Oral SARS-CoV-2 host responses predict the early COVID-19 disease course

Objectives: Oral fluids provide ready detection of Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and host responses. This study sought to determine relationships between oral virus, oral anti-SARS-CoV-2-specific antibodies, and symptoms. Methods: Saliva/throat wash (saliva/TW) were collected from asymptomatic and symptomatic, nasopharyngeal (NP) SARS-CoV-2 RT-qPCR+, subjects (n=47). SARS-CoV-2 RT-qPCR, N-antigen detection by immunoblot and lateral flow assay (LFA) were performed. RT-qPCR targeting viral subgenomic RNA (sgRNA) was sequence confirmed. SARS-CoV-2-anti-S protein RBD LFA assessed IgM and IgG responses. Structural analysis identified host salivary molecules analogous to SARS-CoV-2-N-antigen. Statistical analyses were performed. Results: At baseline, LFA-detected N-antigen was immunoblot-confirmed in 82% of TW. However, only 3/17 were saliva/TW qPCR+. Sixty percent of saliva and 83% of TW demonstrated persistent N-antigen at 4 weeks. N-antigen LFA signal in three negative subjects suggested potential cross-detection of 4 structurally analogous salivary RNA binding proteins (alignment 19-29aa, RMSD 1-1.5 Angstroms). At entry, symptomatic subjects demonstrated replication-associated sgRNA junctions, were IgG+ (94%/100% in saliva/TW), and IgM+ (75%/63%). At 4 weeks, SARS-CoV-2 IgG (100%/83%) and IgM (80%/67%) persisted. Oral IgG correlated 100% with NP+PCR status. Cough and fatigue severity (p=0.0008 and 0.016), and presence of nausea, weakness, and composite upper respiratory symptoms (p=0.005, 0.037 and 0.017) were negatively associated with oral IgM. Female oral IgM levels were higher than male (p=0.056). Conclusion: Important to transmission and disease course, oral viral replication and persistence showed clear relationships with select symptoms, early Ig responses, and gender during early infection. N-antigen cross-reactivity may reflect mimicry of structurally analogous host proteins.

Cochrane assessment of SARS-CoV2 antigen tests (n = 48), including lateral ow assays (LFA), demonstrated varied sensitivity between symptomatic and asymptomatic participants with highest sensitivity closest to symptom onset 10 .While LFA-assessed oral SARS-CoV-2-targeted immune responses can re ect systemic responses 11 , oral biomarkers as prognostic COVID-19 indicators have not been signi cantly explored.Here, RT-qPCR, LFA, and immunoblot were used for oral SARS-CoV-2 detection.
Longitudinal assessment of symptomatic participants suggested oral viral persistence.Disease severity and symptoms were associated with oral SARS-CoV-2 host responses and viral presence.Collectively, these ndings provide novel insights to oral markers of prognosis, persistence, and transmission.

