Prolonged Shedding of Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) at High Viral Loads Among Hospitalized Immunocompromised Persons Living With Human Immunodeficiency Virus (HIV), South Africa

Abstract Background We assessed severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) RNA shedding duration and magnitude among persons living with human immunodeficiency virus (HIV, PLHIV). Methods From May through December 2020, we conducted a prospective cohort study at 20 hospitals in South Africa. Adults hospitalized with symptomatic coronavirus disease 2019 (COVID-19) were enrolled and followed every 2 days with nasopharyngeal/oropharyngeal (NP/OP) swabs until documentation of cessation of SARS-CoV-2 shedding (2 consecutive negative NP/OP swabs). Real-time reverse transcription-polymerase chain reaction testing for SARS-CoV-2 was performed, and cycle-threshold (Ct) values < 30 were considered a proxy for high SARS-CoV-2 viral load. Factors associated with prolonged shedding were assessed using accelerated time-failure Weibull regression models. Results Of 2175 COVID-19 patients screened, 300 were enrolled, and 257 individuals (155 HIV-uninfected and 102 PLHIV) had > 1 swabbing visit (median 5 visits [range 2–21]). Median time to cessation of shedding was 13 days (interquartile range [IQR] 6–25) and did not differ significantly by HIV infection. Among a subset of 94 patients (41 PLHIV and 53 HIV-uninfected) with initial respiratory sample Ct-value < 30, median time of shedding at high SARS-CoV-2 viral load was 8 days (IQR 4–17). This was significantly longer in PLHIV with CD4 count < 200 cells/µL, compared to HIV-uninfected persons (median 27 days [IQR 8–43] vs 7 days [IQR 4–13]; adjusted hazard ratio [aHR] 0.14, 95% confidence interval [CI] .07–.28, P < .001), as well as in unsuppressed-HIV versus HIV-uninfected persons. Conclusions Although SARS-CoV-2 shedding duration did not differ significantly by HIV infection, among a subset with high initial SARS-CoV-2 viral loads, immunocompromised PLHIV shed SARS-CoV-2 at high viral loads for longer than HIV-uninfected persons. Better HIV control may potentially decrease transmission time of SARS-CoV-2.

Immunocompromised persons are thought to shed severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) for a longer duration, increasing time for viral transmission and potentially driving within host viral evolution [1][2][3]. South Africa has 7 500 000 persons living with human immunodeficiency virus (HIV, PLHIV), and there are no systematically collected data on duration of SARS-CoV-2 shedding in PLHIV. We hypothesized that PLHIV may shed SARS-CoV-2 for a longer period of time and at a higher viral load than HIV-uninfected individuals.
SARS-CoV-2 shedding from the upper respiratory tract extends for a mean of 17 days (15.5-18.6) [4]. Unlike severe acute respiratory syndrome coronavirus 1 (SARS-CoV-1), viral shedding of SARS-CoV-2 from the upper respiratory tract peaks on or before symptom onset, allowing some viral transmission to occur before symptom onset in infected individuals [4][5][6][7][8]. Quantifying SARS-CoV-2 viral load (Log 10 RNA copies/mL) is calculated by converting qualitative real-time reverse transcription-polymerase chain reaction (rRT-PCR) cycle threshold values using calibration curves based on quantified E-gene in vitro RNA transcripts [9]. However, studies M A J O R A R T I C L E on SARS-CoV-2 infectiousness indicate that successful virus isolation is most likely from specimens with a rRT-PCR cycle threshold (Ct) value < 34(~7 Log 10 copies/mL); however, not all of these specimens may be positive for viable virus [9,10]. Using this threshold, infectiousness declines significantly 8 days after becoming symptomatic, even though SARS-CoV-2 viral RNA persistence has been shown to occur for months in some individuals [9,11,12].
Various factors have been associated with increased shedding duration, including increased age, male sex, severity of illness and use of corticosteroids [13,14]. There are several published case reports indicating prolonged SARS-CoV-2 transmission in immunocompromised persons, mostly with cancers or autoimmune conditions, and more recently in an immunocompromised person living with HIV [15][16][17][18].
We aimed to evaluate the overall duration of SARS-CoV-2 shedding in upper respiratory tract specimens and duration of high viral load shedding in a cohort of PLHIV and HIVuninfected persons hospitalized with COVID-19 in South Africa. We also estimated SARS-CoV-2 shedding duration in stool and blood specimens, and described serologic responses to SARS-CoV-2 infection.

