SARS‐CoV‐2 innate effector associations and viral load in early nasopharyngeal infection – Liou – 2021 – Physiological Reports

2.1 Samples and study population

Our project was reviewed at the University of Utah by both the Institutional Review Board and the Biosafety Committee. An exemption from informed consent was allowed because patient samples were de‐identified. All samples were handled in a biosafety level (BSL) 2 capable hood (ThermoFisher Scientific, Waltham, MA, USA) using BSL 3 procedures until virus inactivation and were handled with BSL 2 procedures thereafter.

Randomly selected and completely deidentified, residual nasopharyngeal swab samples from patients presenting for diagnosis of symptoms consistent with COVID‐19 during the period of late April through early June of 2020 were included in the study. Clinical testing involved the use of a portion of each sample to test with automated, FDA Emergency Use Authorized real‐time polymerase chain reaction (RT‐PCR) or transcription‐mediated amplification tests for qualitative presence of SARS‐CoV‐2 RNA. We received sample remainders annotated with age, sex, and qualitative nucleic acid amplification‐detection results after being frozen at −80°C for approximately one month.

2.2 Initial extraction of human RNA and proteins

We extracted RNA using Chemagic reagents and Chemagic MSM I extraction platform (Perkin‐Elmer, Billerica, MA, USA) from part of each sample remainder producing sufficient RNA to allow RT‐PCR measurement of reference gene Pol2A (mean CT =32.22, SD =4.86). To exclude potential artifacts due to modified expression of Pol2A with viral infection, we measured ActB and GAPDH as alternative reference points and found no effects due to infection for any of the genes. Pol2A had the lowest standard error of measurements. Thus, for all other mRNA measurements, we used the Pol2A C_T as the reference point to calculate fold change (see Methods for C_T definition and usage).

Protease inhibitor cocktail and equal volume of Hank’s Balanced Salt solution (Sigma Aldrich) were added to the final portion of the thawed patient samples prior to centrifugation (20,000 g for 20 min at 4°C). We carefully aspirated the supernatant for Bead Based Multiplex Immunoassays. Pellets from centrifugation were extracted using All‐Prep Micro kits (Qiagen) in accordance with the manufacturer’s instructions, producing additional RNA suitable for RT‐PCR and a final protein‐containing pellet for Mass Spectrometry.

2.3 Real‐time polymerase chain reaction of viral and human mRNA

For most mRNAs, we had a sufficient samples to study all 40 patients; for selected mRNAs, we were able to study six samples with and six samples without SARS‐CoV‐2 detection. All specific mRNA measurements were based on RT‐PCR employing RNA from a single extraction method to avoid technical sources of noise.

An equal amount of RNA was taken for first‐strand cDNA reverse transcription (ABI High Capacity cDNA Reverse Transcription Kit) and specific amplification in a StepOnePlus (ABI, ThermoFisher Scientific). Gene‐specific primers were designed using the Roche Applied Science Universal Probe Library Assay Design Center. All amplifications were performed using a 2‐step amplification protocol with ABI PowerUp SYBR Green Master Mix as follows: 1 cycle at 50°C for 2 minutes to activate UDG, 1 cycle at 95°C for 2 minutes to release the DNA polymerase then 40–50 cycles with a 3‐second denaturing at 95°C followed by 30‐second annealing and denaturing at 60°C.

A melt curve (dissociation) was performed for every primer to ensure the above amplification conditions resulted in the amplification of a single peak. All of the designed primers gave a single peak upon dissociation after amplification suggesting no nonspecific binding to other genes. Amplification of genomic DNA was prevented by using primers that spanned an intron. The IFN‐α2 gene and IFITM‐1 ISG do not have introns. The primers did, however, give a single peak upon dissociation. All other primers including IFN‐λ spanned an intron.

