Update on ongoing workplace SARS-CoV-2 surveillance clinical study. On March 23, 2020, Oak Crest initiated the ongoing clinical study entitled, “Longitudinal Characterization of COVID-19 Prevalence and Incidence in a Small Working Institution with Both Public Health and Diagnostic Aims”18. The two primary study aims are to: (i) characterize the rate of SARS-CoV-2 acquisition in a small cohort of participants interacting on a daily basis in the workplace; and (ii) determine the ability of these data to manage workplace SARS-CoV-2 exposure and consequences, minimizing further spread as per public health advisories. In addition, related exploratory study goals include characterizing the rate of SARS-CoV-2 acquisition in employee and household members, quantifying antibody-specific responses in blood at baseline (previously exposed) and while on study (to capture asymptomatic/presymptomatic, newly infected individuals), and characterizing viral shedding parameters in saliva and stool samples. The study has remained active to date and sampling of participants has continued three times per week, except for a period between June 9, 2021 to January 3, 2022 where the testing frequency was reduced to twice per week due to low SARS-CoV-2 positivity rates in Los Angeles County. To date, 142 participants have been enrolled and close to 7,000 SARS-CoV-2 RT-qPCR tests (21,000 reactions) have been performed.
The high incidence in Los Angeles County COVID-19 cases between mid-November, 2020 and mid-March, 2021 was reflected in positive RT-qPCR SARS-CoV-2 test results for nine study participants (Table 1). All were unvaccinated against SARS-CoV-2 at that time.
Table 1. Demographics of study participants testing positive for SARS-CoV-2 by RT-qPCR who were followed longitudinally. All participants were unvaccinated against SARS-CoV-2 at the time of the first positive test result.
Characteristic
|
Value
|
No. of participants
|
9
|
Female, no. (%)
|
6 (67)
|
Male, no. (%)
|
3 (33)
|
Age (yrs), median (range)
|
25 (19-53)
|
Race and ethnicity, no. (%)
|
|
Black or African-American
|
0
|
White
|
9 (100)
|
Hispanic
|
6 (67)
|
Non-Hispanic
|
3 (33)
|
Asian
|
0
|
Other
|
0
|
Employees who tested positive immediately were isolated and, consequently, no workplace SARS-CoV-2 transmissions occurred. Four individuals from this positive cohort were available for direct observation and frequent sampling. At some point post diagnosis, all four subjects reported symptoms of headaches, body aches, chills, sore throat, cough, runny nose, nausea, diarrhea, vomiting and/or nasal drip; each subject reported three or more symptoms. These symptoms were consistent with those reported by the remaining five individuals from the positive cohort. No participants were hospitalized. The duration of participants’ identified symptoms ranged from 10 days to > 2 months. During the above period, intensive longitudinal sample collection was performed for participants testing positive for SARS-CoV-2, and the corresponding results are presented below.
Cells were tightly bound in clinical nasal swab specimens when examined by transmission electron microscopy (TEM). Nasal swabs (25-086-PD, Puritan Medical Products, Guilford, ME) from study participants found to be positive for SARS-CoV-2 RNA by RT-qPCR were examined by TEM to investigate how the clinical specimen was associated with the high surface area polyester fiber collection tip, and visualize individual viral particles. Representative TEM images are shown in Fig. 1, demonstrating the close association of cells with swab fibers and the challenges of identifying SARS-CoV-2 virions in clinical nasal swab specimens.
The majority of cells directly attached to swab fibers were lysed and the cytoplasm frequently appeared to be adsorbed between the individual fibrils. Intact cells were commonly found in association with the lysed cell layer (Fig. 1A). Outside of these closely associated cells there was little extracellular material present in the samples. Within the entire cell population, only a small subset showed signs of viral infection evidenced by the presence of unusual double membrane structures that were not observed in epithelial cells of an RT-qPCR-negative specimen. This observation led us to speculate that these membranes represent viral assembly organelles (Fig. 1B). We were unable to unambiguously identify SARS-CoV-2 virions within the cytoplasm as these cells generated numerous vesicles in a wide range of sizes with similar ultrastructural features. We did observe three possible virions in the extracellular space that show features of virions and are of the appropriate diameter (99 nm, 99 nm, and 92 nm, Fig. 1C). However, the extracellular space in these specimens either was empty or filled with cytoplasmic material of lysed cells, rendering this identification inconclusive (Fig. 1D). Unambiguous identification of infected cells and virions in these clinical specimens was complicated further by the presence of different cell types, variable structural preservation due to cell lysis, and confusing cellular structures such as vesicles with “outside-in” ribosomes and extracellular protrusions (Fig. 1E and Fig. 1F).