Study Design and Setting
We conducted a prospective cohort study from 1 May through 31 December 2020 (spanning the first wave and beginning of the second wave of the COVID-19 pandemic in South Africa). Persons hospitalized for symptomatic COVID-19 at one of 20 hospitals situated in 8 of the 9 South African provinces were invited to participate in the study if they met the following inclusion criteria: aged 18 years and older, laboratory confirmed diagnosis of SARS-CoV-2 in the previous 5 days, resided within a 50-kilometer radius of the hospital, and had laboratory confirmation of their HIV status.
Using standardized case report forms, demographic and clinical details were collected at enrollment, daily while in hospital, and at discharge from hospital/cessation of shedding/ death of the participant. Included variables are detailed in Supplementary Material.
At enrollment and every second day thereafter, until cessation of shedding or death, a combined nasopharyngeal/oropharyngeal (NP/OP) swab was collected by trained nursing staff, using 2 flocked nylon plastic shaft swabs (1 for the nasopharynx and 1 for the oropharynx), which were then inserted together into Universal Transport Medium and transported on ice to the laboratory. Sensitivity of the combined NP/OP swab has been shown to be 97% for detection of SARS-CoV-2 in the upper airways [19,20]. A rectal swab or stool was collected from patients at the same time intervals and transported to the laboratory in sealed containers. Results are available in Supplementary Material. Patients were followed up at home if they were still SARS-CoV-2 positive on NP/OP swab on discharge from hospital.
Whole blood and serum specimens were taken on enrollment and at days 7, 14, and 21 post symptom-onset for SARS-CoV-2 testing and serology. Results for SARS-CoV-2 testing of blood samples are available in Supplementary Materials. All specimens were transported to the National Institute for Communicable Diseases (NICD) in Johannesburg for processing and testing.

Laboratory Diagnostics
rRT-PCR for the qualitative detection of nucleic acid from SARS-CoV-2 was performed on NP/OP, stool/rectal swabs and blood specimens using the Allplex™ nCoV 2019 kit (Seegene, Seoul, South Korea). Specimens were considered positive for SARS-CoV-2 nucleic acids if the Ct was < 40 for ≥ 1 of 3 gene targets. A nucleocapsid gene (N gene) Ct value < 30 on NP/OP specimens was used as a proxy for a high viral load based on published data showing a high correlation between low Ct values (using various gene targets), high viral load and increased odds of shedding cultivable virus [10,12,21,22].
Antibody detection against trimeric ectodomain HexaPro spike protein was performed on serum specimens as described previously [23,24]. Absorbance at 450 nm was measured, and specimens with optical density (OD) > 0.4 were considered positive for anti-spike protein antibodies.
SARS-CoV-2 sequencing was performed on the first NP/OP specimens on enrollment from 10 randomly selected participants of 29, who demonstrated high viral load shedding for > 14 days. Clade and lineage assignments were made using the online Nextclade (https://clades.nextstrain.org/) and Pangolin (https:// pangolin.cog-uk.io/) applications, which also enable identification of known variants of concern as well as novel mutations. See Supplementary Methods for more details.

Definitions
Shedding was defined as presence of SARS-CoV-2 nucleic acid in a specimen as detected by a positive SARS-CoV-2 rRT-PCR result. Participants were deemed to have stopped shedding SARS-CoV-2 from the respiratory tract once 2 consecutive NP/OP swabs, taken at least 2 days apart, tested negative for all 3 gene targets on rRT-PCR. Time to cessation of shedding was taken from date of symptom onset to the date of the last rRT-PCR SARS-CoV-2 positive NP/OP swab prior to the two consecutive negative swabs. Participants were deemed to be shedding SARS-CoV-2 virus at high viral loads if their NP/ OP swab N gene Ct value was < 30. Persistence of high viral load was measured in the subgroup of individuals whose first study NP/OP swab had an N gene Ct value < 30 and was measured from date of symptom onset to the last date where NP/ OP swab N gene Ct value was < 30. PLHIV were deemed to be significantly immunocompromised if their CD4 T-lymphocyte count was < 200 cells/µL and not HIV virally suppressed if their HIV viral load measured > 400 copies/mL in the 3 months before hospital admission [25,26]. Severity of COVID-19 was categorized using respiratory rate at time of admission according to the WHO clinical classification of COVID-19 and a quick Sequential (Sepsis-related) Organ Failure Assessment (qSOFA) score [27,28].

