Viral Load of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) in Respiratory Aerosols Emitted by Patients With Coronavirus Disease 2019 (COVID-19) While Breathing, Talking, and Singing

Abstract

Background

Multiple severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) superspreading events suggest that aerosols play an important role in driving the coronavirus disease 2019 (COVID-19) pandemic. To better understand how airborne SARS-CoV-2 transmission occurs, we sought to determine viral loads within coarse (>5 μm) and fine (≤5 μm) respiratory aerosols produced when breathing, talking, and singing.

Methods

Using a G-II exhaled breath collector, we measured viral RNA in coarse and fine respiratory aerosols emitted by COVID-19 patients during 30 minutes of breathing, 15 minutes of talking, and 15 minutes of singing.

Results

Thirteen participants (59%) emitted detectable levels of SARS-CoV-2 RNA in respiratory aerosols, including 3 asymptomatic and 1 presymptomatic patient. Viral loads ranged from 63–5821 N gene copies per expiratory activity per participant, with high person-to-person variation. Patients earlier in illness were more likely to emit detectable RNA. Two participants, sampled on day 3 of illness, accounted for 52% of total viral load. Overall, 94% of SARS-CoV-2 RNA copies were emitted by talking and singing. Interestingly, 7 participants emitted more virus from talking than singing. Overall, fine aerosols constituted 85% of the viral load detected in our study. Virus cultures were negative.

Conclusions

Fine aerosols produced by talking and singing contain more SARS-CoV-2 copies than coarse aerosols and may play a significant role in SARS-CoV-2 transmission. Exposure to fine aerosols, especially indoors, should be mitigated. Isolating viable SARS-CoV-2 from respiratory aerosol samples remains challenging; whether this can be more easily accomplished for emerging SARS-CoV-2 variants is an urgent enquiry necessitating larger-scale studies.

severe acute respiratory syndrome coronavirus 2, SARS-CoV-2, aerosol transmission, airborne transmission, COVID-19

Coronavirus disease 2019 (COVID-19) is caused by the highly transmissible severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Irrespective of symptomatology, patients with COVID-19 can harbor high viral loads of SARS-CoV-2 in their respiratory tracts [1, 2] and emit SARS-CoV-2 RNA into the air [3, 4], which may be culturable under favorable circumstances and collection methods [5]. Although virus emissions from talking and singing have not been measured, these expiratory activities are hypothesized to play a crucial role in virus transmission [6]. A significant proportion of SARS-CoV-2 transmission is estimated to be from asymptomatic individuals [7], and multiple SARS-CoV-2 superspreading events [8–10] suggest that aerosols may be critical in driving the COVID-19 pandemic. Thus, refined public health measures are likely needed to contain the virus, especially in undervaccinated populations.

Respiratory aerosols range from 0.1 to 100 μm in diameter and can be categorized as coarse (>5 μm) and fine (≤5 μm) aerosols, based on where they deposit in the respiratory tract [11]. Coarse aerosols are inhalable and deposit in the upper airways, whereas fine aerosols are respirable and deposit in the lower airways. The amount of infectious virus these size fractions carry and their relative importance to SARS-CoV-2 transmission and infection is not well understood. Experimental studies of nonhuman primates have demonstrated that COVID-19 may be anisotropic [12] as more severe illness results from inhaling infectious aerosols that are 1–3 μm in diameter when compared with direct intranasal and intratracheal inoculation [13]. Other models, however, demonstrate a disease spectrum similar to humans with combined intranasal and intratracheal inoculation [14]. Cynomolgus macaques also shed more SARS-CoV-2 in fine aerosols when compared with coarse aerosols [15]. To better understand how SARS-CoV-2 spreads, and to help refine public health measures in mitigating SARS-CoV-2 transmission, we sought to measure viral loads in coarse and fine respiratory aerosols emitted by patients with COVID-19 during breathing, talking, and singing.