SARS-CoV-2 viral kinetics differ across anatomic compartments. One participant (subject 21) provided nasal, oral, and stool samples for up to 54 days. The SARS-CoV-2 copies per swab, based on RT-qPCR analyses of the samples in all three compartments are shown in Fig. 2.
The viral load kinetic profiles across compartments were heterogeneous and decayed rapidly in nasal and oral swab samples (Fig. 2A and C). However, the maximal SARS-CoV-2 copies (Cmax) were 370 times higher in the nasal swab samples [Cmax(nasal), 1.01×108 copies/swab; Cmax(oral), 2.75×105 copies/swab]. Oral swab and saliva samples collected and pre-purified with the Super SAL2 kit collection device afforded similar SARS-CoV-2 copy numbers (Fig. 2C-D) and decay profiles, but the swab data were less noisy. The SARS-CoV-2 dynamics in stool specimens (Fig. 2B) were different from the other sampled compartments. A bimodal profile was observed and the Cmax was 3,800 times lower than in the nasal swab samples [Cmax(stool), 2.66×104 copies/swab], while the time to reach Cmax (tmax) was 18 days later [tmax(nasal), 2 d; tmax(stool), 20 d]. Despite this large difference in Cmax, subject 21 remained positive for SARS-CoV-2 RNA for close to 40 days in nasal and stool samples, while positivity in oral samples only lasted for ca. two weeks.
Temporal SARS-CoV-2 profiles are heterogeneous. Six study participants were followed longitudinally as their SARS-CoV-2 RT-qPCR status changed from negative to positive. The nasal swab viral copy numbers as a function of time are shown in Fig. 3.
The overall features of the SARS-CoV-2 RNA concentration-time semilog plots are shared across eight participants (Fig. 2A, Fig. 3, Fig. 4A) and follow similar trends. A rapid exponential growth (proliferation) phase is followed by a bimodal decay in SARS-CoV-2 RNA copies, characterized by an initial rapid decay followed by a slow, second decay (clearance) phase. The length of the clearance profile varied widely across participants and, in several cases, the decline in RNA copy numbers was interrupted by small concentration spikes (e.g., Fig. 3F, Day 20), suggestive of multimodal decay kinetics.
The highest measured viral loads for each participant, Cmax(CoV RNA), varied across multiple orders of magnitude and were: subject 21 (Fig. 2A), 1.0×108 copies/swab; subject 38 (Fig. 3A), 5.5×109 copies/swab; subject 48 (Fig. 3B), 3.5×109 copies/swab; subject 63 (Fig. 3C), 4.6×106 copies/swab; subject 83 (Fig. 3D), 5.7×107 copies/swab; subject 84 (Fig. 3E), 4.3×109 copies/swab; subject 85 (Fig. 3F), 1.1×1010 copies/swab; subject 2 (Fig. 4A), 1.3×1010 copies/swab.
Some of the study participants obtained complementary, nasopharyngeal RT-qPCR SARS-CoV-2 tests through CLIA laboratories (arrows in Fig. 2A, Fig. 3A, 3B, 3D, 3E). It is noteworthy that in four out of seven tests (57%), a negative result was obtained with the CLIA test, while a positive test result was obtained in our study on the same sampling day. Differences in the initial stages of sample preparation in the two tests (i.e., the current study and CLIA laboratory) present potential consequences for assay sensitivity. In our study, swab samples were either collected dry (i.e., no preservative added, processed within ca. 40 min of collection) or preserved with RNA Shield (300 µL, Zymo Research, Tustin, CA) and typically processed within 24 h of collection. Extraction of viral RNA from these samples was carried out by first adding lysis buffer (400 µL) followed by column purification of the entire sample volume (i.e., either 400 or 700 µL). In CLIA laboratory SARS-CoV-2 RNA RT-qPCR tests, the swab is preserved in larger volumes of transport media (typically 3-10 mL) and in the subsequent RNA extraction step, an aliquot of the sample is used. It is therefore expected that, for the same sample, lower quantities of viral RNA are analyzed in the CLIA laboratory test than in our study, likely leading to lower sensitivities for the former. These considerations need to be taken into account when comparing cycle threshold (Ct) values, or viral RNA copies per unit volume, across studies. Consequently, we have reported viral loads as SARS-CoV-2 copies per swab above.
SARS-CoV-2 Doubling times in the growth phase generally are remarkably consistent. The regular, repeated, high intensity sampling in our observational study allowed early identification of SARS-CoV-2 positivity, providing an opportunity for measuring the in vivo SARS-CoV-2 doubling time, Td, during the growth phase. Successful mapping of the growth phase is best described by Fig. 4.
The slope of the semilog plot shown in Fig. 4B was used to calculate Td for subject 2 (2.8 h), and a similar approach was employed for three additional participants (subjects 21, 38, and 48) where sufficient early-stage data were available (Fig. 5).
