The contrasting role of nasopharyngeal angiotensin converting enzyme 2 (ACE2) transcription in SARS-CoV-2 infection: A cross-sectional study of people tested for COVID-19 in British Columbia, Canada.

BACKGROUND: Angiotensin converting enzyme 2 (ACE2) protein serves as the host receptor for SARS-CoV-2, with a critical role in viral infection. We aim to understand population level variation of nasopharyngeal ACE2 transcription in people tested for COVID-19 and the relationship between ACE2 transcription and SARS-CoV-2 viral load, while adjusting for expression of: (i) the complementary protease, Transmembrane serine protease 2 (TMPRSS2), (ii) soluble ACE2, (iii) age, and (iv) biological sex. The ACE2 gene was targeted to measure expression of transmembrane and soluble transcripts.
METHODS: A cross-sectional study of n = 424 "participants" aged 1-104 years referred for COVID-19 testing was performed in British Columbia, Canada. Patients who tested positive for COVID-19 were matched by age and biological sex to patients who tested negative. Viral load and host gene expression were assessed by quantitative reverse-transcriptase polymerase chain reaction. Bivariate analysis and multiple linear regression were performed to understand the role of nasopharyngeal ACE2 expression in SARS-CoV-2 infection.
FINDINGS: Analysis showed no association between age and nasopharyngeal ACE2 transcription in those who tested negative for COVID-19 (P = 0•092). Mean relative transcription of transmembrane (P = 0•00012) and soluble (P<0•0001) ACE2 isoforms, as well as TMPRSS2 (P<0•0001) was higher in COVID-19-negative participants than COVID--19 positive ones, yielding a negative correlation between targeted host gene expression and positive COVID-19 diagnosis. In bivariate analysis of COVID-19-positive participants, transcription of transmembrane ACE2 positively correlated with SARS-CoV-2 viral RNA load (B = 0•49, R2=0•14, P<0•0001), transcription of soluble ACE2 negatively correlated (B= -0•85, R2= 0•26, P<0•0001), and no correlation was found with TMPRSS2 transcription (B= -0•042, R2=<0•10, P = 0•69). Multivariable analysis showed that the greatest viral RNA loads were observed in participants with high transmembrane ACE2 transcription (Β= 0•89, 95%CI: [0•59 to 1•18]), while transcription of the soluble isoform appears to protect against high viral RNA load in the upper respiratory tract (Β= -0•099, 95%CI: [-0•18 to -0•022]).
INTERPRETATION: Nasopharyngeal ACE2 transcription plays a dual, contrasting role in SARS-CoV-2 infection of the upper respiratory tract. Transcription of the transmembrane ACE2 isoform positively correlates, while transcription of the soluble isoform negatively correlates with viral RNA load after adjusting for age, biological sex, and transcription of TMPRSS2. FUNDING: This project (COV-55) was funded by Genome British Columbia as part of their COVID-19 rapid response initiative.

Evidence before this study

We conducted a MEDLINE ® search using the MeSH topic terms: “angiotensin converting enzyme 2 expression”, “SARS” and “age”, restricting the search to English-language reports published from January 1st, 2020. The search returned 98 articles, 88 of which reported primary research; these were further filtered by the MeSH qualifiers “epidemiology” and “virology” to provide 43 articles. Search results were further restricted to the MeSH terms: age, virus replication, host pathogen interactions, nasal mucosa, viral RNA load and real-time polymerase chain reaction, which returned 24 results. We read through the abstracts of these twenty-four papers and manually selected n = 4 for full review. This review provided evidence that ACE2 expression is greater in the upper respiratory tract than the lower respiratory tract when measured by single-cell RNA sequencing, immunohistochemistry, and high-sensitivity RNA in-situ mapping. In the upper airway, ACE2 mRNA abundance closely correlates with protein concentration. A reverse genetics study demonstrated that a variable SARS-CoV-2 infection gradient occurs in the respiratory tract, with highest viral loads expected in the upper airway.

Added value of this study

We measured ACE2 transcription in the context of COVID-19 testing to investigate the role of nasopharyngeal ACE2 protein in SARS-CoV-2 infection. Our findings support previous work: the strong correlation we observe between nasopharyngeal ACE2 transcription and SARS-CoV-2 load also suggests an infection gradient across the human airway. Greater viral loads are expected in tissue with high transmembrane ACE2 protein expression. No observed relationship between age and nasopharyngeal ACE2 in COVID-19-negative participants suggests that upper airway ACE2 expression may be independent of the renin-angiotensin-aldosterone-system. We are the first to measure ACE2 transcription in nasopharyngeal specimens from a large sample of COVID-19 patients representing a broad range of ages and clinical manifestations, adjusting for age, sex, and transcription of TMPRSS2 in a multivariable analysis. We demonstrate that nasopharyngeal transcription of soluble ACE2 negatively correlates with viral load, suggesting a potential protective role at the population level.