Materials and Methods
Enrollment.A total of 47 Covid-19 era participants were assessed.Inclusion criteria for entry into the symptomatic longitudinal observational cohort (n = 17) required subjects to be NP SARS-CoV-2 RT-qPCR positive.COVID-19 + patients were recruited after written informed consent and were strati ed with mild, moderate or severe symptoms at the time of presentation based on the NIH criteria (NIH COVID-19 treatment guidelines).For the asymptomatic cohort (n = 30), participants were either asymptomatic, anti-Sars-CoV-2 spike seronegative (n = 15) or asymptomatic, Sars-CoV-2 spike seropositive (n = 15) institution campus dwellers including students, staff and faculty who provided written informed consent.Archived, pre-COVID pandemic saliva and throat wash samples were used as SARS-CoV-2 negative controls.All Biospecimens For RT-qPCR, NPS, saliva and throat wash samples were collected concurrently.The NPS samples were collected in 3ml of viral transport medium (VTM).Unstimulated whole mouth uid (WMF) samples (saliva) and throat wash gargle with 10 ml of normal saline (throat wash) were each collected in sterile 50mL wide-mouthed screw-capped containers.Samples were immediately transported for processing where the NPS, saliva, throat wash, and blood samples were stored at 4ºC and processed within 24 hours.After processing 1 ml aliquots of samples were stored at -80°C.Similarly, 1 ml aliquots of archived, preCovid-19 saliva and throat wash samples have been stored at -80°C for > 5 years.
Nucleic acid extraction and RT-qPCR.Saliva and throat wash from COVID-19 patients were used as a source for the detection of SARS-CoV-2 RNA.Trizol (Life Technologies, Carlsbad, CA) was used to inactivate virus and RNA was extracted according to the manufacturer's instructions.Brie y, 750 microliters of Trizol were added to 250 microliters of oral uid (saliva or throat wash).Following chloroform addition and phase separation, the aqueous phase was collected.Glycogen (2 micrograms) was added to the aqueous phase and RNA was precipitated with isopropanol.RNA pellets were dried and resuspended in water.
Ectopic expression of SARS-CoV-2 proteins in transfected oral keratinocytes.Total RNA was isolated from the saliva of SARS-CoV-2 -infected individuals.RT-qPCR was used to amplify cDNA encoding SARS-CoV-2 Spike (S), Envelope (E) or N-antigen.The generated cDNAs were digest with S I and SalI and cloned into S I/SalI sites of pCMV-myc eukaryotic expression plasmid.Expression in eukaryotic cells would result in the production of viral proteins with an amino-terminal c-myc tag.Immortalized human oral keratinocytes (NOK) were grown in keratinocyte serum-free media (Life Technologies, Carlsbad, CA).Cells were transfected with expression vectors using Fugene-6 (Promega, Madison, WI).Forty-eight hours post transfection, media was removed and subjected to SDS-PAGE and immunoblot analysis to detect myctagged proteins.
To obtain lysates from whole saliva, TW or transfected NOK, protein was isolated from the organic phase obtained using Trizol (Life Technologies, Carlsbad, CA) according to the manufacturer's instructions.
Protein was precipitated after the addition of isopropanol and centrifugation at 12,000 X G. Pellets were washed twice with 95% ethanol/0.3M guanidine-HCl followed by a nal wash with 100% ethanol.Each wash was performed for 20 minutes.Final protein pellets were dried and resuspended in 1X sample buffer (250 mM Tris, 8.0, 2% SDS, 20% glycerol, 50 mM DTT, 0.1% bromophenol blue).Pellets in buffer were incubated at 65° C until dissolved.
Immunoblot analysis of NOK cell lysates and media.Twenty-ve micrograms of whole lysate or 10 microliters of transfected cell culture media were loaded onto precast NuPage 4-12% Bis-Tris gels (Life Technologies, Carlsbad, CA).Proteins were electroblotted to PVDF membrane using western blot transfer buffer (25 mM Tris, 192 mM glycine, 20% methanol) for 2 hours at 200 mamps, constant current.Following transfer, blots were blocked by incubation with 5% nonfat dry milk in 1X PBS containing 0.1% Tween-20 (1X PBS-T) at room temperature.Following blocking, blots were incubated with primary antibody overnight at 4° C. Blots were washed twice with 1X PBS-T followed by incubation with HRPconjugated secondary antibody (Promega) diluted 1:10,000 in PBS-T, 5% milk for 1 hour at room temperature.Blots were washed twice with PBS-T and protein bands were detected by ECL after incubation with ECL Prime (GE Healthcare, Chicago, Il) according to the manufacturer's instructions.Blots were imaged using a GE ImageQuant LAS4000 system.The following primary antibodies were used to detect proteins on western blots: rabbit anti SARS-CoV-2 nucleocapsid (PA5-114448, Life Technologies), mouse anti-c-myc (sc-40, Santa Cruz Biotechnology, Dallas, TX), anti-amylase (A8273, Sigma-Aldrich, St.

