, both sides of the Petri dish (sampling area ~110 cm 2 ) were swabbed following the protocol described above for environmental swabs. Active air samples were collected using two AerosolSense samplers . The AerosolSense sampler works by drawing air into an accelerating slit impactor at a rate of 200 L/minute, causing particles to impact onto a collection substrate. Following the sampling period, the collection substrate was transferred to 1 ml of DNA/RNA Shield using ame-sterilized forceps and transported back to a BSL-2 laboratory. Upon return to the laboratory, the capture media was brie y vortexed, then centrifuged for 2-minutes at 1,500 x g to remove all liquid from the collection substrate. Following centrifugation, the collection substrate was placed into biohazard and discarded appropriately.

Molecular Analysis
All protocols were performed in a Puri er Logic+ Class II, Type A2 biosafety cabinet (LabConco, Catalog RNA control and a no-template control (nuclease-free water) All controls performed as expected.

Statistical Analyses
Analyses were performed using the statistical programming environment R. The correlation between observed C T values and other environmental parameters was conducted through the use of a generalized linear model. One-tailed paired t-test were used to identify statistical differences between categorical variables such as mean C T values and environmental parameters unless otherwise noted. Black solid horizontal line represents median in all box plots in this article. One tailed non-paired t-test was used to identify statistical differences for trials with outdoor air exchange rate of under ~4.5 ACH and above ~9 ACH. equipment, and reagents but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. However, per contractual obligations, the funder had the right to review the nal manuscript for con dential information prior to submission.

Data and code availability
All data and code supporting this study and required to create the analyses are provided in Github, available Rapid deployment modular unit (RDM), a) higher expiratory trials (S1), b) regular trials (S2)     Correlation between surface CT value and mean relative humidity among dehumidi cation, humidi cation, and control trials, d) ) paired comparison of select surface (computer) CT between Dehumidi cation and Humidi cation trials.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.