Statistical Analysis
Data were captured on a real-time data capture (REDCAP) database and transferred onto a password protected Microsoft Access database management system [29]. Statistical analysis was performed using Stata version 14 (StataCorp Inc., College Station, Texas, USA).
We used Kaplan-Meier estimates and Weibull accelerated failure time regression models to measure time from symptom onset to (i) cessation of shedding and (ii) low viral load shedding (at Ct values > 30). The Weibull model was chosen as it allows explanatory variables to proportionately increase or decrease the time to the end-point [30]. Hazard ratios measured the hazard of reaching the end-points with a ratio of < 1 indicating a longer duration of shedding or a longer duration of high viral load shedding. P-values < .2 on univariate analysis were included in all multivariable models, and nonsignificant variables (P-value > .05) were dropped using step-wise manual backward elimination. Patients with missing data were dropped from the models.

SARS-CoV-2 Sequencing of Upper Respiratory Specimens on Enrollment
Eight different SARS-CoV-2 phylogenetic lineages were detected from enrollment NP/OP swabs from the 10 randomly selected participants demonstrating high SARS-CoV-2 viral load shedding for > 2 weeks. These included 2 participants with B.    anti-spike protein antibody by day 7 or 14 compared to HIVuninfected persons. A wide variety of SARS-CoV-2 phylogenetic lineages were detected in the small group of participants who shed SARS-CoV-2 at high viral loads for more than 2 weeks. We showed prolonged duration of high viral load SARS-CoV-2 shedding in PLHIV with low CD4 cell counts. Prolonged influenza virus shedding has been described in immunocompromised PLHIV, and it has been associated with viral evolution and antigenic drift within individual immunocompromised patients, similar to that occurring in communities [31][32][33][34]. Further studies are needed to explore SARS-CoV-2 intra-individual adaptations in a similar manner to influenza, particularly in PLHIV.
The national prevalence of HIV infection in South Africa is 13%, and it is estimated that 4.4 million of the 7.5 million PLHIV in South Africa are not HIV virally suppressed. (UNAIDS https://www.unaids.org/en/regionscountries/countries/southafrica). If immunocompromised PLHIV shed SARS-CoV-2 for longer at higher SARS-CoV-2 viral loads, this could potentially impact the risk of viral evolution and transmission, especially in South Africa. Alternately, PLHIV with CD4 counts of ≥200 cells/µL shed for a similar time and at a similar SARS-CoV-2 viral load than HIV-uninfected persons. This finding highlights the importance of early HIV diagnosis and treatment with the goal of HIV viral suppression, thus improving the health of PLHIV and minimizing transmission risk of SARS-CoV-2 in the general population. The study was conducted prior to any COVID-19 vaccinations being available in South Africa. Now that vaccines are available, prioritizing COVID-19 vaccination of PLHIV may be another strategy for minimizing SARS-CoV-2 spread.
Recently, accelerated SARS-CoV-2 viral evolution was described in an immunocompromised host with persistent SARS-CoV-2 infection, as well as in a PLHIV host with poor HIV viral suppression [17,35]. SARS-CoV-2 viral mutations occur commonly within hosts and within populations and prolonged high viral load shedding provides a good opportunity for virus evolution and transmission of potential viral mutations. Between June and October 2020, 44 different SARS-CoV-2 lineages were detected in South Africa, with the B .1, B.1.1, B.1.1.448, B.1.1.54, and C.1 lineages accounting for 59% of all sequences (n = 1705) [36]. In this study the multiple SARS-CoV-2 phylogenetic lineages detected reflect the lineage diversity during the earlier waves in South Africa, indicating that persistent shedding was not dominated by a specific lineage or variant. All lineages detected in these individuals had been identified in national genomic surveillance efforts, with 7 of the 8 being lineages first detected in South Africa [36]. Going forward we aim to investigate within host viral evolution by sequencing sequential specimens of the participants with prolonged high viral load shedding, many of whom were immunocompromised PLHIV.  Obese patients in our study showed a 45% increase in median time of SARS-CoV-2 shedding compared to their nonobese counterparts. However, obesity was not associated with prolonged duration of high viral load SARS-CoV-2 shedding. Recent studies indicate that obesity may be associated with prolonged SARS-CoV-2 shedding, in a similar manner to prolonged shedding seen in obese adults with influenza A virus [37,38]. See Supplementary Material for further discussion.
