Index patients and protocol for measurement safety
Patients were voluntary participants with an RT-PCR-confirmed symptomatic COVID-19 infection between 1.7.2020–16.3.2021. None of the participants had been vaccinated. As infectivity has been observed to be highest in early disease, the patient with the fewest symptom days was selected as the index patient39, with the exception of collection 13, where all the patients in the room had been symptomatic for over 10 days and the patient with the freshest positive PCR result (P26) was selected (Table S1). In one case, the index patient was the same in two collections, and in two other collections, two index patients were included. All research personnel conducting the sampling followed aerosol safety protocols and precautions and no infections were detected. All procedures that involved human participants, including environmental sampling, were conducted in accordance with the ethical standards of the institutional or national research committee and the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards. The Ethics Committee of Helsinki University Hospital approved the study protocol (HUS/1701/2020). All respondents provided written informed consent prior to their participation.
Measurement environment
In total, 31 collections were performed, 24 of them on the cohort COVID-19 ward at the HUS Surgical Hospital, Helsinki, Finland. Of these, 22 were carried out on a relatively large ward with a completely open patient room area (146.4 m2, height 4.5 m, total air volume 655.25 m3, and supply air of 421 l/s, 1.67 air changes per hour (ACH)), and a maximum of 13 confirmed SARS-CoV-2 patients at a time. One collection was performed in the intensive care unit (ICU) (operating room mechanical ventilation, >30 ACH) and one in a single patient room (mechanical ventilation, 1.67 ACH). The layout of the rooms is illustrated in Figure 1. The infection prevention and control protocols on the COVID ward included hand hygiene, universal masking for staff (FFP2/3 for ICU and surgical masks for the COVID ward), guidance on social distancing (2 m), and personal protecting equipment (PPE) following droplet precautions. The room was cleaned twice a day. Additionally, seven collections were performed in patients’ homes in normal rooms where the patients spent time during illness. The measured rooms were circa 15-30 m2 in area with a normal room height of 2.5 m and circa 0.5 ACH.
Cell lines
Vero E6 cells (VE6) and their TMPRSS2-expressing clone VE6-TMPRSS2-10 (VE6T)58 were grown as previously described59. To inhibit fungal growth, 0.205 µg/ml of amphotericin B (Fungizone, Thermo Scientific) was added to the medium of the cells that were taken to the hospital for aerosol collections.
Sampling protocols for air sampling
Seven different air collection methods were used. Details of the collections and samples are presented in Table S3. A Dekati PM10 cascade impactor (20 l/min air flow) with three stages (>10, >2.5, and >1 µm), intended to ascertain the particle distribution according to aerodynamic size (PM10, PM2.5, PM1, and a backup filter for particles <1 µm), was used in eleven collections. The impaction stages of PM10, PM2.5, and PM1 were fitted with 25-mm-diameter cellulose acetate membrane filters (CA filter, GE Healthcare Life Sciences) and the backup plate with a 40-mm CA filter. The collector was placed within 1–2 m from the patient and particles were collected for 2–4 hours. After sampling, filters were immediately placed in 2 ml (25-mm filter) or 3 ml (40-mm filter) of minimal essential Eagle’s medium (MEM, Sigma-Aldrich).
The BioSpot 300p bioaerosol sampler prototype (Aerosol Devices Inc.) has a flow rate of 8 l/min and a mechanism that allows water to condense on aerosol particles from as small as 5–10 nm to 20 µm in diameter and minimize the stress when the sample is impacted onto the surface with the collection medium. To increase the sample collection rate, the biosampler is equipped with eight wicking tubes fitted with three nozzle jets to secure gentle transfer of the sample. This sampler was used in 8 collections for 1.5–4 hours within a distance of 1–2.5 m from the patient and the sample was collected in 1–2 ml of MEM.