Implications of all the available evidence

Considering all available evidence, ACE2 protein may play a dual, contrasting role in SARS-CoV-2 infection of the upper airway. Transmembrane ACE2 transcript positively correlates with SARS-CoV-2 viral load, while soluble ACE2 transcript shows a negative association. Total nasopharyngeal ACE2 transcript does not correlate with age, as would be expected in the lower respiratory tract. Genetic and environmental factors such as ACE2 single nucleotide polymorphisms or smoking, and whether they regulate the expression of transmembrane and soluble ACE2 protein in healthy nasopharyngeal tissue, require further investigation.

Alt-text: Unlabelled box

Introduction

In December 2019, clusters of viral pneumonia were reported in Wuhan, China. A novel highly pathogenic human coronavirus was isolated and named severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the etiological agent of coronavirus disease (COVID-19) [1]. SARS-CoV-2 utilizes the same host receptor and protease for cell entry as the human coronaviruses SARS-CoV and HCoV-NL63. The receptor angiotensin converting enzyme 2 (ACE2) mediates cellular entry, while transmembrane serine protease 2 (TMPRSS2) serves as a complementary host factor [2,3]. TMPRSS2 protein contributes to SARS-CoV-2 cell entry by cleaving the viral spike protein into a conformational form necessary for membrane fusion [2]. Unlike ACE2, TMPRSS2 expression occurs more stably across upper airway tissue and alternative enzymes, such as cathepsin B/L or furin, may perform its role in viral infection interchangeably [4,5]. In human physiology, ACE2 has a cardiovascular protective and anti-inflammatory role, as a constituent of the renin-angiotensin-aldosterone-system (RAAS) [6]. Interestingly, expression of ACE2 in the nasopharyngeal tract exceeds that in alveolar tissue, explaining initiation of SARS-CoV-2 infection in the upper respiratory tract [4]. Transcription of ACE2 produces at least two dominant mRNA transcript variants responsible for translation into soluble and membrane-bound protein isoforms [7], [8], [9], [10]. The soluble variant (cAug10) shares the same start codon as the transmembrane variant (aAug10), but terminates at nucleotide position 1665 or amino acid position 555 [9]. Although the transmembrane isoform has been shown to be crucial for viral entry into host cells, the role of soluble ACE2 remains uncharacterized, though evidence exists that it may protect against SARS-CoV-2 infection [11,12]. Production of soluble ACE2 also occurs post-transcription via shedding of ACE2 from the cellular surface. The metallopeptidase domain 17 (ADAM17) cleaves membrane anchored ACE2 downstream of the cAug10 transcript termination site between amino acid positions 716 and 741 [13]. Therefore, there are at least two independent mechanisms of soluble ACE2 protein production, alternative spicing and ectodomain shedding. The relationship between alternative splicing and ectodomain shedding to produce soluble ACE2 protein is not well understood in human physiology or in the context of COVID-19 infection; the two mechanisms could work mutually or competitively to regulate expression of soluble ACE2.

To understand the importance of ACE2 expression in SARS-CoV-2 infection, we performed a cross-sectional study of people tested for COVID-19 in British Columbia, Canada. The study aims to investigate the relationship between (i) nasopharyngeal transcription of ACE2 and age in COVID-19-negative participants, (ii) nasopharyngeal transcription of host genes by COVID-19 test result, and (iii) nasopharyngeal transcription of transmembrane ACE2 and viral RNA load in those who tested COVID-19-positive adjusting for age, biological sex, transcription of soluble ACE2 and TMPRSS2. Our study builds on previous work characterizing ACE2 transcription in older, hospitalized COVID-19 patients, drawing from a larger sample across broader ages and clinical manifestations with adjustment for important covariates [14]. These findings increase our knowledge of host genes involved in SARS-CoV-2 infection and may allow for assessment of baseline COVID-19 risk.