METHODS

Patient Recruitment and Data Collection

Participants were recruited from February through April 2021 at the National Centre for Infectious Diseases in Singapore. During this outbreak phase, as per national public health policy, all persons in Singapore with confirmed SARS-CoV-2 infection, regardless of symptom or clinical status, were admitted for inpatient isolation and evaluation before transfer to designated isolation facilities. All newly admitted patients were screened based on the following inclusion criteria: age 21 years or older and positive for COVID-19 via reverse transcription–quantitative polymerase chain reaction (RT-qPCR). Basic demographic data were recorded. Symptom data were collected based on a list of 7 prespecified symptoms. For asymptomatic individuals, the day of diagnosis was recorded as day 1 of illness. Cycle threshold (Ct) values of clinical respiratory samples and SARS-CoV-2 serology test results were obtained from medical records. Virus genome sequence data were obtained from National Public Health Laboratory records. This study was approved by the National Healthcare Group Domain Specific Review Board, reference number 2020/01113. Written informed consent was obtained from all study participants.

Expiratory Sample Collection

Expiratory samples were collected using the G-II exhaled breath collector, described in detail by McDevitt et al [16]. Briefly, study participants were seated facing the truncated cone-shaped inlet, with air drawn continuously (130 L/minute) around the subject’s head and into the sampler (Figure 1). The cone served as a capture-type ventilation hood, which allowed the collection of expiratory particles with minimal fugitive emissions. Participants were asked to perform 3 separate expiratory activities on the same day: 30 minutes of tidal breathing, 15 minutes of talking, and 15 minutes of singing. For the talking activity, participants were asked to repeat passages read to them from the children’s book, “Green Eggs and Ham” by Dr Seuss. For the singing activity, participants were asked to sing “Happy Birthday,” “ABC song,” “Twinkle, Twinkle, Little Star,” and “We Wish You a Merry Christmas” with background music. Aerosols were collected in 2 size fractions, namely coarse (>5 μm) and fine (≤ 5μm). The coarse fraction was collected by impaction on a polytetrafluoroethylene surface. The polytetrafluoroethylene impactor was swabbed thrice, end-to-end, with a flocked swab first dipped in 1× phosphate buffered saline (PBS) solution with 0.1% bovine serum albumin (BSA). The swab was rotated during swabbing to ensure that all surfaces of the flocked tip were in contact with the impactor for optimal retrieval of coarse particles. The flocked swab was then placed in a 15-mL conical tube containing 1 mL of 1× PBS with 0.1% BSA. Fine particles were collected by condensation growth and impaction on a steel surface into a reservoir of 1× PBS with 0.1% BSA and collected into 50-mL conical tubes. Condensation growth was achieved by injecting a small amount of steam into the already humid inlet air and breath and immediately cooling the airstream in a heat exchanger held at −2°C to achieve supersaturation conditions sufficient to grow fine particles ≥0.05 µm in diameter to ≥1.0 µm. In between each activity, the G-II was decontaminated with 10% bleach, rinsed with water, and wiped dry.

Figure 1.

Schematic representation of expiratory sample collection using the G-II exhaled breath collector inside the COVID-19 patient room. Abbreviation: COVID-19, coronavirus disease 2019.

Sample Processing and Laboratory Analyses

Samples were transported to and processed in the National University of Singapore Biosafety Level 3 Laboratory on the same day as collection. See Supplementary Materials for detailed laboratory methods. Coarse-fraction swab samples were vortexed and transferred into 1.5-mL screw-capped tubes. Fine-fraction samples were concentrated with centrifugal ultrafiltration and topped up to 1.6 mL with media. Vero E6 cells were used to culture fine-fraction samples on the same day of processing. Coarse-fraction samples were not cultured as the impaction method was not designed for culture analysis [16]. RNA was extracted from each expiratory sample using the QIAamp MinElute Virus Spin Kit (Qiagen, Germany) according to the manufacturer’s instructions. The Centers for Disease Control and Prevention (CDC) N1 assay (Integrated DNA Technologies, Singapore) was performed for the detection of SARS-CoV-2. All samples were analyzed in duplicate. Viral RNA copies were calculated from a standard curve constructed with the N gene positive control plasmid (Integrated DNA Technologies, Singapore).

Statistical Analyses

Data analyses were completed using STATA version 13.0 (StataCorp, College Station, TX, USA). Fisher’s exact test was used to compare categorical variables, while Mann-Whitney U test was used to compare continuous variables between patients with and without detectable virus to identify variables associated with viral shedding in respiratory aerosols. Kruskal-Wallis test was used to compare median viral loads of different respiratory activities within the subgroup of patients with detectable virus in respiratory aerosols. All statistical tests were 2-sided and a P value of less than .05 was considered significant.