As a more portable solution for personal area air sampling, a standard 25-mm gelatin (Sartorius Stedim Biotech) or mixed cellulose ester (MCE) filter equipped in the Button sampler with a Gilian 5000 air sampling pump, 4 l/min air flow, and a porous curved surface inlet was used in 9 collections. The Button sampler collects particles smaller than 100 µm60. The stability of SARS-COV-2 on two filter materials was compared under laboratory conditions to select the more optimal filter type and to optimize the collection time (details in supplementary material). Samples were collected for 10–30 min from patient’s breathing area. Depending on the health status, a conversation was prompted to increase the output of aerosols. The collection filter was removed into 3 ml of MEM immediately after collection ended.
Three Andersen cascade impactors (400 W pump and 28.3 l/min flow rate) were used simultaneously in six collections. The impactors consist of six stages with size cut points of: 1) >7 µm, 2) 4.7–7.0 µm, 3) 3.3–4.7 µm, 4) 2.1–3.3 µm, 5) 1.1–2.1 µm, and 6) 0.65–1.1 µm. To ensure the correct volume flow rate, each Andersen impactor was fitted with a TSI flow meter. Samples were collected using Petri dishes (94/16 MM) with 15 ml of cell medium for 10, 20, and 30 min. The medium was transferred onto VE6T cells grown on 100/20 MM cell culture dishes either immediately after collection in the hospital (collections 24, 25, 27, and 29) or later in the laboratory (collection 31).
To evaluate the real-time particle number concentration during the hospital collections and to gather additional air samples, a Dekati eFilter was used in two collections. The eFilter monitors changes in real-time particle concentration by utilizing a small diffusion charger powered by an inner chargeable battery. The charge changes were automatically translated into a signal, which was recorded on a data card. When postprocessing the data, the raw charge signal was further converted to represent particle number concentrations using a conversion factor (411 cm^-3 fA^-1) provided by the manufacturer. A count median diameter (CMD) of 60 nm and a geometric standard deviation (GSD) of 1.5 were assumed61,62. In addition, the eFilter simultaneously collected samples on a 47-mm gelatin filter using an external pump. After sample collection, the gelatin filter was transferred into 6 ml of MEM. The eFilter was fitted with the same EPA-designed inlets as the Andersen cascade impactors. The particle size cut point of the inlets was approximately 12 µm, with an air volume flow rate of 28.3 l/min. The duration of sample collection was 30 minutes at a similar distance from the patient as with the Andersen’s cascade impactors.
Passive air samples were collected either directly on VE6 cells (2 collections) or VE6T cells (9 collections) grown on 100/20 MM (collection 22) or 35/10 MM (other collections) cell culture dishes or on empty 35/10 MM Petri dishes containing 1 ml of growth medium (10 collections). Open dishes were positioned at different proximities from the patient for 30–60 min, and the patient was encouraged to perform an aerosol-producing activity such as talking. The ability of SARS-CoV-2 to infect cells at room temperature was confirmed, and major differences in the culture sensitivity of these two collection methods were excluded under laboratory conditions (see supplementary methods for details).
Living cells were transported to the laboratory in a warm environment with heat accumulators warmed to 37°C. One plate was used as a negative control to ensure that the cells survived the transport. Other samples were transported with cold accumulators and handled during the same or next day.
Sampling protocols for surface sampling
Altogether, 252 surface samples in 26 collections for PCR testing were taken from surfaces in possible direct or indirect contact with the patient (Supplementary Table 3) with pre-wetted Dacron swabs (Copan, 25 collections), a nitrile glove (1 collection), gauze (1 collection), or by pipetting the sampling liquid up and down on the surface a few times and transferring it into a sampling tube (3 collections). Swabs were placed into 1 ml of PBS. In 22 collections (212 samples), an additional sample was taken for virus culture, which was placed in 250 µl or 1 ml of MEM. Samples for PCR and culturing were taken immediately next to each other. Surfaces were divided into four surface groups (high-touch surfaces, low-touch surfaces, toilet surfaces, and other surfaces) for statistical analyses.