Methods

Study design and participants

We performed cross-sectional sampling of people tested for COVID-19 at the British Columbia Center for Disease Control Public Health Laboratory (BCCDC-PHL) from 24/3/2020–9/5/2020. At the time of sampling, provincial health guidelines required a clinical indication for referral of a COVID-19 test. Inclusion criteria were applied to select study participants whose diagnostic specimens were: tested centrally at the BCCDC-PHL, the individual's first test administered by personal health number, collected by nasopharyngeal swab, suspended in Hologic Aptima™ media, negative for concurrent Influenza A, B or Respiratory syncytial virus infection, stored at −80 °C following RNA extraction, and for whom host gene transcription was successfully measured by quantitative reverse-transcriptase polymerase chain reaction (qRT-PCR). People meeting the inclusion criteria (n = 444), were excluded (n = 16) if samples had a SARS-CoV-2 E gene cycle threshold (Ct) value of >=38 by qRT-PCR. COVID-19-positive cases (n = 212) were matched in a 1:1 ratio with those that tested negative for COVID-19 by age and biological sex (Figure S1). Demographic variables of age and biological sex were drawn from public health laboratory data. Laboratory methods were performed in a College of American Pathologists accredited laboratory with externally validated qRT-PCR assays [15], [16], [17], [18].

Ethics statement

Ethical approval for the study was obtained from the University of British Columbia human ethics board (H20-01110). Written informed consent was not required as per ethics board approval. Participant data was de-identified prior to analysis; the results of this non-interventional observational study were not linked back to any identifying patient records. The study was deemed to be of minimal risk.

Procedures

Nasopharyngeal samples collected in Hologic Aptima™ media were stored at 4 °C before RNA extraction using the Viral RNA isolation kit on the MagMAX-96™ platform (ThermoFisher) [19].

Host and viral gene targets were assayed by qRT-PCR on the Applied Biosystems 7500 Real-Time PCR platform using TaqMan FastVirus 1-step polymerase (ThermoFisher). Total reaction volumes equaled 20ul, with 5ul of RNA template, 1ul of 20x primer/probe, 5ul Fast Virus and 9ul of nuclease free water per reaction. Cycling conditions were set to: 50 °C for 5 min, 95 °C for 20 s followed by 40 cycles of 95 °C for 15 s and 60 °C for 1 min. A multiplex qRT-PCR reaction targeting SARS-CoV-2 envelope (E) and host ribonuclease P (RNaseP) was used to diagnose acute viral infection by presence of viral RNA [20]. Expression of RNaseP was used to measure sampling variation, ensuring sample quality [20]. Participants were diagnosed as COVID-19-positive with an E gene Ct value of <38. The E gene Ct values were transformed to genome equivalents per milliliter by making a 5-fold, 1:10 standard curve of SARS-CoV-2 synthetic RNA- MN908947•3 (Twist Bioscience) (Table S1) [21]. Commercially available primer probe sets were used to amplify the host gene targets TMPRSS2 [18] (Hs00237175_m1) (ThermoFisher) and ACE2 [16] (Hs01085333_m1, HS01085340_m1) by multiplex qRT-PCR with a glyceraldehyde 3-phosphate dehydrogenase (GAPDH) control [17] (Hs02758991_g1) (ThermoFisher). The full length ACE2 reference sequence (NC_000023.11) was downloaded from the National Centre for Biotechnology Information GenBank and annotated to show the (Hs01085333_m1, HS01085340_m1) qRT-PCR targets using Geneious Prime 2021.3 (Figure S3) [47]. Relative gene transcription was calculated between TMPRSS2, ACE2 and GAPDH using the 2−ΔΔCt method [22]. Transmembrane ACE2 was defined from the Hs01085333_m1 (ThermoFisher) gene target, which exon spans the transmembrane domain. Soluble ACE2 was defined as the absolute difference in transcript expression between the HS01085340_m1 (ThermoFisher) and Hs01085333_m1 (ThermoFisher) gene targets (HS01085340_m1- Hs01085333_m1) (Figure S3). Subtraction of the Hs01085333_m1 target from HS01085340_m1 is mathematically equivalent to the use of Hs01085333_m1 instead of GAPDH in the 2−ΔΔCt equation to calculate the relative expression of HS01085340_m1. The HS01085340_m1 target does not include the transmembrane domain (Figure S3).

Statistical analysis

Sample size estimation was performed in the design stage of the study. Recruitment of n = 424 participants, or n = 212 participants per COVID-19 test result, provides evaluation of a Cohen minimum effect size (h = 0•20), assuming: alpha=0•05 and power=0•80 for comparison between groups. For comparison within groups, n = 212 provided 0•85 power assuming a minimum effect size (h = 0•20), standard deviation (σ =1), and alpha of 0•05 [23].