RESULTS

Twenty-three patients were enrolled in the study, including 1 patient who withdrew before sample collection. Among the 22 remaining participants, 19 (86%) were male, with a median age of 38 years (range: 23–66 years). Five (23%) were asymptomatic (never developed symptoms). Thirteen (59%) emitted detectable levels of SARS-CoV-2 RNA in respiratory aerosols (Table 1), including 3 asymptomatic patients and 1 presymptomatic patient. SARS-CoV-2 copies emitted per expiratory activity per participant (30-minute breathing, 15-minute talking, or 15-minute singing) ranged from 63 to 5821 viral N gene copies. Age, sex, virus variant type, clinical symptoms, presence of SARS-CoV-2 antibody at diagnosis, and Ct value of clinical sample at diagnosis were not significantly different between patients with and without detectable viral RNA in respiratory aerosols (Table 2). However, the median day of illness was significantly different: patients with detectable viral RNA in aerosols were earlier in the course of illness (median day of illness: day 3 vs day 5; P = .025). The highest emitters (participants 12 and 16) were sampled on day 3 of illness and accounted for 52.4% of the total viral load captured in our study.

Table 1.

SARS-CoV-2 in Respiratory Aerosols Emitted by Patients With COVID-19 in Singapore, February–April 2021

					Aerosolized SARS-CoV-2 RNA Copies Emittedc
Participant	Symptoms	Day of Illnessa	Clinical Ct Valueb	SARS-CoV-2 Serology	Breathingd	Talkinge	Singingf	Total	SARS-CoV-2 Variant
1	Sore throat, rhinorrhea, anosmia, fever	6	14.3	Positive	ND	ND	ND	…	Failed WGS
2	Rhinorrhea, anosmia	7	16.6	Negative	ND	ND	ND	…	Alpha (B.1.1.7)
3	Sore throat, chronic cough	9	30	Negative	ND	ND	ND	…	Non-VOC/VOI
4	Rhinorrhea, anosmia, cough, SOB	2	19.4	Negative	ND	417	ND	417	Alpha (B.1.1.7)
5	Asymptomatic	5 (day of diagnosis)	22.4	Negative	ND	234.5	135.2	369.7	Non-VOC/VOI
6	Sore throat, rhinorrhea	1	13.2	Negative	ND	79.9	713.6	793.5	Beta (B.1.351)
7	Asymptomatic	3 (day of diagnosis)	32.9	Positive	ND	ND	ND	…	Failed WGS
8	Slight sore throat and rhinorrhea (due to swab test), fever	5	16.5	Negative	ND	ND	ND	…	Non-VOC/VOI
9	Rhinorrhea, cough	3 (day of diagnosis; 2 days pre–symptom onset)	15.4	Positive	ND	908.2	ND	908.2	Beta (B.1.351)
10	Sore throat	4	16.1	Negative	63.5	310.9	1811.7	2186.1	Non-VOC/VOI
11	Rhinorrhea, fever	8	17	Negative	ND	ND	154.4	154.4	Beta (B.1.351)
12	Fever, dry throat	3	15.4	Negative	227.6	4336	4277.9	8841.5	Alpha (B.1.1.7)
13	Fever	2	19.2	Negative	140.9	733	ND	874	Beta (B.1.351)
14	Fever, dry cough	4	15.1	Negative	ND	ND	ND	…	Alpha (B.1.1.7)
15	Fever	5	16.8	Negative	442.1	1356.5	978.8	2777.5	Kappa (B.1.617.1)
16	Fever	3	14.7	Negative	224.2	1373.3	5821.4	7419	Beta (B.1.351)
17	Asymptomatic	2 (day of diagnosis)	14.5	Positive	ND	ND	143.6	143.6	Kappa (B.1.617.1)
18	Asymptomatic	3 (day of diagnosis)	15.3	Negative	550.3	477.9	1216.1	2244.3	Kappa (B.1.617.1)
19	Asymptomatic	3 (day of diagnosis)	14.4	Negative	ND	ND	ND	…	Beta (B.1.351)
20	Diarrhea, intermittent blocked nose	5	19.5	Positive	ND	ND	ND	…	Beta (B.1.351)
21	Sore throat, fever, body ache	5	16	Negative	310.5	2428.7	1162.3	3901.4	Delta (B.1.617.2)
22	Rhinorrhea, fever, cough	9	17.7	Negative	ND	ND	ND	…	Beta (B.1.351)

Abbreviations: COVID-19, coronavirus disease 2019; Ct, cycle threshold; ND, none detected; PCR, polymerase chain reaction; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; SOB, shortness of breath; VOC, variant of concern; VOI, variant of interest; WGS, whole-genome sequencing.

