Intronic regulation of SARS-CoV-2 receptor (ACE2) expression mediated by immune signaling and oxidative stress pathways

Summary The angiotensin-converting enzyme 2 (ACE2) protein is a key catalytic regulator of the renin-angiotensin system (RAS), involved in fluid homeostasis and blood pressure modulation. ACE2 also serves as a cell-surface receptor for some coronaviruses such as SARS-CoV and SARS-CoV-2. Improved characterization of ACE2 regulation may help us understand the effects of pre-existing conditions on COVID-19 incidence, as well as pathogenic dysregulation following viral infection. Here, we perform bioinformatic analyses to hypothesize on ACE2 gene regulation in two different physiological contexts, identifying putative regulatory elements of ACE2 expression. We perform functional validation of our computational predictions via targeted CRISPR-Cas9 deletions of these elements in vitro, finding them responsive to immune signaling and oxidative-stress pathways. This contributes to our understanding of ACE2 gene regulation at baseline and immune challenge. Our work supports pursuit of these putative mechanisms in our understanding of infection/disease caused by current, and future, SARS-related viruses such as SARS-CoV-2.


INTRODUCTION
The angiotensin-converting enzyme 2 (ACE2) protein has been highly studied as a key catalytic regulator of the renin-angiotensin system (RAS), involved in fluid homeostasis and blood pressure modulation (Lavoie and Sigmund, 2003). ACE2 control on this system occurs both directly (i.e., by lowering levels of angiotensin II) and indirectly (i.e., via alternative cleavage products) inhibiting the self-damaging effects of RAS overactivation, including vasoconstriction, fibrosis, and excessive inflammation ). The RAS system functions across different organs (Lavoie and Sigmund, 2003), and similarly, ACE2 is expressed throughout the body (Aguiar et al., 2020;Hikmet et al., 2020) where it mediates its protective effects and impacts tissue function . This activity has prompted the pursuit of ACE2 as a clinical target for protection and treatment against cardiovascular disease, diabetes mellitus, and acute lung damage Wang et al., 2020).
In addition to its important physiological role as a broadly expressed membrane-bound protein , ACE2 serves as a cell-surface receptor for some viruses-most notably, coronaviruses such as SARS-CoV (Li et al., 2003) and SARS-CoV-2 Wang et al., 2020;Zhou et al., 2020). Protein overexpression studies have demonstrated that ACE2 facilitates SARS-CoV-2 infection (Zhou et al., 2020), while mice with engineered human ACE2 are susceptible to infection (Yinda et al., 2021), and it has been suggested that the distribution of ACE2 receptor expression across different tissues contributes to differential virus susceptibility (e.g., lung tissue and alveolar cells) (Zhang et al., 2020a). The tissue expression of ACE2 may also explain the wide-ranging symptoms of COVID-19 in patients (Clerkin et al., 2020), though alternative means of viral entry has been suggested (Aguiar et al., 2020). During infection, ACE2 proteins bound by SARS-CoV-2 particles are endocytosed which, along with increased ADAM17 activity and upstream transcriptional changes, lead to a depletion of cell-surface ACE2 localization and reduced angiotensin catalytic activity (Clerkin et al., 2020;Wang et al., 2020). It has been suggested that reduced expression of ACE2 may lead to an imbalance of the RAS system in patients with COVID-19, which may represent a major pathological outcome of viral infection Lanza et al., 2020).
Given that levels of ACE2 expression may impact viral susceptibility (Aguiar et al., 2020;Devaux et al., 2020), and that subsequent changes to expression is a likely pathogenic mechanism of disease Lanza et al., 2020), an improved understanding of how ACE2 expression is regulated at the genomic and transcriptional level may help us understand not only how the effects of pre-existing conditions (e.g., chronic obstructive pulmonary disease, (COPD)) may manifest with increased COVID-19 incidence but also the mechanisms that regulate ACE2 levels following viral infection (Bhalla et al., 2020;Ni et al., 2020a;Wang et al., 2022). In this study, we first perform bioinformatic analyses of several datasets to generate hypotheses about ACE2 gene-regulatory mechanisms in the context of immune signaling and chronic oxidative stress. We next identify putative non-coding regulatory elements within the intronic regions of the ACE2 gene as potential determinants of ACE2 expression activity. We then perform functional validation of our computational predictions via targeted deletion of the identified ACE2 cis-regulatory elements in the context of immunological stimulation and oxidative stress conditions. Our results demonstrate the presence of intronic ACE2 regulatory elements responsive to both immune signaling and oxidative-stress pathways, contributing to our understanding of how expression of this gene may be modulated at both baseline and during immune challenge. Furthermore, our work supports the further pursuit of these putative mechanisms in our understanding, prevention, and treatment of infection and disease caused by ACE2-utilizing viruses such as SARS-CoV, SARS-CoV-2, and future emerging SARS-related viruses (Wells et al., 2021).