2.4 Bead based multiplex immunoassays

Cytokine analyses of patient samples were performed using a commercially available enzyme‐linked immunosorbent assay (LEGENDplex Human Anti‐Virus Response Panel 13‐Plex with Filter Plate, BioLegend, San Diego, CA). This bead‐based multiplex assay allowed for the simultaneous quantification of interleukins (IL‐1β, IL‐6, IL‐8 [or CXCL8], IL‐10, IL‐12p70); interferons (IFN‐α2, IFN‐β, IFN‐γ, IFN‐λ1, IFN‐λ2,3), TNF‐α, IP‐10 (or CXCL10), and GM‐CSF in patient samples using a flow cytometric approach. All standards and samples were assayed in duplicate using manufacturer recommended protocols. Incubation steps were conducted at room temperature with constant agitation (500 rpm), and shielded from exposure to light. Performing the assay in standard 96‐well filter plates facilitated thorough washing of samples and required the use of a MultiScreen Vacuum Manifold (EMD Millipore Corporation, Billerica, MA) alongside a uniform vacuum source. Following the final wash of combined sample and biotinylated detection antibodies, bound proteins of interest were re‐suspended in 0.008% final concentration of EM‐grade glutaraldehyde (Electron Microscopy Sciences, Hatfield, PA, USA, Cat #16216) for 48 hours at 4°C to inactivate SARS‐CoV‐2, adapting a protocol previously investigated for SARS‐CoV (Darnell et al., 2004). Following incubation, samples were washed a final time and transferred to a polystyrene 96‐well plate with conical bottoms. Flow cytometric analysis of cytokines was performed using a BD FACSCanto II system (BD Biosciences, San Jose, CA) at the University of Utah Flow Cytometry Core (Salt Lake City, UT) and analyzed using LEGENDplex software (BioLegend).

2.5 Mass spectrometry

2.6 Calculations and statistical analysis

For all mRNA, we calculated fold‐change for each sample (FCsample) after measuring the fractional number of polymerase chain reaction doubling cycles required so that SYBR Green fluorescence exceeded the threshold for detection (C_T). We used the following formula:

where ΔC_Tsample was the number of doubling cycles to detect each mRNA minus the number of doubling cycles to detect mRNA from the Pol2A reference gene for each sample, and the ΔC_Tmedian was the median ΔC_Tsample for samples without detection of SARS‐CoV‐2. The incorporation of C_T for Pol2A mRNA in the calculation indexes the measurement so that samples with different efficiencies of recovery of mRNA containing cells between testing individuals are standardized.

For IFIT‐3 fold change values from DIA mass spectrometry, FCsample was calculated:

where AUC_IFIT3 is the area under the curve (AUC) for peptides identified as part of IFIT‐3, and median(AUC_{virus negative IFIT3}) is the median AUC for IFIT‐3 from samples without detection of SARS‐CoV‐2. The other prespecified ISG proteins were undetectable by DIA mass spectrometry, and thus no fold‐change calculation was possible.

We calculated summary statistics. We examined associations between different biomarker measurements by calculating Spearman’s rank correlation statistic to better understand potential dependencies. We used log transformations of all mRNA and protein measurements in our statistical calculations because of the log‐normal nature of our results. Others using the methods that we employed, however, often report results using either natural or base 2 logs. Because the bulk of our results are commonly reported using natural log values, we standardized on those for reporting. The effect is to slightly change results normally reported using base 2 logs by a proportion equal to natural log of 2 (or 0.698). This usage has no effect on interpretations of results.

Using SARS‐CoV‐2 infection status as the independent variable, we performed linear regression with natural log(FC) of each prespecified mRNA or natural log(FC_IFIT3) as the dependent variable because we seek to understand the biological effects of infection. Each univariable model was adjusted with age and sex with backward selection to understand the impacts on model fits. We performed univariable linear regression with natural log(FC) or natural log(protein concentration [pg/ml]) as the dependent variables and natural log(FC_{viral N1 protein mRNA}) as the independent variable to understand associations with viral load, adjusting with age and sex as above.

We performed sensitivity analyses of significant associations. For each dependent biomarker with significant associations with infection status or viral load, we selected all other biomarkers reported to have significant correlations as additional adjustment variables. Using these adjustment biomarkers one at a time, we assessed the impact on the estimates for infection status and viral load for each significant association.

We assigned 50 as the C_T value for undetectable mRNA. For undetectable proteins by bead‐based multiplex immunoassay, we assigned the minimum detection value. These assignments enable quantitative analysis without treating the values as missing. Results were similar when the analysis was restricted to raw data derived from the six infected and six uninfected samples with the highest recovery of RNA. All calculations and statistical modeling were performed using the R statistical system (R Core Team, 2020).

3.1 Study population

We evaluated 40 samples from individuals, evenly divided into 20 positive and 20 negative detection results for SARS‐CoV‐2. Samples were deidentified but annotated by age (median 46.5 years, range 11–90) and sex (17 females, 42.5%). Older patients were more likely to be male and negative for SARS‐CoV‐2 detection (Figure 1 and Tables 1‐3). This limited demographic information suggested that further evaluation of statistical relationships in our sample set required testing adjustments for age, sex, or both to avoid confounding.

