A Longitudinal Study of the Epidemiology of Seasonal Coronaviruses in an African Birth Cohort

A longitudinal study of African infants showed that endemic coronavirus infections were common and weakly associated with lower respiratory tract infection (LRTI), particularly coronavirus-OC43, which was also associated with hospitalized LRTIs.

Symptomatic or severe illness appears to be uncommon in children following SARS-CoV-2 infection, with children accounting for only approximately 2% of those experiencing disease in the current coronavirus disease 2019 (COVID- 19) pandemic [6,7]. Children may nevertheless be important as reservoirs of infection or sources of transmission of SARS-CoV-2, but there are very limited data [8,9]. Moreover, preexisting cross-reactive cellular immune responses to SARS-CoV-2, presumably due to prior infection with sHCoV, have been demonstrated and may afford some protection against COVID-19 [10,11]. We, therefore, aimed to describe the incidence of and risk factors for sHCoV infection and the association between sHCoV and LRTI in an African birth cohort. METHODS We conducted a nested case-control study of infants enrolled in a South African birth cohort, the Drakenstein Child Health Study [12], undertaken at 2 public primary health clinics in low-income communities in a peri-urban area near Cape Town. Pregnant women were enrolled during their second trimester, and infants followed from birth. All births occurred at a single hospital, from June 21, 2012, to August 12, 2015. Follow-up was at 6, 10, and 14 weeks and 6, 9, and 12 months. Where caregivers gave approval, a subset of infants ("intensive cohort") was followed up fortnightly over the first year of life. Infants were given immunizations at primary care clinics including 4 doses of a 5 vaccine combination (diphtheria, tetanus, acellular pertussis, Haemophilus influenzae type b, and inactivated polio vaccine) at 6, 10, and 14 weeks and 18 months; measles vaccine at 9 months and 18 months; and 13-valent pneumococcal conjugate vaccine at 6 weeks, 14 weeks, and 9 months.
LRTI or pneumonia surveillance took place at local clinics and the nearby district hospital [13]. Mothers were counseled regarding respiratory symptoms and advised to contact study staff whenever symptoms developed. Primary healthcare nurses and study staff were trained to recognize the World Health Organization (WHO)-defined pneumonia. Measurement of risk factors (nutrition, environment, vaccinations received, and child and maternal factors) and anthropometry were carried out at each study visit, as described previously [12], including maternal smoking or passive smoke exposure measured by self-report.
Infants were eligible for inclusion in this analysis if they were part of the intensive cohort (nasopharynx sampled fortnightly). LRTI cases were defined as clinical episodes meeting the WHO case definition for pneumonia or severe pneumonia during the first year of life, excluding congenital episodes (occurring prior to discharge from hospital after birth). Controls (without WHO-defined pneumonia, but who may have had other respiratory symptoms) were incidence-density matched to cases (1:1) by birth date (to within 2 weeks), age of presentation (to within 2 weeks), and site of enrollment. Controls could be sampled more than once at different ages, but this occurred infrequently.
Ethics approval was obtained from the University of Cape Town, Faculty of Health Sciences Research Ethics Committee. Mothers provided written informed consent.

Microbiological Sampling and Testing
Nasopharyngeal swabs were collected fortnightly over infancy for the intensive cohort. In addition, all infants had nasopharyngeal swabs taken at the time of LRTI. In a subset of infants with LRTI, swabs could not be collected due to the unavailability of study staff at the time of presentation. Nasopharyngeal swabs were transferred into nucleic acid preservation medium (PrimeStore, Longhorn Vaccines and Diagnostics, San Antonio, TX, USA), transported on ice, and frozen at -80°C for batch testing.
We tested swabs from the time of LRTI or the matched time point in controls, as well as all available swabs collected in the 90 days preceding the LRTI event (or the matched time period for controls). Swabs underwent mechanical lysis on a Tissuelyzer LT (Qiagen, Hilden, Germany) followed by total nucleic acid extraction (QIAsymphony Virus/Bacteria Mini Kit, Qiagen, Hilden, Germany). Quantitative, multiplex, realtime PCR (qPCR) with FTDResp33 (Fast-Track Diagnostics, Esch-sur-Alzet, Luxembourg) identified potential respiratory pathogens including respiratory viruses (HCoV-NL63, -229E, -OC43, -HKU1; influenza A, B, and C; parainfluenza 1, 2, 3, and 4; human metapneumoviruses A and B; rhinovirus; respiratory syncytial viruses A and B; adenovirus; enterovirus; parechovirus; bocavirus; and cytomegalovirus) and bacteria (Streptococcus pneumoniae, H. influenza, Staphylococcus aureus, and Moraxella catarrhalis). Standard curves were derived using standards supplied by the manufacturer.