De-identified data reporting participant COVID-19 test result, site of sample collection, type of collection media, accessioning laboratory, age and biological sex were accessed from public health laboratory records. Matching of COVID-19-positive people, who met the described inclusion and exclusion criteria, to COVID-19-negative participants by age and biological sex in a 1:1 ratio was performed by a nearest neighbor algorithm, caliper = 0 SD (n = 424) [24]. Bivariate analyses were performed between: age, biological sex, viral RNA load, TMPRSS2 transcription, soluble or transmembrane ACE2 transcription, and COVID-19 test result. The balance of covariates between test groups was examined post-matching by the standardized mean difference (SMD) [25]. Parametric statistical tests were used given the large sample size of the study [26].

The relationship between age and nasopharyngeal transcription of transmembrane ACE2, soluble ACE2, and TMPRSS2 was examined in unmatched COVID-19-negative participants over the age of 18 and less than 98 years by both linear regression and categorization of age into 10-year intervals. Differences in ACE2 and TMPRSS2 transcription by age category were tested by analysis of variance. The bivariate relationship between transmembrane ACE2 expression and biological sex was analyzed by using a t-test in unmatched COVID-19-negative participants.

Differences in mean transcription of host genes by COVID-19 diagnosis was further examined in the matched sample by a two-tailed, paired t-test assuming non-equal variance [27]. Correlation between host gene transcription and viral RNA load was analyzed by simple linear regression. Multivariable analysis was performed by multiple linear regression, variable importance was assessed by the partial F-test [28]. Collinearity was examined by the variable inflation factor with a cut-off of 10 [29]. The common cause criterion was applied to control for measured confounding [30]. Effect modification was reported when found statistically significant and supported conceptually [8]. All analysis was performed in R version 3.6.3 using the packages: car, ggsci, tidyverse, dataexplorer, ggpubr, lmtest, publish, forcats, matchit, tableone, emmeans and effects [31].

Role of the funding source and data stewards

The funder and data stewards played no role in the study design, analysis, or interpretation of the results. As such, interpretation of the results does not reflect the views of the funding organization or data stewards. The research was performed at the University of British Columbia and British Columbia center for Disease Control.

Results

The analytic dataset contains age- and biological sex-matched participants tested for COVID-19 in British Columbia from 24/3/2020–9/5/2020 (n = 424; Figure S1). Participant characteristics are shown in Table 1. The mean age of COVID-19-positive participants was 61•69 years, 47•60% were biologically male. In COVID-19-negative participants, the mean age was 62•18 years, 48•60% were biologically male. Age (SMD=0•020) and biological sex (SMD=0•019) were balanced between COVID-19-negative and -positive groups. Viral RNA (E gene Ct) was only detected in COVID-19-positive participants, with an average Ct value of 28•24. Mean relative transcription of transmembrane ACE2, soluble ACE2, and TMPRSS2 statistically differed by COVID-19 test result; on average, transcription of each gene was lower in samples from COVID-19-positive participants (Table 1).

Table 1

Characteristics of analytic data stratified by COVID-19 test result (n = 424).

			COVID-19 Test Result
Variable Name	Level	Total (n)	Negative	Positive	SMD a	P-Value b
		424	212	212
E Gene Ct (mean [SD])		424	.. c	28•24 [7•18]
Viral RNA Load (mean [SD]) d		424	..	5•14 [2•46]
Age (mean [SD])		424	62.18 [24.29]	61.69 [23.49]	0.020	0.83
Biological Sex (n [%])					0.019	0.92
	Male	204	103 [48.60]	101 [47.60]
	Female	220	109 [51.40]	111 [52.40]
Transmembrane ACE2 (mean [SD])		424	(0.00) [1.08]	(−0.61) [1.90]		0.00012
Soluble ACE2 (mean [SD])		424	(0.00) [0.62]	(−0.89) [1.47]		<0.0001
TMPRSS2 Total (mean [SD])		424	(0.00) [0.79]	(−1.35) [1.60]		<0.0001

Participants who tested positive for COVID-19 by qRT-PCR were matched with COVID-19-negative participants by age and biological sex, balance was checked between groups post matching by calculating the standardized mean difference (SMD) (cutoff=0.20). Relative transcription of ACE2 and TMPRSS2 was normalized to the negative group by the 2−ΔΔCt method.

the standardized mean difference (SMD) was calculated to determine balance of age and biological sex between participants who tested negative or positive for COVID-19 post-matching (cutoff=0.20).

p-values are reported for parametric tests used for continuous (t-test) and categorical variables (x^2 test).

no SARS-CoV-2 E Gene Ct values were reported from qRT-PCR for COVID-19-negative participants.

measured as Log10 genome equivalents/ml (Log10 GE/ml), viral RNA load data is not available from patients who tested negative.