Samples included in our study were randomly selected from those collected from April‐June of 2020 from patients who may have come from nine states within the Mountain West of the United States. During this period, positive results were reported for about 9–10% of tested patients. Among the positive test patients, about 10% eventually required hospitalization for COVID‐19 with less than 50% of those hospitalized suffering respiratory failure, ARDS, or succumbing to severe disease. While we know this context for our samples, the specific outcomes for individual patients in our study remain unknown.

3.2 Viral load

Using RT‐PCR, we estimated fold‐change in mRNA expression of SARS‐CoV‐2 small envelope protein E1 (Odds Ratio [OR] =10.8 × 10⁶, 95% Confidence Interval [CI] =8.37 × 10⁵–1.40 × 10⁸, p < 0.001) and nucleocapsid protein N1 (OR =5.1 × 10⁷, CI =4.5 × 10⁶–5.9 × 10⁸, p < 0.001) relative to expression in patients without infection. We selected primers (Udugama et al., 2020) for E1 originally from Charité, Germany (Corman et al., 2020) and N1 from the US CDC (Lu, Wang, et al., 2020). Both E1 and N1 mRNA fold changes gave virtually total discrimination between patients with and without infection diagnosed by clinical testing for SARS‐CoV‐2 infection using qualitative RT‐PCR.

3.3 Viral entry

We measured two human protein transcripts important for understanding SARS‐CoV‐2 cell entry, angiotensin‐converting enzyme‐2 (ACE‐2), which is essential for entry of SARS‐CoV‐2 and SARS‐CoV (Hoffmann et al., 2020; Shang et al., 2020), and transmembrane protease, serine‐2 (TMPRSS‐2) which enhances cell entry up to a thousand‐fold (Heurich et al., 2014; Hoffmann et al., 2020). ACE‐2 mRNA was increased threefold in patients with infection, and the fold‐change results were strongly associated with viral load. TMPRSS‐2 mRNA expression, however, was not associated with infection nor viral load (Figure 2).

3.4 Viral detection signaling

We examined transcription signals for two genes in the signaling pathway downstream of viral detection important for IFN responses, TNF‐associated factor‐binding kinase‐1 associated with inhibitor of NFκB (TBK‐1) and Stimulator of IFN genes‐(STING)‐1 for six patients with positive detection of SARS‐CoV‐2 and six patients with negative detection. The mRNA expressions of TBK‐1 and STING‐1 were not associated with infection.

3.5 Inflammatory responses to SARS‐CoV‐2 infection

We found increased mRNA expression of IL‐8, IFN‐γ‐induced protein‐(IP)‐10, and TNF‐α in SARS‐CoV‐2‐infected individuals (Figure 3A‐C). Moreover, we found that there was a strong association between viral load and the level of mRNA expression of these innate immune effector molecules (Figure 3F‐H).

Viral entry, detection, and signaling may lead to a systemic inflammatory response via NFκB activity which may potentially be augmented by TNF receptor‐associated factor‐(TRAF)‐1 activity (Edilova et al., 2018; Lalani et al., 2018). We found a large reduction in TRAF‐1 mRNA (Figure 3D). The reduction in TRAF‐1 was inversely associated with expression of both viral protein E1 mRNA (OR =0.945, CI =0.906–0.986, p = 0.025) and N1 mRNA (OR =0.947, CI =0.907–0.989, p = 0.034, Figure 3E). We found NFκB‐1 and NFκB‐2 mRNA transcripts were not significantly changed compared to uninfected status (Table 4). Other downstream immune effectors, granulocyte‐macrophage colony‐stimulating factor (GM‐CSF), IL‐6 and IL‐10 mRNA were not increased (Table 4).

Protein measurements using bead‐based multiplex immunoassays (BioLegend) for systemic inflammatory markers matching many of the mRNAs measured (plus IL‐1b and IL‐12p70) revealed no significant changes with infection and no significant associations with viral load with one exception. IP‐10 protein was increased nearly fourfold above measured control values (Table 5), and the log of concentration was strongly associated with viral load (OR =1.09 per unit of log unit of viral N1 mRNA fold‐change, CI =1.06–1.12, p < 0.001).

3.6 IFN responses to infection

We found five to sixfold increases in expression of both IFN‐λ1 and IFN‐λ2 mRNA among patients with detection of SARS‐CoV‐2 (n = 20) compared to patients without detection of virus (n = 20) (Table 6, Figure 4A,C). There were no other significant increases in IFN mRNA. The increases in IFN‐λ1 and IFN‐λ2 mRNA production were strongly associated with viral load (Table 7 and Figure 4E,F).