Upregulation of ACE2 gene expression in healthy individuals is associated with immune signaling and viral infection
To first examine patterns of baseline ACE2 gene expression, we analyzed microarray expression datasets from a cohort of healthy, never-smokers (N = 109) (see Table S1 for accessions). In these individuals, ACE2 was co-expressed with a gene set that is most significantly enriched in immune signaling and virus perturbations ( Figure 1, Table S1). The top transcription factors associated with these genes included IRFs and STATs (e.g., IRF1 and STAT1). Consistent with this finding, both IRF1 and STAT1 genes were also among the top 200 ACE2-correlated genes. Other genes that were associated with these enriched ''immuneresponse'', and ''viral response'' terms, and co-expressed with ACE2, include IFI16, IFI44, IFI35, NLRC5, and TLR3. These findings suggest that in healthy never-smokers, ACE2 may be a component of an immune signaling pathway, specifically relating to viral sensing and response and potentially mediated by IRF and STAT transcription factors.
Given the above findings on viral sensing, we next looked for the presence of these co-expressed immune response genes in an aggregated resource of COVID-19 RNA-seq studies Park et al., 2022). We found that all of the genes highlighted above were also significantly upregulated in infected and diseased patients. More broadly, we also found that interferon signaling genes (e.g., STAT1) were upregulated in infected samples ( Figure S1). This suggests that our association between ACE2 co-expression and immune signaling in healthy microarray expression datasets also reflects the importance of immune signaling (possibly via IRF and STAT regulation) during SARS-CoV-2 infection and disease.
During our analyses, we also detected as co-expressed with ACE2, DUOX2, a known response factor to reactive oxygen species (ROS) (Ewald, 2018). This suggests that oxidative stress may be another important mechanism that regulates ACE2 expression (see below). Also of interest are genes that help identify celltype-specific regulation of ACE2, along with its co-regulated gene network. The third top-correlated gene with ACE2 in healthy non-smokers was MUC13, an epithelial mucin known to be expressed in the large intestine and trachea (Williams et al., 2001) as well as in goblet cells, which are all proposed sites of SARS-CoV-2 replication .
We next examined a cohort of N = 136 individuals with asthma to investigate whether the observed associations (e.g., between ACE2 co-expression and inflammatory signaling) persisted in individuals with chronic inflammatory lung disease ( Figure S2). ACE2 co-expression in asthmatic individuals was also associated with immune signaling, antiviral responses, and IRF and STAT transcription factors. The top ACE2correlated gene in asthmatics was CD47, which is involved in the regulation of interferon gamma. iScience Article Consistent with this finding, ACE2 and CD47 are both co-expressed with the interferon-inducible gene IFI44, whose expression is regulated by IFN-g exposure (Zeng et al., 2006). Interestingly, IFI44 has been suggested as a key target for controlling the cytokine storm-induced immunopathology observed in patients with influenza virus and high pathogenic coronavirus infections (Dediego et al., 2019). Based on our microarray expression analyses, we hypothesized that ACE2 transcriptional regulation is associated with an immune signaling pathway involving IRF and STAT factors.
An important limitation of microarray data concerning ACE2 is the inability to discriminate between fulllength ACE2 and the recently discovered short-length isoform dACE2 (Onabajo et al., 2020). Therefore, the relative contribution of full-length ACE2 versus short-form dACE2 to these expression profiles remains unclear. We therefore sought more explicit, experimental interrogation of ACE2 gene regulation by considering the cis-regulatory landscape of the ACE2 locus. iScience Article Identification of functional intronic ACE2 regulatory elements with STAT1 and IRF1 binding sites Gene expression is controlled by regulatory sequences bound by transcription factors. We next examined the regulatory region encompassed by ACE2, compiling chromatin-accessibility datasets (i.e., DNase-I Hypersensitivity Sites, (DHS)) from in vitro and adult in vivo lung samples from the ENCODE project (Davis et al., 2018) ( Figure 2). We identified six intronic putative regulatory regions overlapping either cell-line or primary tissue DHS signals, a number of which also possess potential binding motifs for STAT1 and IRF1. We then refined this list to three intronic putative regulatory elements (Regions 1,4, and 5 in Figure 2) that overlap DHS data from both lung cell and tissue data and which contained either predicted STAT1 and IRF1 binding motifs and/or aggregated ChIP-seq datasets for each factor (see STAR Methods). These predicted factor binding motifs may be directly bound by STAT1 (Regions 1 and 5), and one possibly bound by IRF1 (Region 5). Furthermore, these putative regulatory elements were also identified in a previous study of regulatory activity in the ACE2 locus (Lee et al., 2021).