^aOn aerosol sample collection day; for symptomatic patients, day 1 of illness was defined as the day symptoms began; for asymptomatic and presymptomatic patients, day 1 of illness was defined as the day of diagnosis (day of the first PCR-positive clinical sample).

^bPCR Ct value from patient’s diagnostic sample.

^cViral N gene copies per expiratory activity.

^dThirty minutes of tidal breathing.

^eFifteen minutes of talking with brief pauses.

^fFifteen minutes of continuous singing.

Table 2.

Comparison of Variables Between Patients With COVID-19 With and Without Detectable Virus in Respiratory Aerosols

Variable	Participants With Positive Aerosol Detection (n = 13)	Participants With Negative Aerosol Detection (n = 9)	P
Age, years	36 (31–47)	43 (33–47)	.84
Female sex, n (%)	3 (23.1)	0 (0.0)	.24
PCR Ct value of clinical sample	16 (15.3–17)	16.6 (15.1–19.5)	.48
Positive SARS-CoV-2 serology,a n (%)	2 (15.4)	3 (33.3)	.61
Variant type (WHO classification), n (%)			.74
Non-VOC/VOI	2 (15.4)	2 (28.6)
Alpha (B.1.1.7)	2 (15.4)	2 (28.6)
Beta (B.1.351)	5 (38.5)	3 (42.9)
Kappa (B.1.617.1)	3 (23.1)	0 (0.0)
Delta (B.1.617.2)	1 (7.7)	0 (0.0)
Day of illness on samplingb	3 (2–5)	5 (4–7)	.025
Presence of symptoms, n (%)	10 (76.9)	7 (77.8)	>.99
Sore throat	3 (23.1)	3 (33.3)	.66
Rhinorrhea	4 (30.7)	4 (44.4)	.66
Anosmia	1 (7.7)	2 (22.2)	.54
Fever	6 (46.2)	4 (44.4)	>.99
Cough	2 (15.4)	3 (33.3)	.61
Dyspnea	1 (7.7)	0 (0.0)	>.99
Diarrhea	0 (0.0)	1 (11.1)	.41
Total number of symptoms	1 (1–2)	2 (1–3)	.25

Values are stated as number (percentage of column) for categorical variables and median (interquartile range) for continuous variables. Categorical variables were compared using Fisher’s exact test and continuous variables were compared using Mann-Whitney U test. Abbreviations: COVID-19, coronavirus disease 2019; Ct, cycle threshold; PCR, polymerase chain reaction; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; VOC, variant of concern; VOI, variant of interest; WHO, World Health Organization.

^aAt time of diagnosis.

^bFor symptomatic patients, day 1 of illness was defined as the day symptoms began; for asymptomatic and presymptomatic patients, day 1 of illness was defined as the day of diagnosis (day of the first PCR-positive clinical sample).

Six participants (27%) emitted detectable levels of SARS-CoV-2 RNA from all the expiratory activities. Two (9%) emitted detectable levels only from fine speech aerosols. Another 2 emitted detectable levels only from singing. No patients were observed to have sneezed during sample collection; however, 2 participants were observed to be coughing. Participant 4, who emitted 417 RNA copies in fine speech aerosols, was coughing during talking and singing. Participant 22 coughed frequently during all 3 activities but did not emit detectable viral RNA. Altogether, most SARS-CoV-2 RNA copies were emitted by singing (53%), followed by talking (41%) and breathing (6%) (Table 3).

Table 3.

Sum Total of Viral RNA Loads Emitted in Coarse and Fine Respiratory Aerosols, for a Subgroup of Patients With COVID-19 With Detectable SARS-CoV-2 in Respiratory Aerosols

	Coarse Fraction	Fine Fraction	Total (% of column)
Three expiratory activities	4527.3 (14.6)	26 503 (85.4)	31 030.3*
Breathinga	897 (45.8; 2.9)	1062.3 (54.2; 3.4)	1959.3 (6.3)
Talkingb	868.4 (6.9; 2.7)	11 787.5 (93.1; 38)	12 655.9 (40.8)
Singingc	2762 (16.8; 9)	13 653.2 (83.2; 44)	16 415.5 (52.9)

All values are expressed as viral N gene copies (percentage of row) unless otherwise noted. n = 13.