Analysis
The sHCoV result was defined as positive if any of HCoV-229E, -NL63, -OC43, and -HKU1 were positive by qPCR. LRTI episodes were stratified into those in whom the swab collected at LRTI was positive for any of the 4 sHCoV (sHCoV+ cases) and those in whom the swab was negative for all sHCoVs (sHCoV− cases). Controls were similarly stratified based on whether the swab collected at the time point corresponding to LRTI in the matched case tested positive or negative for any sHCoV (sHCoV+ controls or sHCoV− controls).
Analysis of risk factors was done on a per-child basis [14]. In order to account for intraindividual clustering, we used generalized estimating equation (GEE) models to explore the risk factors for LRTI by the sHCoV subgroup. In each of the GEE models, a binomial distribution was considered with a logit link function. Data were analyzed using STATA 15.1 (STATA Corporation, College Station, TX, USA). Weight-for-age at birth Z-scores were derived from the Fenton's growth standards [15]. Socioeconomic status was evaluated in quartiles [12]. Prematurity was defined as birth at <37 weeks of gestation.
All statistical tests were 2-sided at α = 0.05. Kruskal Wallace, Wilcoxon rank-sum test, Chi-square, or Fisher's exact was used for crude comparison, as appropriate. Child follow-up over the 90-day period of sampling was used to calculate person-time at risk, with a not at risk period of 2 weeks after an LRTI.
Acquisition of sHCoV was defined by first sHCoV+ sample, a change of sHCoV species, and a sHCoV+ test following 2 consecutive sHCoV− tests or more than 2 months after the last positive test. The acquisition was assumed to start at the midpoint between the last of 2 negative consecutive tests (missed visits were regarded as having negative tests) and the first sHCoV+ test. If the first swab collected for a child was sHCoV+, the acquisition was assumed to start 7 days prior. Clearance was assumed to be the midpoint between the last sHCoV+ sample and the first of 2 consecutive negative (or missing) results. If carriage extended past the final sample collected, the acquisition was assumed to end in 7 days after the final sample collection date. The duration between acquisition and clearance was considered a carriage episode.
Overall, 1143 live births occurred in 1137 mothers, of whom 885 (77%) were enrolled into the intensive cohort with 2 weekly nasopharyngeal swab collection ( Figure 1). Cohort retention was 89% at 1 year. Among the 885 infants in the intensive cohort, there were a total of 464 LRTI events in 290 infants over the first year of life (174 LRTI events were recurrent), which were matched to 464 equivalent time points from 315 control infants. From the 464 LRTI events, 432 nasopharyngeal samples were available (samples were not available from 32 events), from 268 infants at the time of LRTI. An additional 1691 samples were available from LRTI cases from the period 90 days preceding LRTI, while 2164 samples were available from controls over the equivalent period. More samples were collected in autumn (1484 [31%]) and winter (1500 [32%]) than from spring (959 [20%]) or summer (808 [17%]).
Characteristics of infants experiencing LRTI (cases) and matched controls, stratified by sHCoV status, are given in Table 1. Eighteen percent of infants were born prematurely, 27% were HIV exposed, but all were HIV uninfected at birth. The prevalence of maternal smoking in pregnancy was 25%. Duration of exclusive breastfeeding was a median (interquartile range [IQR]) of 1.4 (0.5-3.2) months. Median ages were similar among cases and controls, irrespective of sHCoV status. Vaccine coverage was excellent: 100% for the primary series and 99% at 9 months.