Other sampling protocols
Saliva samples were taken from 26 index patients either with a Dacron swab (collections 5–9) or by spitting into a Falcon tube (from collection 10 onwards). Ten additional saliva samples were collected from other patients from the ward in four collections and from seven healthy family members of home-treated patients. If possible, patients were asked to rinse their mouth before sampling. In collection 23, the index patient and a healthy family member also took follow-up saliva samples until 12 days from the start of the patient’s symptoms. In collection 26, follow-up saliva samples were taken from patients until days 14–17 from the start of symptoms.
Nasopharyngeal samples from consenting patients were taken and sent to HUSLAB for a fresh diagnostic PCR63,64. Serum samples from consenting patients were taken within a day from sampling and tested for SARS-CoV-2 IgG antibodies with two different tests65. Serum samples (dilutions 1:10 to 1:640) were studied with the microneutralization assay66. Blood lymphocyte and eosinophil counts, and plasma CRP from consenting patients were measured within a day of sampling, and plasma ferritin, ALP, ALT, D-dimer, and fibrinogen levels within three days. The respiration rate and SpO2 levels were measured during the same day (Table S1).
Since the first cases caused by variants of concern (VoC) were detected in Finland at the end of December 2020, they were determined from all patients as a part of routine diagnostics. This information was used to compare the results between VoC strains (mainly alpha in Finland) and non-VoC strains. Virus strains of collections 1–22 (P1–P45) were considered as non-VoC, as they were collected before the first cases were reported in Finland.
RNA extraction and PCR protocols for air, saliva, and culture medium samples
Trizol (Invitrogen) was used to extract RNA from all saliva samples and from air and culture medium samples of collections 1–23 according to the manufacturers’ instructions. A 200-µl sample was added to 800 µl of Trizol reagent and a resuspension volume of 50 µl was used. RNA was extracted from air and culture medium samples of collections 24 onwards with a QIAcube HT system and QIAamp 96 Virus QIAcube HT kit (QIAgen) using off-board lysis.
All samples were tested with two different RT-PCRs, N Charité67 and N1 US CDC68, using TaqMan Fast Virus 1-Step Master Mix (ThermoFisher), a 20-µl reaction volume, and fast cycling mode (annealing temperatures 55°C (N1 US CDC) and 58°C (N Charité)). The primer and probe concentrations of N Charité were according to the original publication, and those of N1 US CDC were 500 nM of both primers and 125 nM of probe (Table S7). PCRs were performed using a Stratagene Mx3005P instrument (Agilent Technologies) with a Ct cut-off value of 0.04. The results were considered positive if both PCRs were positive with a Ct value under 40 or if one PCR was positive with a Ct value under 38. Samples with Ct values over 38 in one PCR and no Ct with the other one were treated as negative, even though the possibility of them being very weak positives could not be excluded. RNA extracted from the Fin/20 strain66 culture was used as a positive control and nuclease-free water as a negative control.
The N gene transcript for qPCR was prepared as follows: the target region (352–712, 360 bp) was amplified from SARS-2 RNA, Wuhan strain, and cloned into pGEM-T cloning vector (Promega, Madison, USA) under control of the SP6 promoter. The presence of the insert was verified by sequencing and restriction enzyme analysis. After linearization of the plasmid by digestion with AscI (Thermo Fisher Scientific, USA), RNA was generated using the RiboMAX™ Large Scale RNA production system with SP6 polymerase (Promega, Madison, USA) according to the manufacturer's instructions. The transcribed RNA was then treated with DNAse I and purified with the RNeasy Mini Kit (QIAGEN, Hilden, Germany). Finally, RNA was quantified by spectrophotometry and the RNA copy number was calculated based on its concentration, length, and molecular weight. qPCR was performed with N Charité PCR by including a dilution series from 10 to 109 copies/reaction in triplicate.