Relationship between ACE2 expression, age, and sex in unmatched COVID-19-negative participants

To assess ACE2 expression in uninfected nasopharyngeal tissue, we compared ACE2 transcription and age in unmatched COVID-19-negative participants between 19 and 98 years of age. No relationship was found between nasopharyngeal transmembrane ACE2 transcription and age with age as a continuous variable in linear regression (P = 0•076). This finding was reproduced when age was categorized into ten-year intervals (ANOVA, P = 0•092) (Fig. 1). Similarly, no association was observed between age and soluble ACE2 (ANOVA, P = 0•081) or TMPRSS2 (ANOVA, P = 0•173) transcription. We also assessed the relationship between sex and transmembrane ACE2 transcription but found no relationship in unmatched COVID-19-negative participants (t-test, P = 0•832).

Fig. 1

Relationship between age and nasopharyngeal transmembrane ACE2 transcription in unmatched COVID-19-negative participants. Boxplots of transmembrane ACE2 transcription by 10-year age categories in unmatched COVID-19-negative participants between the ages of 19 and 98 (n = 198); boxes represent the Q1-Q3 interquartile range, whiskers represent 1.5x the Q1 or Q3 and horizontal lines the median transmembrane ACE2 transcription by age category. Participants who tested negative younger than 19 or older than 98 were excluded based on n<10 observations per age group. No difference was detected in mean transmembrane ACE2 transcription among age categories (ANOVA, P = 0•092).

Nasopharyngeal transcription of ACE2 and TMPRSS2 by COVID-19 test result

Bivariate analysis showed a significant mean difference in transmembrane ACE2, soluble ACE2 and TMPRSS2 transcription between matched COVID-19-negative and -positive participants (Fig. 2). These differences were further examined to detect non-equal variance for all host gene targets by COVID-19 test result. Assuming non-equal variance, paired t-tests showed lower mean transcription of transmembrane ACE2 (P = 0•00,012, Fig. 2a), soluble ACE2 (P<0•0001, Fig. 2b) and TMPRSS2 (P<0•0001, Fig. 2c) in participants who tested COVID-19-positive in comparison to -negative (Fig. 2).

Fig. 2

Relative nasopharyngeal transcription of targeted host genes by COVID-19 test result. Gene transcription is portrayed in kernel density plots stratified by COVID-19 test result. Probability densities of relative host gene transcription are shown by positive (red, transmembrane ACE2; blue, soluble ACE2; yellow, TMPRSS2) and negative COVID-19 test results (gray). Levene's test was used to detect non-equal variance in gene transcription for all host targets between COVID-19-negative and -positive participants. A two-tailed, paired t-test was used to examine mean difference in host gene transcription by COVID-19 test result assuming unequal variance: (a) transmembrane ACE2 (P = 0•00,012), (b) soluble ACE2 (P<0•0001) and (c) TMPRSS2 (P<0•0001).

Host gene transcription and viral RNA load in COVID-19-positive participants

The relationship between transmembrane ACE2 transcription and viral load was assessed in nasopharyngeal specimens from COVID-19-positive participants. We estimated viral load by qRT-PCR quantification of viral genome copies, a commonly-used surrogate, which has been associated with the presence of culturable SARS-CoV-2 from nasopharyngeal specimens [32], [33], [34], [35], [36]. Nasopharyngeal transcription of transmembrane ACE2 positively correlated with viral RNA (linear regression, B = 0•49, R2=0•14, P<0•0001). Transcription of soluble ACE2 negatively correlated with viral RNA (linear regression, B= −0•85, R2= 0•26, P<0•0001). No correlation was found between TMPRSS2 transcription and viral RNA (linear regression, B= −0•042, R2=<0•10, P = 0•69) (Figure S2) by bivariate analysis. Multiple linear regression estimated that a one-unit change in transmembrane ACE2 transcription increases viral RNA load by 0•89 Log10 GE/ml (95%CI: 0•59 to 1•18) adjusting for age, biological sex, transcription of TMPRSS2, and soluble ACE2 (Table 2). Biological sex could have been dropped from the model as evident from comparison between nested models (partial F-test, P = 0•76), but was kept for validity. A partial F-test indicated effect modification between transmembrane and soluble ACE2 transcription (P = 0•010) (Table 2). No effect modification was observed between transcription of transmembrane ACE2 and TMPRSS2 (partial F-test, P = 0•23). Transcription of soluble ACE2 decreases viral load (Β= −0•099, 95%CI: [−0•18 to −0•022]) (Table 2g). The association between transmembrane ACE2 transcription and viral RNA in nasopharyngeal tissue differs by the concomitant level of soluble ACE2 transcribed (Fig. 3). Effect modification was visualized by categorizing soluble ACE2 by the relative mean transcription +/- 1 | 2 standard deviations of all study participants (low= −1•66 | −2•86, mean= −0•44 | −0•44, high=0•77 | 1•98, SD=1•21) (Table 2, Fig. 3 & Figure S4).