Louis, MO).
Overexpression of His-tagged SARS-CoV-2 Nucleocapsid in E. coli.The NcoI/SalI fragent from pCMV-mycN containing the entire coding region of N was removed and inserted into the NcoI/SalI sites of pET30a.The resulting plasmid, containing the SARS-CoV-2 nucleocapsid (N) coding region in frame with the 6X histidine tagged coding region of pET30, was used to transform E. coli BL21 (DE3) bacteria.Bacteria with pET30a empty vector were used as a source of negative control his-tagged protein.Bacteria containing pET30a will produce an ~ 9 kD his-tagged protein after induction with IPTG.Single colonies were grown in LB medium containing kanamycin (50 micrograms/ml) to an OD A600 of 1.0.To induce gene expression, IPTG (1 mM nal concentration) was add and cultures were incubated at 30° C for 2 hours.Following centrifugation of cultures, bacterial pellets were resuspended in 1 ml NPI-10 (300 mM NaCl, 50 mM sodium phosphate, pH8.0, 10 mM imidazole, 1 mg/ml lysozyme).Lysates were incubated with NTA-Ni agarose beads (Life Technologies) to purify his-tagged proteins.Beads were washed with NPI-20 (300 mM NaCl, 50 mM sodium phosphate, pH8.0, 20 mM imidazole).Protein was eluted by resuspending beads in 250 microliters NPI-500 (300 mM NaCl, 50 mM sodium phosphate, pH8.0, 500 mM imidazole).Eluates were subjected to SDS-PAGE followed by Coomassie staining to assess retrieval and protein purity.Protein concentration was determined by Bradford assay.Lateral Flow Assay.Detection of SARS-CoV-2 N-antigen or anti-SARS-CoV-2 spike protein IgG and IgM was accomplished using lateral ow chambers (BioMedomics, Research Triangle Park, NC).For Nantigen 50 microliters of sample (saliva or throat wash) was diluted with 50 microliter of lysis buffer (supplied by the manufacturer).Samples in lysis buffer were incubated at room temperature.Following incubation, 80 microliters of sample was applied to the lateral ow chamber.Bands were visualized and quantitated using ImageJ software.The control band for each detection cartridge was used to normalize antigen-speci c band.For detection of anti-spike IgG/IgM, 50 microliters of sample (saliva or throat wash) was added directly to the lateral ow chamber.Two drops of COVID-19 IgG/IgM rapid test buffer (supplied by manufacturer) were added and bands were allowed to develop.Control bands were used normalize IgG and IgM intensity.Quantitation was determined relative to signal produced by pre-COVID-19 archived saliva and throat wash samples.

NCBI Structural analysis
The crystal structure of the N-terminal RNA binding domain of SARS-CoV-2 N antigen (PDB ID: 6M3M) was compared to other publicly available crystal structures in the Molecular Modeling Database (MMDB) using Vector Alignment Search Tool (Vast+) within the Domains & Structures module on the National Center for Biotechnology Information website (https://www.ncbi.nlm.nih.gov).Vast + makes geometric structural comparisons between macromolecules in the absence of sequence similarity.The 3D structures of superimposed biological assemblies were visualized using the web-based 3D viewer iCn3D version 4.3.1 to support the visualization style 11 .

Statistical methods
Given the small sample sizes, non-parametric tests were used to test correlations and compare data.Kendall rank correlation and Mann -Whitney U tests were used to test the correlation between the independent and response variables.The Kendall rank correlation test assessed the existence of monotonic relationships of ordinal or continuous variables using the ranks of the data.The Mann -Whitney U test was used where the independent variables were binary and was used to determine whether the two groups were from the same population when the groups were independent.Mann -Whitney U and Wilcoxon signed rank tests were used to compare response variables between two groups.When the groups were dependent, Wilcoxon signed rank tests were used to assess, differences in the medians of matched groups.