Abbreviations: COVID-19, coronavirus disease 2019; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2.

^aThirty minutes of tidal breathing.

^bFifteen minutes of talking with brief pauses.

^cFifteen minutes of continuous singing.

*Overall total.

Viral loads in respiratory aerosols differed significantly between the 3 activities, with 7 participants emitting more virus from talking than singing. Comparing patients with detectable SARS-CoV-2 RNA in aerosols (n = 13), the median number of viral N gene copies generated during singing was 713.6 (interquartile range [IQR]: 135.1–1216.1) compared with 477.9 (IQR: 234.5–1356.6) for talking and 63.5 (0–227.6) for breathing (Kruskal-Wallis test, P = .026). Further comparison revealed that this difference remained significant for fine aerosols but not for coarse aerosols (Table 4). Altogether, fine aerosols (≤5 µm in diameter) constituted 85.4% of the total viral RNA load detected in our study.

Table 4.

Median Viral RNA Loads Emitted for Each Expiratory Activity, in a Subgroup of Patients With COVID-19 With Detectable SARS-CoV-2 in Respiratory Aerosols

	Breathing	Talking	Singing	P
Total number	63.5 (0–227.6)	477.9 (234.5–135.6.5)	713.6 (135.2–1216.1)	.026
Fine fraction	0 (0–0)	417.0 (191.2–979.5)	366.4 (93.9–1078.1)	.013
Coarse fraction	0 (0–159.9)	0 (0–77.8)	38.4 (0–508.4)	.36

All values expressed as viral N gene copies per expiratory activity (30-minutes breathing, 15-minutes talking, 15-minutes singing) in median (interquartile range). n = 13. Medians across 3 groups were compared using Kruskal-Wallis test.

Abbreviations: COVID-19, coronavirus disease 2019; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2.

SARS-CoV-2 Variants

Sixteen participants (73%) were infected with a SARS-CoV-2 variant of concern (VOC) or variant of interest (VOI) during our study (Table 1). Due to the small number of non-VOC/VOI variants, aerosol-shedding patterns related to variant type could not be determined.

SARS-CoV-2 Culture

Virus cultures were negative after 2 consecutive passages. Vero E6 cells infected with a known SARS-CoV-2 isolate (positive control) displayed distinct Cytopathic effect (CPE), whereas uninfected Vero E6 cells (negative control) remained as a healthy cell monolayer.

Discussion

Our study demonstrates that SARS-CoV-2 can be aerosolized in the absence of coughing, sneezing, and aerosol-generating medical procedures. More than half of our study participants emitted detectable levels of SARS-CoV-2 RNA in respiratory aerosols, including 3 asymptomatic patients and 1 presymptomatic patient. Patients earlier in illness were more likely to emit detectable levels of virus, which is congruent with studies demonstrating higher viral loads in clinical samples in early illness [17]. Two participants sampled on day 3 of illness accounted for 52.4% of the total viral load captured in our study, which aligns with studies on overdispersion [18] and the predominance of superspreading events in transmission dynamics. Although the overall viral RNA loads were relatively low, they differed significantly between breathing, talking, and singing, with singing generating the most virus in aerosols and breathing the least. However, 7 participants emitted similar or more RNA copy numbers from talking when compared with singing. Although voice amplitude was not measured in our study, SARS-CoV-2 aerosol-shedding models demonstrate similar aerosol emission rates for talking loudly and singing [18]. Overall, 85% of the total viral load was emitted in fine aerosols (≤5 µm in diameter) when compared with coarse aerosols (>5 µm in diameter), which is consistent with the observation that smaller particles (0.65–4.7 µm) account for 77–79% of total virus particles shed by experimentally infected cynomolgus macaques [15]. Our results demonstrate the potential for fine respiratory aerosols to play an important role in community transmission of SARS-CoV-2, which is in agreement with other expert views suggesting that SARS-CoV-2 transmission events are driven by the airborne route [19], and could explain the difficulty in containing the virus. Our results support the calls for proper respiratory protection (ie, universal masking and N95, FFP3 respirators or equivalent for healthcare and frontline workers), airflow patterns, ventilation, filtration, and safe airborne disinfection, particularly in indoor environments [20], such as schools, to reduce exposure to SARS-CoV-2 in fine aerosols—albeit live virus could not be isolated.