Despite increased mRNA expression for some of the IFNs, protein measurements showed reductions in IFN‐α2, IFN‐γ, and IFN‐λ2,3 in patients with viral infection that averaged 66%, 49%, and 40%, respectively, relative to control patients (Table 8).

3.7 ISG responses to infection

We prospectively selected four ISGs to evaluate because of their importance in defense against RNA viruses (Sadler & Williams, 2008): GTP‐binding Myxovirus protein (MX‐1) (Haller et al., 2015; Hefti et al., 1999; Kochs & Haller, 1999; Kochs et al., 2002; Verhelst et al., 2013). IFN‐induced proteins with tetratricopeptide repeats (IFIT) (Diamond & Farzan, 2013), IFN‐induced transmembrane protein (IFITM) (Diamond & Farzan, 2013; Perreira et al., 2013), and Tetherin (BST‐2) (Blanco‐Melo et al., 2016; Wang et al., 2019). MX‐1 and Tetherin mRNA were both increased in infected patients with moderate statistical significance, whereas two IFIT mRNAs were greatly and significantly increased (Table 9). Focused proteomic examination of proteins extracted from samples using DDA mass spectrometry detected an association between positive clinical testing for SARS‐CoV‐2 and IFIT‐1, IFIT‐3, and Tetherin proteins with a borderline finding for MX‐1 protein (Table 10). Proteomic examination using DIA mass spectrometry, which has less sensitivity but better specificity and precision, detected only a large increase in IFIT‐3 protein that was associated with clinical infection detection and increasing viral N1 mRNA (Figure 5). Transcript and proteome results are based on the same six positive and six negative samples for which we had sufficient mRNA remaining after other studies.

3.8 Model adjustments with older age

The increase in IP‐10 mRNA with infection was lower in older individuals (Figure 3I). All other things being equal, 10 additional years in age were associated with an approximately overall 70% reduction in IP‐10 mRNA compared to controls, whereas 25 additional years in age were associated with an overall 90% reduction, producing a remarkable counter effect to the infection itself.

We found a borderline significant effect for IFN‐λ2,3. Patients with older ages have slightly lower IFN‐λ2,3 per additional year of age with SARS‐CoV‐2 infection. This small per year effect was associated with a 10% lower IFN‐λ2,3 on average for every additional 10 years of age and about 23% lower IFN‐λ2,3 for an additional 25 years of age in addition to the approximately 40% reduction associated with SARS‐CoV‐2 infection (Table 8).

3.9 Correlations between IFN, inflammatory and ISG measurements

We found no strong and significant correlations between any protein and its transcript suggesting widespread translation blockade or rapid protein degradation with two exceptions. Both IP‐10 mRNA and protein were directly associated with viral load (N1 protein mRNA), and the correlation between mRNA and protein was moderate with strong significance, p < 0.001 (Table 12). IFIT‐3 mRNA and protein were strongly associated with viral load (Figure 5E‐F) but somewhat weakly correlated with each other (Figure 5G and Table 12). In contrast, we found strong correlations within IFN and inflammatory proteins (Table 11) and within IFN gene, inflammatory signaling gene and ISG transcripts (Table 13). Correlations within IFN types, for example, among IFN‐λ subtypes were exceptionally strong which corresponds to the biology of the IFNs.

3.10 Sensitivity testing

We performed sensitivity testing for all significant associations between IFN, inflammatory and ISG measurements with viral load (N1 protein mRNA) by re‐examining the relationships after exclusion of patients without infection by SARS‐CoV‐2. In every case, we found similar relationships between each biomarker and viral load, increasing the confidence in our findings.

Because of the high degree of correlation between some biomarkers (Tables 11‐13), for example, between IFN‐β1 and IFNα2 transcripts (Spearman correlation coefficient =1.00, p < 0.001), we examined the effect of adding a second biomarker as an adjustment to the relationship between each significant biomarker (p < 0.05) with the clinical diagnosis of infection (Tables 4‐6, 8 and 9) or with viral load (Table 7 and individually reported results for ACE2 mRNA, IP‐10 mRNA and IP‐10 protein in text). In every case, we found similar results for the association between the biomarkers reported and infection status or viral load. A number of the biomarker measurements tested as adjustment variables appeared to have independent significant effects suggesting that significant and independent multivariable associations exist, however, our study is too small to report those results with confidence.