Table 2

Unadjusted, adjusted and effect modification inclusive linear regression models of SARS-CoV-2 viral RNA load.

Model	Effect Estimate (Β)	95% CI for Β	P-Valuef
Transmembrane ACE2 Transcription a	0.49	(0.33 to 0.66)	<0.0001
Age and Sex Adjusted b	0.36	(0.19 to 0.53)	<0.0001
Covariate Adjusted c	0.89	(0.59 to 1.18)	<0.0001
Effect Modification d,e	–	–	0.010g
Modification was Evaluated at +/- 1SD and 2SD	1SD | 2SD	1SD | 2SD	–
Low Soluble ACE2 Transcription	0.79 | 0.91	(0.49 to 1.08) | (0.62 to 1.19)	–
Mean Soluble ACE2 Transcription	0.67 | 0.67	(0.34 to 1.00) | (0.34 to 1.00)	–
High Soluble ACE2 Transcription	0.55 | 0.43	(0.16 to 0.94) | (−0.028 to 0.88)	–

Regression of SARS-CoV-2 viral RNA load by transmembrane ACE2, soluble ACE2, TMPRSS2 transcription, age, and biological sex. Columns report beta coefficients (Β), 95% confidence intervals (95% CI) and P-values for transmembrane ACE2.

the unadjusted effect of transmembrane ACE2 transcription on SARS-CoV-2 viral load (Log10 GE/ml).

the effect of transmembrane ACE2 transcription adjusted by age and biological sex on SARS-CoV-2 viral load (Log10 GE/ml).

the effect of transmembrane ACE2 transcription adjusted by: age, biological sex, soluble ACE2, and TMPRSS2 expression on SARS-CoV-2 viral load (Log10 GE/ml).

the effect of transmembrane ACE2 transcription adjusted by: age, biological sex, soluble ACE2, and TMPRSS2 transcription; including effect modification by soluble ACE2 transcription or how the effect of transmembrane ACE2 transcription changes by the level of soluble ACE2 transcription.

soluble ACE2 transcription was categorized as: low, mean and high and reported for +/- 1 or 2 SD or greater, simple slopes were calculated across transmembrane ACE2 transcription from: −7 to 5 by 0.5.

P-values are reported from the results of a partial F-test to evaluate nested models.

soluble ACE2 effect modification was also assessed as a continuous value using multiple linear regression (Β= −0.099, 95%CI: [−0.18 to −0.022], P = 0.011).

Fig. 3

The association between transmembrane ACE2 transcription and SARS-CoV-2 RNA load in nasopharyngeal tissue differs by the amount of soluble ACE2 transcription. Soluble ACE2 transcription was categorized into low, mean and high levels to demonstrate the relationship. The low category (gray) represents soluble ACE2 transcription one standard deviation or greater below the mean transcription (n = 74). The mean category (orange) codes for the mean soluble ACE2 transcription at zero standard deviations. The high category (blue) indicates soluble ACE2 transcription one standard deviation or greater above mean transcription (n = 80). Shaded areas represent 95% confidence intervals, solid lines represent Β-coefficients for simple slopes derived from multiple linear regression. Figure S4 visualizes the corresponding effect of soluble ACE2 transcription +/- 2 SD or greater from the mean; estimates are reported in Table 2.

Discussion

In the described study, we measured nasopharyngeal ACE2 and TMPRRS2 gene transcription by qRT-PCR in a cross-sectional sample of n = 424 participants tested for COVID-19 in British Columbia. Analysis was performed to understand: the relationship between ACE2 transcription and age in unmatched COVID-19-negative participants, differences in host gene transcription between matched COVID-19-negative and -positive participants, and the role of soluble ACE2 transcription in SARS-CoV-2 infection in those who tested positive for COVID-19.