Results
Assessment of the SARS-CoV-2 N-Antigen Lateral Flow Assay.N-antigen LFA was validated using lysates from immortalized normal oral keratinocytes (NOK) transfected with a myc-tagged SARS-CoV-2 N-antigen expression vector.N-antigen was detected by anti-myc antibody immunoblot (Fig. 1A) or LFA (Fig. 1B).Bacteria-produced, his-tagged N-antigen was puri ed (Fig. 1C), used for N-LFA quanti cation, and migrated at 55 KD, matching the predicted molecular weight of SARS-CoV-2 N-antigen, while pET30 vector negative control ran at ~ 9 KD (Fig. 1C).LFA limit of detection was determined using two-fold serial dilutions of preCOVID-19 saliva spiked with recombinant N-antigen and detected his-tagged N-antigen at 43-680 picograms (pg).A faint band at 85 pg and robust detection > 170 pg demonstrates LFA's semiquantitative nature (Fig. 1D).No band was detected in saliva spiked with pET30 recombinant protein, indicating signal speci city.LFA allowed longitudinal detection of N-antigen in saliva from a representative NP-RT-qPCR + participant at symptom onset 14, and 28 days, suggesting active persistent infection, while signal was undetected in two PreCOVID-19 saliva samples (Fig. 1E).
Assessment of the SARS-CoV-2 Anti-Spike RBD IgM/IgG Lateral Flow Assay).Recombinant anti-Spike RBD IgG/IgM in LFA assays detected anti-SARS-CoV-2 Spike-speci c IgM and IgG oral immune responses.Two-fold dilution allowed immunoglobulin LFA validation.Mixtures of Spike RBD-speci c recombinant human IgM/IgG in preCOVID-19 saliva detected analytical sensitivity down to 0.63 microliters for IgM, however, IgG was undetected after a single 2-fold dilution (Fig. 1F).Archived saliva, spiked with anti-SARS-CoV-2 antibody demonstrated IgM signal but not IgG, perhaps re ecting IgM detection of a pre-2019 human coronavirus.Unspiked demonstrated no signal (Fig. 1G).A symptomatic participant demonstrated IgG at baseline and at 14 days but not IgM (Fig. 1G).

Viral metrics in the oral uids over time and between genders
LFA test and control band intensity was measured by ImageJ to provide relative measures of N-antigen and immunoglobulin presence in oral uids.An insigni cant downward trend in salivary N-antigen detection from baseline to 28 days, suggested persistence in oral uids (Wilcoxon Signed Rank Test) (Fig. 6A).While quantitatively less N-antigen was detected in symptomatic TW at 28 days compared to baseline (relative band intensity, median = 2070 vs 2310), this decrease was not signi cant (Wilcoxon Signed Rank Test, p = 0.9).In the symptomatic group, signi cant differences were detected between preCOVID-19 archived saliva, baseline, and 28 days (Wilcoxon Rank Sum Test, p = 0.003) (Fig. 6B).At baseline, 56% of the symptomatic were N-antigen positive in saliva and 91% were N-antigen positive in TW by LFA.Using NP-RT-qPCR as the gold standard, LFA assay sensitivity was 0.563 (CI, 0.299, 0.802) in saliva and 0.91 (CI, 0.587,0.998) in TW (Fig. 6C).Salivary IgG levels were signi cantly different between baseline and preCOVID-19 archived samples (Wilcoxon Rank Sum Test, p = 0.005).IgG trended upward over time and IgM levels, approximately a log lower than the IgG levels, remained the same over 28 days (p = 0.625) (Fig. 6A).While gender differences were not detected in oral N-antigen or TW IgG, females demonstrated higher levels of IgM than males (Mann-Whitney U test, p = 0.056) (Fig. 6D).SARS-CoV-2 oral outcomes, COVID-19 symptoms and symptom severity, and oral persistence.Selfreported COVID-19 symptom severity at baseline (absent, mild, moderate, or severe) and presence of SARS-CoV-2 oral N-antigen or antibody in saliva/TW were assessed.COVID-19 symptom (weakness, muscle ache, nausea, loss of taste/smell and upper respiratory tract symptoms which encompassed cough, shortness of breath, sore throat, nasal obstruction, nasal discharge) presence or absence at baseline was reported by participants as yes/no.Kendall Rank Correlation tests determined associations between severity of cough and fatigue with salivary IgM (p = 0.008 and 0.016 respectively) (Fig. 7A and  B).At baseline, salivary IgM was consistently associated with weakness (p = 0.037), muscle ache (p = 0.019), nausea (p = 0.005) and upper respiratory symptoms (p = 0.017), but not loss of taste/smell (p = 0.458) (Mann-Whitney U test).Oral N-antigen and IgG detection were not associated with COVID-19 symptoms in oral uids (Mann-Whitney U test) (Fig. 8A).While longitudinal assessment of oral antigen and immunoglobin responses detected no clear directional trends, there was consistent detection of IgM, IgG, and oral N-antigen at baseline and over time (n = 11) (Fig. 8B).