In order to test the functionality of these three regulatory elements on ACE2 regulation, we designed CRISPR guide RNAs (sgRNAs) to target and delete each element. We also designed sgRNAs to target predicted STAT1 and IRF1 motifs in Region 4 (Table S2). We tested our targeting strategy in vitro on a human lung epithelial cell line (Calu-3). To rule out potential off-target effects, we first confirmed that transfection of sgRNA plasmids did not disrupt the expression of nearby genes. Expression levels of nearby TMEM-27 and BMX1 were not significantly altered with deletion of any element or putative binding site (Table S2).
Using full-length isoform-specific primers, we next assessed levels of full-length ACE2 transcripts using qPCR in wild type and CRISPR-deleted cells. We found that full-length ACE2 expression was significantly decreased with deletions of each individual element, as well as the targeted binding sites within Region 4 ( Figure 2C and Table S2). A recent study identified that the dACE2 isoform is regulated upon SARS-CoV-2 infection (Onabajo et al., 2020). Interestingly, using primers specific to dACE2, we found that its transcript levels were also significantly decreased in our deletion experiments, and to a greater degree compared to full-length ACE2 (Table S2). These results indicate that, in the absence of additional perturbation (i.e., above transfection), each of the three candidate intronic regulatory sequences we tested acts as an enhancer specifically for ACE2.
Our bioinformatic analyses of RNA-expression datasets and subsequent motif/ChIP-seq scans suggest that ACE2 expression is regulated by an immune signaling pathway, possibly through STAT1 and IRF1 binding activity intronic to the ACE2 locus. We therefore tested the effects of deleting these putative immuneresponsive elements and specific binding sites in the context of immune signaling. Type I interferons (IFNs), such as IFN-a are our first line of defense against invading viruses . We used IFN-a treatment to induce intracellular immune signaling pathways that would occur during viral infections (Taniguchi and Takaoka, 2002) (see STAR Methods). We first performed this experiment on wild-type cells and found that this treatment led to a moderate increase of full-length ACE2 transcripts only after 48H (Figure S3A), with dACE2 levels increasing strongly and significantly at both time points, consistent with previous studies (Lee et al., 2021;Onabajo et al., 2020) (Table S2). We independently confirmed this finding at the protein level ( Figure S3B), and further found that additional potent inducers of immune signaling, such as poly(I:C) treatment and direct infection with SARS-CoV-2, did not lead to significant upregulation of fulllength ACE2 at the protein level.
We next performed IFN-a treatment in the context of enhancer deletion. We observed a significantly decreased effect of CRISPR element deletion on full-length ACE2 gene expression reduction with IFN-a stimulation compared to expression changes in the absence of stimulation ( Figure 2D and Table S2). This was observed across the majority of our element and sub-element (i.e., motif) deletions. This attenuated downregulation was also observed for dACE2 across stimulation-deletion experiments (Table S2). These findings suggest that these enhancer elements may be in part responsive to immunological stimulation (via IFNs) and play a role in a more complicated, potentially redundant, regulatory mechanism for ACE2 expression (see Discussion).
ACE2 gene expression in lung epithelial cells is correlated with smoking and COPD disease status and associated with an NRF2 antioxidant response iScience Article Vardavas and Nikitara, 2020), more recent studies have cast doubts as to the strength and significance of this relationship (Cattaruzza et al., 2020;Lippi and Henry, 2020). The relationship between smoking history and respiratory viral infection disease severity has been suggested to be more complicated (Zhao et al., 2020). It is also worth noting that ACE2, in addition to being the primary receptor for SARS-CoV-2 infection  iScience Article Zhou et al., 2020), serves an important biological role in multiple tissues , and is present in lung epithelium (Aguiar et al., 2020;Hikmet et al., 2020). Thus, shifts in basal expression levels of this protein, especially over time, may contribute to lung dysfunction in an indirect, more complex manner than can be measured using metrics such as COVID-19 disease severity. Given this possibility, we next assessed expression patterns of bronchial brushing datasets from current and previous smokers, focusing on ACE2 and other co-expressed genes (e.g., DUOX2).