Among participants who tested negative for COVID-19, we found no relationship between nasopharyngeal ACE2 transcription and age in a group of adults between 19 and 98 years of age. This result agrees with findings from other efforts to characterize the relationship between age and nasopharyngeal ACE2 expression at the population level. Bunyavanich et al. reported a difference in nasal ACE2 expression levels between children younger than ten years old and young adults from eighteen to twenty-four years old [37]. Children younger than ten years old were found to have on average lower nasal ACE2 expression [37]. No difference in nasal ACE2 expression was evident between young adults, and adults twenty-five or more years in age which agrees with the findings of our study [37]. Lack of a relationship between nasopharyngeal ACE2 transcription and progressive aging suggests that expression of ACE2 protein in the upper airway may not correlate with expression of ACE2 protein in the RAAS [38]. The physiological importance of ACE2 protein in the RAAS suggests that its expression would increase in response to age as older people have higher prevalence of cardiovascular disease or similar comorbidities such as type 2 diabetes [39]. If nasopharyngeal expression of ACE2 protein occurs independently of the RAAS pathway, future studies should ascertain what factors, if any, regulate expression of ACE2 protein in nasopharyngeal tissue and variation of ACE2 expression in COVID-19-negative people over time. Some populations, such as young children, may have differential levels of nasopharyngeal ACE2 protein expression, which protect against SARS-CoV-2 viral infection [37,40].

Participants who tested positive for COVID-19 were matched to those who tested negative by age and biological sex to estimate the direct relationship between host gene expression and SARS-CoV-2 test result. The matched analysis demonstrated that mean transcription of transmembrane ACE2, soluble ACE2, and TMPRSS2 decreased in COVID-19-positive participants, while variation increased. Unfortunately, the limitation of reverse-causality in cross-sectional study design prevents us from understanding whether the presence of SARS-CoV-2 in positive samples affects host gene transcription or host gene transcription puts people at risk of viral infection. However, if we make the plausible assumption that at least a single round of viral replication must occur in order for qRT-PCR to generate a positive test result, it suggests that the observed variation in host gene expression of participants who test positive for COVID-19 results from viral replication [41]. Numerous studies have demonstrated that coronavirus replication disrupts cellular transcription, as resources required by the cell to produce mRNA are instead sequestered by viral genome replication [42]. Reduced transcription of ACE2 by SARS-CoV, but not HCoV-NL63, in an in vitro model of Vero E6 cell infection was previously suggested as a pathological mechanism [3]. Putative SARS-CoV-2 disruption of ACE2 expression could partially explain the apparent association between hypertension, type 2 diabetes, and severe COVID-19 [43]. Answering these questions, requires a longitudinal study design to test the temporal effect of SARS-CoV-2 replication on the expression of ACE2 and other implicated host genes.

The sample was then restricted to participants who tested positive for COVID-19 to investigate the association between nasopharyngeal ACE2 transcription and SARS-CoV-2 RNA load. Bivariate and multivariable analysis both suggest that ACE2 plays a dual role in SARS-CoV-2 infection. Transmembrane ACE2 positively correlates with viral RNA load, while soluble ACE2 may limit viral infection. Interestingly, effect modification was observed between transmembrane and soluble ACE2 transcription, implying that the proportion between these molecules may have more importance than absolute expression of either transcript individually. For example, between two people with similar above-average nasopharyngeal transmembrane ACE2 expression, we would expect a higher viral RNA load in the one with lower soluble ACE2 expression. These findings have interesting potential for clinical application, considering RT-qPCR quantification of ACE2 transcripts can be easily integrated into existing COVID-19 molecular diagnostics. Routine clinical measurement of ACE2 may provide a useful predictive marker of peak viral load in patients, disease severity, or SARS-CoV-2 transmissibility after the onset of symptoms.

Though we are not equipped to characterize the mechanism by which transcription of soluble ACE2 protein plays a protective role in SARS-CoV-2 infection, identification of this effect at the population level warrants further investigation of the underlying molecular process. Previous work has postulated that soluble ACE2 protein restricts SARS-CoV/SARS-CoV-2 infection by acting as a decoy substrate [44]. In a study of virus-ACE2 dynamics in engineered human tissue, virus particles bound to soluble ACE2 protein instead of binding to transmembrane ACE2 protein and, consequently were unable to infect susceptible cells [44]. Moreover, recombinant soluble ACE2 protein was recently administered to a COVID-19 patient requiring ventilation [11]. In the case report, administration of soluble ACE2 protein decreased viral RNA load in the patient's plasma, tracheal aspirate and nasopharyngeal specimens. Additionally, soluble ACE2 protein did not interfere with production of immunoglobulin G and assumed its cardiovascular protective role in the RAAS [11].