Discussion
In this study, we consistently detect SARS-CoV-2 antigen and antibody in oral uids during symptomatic COVID-19.N-antigen detection was immunoblot-and sgRNA-con rmed in NP-RT-qPCR + subjects providing signi cant implications for oral transmission.SARS-CoV-2-targeted oral IgG responses were highly correlated with nasopharynx positivity and oral N-antigen detection.Oral IgM levels indicated both symptom presence and severity.
This study, and others, detected salivary SARS-CoV-2 RNA with distinct viral shedding dynamics compared to NP 1,21 .Prior assessments detected a 3-fold lower positive detection rate in saliva than NPS 21 , leading some to question the role of oral virus.Here, cumulative data suggest oropharyngeal viral replication.SARS-CoV-2 was detected in oral uids from NP-RT-qPCR + participants using several RT-qPCR-based viral detection methods, 1) three distinct primer pairs targeting 3 regions of the viral genome, 2) absolute RNA copy number determination using a standard curve, and 3) sgRNA, shown to be a marker of active replication during early symptomatic infection 22,23 .Orf3A sgRNA was detected in 82% of those tested by sgRTqPCR, (fast, sensitive, economically feasible, and reliable) and is a viable marker of viral replication based on its contribution to viral titer and disease in hACE + mice 24 .SARS-CoV-2 protein was consistently detected by two distinct methods, LFA and immunoblot.We are unaware of other studies demonstrating immunoblot con rmed viral antigen detection in oral uids.Importantly, persistent Nantigen detection provides signi cant implications for continued potential oral transmission (Fig. 3F,8B).
One critique of rapid LFA was false positives.Here, we detect positive LFA results in two seropositive (1A,4A), and three uninfected subjects (16A,17A,18A).During asymptomatic infection it is impossible chronical where subjects are in their infection cycle.Seropositive/NP-subjects could demonstrate oral infection (1A,4A) or oral infection may subside in seropositive/NP+ (8A).Alternatively, cross-reactive host proteins might be detected.Analysis determined potential cross-reactivity with structurally analogous host RNA binding proteins (Fig. 5).17S U2 snRNP and RBM7, are salivary proteome members and two others, LINE1 Orf1p and hnRNP H, have salivary proteome isoforms (LINE1 Orfp1 and hnRNP A2/B1, hnRNP M, hnRNPK) 25 .This raises the possibility that N-antigen-speci c LFA can detect conformationallysimilar proteins, perhaps re ecting the potential for N-antigen mimicry through RNA binding domains.An example implication is SARS-CoV-2 host genome integration.The LINE1 ORF1p analogue mediates retrotranposable element genome integration 26 .RNA virus sequences have been detected across vertebrate genomes, with several integration signals consistent with LINE retrotransposon germline integration of viral cDNA copies 27 .Recently, subgenomic sequences, derived from SARS-CoV-2's 3' end, were shown to be integrated into host cell DNA 28 .Hence, host interactions with N-antigen or similar actions by N-antigen might facilitate SARS-CoV-2 genome integration.Anti-spike RBD-speci c antibodies during natural infection and vaccination were detected in saliva 29,30 with temporal kinetics that re ect blood.While anti-spike-RBD IgG levels were negatively correlated with salivary viral load in the Silva study 31 , here in 16/16 symptomatic and 1/1 asymptomatic participants, anti-spike-RBD IgG was 100% positively correlated with SARS-CoV-2 NP-RT-qPCR + status, and symptomatic individuals consistently demonstrated salivary N-antigen by immunoblot.The NP + status and IgG relationship was also shown by Pisanic et,al.with IgG positive responses in 24/24 RT-qPCRcon rmed COVID-19 cases 11 .Together, this suggests that salivary IgG detection indicates concurrent nasopharynx positivity.Detection of simultaneous oral virus and oral host immune responses at baseline suggests a highly infectious pre-symptom state followed by concurrent symptoms, persistent oral infection, and virus-targeted host responses.