Comparison of sHCoV+ and sHCoV− Cases and Controls at the Time of LRTI
Among the LRTI events, 46 of the 432 (10.6%) cases tested positive for any sHCoV at the time of LRTI (sHCoV+ cases), while, among the controls, 38 of the 464 (8.2%) cases tested positive at the matched time point (sHCoV+ controls) P = .207 (Table 2).
Risk factors for LRTI among participants who were sHCoV+ (model B) were similar to those identified when comparing all cases with all controls. However, sHCoV+ cases were more likely to have been born in winter than sHCoV+ controls (adjusted OR: 5.3; 95% CI: 1.1-25.0) or sHCoV− cases (adjusted OR: 2.8; 95% CI: 1.1-7.3). Season of the episode could not be evaluated in the GEE models together with season of birth, but the distribution of participants by category was similar (Table 1). We further explored the proportion of infants with positive sHCoV tests by season ( Figure 2). Among infants with LRTI, the proportion with a positive test for sHCoV was higher in spring and summer (13.7%) than autumn and winter (8.1%); however, this difference was not statistically significant, after adjusting for multiple episodes per participant (P = .065).
The symptoms of sHCoV+ cases and sHCoV− cases were similar, with cough, blocked nose, and diarrhea nonsignificantly more common in sHCoV+ cases. Seasonal HCoV+ cases were not more likely to be hospitalized (8/46, 17%) than sHCoV− cases (98/386, 25%) (Supplementary Table 1). Since HCoV-OC43 was most clearly associated with LRTI, we compared the clinical presentation of cases positive for HCoV-OC43 with cases positive for the other sHCoVs (Supplementary Table 2). Strikingly, all 8 infants hospitalized with LRTI who had a positive test for sHCoV were HCoV-OC43 positive.
Among controls, there was no difference in the prevalence of symptoms of upper respiratory tract infection at the time of specimen collection between sHCoV+ controls and   Table 3). The majority (61%) of sHCoV+ controls had no symptoms. There was a small but significant difference in the viral load between infants (cases or controls) with symptoms (6.6 log 10 copies/mL) and those without symptoms (6.3 log 10 copies/mL), P = .03.
The viral load (measured by quantitative PCR) of sHCoV in the nasopharynx did not change in the period leading up to LRTI and was not substantially different between infants hospitalized with LRTI, infants with ambulatory LRTI, and controls (Supplementary Figure 1). Sequential infection, or coinfection, with different sHCoV over the 90-day interval was observed in 16 infants (9 cases and 7 controls), of whom 8 had detection of 2 different sHCoVs in the same sample.  Only one sHCoV covariate could be included simultaneously in the model.
The detection of nasopharyngeal commensal bacteria (S. pneumoniae, S. aureus, H. influenzae, or M. catarrhalis) was similar when comparing cases and controls, except for H. influenzae, which was more likely to be detected in cases (Table 5). There were no differences in the detection of commensals between sHCoV+ cases and sHCoV− cases, except the odds of detecting S. pneumoniae in sHCoV+ cases vs sHCoV− cases were 1.8 (95% CI: 0.9-3.6). There were no differences in bacterial load for each of the 4 bacterial species, between cases and controls or in relation to the sHCoV status (Supplementary Figure 2).
Rhinovirus and cytomegalovirus were commonly co-detected with sHCoV (Supplementary Table 4). While RSV, influenza A and B, parainfluenza viruses, and adenovirus were all more commonly detected in cases compared with controls, there were no significant differences in the detection of any viruses in sHCoV+ cases compared with sHCoV− cases.