Soluble ACE2 protein may also influence the viral use of Neuropilin-1 as a co-receptor or alternative receptor for SARS-CoV-2. Given that the role of Neuropilin-1 in cell entry and infectivity may depend upon high SARS-CoV-2 viral load [45], it is conceivable that accessory expression of soluble ACE2 protein may limit SARS-CoV-2 tropism.

Strengths and limitations

We acknowledge that the described study has several limitations. In our sample, the age distribution possesses a left skew, which prevents us from examining the relationship between age and nasopharyngeal expression of ACE2 in participants younger than 19 years of age. The difference in mean host gene expression observed between participants who tested negative or positive for COVID-19 may be due to reverse causality, which cannot be resolved in a cross-sectional study design. Unmeasured confounding from comorbidities or patient drug prescriptions may bias the estimated effect of transmembrane ACE2 transcription on viral RNA load. Peters et al. reported increased expression of transmembrane ACE2 in asthma patients with type 2 diabetes and decreased expression in those prescribed inhaled corticosteroids [38]. We describe associations between transmembrane, soluble ACE2, and SARS-CoV-2 RNA load, which suggest the dual role of ACE2 transcripts in viral infection. However, we cannot provide evidence of the mechanism responsible for this association.

We also acknowledge the limitations of using quantitative polymerase chain reaction (qPCR) -based methods as a surrogate for gene expression and viral load. The measure of viral RNA load may overestimate infectious viral titre and viral RNA may be isolated in the absence of infectious virus; however, laboratory studies have demonstrated the presence of culturable virus in nasopharyngeal specimens across a broad range of Ct values [32], [33], [34], [35], [36]. Similarly, we acknowledge that the abundance of host gene transcripts does not reflect post-transcriptional regulation of expression. However, experimental studies have found a strong concordance between ACE2 transcript expression and protein concentration in upper airway tissues [46]. Notwithstanding these limitations, qPCR-based methods are both sensitive and, practical for high-throughput use in clinical laboratories with pandemic-imposed resource constraints, while easily integrated into existing COVID-19 molecular diagnostics.

Future studies are required to validate our findings at the molecular level. Importantly, the correlation between alternatively spliced ACE2 transcript and soluble ACE2 protein should be investigated in nasopharyngeal tissue of COVID-19 -negative and -positive specimens.

We have characterized nasopharyngeal ACE2 transcription in a cross-sectional sample of people tested for COVID-19 in British Columbia, Canada. Analysis shows no relationship between age and nasopharyngeal ACE2 transcription in COVID-19-negative participants 19–98 years old. Transcription of nasopharyngeal ACE2 and TMPRSS2 differs between COVID-19-positive and -negative groups, with lower mean values and greater variation observed in COVID-19-positive participants. The role of nasopharyngeal ACE2 in SARS-CoV-2 infection of the upper-airway may be differentiated by gene isoform expression; transmembrane ACE2 isoform transcription positively correlates with viral RNA load, while soluble ACE2 isoform transcription shows a negative association. The contrasting role of transmembrane and soluble ACE2 transcription is positioned to be of importance for prevention and care of COVID-19. Individuals with high endogenous soluble ACE2 expression may be less likely to transmit the virus to others or develop a lower respiratory infection as they clinically present with lower viral load. Similarly, we would expect treatment with recombinant soluble ACE2 protein to decrease viral load and, thereby, the severity and transmissibility of COVID-19 disease.

Contributions

AMN, DDWT, KSK, NAP, ANJ, MK, DMP, and IS conceived, designed, and obtained funding for the project. AMN wrote the manuscript, editing was provided by Dr. Karen Chu. AMN, DDWT and KSK performed the experiments. CDL, HS and AMN obtained and cleaned laboratory data. AMN, KSK, DDWT, CAB, CDL, and HS analyzed the data under the direction of CS, NAP, ANJ, MK, DMP and IS. AMN, DDWT, KSK and IS verified the raw data. All authors interpreted the data, read the manuscript and provided their approval to publish.

Declaration of Competing Interests

The authors have no conflicts of interest to declare.