Three out of the four participants had remarkably similar doubling times, while one participant (subject 48) exhibited considerably slower SARS-CoV-2 growth (Fig. 5). The Td value for subject 48 was calculated from three timepoints spanning 24 h (Fig. 3B). Subject 48 also took longer to clear the virus than any of the other participants (Fig. 3B).
Viral dynamics model analysis. A Bayesian model8 was used to estimate individuals' peak viremia, and duration of the viral proliferation and clearance phases based on the experimental data shown above. We removed series of three or more consecutive negative tests (Ct = 40) to avoid overfitting to these trivial values. Viral load was log10-transformed and a piece-wise linear model was fitted using control points for time of infection, time and height of peak viral load, and time to infection clearance. The control points were inferred with the Hamiltonian Monte Carlo method using Stan (version 2.21.0) in R studio (version 1.2.5033)20.
For the main analysis we used the priors (informative, uninformative, and biased set) as described in Kissler et al.8,21. The term “informative priors" literally refers to “prior beliefs”; i.e., what was known before the experiment. The priors also can be uninformed, also known as “uniform”, where no assumptions from prior experiments are made, or strongly biased. We also conducted a sensitivity analysis using all three approaches to assess the robustness to the choice of priors (vide infra).
Informative prior settings were used as in Kissler et al.8:
μωρ ~ Normal(2.7, 14/6) [0.25, 14]
μωr ~ Normal(7.4, 30/6) [2, 30]
μδ ~ Normal(20, 40/6) [0, 40]
where wp is the duration of the proliferation stage (constrained to 0.25-14 days), wr is the duration of the clearance stage (constrained to 2-30 days to prevent inferring unrealistic values), d (constrained to 0-40) is the absolute difference in Ct between the limit of the detection and the lowest Ct. The estimated trajectory using the informative priors for eight individuals is shown in Fig. 6.
For sensitivity analysis, we conducted a similar analysis (Fig. 7) with the following set of uninformative priors settings:
μωρ ~ Normal(14/2, 14/6) [0.25, 14]
μωr ~ Normal(30/2, 30/6) [2, 30]
μδ ~ Normal(40/2, 40/6) [0, 40]
We then applied the model using the following set of biased prior settings:
μωρ ~ Normal(0, 14/6) [0.25, 14]
μωr ~ Normal(0, 30/6) [2, 30]
μδ ~ Normal(20, 40/6) [0, 40]
The same parameter constraints were used for each set of priors. Viral trajectories estimated for different priors are compared in Table 2.
Table 2. Estimated viral trajectories for different priors presented as means with the 95% confidence interval (CI) in brackets.
|
Informative priors
|
Uninformative priors
|
Biased priors
|
Clearance time (d)
|
15.64
(13.09, 18.81)
|
17.16
(14.54, 21.29)
|
15.25
(12.73, 16.00)
|
Peak Ct value
|
10.49
(6.93, 14.80)
|
10.43
(6.79, 15.12)
|
10.52
(7.02, 14.93)
|
Proliferation time (d)
|
3.07
(0.68, 7.08)
|
5.95
(1.42, 9.91)
|
2.42
(0.54, 6.91)
|
Table 2 shows the results of the model analysis under three sets of different priors. The estimated values are fairly consistent under each prior, but it should be noted that our sample size was limited. We observed a longer mean clearance time than reported by Kissler et al.8 that could be due to our extended and more frequent sampling period. Further, the mean peak Ct value was much lower in our sampled population (i.e., higher viral loads) compared to the reference study8.
Kinetics of innate immune activation and responses differ among participants. Longitudinal plasma cytokine concentrations for participants testing positive for SARS-CoV-2 are shown in Fig. 9. The data are normalized temporally to the day of the first positive SARS-CoV-2 RT-qPCR test result (Day 1). Additional, related cytokine data are included in Supplementary Fig. 1.
Overall, the cytokine/chemokine concentrations declined from the first day of SARS-CoV-2 RT-qPCR positivity, but the temporal profiles varied across participants and cytokines. Interleukin-1 beta (IL-1β) and interleukin-1 receptor antagonist (IL-1Ra) plasma concentration timeseries did not follow a consistent trend across participants, but appeared to peak around Day 20 prior to declining for two individuals (subjects 21 and 38, Fig. 9A-B). Interleukin-8 and -12 plasma concentrations generally declined over time for all participants (Fig. 9C-D). Concentrations of the chemokine interferon gamma-inducible protein-10 (IP-10, also known as CXCL10) generally decreased slowly over time, except in subject 83 where they remained constant (Fig. 9E). The most disparate temporal concentration profiles for the measured inflammatory markers were observed for soluble CD40L (sCD40L, Fig. 9F), shed by activated T lymphocytes and platelets. In two participants (subjects 21 and 83), the plasma concentrations remained relatively constant over 60 days. In two other participants (subjects 38 and 48), the concentrations rose sharply within the first three weeks and then declined rapidly. The final two participants (subjects 2 and 84, barely visible behind the plot from subject 83) only had quantifiable concentrations in the first 25 days.