We analyzed a dataset of 159 healthy current smokers versus healthy former smokers, and identified the top 200 ACE2-correlated genes ( Figure 3, Table S1). Expression patterns for ACE2 suggest current smoking status is associated with increased ACE2 levels, consistent with previous observations (Brake et al., 2020;Cai et al., 2020;Leung et al., 2020aLeung et al., , 2020b, and that this also accounts for the increased ACE2 in patients with COPD ( Figure 3A). Expression patterns for ACE2-correlated genes alone were able to effectively distinguish smokers from non-smokers ( Figure 3B). Functional enrichment analysis showed that in this dataset, genes co-expressed with ACE2 are significantly associated with the NRF2 pathway, oxidative stress, glutathione metabolism, and TGF-b regulation of the extracellular matrix. NRF2 is a key transcription factor that regulates the oxidative stress response in the lung (Gebe et al., 2010;Rangasamy et al., 2004). Consistent with this, according to both ChIP-seq data and gene expression perturbation data, NRF2 (NFE2L2) was the top transcription factor identified as a likely regulator of these genes; for example, the NRF2-regulated antioxidant gene NQO1 was the fourth ranked ACE2-correlated gene in this dataset. ACE2-correlated genes also overlapped significantly with genes upregulated by the transcription factor ETS1 (GSE21129) (Verschoor et al., 2010); ETS1 is an important regulator of ROS (reactive oxygen species) in response to angiotensin II, linking it to ACE2 function (Ni et al., 2007). Moreover, ETS1 expression is induced by ROS exposure through an antioxidant response (Wilson et al., 2005). Thus, ACE2 expression in smokers appears to be associated with oxidative stress gene regulation, likely mediated by NRF2 and ETS1.
To verify these trends, we repeated the same analyses with a second cohort dataset from a different microarray platform associated with 345 healthy smokers versus healthy non-smokers ( Figure S4). Notably, this dataset consists predominantly of younger individuals (age <50), whereas the first dataset includes predominantly older individuals (>50). ACE2 co-expression patterns were highly correlated between the two independent datasets providing support that these are robust signals. As with the first analysis, genes co-expressed with ACE2 showed the strongest associations with NRF2 gene targets. Moreover, NRF2 and ETS1 formed the top three overlapping datasets according to enrichments for transcription factor perturbation datasets ( Figure S4C). This dataset also included a larger number of patients with COPD; that we observed similar patterns of ACE2 expression and co-expressing genes in this analysis may suggest similar effects on smoking and disease status on ACE2 regulation (see Discussion).

Identification of functional intronic ACE2 regulatory elements with possible antioxidantresponse element (ARE) activity
Our microarray data analyses led to a predicted association between oxidative stress and ACE2 levels, which prompted us to consider the existence of antioxidant-response elements (AREs within the ACE2 locus ( Figure 4A). We performed an unbiased analysis of the regulatory regions intronic to ACE2 ( Figure 4B). Our analysis identified AP-1 as the top enriched transcription factor binding site in the locus, suggesting that AP-1 may be an additional regulator of ACE2. Klatt and colleagues have shown AP-1-c-Jun subunit binding to DNA is dependent on the cellular GSH/GSSG ratio, a marker of cellular ROS levels (Klatt et al., 1999). Looking at the top ACE2-correlated genes across all datasets, we identified significant coexpression between ACE2 and FOSL2, as well as ACE2 and JUN. This suggests that AP-1 may be a transcription factor involved in oxidative stress-mediated regulation of ACE2 levels.