Most participants developed anti-spike protein antibodies by day 14, with fewer immunocompromised PLHIV developing antibodies compared to HIV-uninfected persons. Overall, lower median anti-spike protein antibody titers were detected in PLHIV. Our study showed 14 individuals (4 HIV-uninfected  persons and 10 PLHIV-8 of whom had low CD4 count) who did not develop antibodies within the initial 21 days' follow-up period, half of whom showed prolonged shedding (including 5 PLHIV with CD4 count < 200 cells/µL). Other studies have found similar findings in immunocompromised individuals with prolonged viral shedding having negative seroconversion [15,16].
The cohort of participants from 20 public hospitals is likely representative of the hospitalized COVID-19 South African adult population; however, the analysis did not account for within-hospital correlations due to case mix, treatment strategies, or interhospital differences in quality of care [39]. Thrice weekly collection of NP/OP swabs to detect ongoing SARS-CoV-2 viral shedding may result in underestimated viral shedding due to interval censoring of the data. A major limitation of our study is the use of qualitative Ct values as a proxy for viral load. Although all specimens were tested in the same laboratory, using standardized technique and assays, quantitative RT-PCR was not performed using calibration curves to determine viral load. In addition, we were unable to perform viral culture to confirm shedding of infectious virus at proxy Ct values < 30. Our findings may not be generalizable to outpatients with mild COVID-19, as hospitalized COVID-19 patients may represent more severe disease and therefore shed for a longer duration. Nonenrollment of participants due to unavailability of HIV test data (37% (702/1875) of nonenrollments) may have introduced bias due to potential differences in persons willing to undergo HIV testing versus those who are unwilling. Previous South African studies have shown similar unknown HIV results, as opt-out policies on HIV testing are not routinely followed [39]. Although sensitivity analyses in our study did not show much effect when excluding patients who died (data not shown), it should be noted that in-hospital mortality in our study was low (8%) compared to 23% in Jassat et al, as patients too ill to consent would have been excluded [39]. Follow-up of participants' post-hospital discharge was challenging due to stigma of COVID-19 in the communities, safety concerns and travel limitations. Fifty-eight participants were lost to follow-up prior to cessation of shedding; however, the accelerated time failure Weibull regression analysis included patient data up until the patient was lost to follow-up, thereby utilizing all data generated.
Immunocompromised persons living with HIV shed SARS-CoV-2 at a higher viral load for a longer duration, and have lower anti-spike protein antibody levels than HIV-uninfected persons, which could lead to increased transmission and ongoing viral evolution of SARS-CoV-2, unless HIV control is achieved. During the COVID-19 pandemic, access to HIV testing and treatment with an aim for HIV viral suppression remains extremely important. In high HIV-prevalence settings better HIV control may potentially decrease the transmission period of SARS-CoV-2 in communities.

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.
Supplementary Figure 1: Mosaic of relative SARS-CoV-2 C t -values of nasopharyngeal/oropharyngeal swabs taken every second day from persons living with HIV and HIV-uninfected persons who were hospitalized with COVID-19, South Africa, May through December 2020 (n = 257).
Supplementary Table 1: Reasons for non-enrolment by HIV status amongst persons screened for enrolment into the SARS-CoV-2 shedding study in South Africa Supplementary Table 2: Accelerated Weibull Regression for duration of SARS-CoV-2 N gene detection by RT-PCR among hospitalized persons living with HIV, South Africa Supplementary Table 3: Accelerated Weibull Regression for time taken from onset of symptoms to reach SARS-CoV-2 RT-PCR N gene C t -value of > 30 among a subset of hospitalized persons living with HIV with laboratory confirmed SARS-CoV-2 infection and initial study SARS-CoV-2 RT-PCR N gene C t -value of < 30, South Africa Supplementary