While it has been previously shown that patients with COVID-19 can emit infectious virus-laden aerosols into their environments [5, 21], most environmental SARS-CoV-2 sampling studies have been unable to mechanically retrieve and isolate viable virus from ambient air in the vicinity of patients with COVID-19 [22]. Hence, the infectious proportion of virus emitted from patient expiration remains unclear. In our study, the inability to isolate viable virus from respiratory aerosol samples collected directly from patients (not from their environments) is likely related to the low viral load in our samples compared with those generally found in culturable clinical samples. Our study was limited in that respiratory swabs were not collected on the day of aerosol sampling for comparison of culturability. However, studies have reported that, for clinical SARS-CoV-2 samples, viral loads of 10⁵ to 10⁶ genome copies/mL are required for isolation of SARS-CoV-2 in vitro [23]. Our sampling methodology yielded viral RNA loads below 10^3.8 genome copies per sample, suggesting that increased sampling duration is needed to reach culturable virus levels. However, critical mutations in certain SARS-CoV-2 variants can augment virus infectivity [24]—for example, some patients infected with the Delta variant demonstrate higher viral loads in their respiratory swabs [25]. These SARS-CoV-2 variants, especially Delta [25], can cause a higher secondary attack rate than older strains [26] and may be more successfully cultured from aerosol samples in future studies, especially if patients are sampled during the short window of enhanced viral shedding [27]. More studies are warranted to test this hypothesis given that only 4 study participants were infected with non-VOC/VOI variants and only 1 patient with Delta. Thus, aerosol-shedding patterns between early and new SARS-CoV-2 strains could not be compared. Additionally, for virus culture in our study, we did not use Vero E6 cells expressing the transmembrane serine protease 2 (TMPRSS2), which can bind and cleave SARS-CoV-2 spike protein more efficiently and facilitate early surface-mediated cell entry and viral fusion [28, 29]. Although SARS-CoV-2 from saliva and respiratory swabs can be isolated using classical Vero E6 cells, a more sensitive culture assay using Vero E6 TMPRSS2 cells may be superior for culturing virus from patient aerosol samples. Human bronchial epithelial cells may also be more susceptible to infection with wild-type viruses than Vero cells [24]. Further efforts to identify optimal culture methods for exhaled breath and environmental samples are warranted.

We observed that patients earlier in illness were more likely to emit detectable levels of virus in aerosols, which is in line with a recent nonhuman primate model indicating that SARS-CoV-2 aerosol shedding is substantially reduced 4 days postinfection when compared with 2 days postinfection [15], and concurs with the higher viral loads and greater infectivity observed in human clinical samples collected early in illness [17]. Additionally, neutralizing antibodies start to appear in patients with COVID-19 at 5 days post–symptom onset [30], which may reduce and neutralize virus that is shed, preventing isolation in cell culture. Although 17 participants (77%) were seronegative at diagnosis (Table 1), a serology test nearer to the sampling day would have been a better indicator of infectiousness during aerosol sampling. Although 12 (55%) were sampled with the G-II machine within 5 days post–symptom onset (plus participant 9, sampled 2 days pre–symptom onset), we failed to isolate viable virus, suggesting that participants might need to be sampled at an earlier stage of infection, or for longer durations. Furthermore, 2 participants sampled on day 3 of illness accounted for 52% of the total viral load captured in our study, which aligns with a recent model of SARS-CoV-2 aerosol shedding demonstrating broad heterogeneity among cases [18]. Recent data also suggest that only 2% of infected individuals carry 90% of the total viral load circulating in a population at any given time [27]. This implies that only 1 in 50 active cases at any given time would be expected to have high viral loads in exhaled breath. The likelihood of capturing such cases was limited by our small sample size. The small sample size of our cohort was also further limited by a skewed demographic toward younger males. Thus, researchers must work with contact tracers to proactively isolate and strategically sample large numbers of close contacts of individuals recently infected with SARS-CoV-2 to capture the most accurate data on viral shedding in the community across all demographic ranges, for which research gaps remain.