DISCUSSION
Our knowledge of the epidemiology of sHCoV has largely been derived from case series or from cross-sectional studies of pneumonia etiology. Here, we present detailed data on the epidemiology of sHCoV infections in infants participating in a birth cohort in an African population. Seasonal HCoV infection was common among infants followed through the first year of life (1.63 and 2.05 events per child year among controls and cases, respectively).
There was a weak but significant association between sHCoV and LRTI, consistent with the literature that these viruses uncommonly cause LRTI in immunocompetent children [16]. However, this varied between species, with HCoV-OC43 more strongly associated with LRTI, including hospitalized LRTI. HCoV-OC43 is phylogenetically the most closely related of the sHCoV to the SARS group of coronaviruses [17]. The number of cases with HCoV-229E was small, precluding the definitive assessment of association with LRTI; however, the suggestive results indicate the need for more studies.
The risk factors for sHCoV-associated LRTI were similar to those for all LRTI. However, while there was no seasonal pattern to sHCoV infection, birth in winter was significantly associated with sHCoV detection among cases and with the presence of LRTI among infants positive for sHCoV. Season of birth has previously been associated with poor birth outcomes [18] and nutritional indicators in children [19,20], which may be related to dietary or environmental factors [19]. Children born in autumn or winter have an increased risk of wheezing or asthma  in later life, which may be partly mediated by respiratory tract infection [21].
In 16 infants, we identified coinfection with more than one sHCoV simultaneously, or within short intervals of each other, suggesting that immunity against one sHCoV does not give complete cross protection against other sHCoV species.
There were no symptoms or signs that distinguished children infected with sHCoV among cases. The majority of sHCoV+ controls were asymptomatic. This is consistent with reports of seasonal coronavirus in children that have described nonspecific symptoms or signs [6]. However, in our study, the presence of symptoms was associated with higher viral load, and infants with HCoV-OC43-associated LRTI had more severe illness, requiring hospitalization. The apparent low pathogenicity of sHCoV does not allow us to draw any similar conclusions about SARS-CoV-2.
Since sHCoV status and sHCoV viral load were only weakly associated with LRTI, it is probable that LRTI in many sHCoV+ cases was due to other pathogens. Indeed, viruses that are significantly associated with LRTI, such as RSV, influenza, and parainfluenza viruses, were frequently detected. There was no significant association between other viruses and the presence of sHCoV in cases.
Data from in vitro and animal studies suggest that interactions between human coronaviruses and bacteria may be important. HCoV-NL63 infection enhances the adherence of S. pneumoniae to primary human epithelium cultures [22]. Lipoteichoic acid, from the cell wall of S. aureus, stimulates the secretion of proinflammatory cytokines, which increased susceptibility to coronavirus in a swine model [23]. We, therefore, explored co-detection with bacteria. We identified a potential association between S. pneumoniae and sHCoV infection among cases (OR: 1.8; 95% CI: 0.9-3.6), which requires further study. Interactions between S. pneumoniae and influenza A virus are thought to have been important in the 1918 influenza pandemic [24], while co-detection of S. pneumoniae with RSV has been associated with increased clinical severity compared with detection of RSV alone [25].
Limitations of our study which might affect generalizability are the use of only 2 study sites; however, these sites are broadly representative of many childhood populations in lowand middle-income countries. Only the first year of life was studied; however, this is the highest risk period for LRTI in children. Since we only tested samples from the 90 days prior to LRTI event, we may have underestimated the duration of carriage. We used the WHO definition of pneumonia, which is based on clinical symptoms alone, and may lack specificity. Strengths include detailed longitudinal sampling, testing for a wide range of LRTI pathogens, measurements of a range of risk factors, and the community-based cohort study with high cohort retention.
In summary, sHCoV infections were common in infants in this setting and associated with LRTI, particularly HCoV-OC43. Seasonality of birth was associated with sHCoV+ LRTI. Our finding in relation to the co-detection of S. pneumoniae suggests that interactions of human coronaviruses with bacteria in the pathogenesis of LRTI require further study.

Supplementary Data
Supplementary materials are available at the Journal of the Pediatric Infectious Diseases Society online (http://jpids.oxfordjournals.org). Abbreviations: OR, odds ratio; 95% CI, 95% confidence interval; nHCoV+, defined as positive if positive for any of the 4 endemic coronaviruses. Models adjusted for age at episode/reference.