RNA extraction and PCR protocols for surface samples
RNA was extracted with the NucliSENS miniMAG kit (Biomerieux). Process control virus (mengovirus) was added to at least half of the samples. Tubes containing PBS and swabs were mixed by vortexing and swabs were moved to 1 ml of high pH tris-glycine-beef extract buffer (TGBE, pH 9.5). The tubes were vortexed again and agitated at 250 rpm for 5 minutes in an orbital shaker (IKAKS 2060 basic, Patterson Scientific, UK), and the swabs were moved into a tube with 4 ml of lysis buffer, vortexed and agitated at 250 rpm for 10 minutes. PBS, TGBE, and lysis buffer were then combined, vortexed, and incubated for 10 minutes. PBS without process control virus was included as a negative control and PBS with process control virus as a positive control. The rest of the extraction was carried out according to the NucliSENS miniMAG kit instructions. The samples were further treated with the OneStep PCR Inhibitor Removal Kit (Zymo Research, CA, USA) according to the manufacturer’s instructions.
Samples were tested for SARS-CoV-2 with modified versions of N Charité67 and N1 US CDC PCRs68 and for process control virus69. The RT-PCR was carried out using a QuantiTect Probe RT-PCR kit (Qiagen, USA). Reaction mixes included 10 µl of 2X QuantiTect Probe RT-PCR Master Mix, 0.2 µl of QuantiTect RT mix, 0.6 µM of forward and 0.8 µM of reverse primer, 0.2 µM of probe for N Charité PCR primers, and 5 µl of RNA template, and the volume was adjusted to 20 µl with water. For US CDC PCR, final concentrations of 0.5 µM for both primers and 0.2 µM for the probe were used. For mengovirus PCR, 1 µM of both primers and 0.2 µM of probe were used. N Charité and N1 US CDC runs included one 10−4 dilution of SARS-CoV-2 RNA extracted from cell-grown virus as a standard positive control and one or two blanks as a standard negative control, and the reactions were performed in duplicate whenever the sample amount was sufficient. A Rotor Gene 3000 (Qiagen) real-time PCR cycler was used. The cycling conditions were reverse transcription for 30 min at 53°C, a denaturation step at 95°C for 15 min, followed by 45 cycles of amplification/denaturation at 95°C for 15 s, annealing at 58°C for 45 s, and extension at 72°C for 45 s. The results were analyzed with the thermocycler software Rotor-Gene 6.0.31 (Qiagen, USA) using similar criteria as with other samples described above.
Culturing protocols
Samples were initially cultured in VE6 cells (collections 1–18), which were changed to VE6T cells after reports of these being more sensitive (collections 19–31)61. Air samples that were collected directly on cells were cultured as such, and the rest of the air and surface samples and 75 µl of saliva were used for culturing in 6-well plates. Medium was added to the final volume of 3 ml (saliva) or 2 ml (other samples). E-filter samples were cultured in two wells (3 ml/well). Samples were cultured at 37°C for 10–14 days and checked for cytopathic effect (CPE). A 200-µl sample of culture medium was taken from those samples that had unclear results based on microscopic observation or possible CPE and tested with N Charité PCR. Culturing was considered positive if CPE was detected and the Ct value of PCR performed from the culture media was under 20. If Ct value was higher, it was judged to be caused by original (possibly noninfectious) virus in the sample instead of virus growth. All virus culturing was performed in a BSL3 laboratory. Optimization of the culturing protocols is described in more detail in the supplementary material.
Statistical tests
Statistical tests were carried out with SPSS IBM Statistics version 27. When comparing means between two independent groups, data were first tested for normality with the Shapiro–Wilk test before testing them either with the independent-samples t-test or a non-parametric test (independent-samples Mann–Whitney U-test for two groups and independent-samples Kruskal–Wallis test for more than two groups). For categorical data, the Fisher–Freeman–Halton exact test was used. Air and surface results of collections were compared with McNemar’s test. Spearman's rank correlation coefficient was used for correlation testing. Mean values and standard deviations (normally distributed data), medians and interquartile ranges (non-normally distributed data), or percentages (categorical data) of compared subgroups, test statistics, p-values, and effect sizes (Cohen’s d for the t-test and z/ \(\sqrt{N}\) for the Mann–Whitney U-test) are reported in Table S2. P-values below 0.05 were considered statistically significant. Air or surface collections were considered positive if at least one of the samples from the collection was PCR positive. Individual data points that were added to the boxplot figures were jittered in all dimensions using a uniform distribution.