There were no notable temporal changes in the plasma concentrations of the remaining inflammatory markers measured (Supplementary Fig. 1). Human interferon alpha-2 (IFNα2) and interferon gamma (IFNγ) only were detected in a small subset of samples. Interleukin-1 alpha (IL-1α) and interleukin-2, -4, -6, -10, -13, -15, and -17A (IL-2, IL-4, IL-6, IL-10, IL-13, IL-15, and IL-17A) concentrations either were below the limit of quantification in most samples (IL-2, IL-6, IL-15), or were only observed in a minor subset of samples. Monocyte chemoattractant protein-1 (MCP-1) and macrophage inflammatory protein-1 alpha and beta (MIP-1α and MIP-1β) plasma concentrations remained constant over time in the majority of samples. The concentration of tumor necrosis factor alpha (TNF-α) was not quantifiable in most samples except for subject 48, where a concentration maximum was observed at Day 16, followed by a concentration plateau until the last sample on Day 74 (Supplementary Fig. 1).
Neutralizing antibody temporal decay profiles differ widely across participants. Longitudinal humoral responses against SARS-CoV-2 in the study cohort are shown in Fig. 10. While the IgG concentrations generally were the most stable, the temporal decay profiles are strikingly different across participants. Robust antibody responses were measured for all three analytes (IgG, IgA, and IgM) over the ca. 3-month period following the first SARS-CoV-2 positive RT-qPCR test result, with the exception of subject 48 (Fig. 10D). For this participant, IgG responses were lower than for the rest of the cohort and were delayed. IgA Concentrations were only quantifiable at one timepoint (Day 28).
When antibody concentrations declined over time, the corresponding half-lives, t1/2, could be calculated using a one-phase decay model (Fig. 11). Due to the stability of the IgG serum concentrations over the ca. 90-day window of analysis, only t1/2 values for IgA and IgM could be calculated. When the IgA and IgM t1/2 values were compared using a Wilcoxon matched-pairs signed rank test, the datasets were found not to be significantly different (P > 0.9999). When an unpaired analysis was performed on the data using a Mann-Whitney test, the same result was obtained (P > 0.9999).
An unusually mild case of COVID-19. One participant (subject 82) developed asymptomatic COVID-19, characterized by low SARS-CoV-2 viral loads in nasal swab samples [Fig. 12A; Cmax(CoV RNA), 4.8×103 copies per swab]. The sudden shift to SARS-CoV-2 RT-qPCR positivity was preceded by eleven consecutive negative test results (green stars, Fig. 12A). Serum antibody concentrations over the same time period only could be quantified for IgG and these remained stable throughout (Fig. 12B; median [IgG], 162 ng mL-1). Interestingly, the IgG measurement on Day 6 (144 ng mL-1) corresponds to the period of SARS-CoV-2 negativity by RT-qPCR.
Cellular responses to possible SARS-CoV-2 exposures across the study cohort provided additional insights. Blood samples for analysis of possible SARS-CoV-2-targeted cellular immune responses were collected on April 1, 2021 from participants who contracted COVID-19 during our study (vide supra, Fig. 13) and others who remained RT-qPCR negative for SARS-CoV-2 RNA throughout (i.e., from March 23, 2020 to April 1, 2021, Fig. 14). The CD8+ T cell responses were studied using IFN-γ ELISpot assays targeting pools of overlapping peptides spanning spike (12 pools), nucleocapsid (4 pools), matrix (2 pools), and envelope (1 pool) proteins.
The overall response patterns in Fig. 13 indicate the highest targeting of nucleocapsid protein regions, followed by similar targeting frequencies for spike and matrix. No envelope targeting was observed. The most diverse response to the peptide pools was observed for subject 38 (Fig. 13C), while subject 48 only had one response (peptide S3, Fig. 13D) that met the criteria for positivity.
Surprisingly, four participants who remained negative for SARS-CoV-2 RNA in nasal swab samples between March 23, 2020 and the date of blood collection on April 1, 2021 exhibited cellular responses where at least one peptide probe met the criteria for positivity (Fig. 14). These results suggest possible exposure to SARS-CoV-2 prior to the start of our clinical study on March 23, 2020. For these samples, nucleocapsid and spike protein region targeting were comparable and no targeting for matrix and envelope was observed. Subject 24 had the most individual peptide responses meeting the criteria for positivity (Fig. 14C, S1, S7, S8, N2).