To our knowledge, this is the first study to quantify SARS-CoV-2 in aerosols generated by singing. Our results support existing laboratory simulation data [31, 32], and can explain the many airborne SARS-CoV-2 outbreaks involving singing [8, 9, 33–35]. Higher concentrations of aerosols are generated by singing compared with talking, with loudness having a large effect on the number of aerosols produced [31, 32, 36]. However, there was high person-to-person variation in virus emission between expiratory activities in our study. Individuals who generate an above-average amount of aerosols (known as “super-emitters”) also exist, but it is unclear what causes this phenomenon [37]. Interestingly, a small number of individuals produce more aerosols from breathing when compared with talking [32], which may partially explain the asymptomatic participant in our study who emitted more SARS-CoV-2 from breathing than talking. The physiological or experimental reasons underlying these observations are unclear.

Our results underscore the importance of reducing exposure to fine respiratory aerosols through nonpharmaceutical interventions (NPIs), such as universal masking, physical distancing, and increased room ventilation during the COVID-19 pandemic. Additionally, portable high-efficiency particulate air (HEPA) cleaners in indoor environments can reduce exposure to exhaled respiratory aerosols by up to 90% in combination with universal masking, and up to 65% without universal masking [38], indicating that a multilayered approach of control measures is most effective at decreasing the risk of airborne SARS-CoV-2 transmission. Other NPIs include upper-room ultraviolet air disinfection and the use of fans to control airflow patterns within a space. In singing situations, safe distancing among singers and averting and filtering airflow from choir to audience (eg, by deploying air curtains) are important considerations. For situations involving talking, determining airflow patterns and minimizing exposure through seating and furniture configurations, distancing, and air-movement alteration (such as fans, including desk fans) would be practical options [39, 40].

Conclusions

Fine aerosols (≤5 μm) produced by talking and singing contain more SARS-CoV-2 than coarse aerosols (>5 μm) and may play a significant role in SARS-CoV-2 transmission. Thus, exposure to fine aerosols should be mitigated, especially in indoor environments where airborne transmission of SARS-CoV-2 is most likely to occur. While patients with COVID-19 early in the course of illness are likely to shed detectable levels of SARS-CoV-2 RNA in respiratory aerosols, culturing SARS-CoV-2 from these patient aerosol samples remains challenging. Person-to-person variation in virus emission is also high. Careful focus is needed on sampling methodology and duration, infectiousness of patients during sampling, and virus culture methodology. Whether isolating viable virus in respiratory aerosols can be more easily accomplished from sampling patients infected with emerging SARS-CoV-2 variants is an urgent enquiry for future investigations. Reducing airborne transmission by altering or averting direct airflow exposure in singing and speech situations indoors may be important practical options to adopt.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. The authors thank Tan Chorh Chuan and Leo Yee-Sin for their administrative support and Julian Tang for his virology insights. They also thank Chandra Sekhar and David Kok Wai Cheong for assembling the G-II machine and Somayeh Youssefi, Jacob Bueno De Mesquita, and Jovan Pantelic for guiding the G-II assembly and operation. They thank Raymond Lin and Lin Cui for sharing whole-genome sequencing data. They also thank Margaret Soon, Phoon Long Yoke, Loh Kyun Yen, Pang Jia Xin, and all nursing, infection-control, and operational staff at the National Centre for Infectious Diseases for their support. Last, they thank the NUS BSL-3 Core Facility team for their support in BSL-3 procedures. They gratefully acknowledge the University of Maryland for the availability of the G-II machine for the study.

Financial support. This work was supported by the Singapore National Medical Research Council (MOH-000443 to K. W. T., K. K. C., and M. C., and NMRC/CG/M009/2017 NUH/NUHS to J. J. H. C.) and the National University of Singapore (NUS Reimagine Research Grant to J. J. H. C.).

Potential conflicts of interest. P. A. T. reports receiving grants from Roche, Arcturus, Johnson and Johnson, and Sanofi Pasteur, and personal fees from AJ Biologicals, outside the submitted work. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Author notes

K. K. C., D. J. W. T., K. S. T., and S. W. X. O. contributed equally to this work.

T. S. T., M. H. K., Y. Q. C., H. N., and T. M. M. contributed equally to this work.