Relationships between IgG, VL, and disease course were not detected 21,32 .However, we detected statistically signi cant relationships between salivary IgM and multiple symptoms including weakness, muscle ache, nausea, and upper respiratory symptoms but not taste/smell.Relative IgM levels were signi cantly associated with degree of fatigue and cough severity suggesting that oral IgM signi es an active early immune response associated with milder disease course.
Gender differences are detected in COVID-19-related morbidity and mortality.Here, gender differences were detected, with women having consistently higher oral IgM levels than men.Systematic review of gender differences determined that women were less likely to present with severe disease and be admitted to the intensive care unit (ICU) than men (OR 0.75 [0.60-0.93]p < 0.001 and OR 0.45 [0.40-0.52]p < 0.001, respectively 33 .While higher smoking rates and reluctance to seek health care may contribute to male disease, the well-described evolution of greater humoral immunity in females could contribute to these differences 34,35 .Here, women had markedly higher IgM levels, fewer symptoms, and milder symptom severity.Differences in IgM production may provide biologic underpinning, contributing to the COVID-19-related morbidity and mortality gender gap.The SARS-CoV-2 nucleocapsid mimics host RNA binding proteins that are expressed within the salivary proteome and may be responsible for cross reactivity in LFA assays.VAST+ was used to generate RNA binding protein structures demonstrating 3D similarity to the SARS-CoV-2 N-antigen.Simultaneous alignment generated molecular protein pairs between N-antigen ( rst column) and host RNA binding proteins (second column).Non-aligned N-antigen amino acids are rendered in green, conserved aligned amino acids are rendered in red and aligned, non-conserved amino acids are rendered in blue.Nonaligned host protein amino acids proteins are rendered in olive green, conserved aligned amino acids are rendered in pink and aligned, non-conserved amino acids are rendered in light blue.The spatial arrangement of overlapping amino acids can be seen in the merged sequence alignment (third column).The number of structurally aligned amino acids (conserved and non-conserved) is shown.The root mean square deviation (RMSD) is used as a measurement between atoms in the backbone of the molecular structures and is listed in angstroms above each merged image and detection of the host protein within the salivary proteome is displayed as yes in the fourth column while detection of a related isoform within the salivary proteome is displayed as ISO.
samples were obtained and stored under IRB approval (Symptomatic participants, UNC IRB study #20-0792; Asymptomatic participants, UNC IRB study #20-1771; Archived, PreCovid participants, UNC IRB study #07-1431).All experimental protocols used to assess specimens were performed in accordance with approval by the University of North Carolina IRB (Symptomatic participants, UNC IRB study #20-0792; Asymptomatic participants, UNC IRB study #20-1771; Archived, PreCovid participants, UNC IRB study #07-1431).All methods used to assess specimens were performed in accordance with the relevant guidelines put forth by the University of North Carolina.
These ndings provide novel insights to oral SARS-CoV-2 infection.Oral IgM detection predicted milder disease and female predilection.The potential for cross-detection of N-antigen and structurally similar host RNA binding factors, suggests viral mimicry.Oral persistence has signi cant implications for viral transmission in the absence of mask use and vaccination.

Figures
Figures

Figure 1 Assessment
Figure 1

Figure 2 Study
Figure 2

Figure 3 Oral
Figure 3