Given the observed motif enrichments for AP-1 and NRF2 transcription factor binding sites in the ACE2 locus, we next looked at the individual predicted motif hits for AP-1 and NRF2 factors within the six putative regulatory regions defined above ( Figure 4B). Three of these enhancers (Regions 1, 4, and 5) contained predicted AP-1 and NRF2 binding sites and were active in both in vivo and in vitro lung datasets. Two of these regions (Region 1 and 4) also overlapped ChIP-seq datasets for FOS and JUN factors, important co-factors associated with AP-1 complex (Rauscher et al., 1988) and NRF2 (Jeyapaul and Jaiswal, 2000) enhancer binding, respectively. We next sought to delete these elements in the context of oxidative stress, with the expectation that, if these elements act as AREs, that the effects of deletion should be magnified during an ROS response. We first examined ACE2 transcript levels after exposing wild-type Calu-3 cells to hydrogen peroxide (0.5mM), a potent ROS (Boardman et al., 2004) (see STAR Methods). We found that exposure to hydrogen peroxide led to significant decreases in expression of both full-length ACE2 and dACE2 ( Figure S5, Table S2). Previous mouse studies have also demonstrated a downregulation of ACE2 levels and activity following acute ROS exposure (Fang et al., 2019). However, ACE2 levels are upregulated with chronic oxidative stresses (Gebe et al., 2010;Smith et al., 2020); this may suggest a more complicated regulation of ACE2, possibly as a function of time (see Discussion).  iScience Article We next deleted each regulatory element and assessed ACE2 expression in the presence/absence of hydrogen peroxide treatment (see STAR Methods). We observed a significant decrease in basal levels of full-length ACE2 transcripts for the majority of regions/sites deleted in the absence of external stimulus (Table S2). In the context of exogenous oxidative stress, we observed a further significant decrease only for Region 1, while all others trended downward (Table S2). For this first region, the magnitude of downregulation was significantly greater under oxidative stress when compared to the unstimulated deletion change ( Figure 4C), while for other regions magnitudes were similarly larger despite the lack of significance (potentially due to the increased variability in expression observed in ROS-stressed cells). This first region contains predicted NRF2 and AP-1 motifs, and also overlaps with both FOS and JUN ChIP-seq signals, possibly explaining the increased effect of deletion under ROS conditions. Finally, we again saw these differences to be accentuated when considering levels of the dACE2 transcript (Table S2).

DISCUSSION
Regulation of ACE2 expression at the transcriptional level may impact susceptibility to viral infection. Subsequently, changes to ACE2 expression during viral infection can lead to an imbalance in renin-angiotensin system (RAS) signaling contributing to the manifestation of clinical symptoms such as excessive inflammation (Mahmudpour et al., 2020), myocardial injury (Clerkin et al., 2020), and lung injury (Ni et al., 2020b). It has recently been suggested that dysregulation of the RAS system during initial infection may be partially responsible for the activation of the cytokine storm observed in some severe cases of COVID-19 (Mahmudpour et al., 2020), wherein the loss of surface ACE2 promotes the release of inflammatory cytokines via enhanced Ang II signaling. Elevated Ang II has also been suggested to mediate consequences of RAS imbalance tied to COVID-19 (Lanza et al., 2020), including severe hypoxia and lymphopenia. These findings indicate that targeting transcriptional inhibition of ACE2 expression may be a therapeutic avenue for prevention of severe COVID-19 infection (Qiao et al., 2020), while counteracting infection-induced ACE2 downregulation may act as a therapeutic treatment to reduce disease severity (Chatterjee and Thakur, 2020;Zhang et al., 2020b).
In the first part of our study, we utilized microarray expression datasets from healthy non-smokers and identified genes whose expression patterns significantly correlated with that of ACE2. Groups of correlated genes may suggest shared upstream regulators; gene set enrichment analyses indicated that ACE2 and correlated genes may be under the control of immune signaling pathways integrating on the STAT and IRF families of transcription factors-namely, STAT1 and IRF1. These results were also observed when performing a separate analysis on asthmatics individuals ( Figure S2). Moreover, expression data from COVID-19-infected samples further substantiated this association with immune signaling pathways ( Figure S1). Interestingly, a recent study of microarray datasets from SARS-CoV-2 infected cells suggested the importance of JAK-STAT signaling (Luo et al., 2021). In that study, the authors found that JAK-STAT components (e.g., STAT1) were co-expressed with ACE2 in SARS-CoV-2-infected cell lines, and suggested that ACE2 could act to regulate JAK-STAT activity. Given our computational and experimental findings, we suggest that immune signaling pathways such as JAK-STAT act upstream to regulate ACE2. This hypothesis has been further corroborated by an independent study, which found that the transcriptional response of ACE2 (as well as dACE2) to interferon stimulation was mitigated by pharmacological inhibition of JAK (Lee et al., 2021). Interestingly, this study also found that the expression of genes identified in our correlated expression analysis, including STAT1, IFI44, and IRF1, were also under control of immune signaling pathways involved in interferon response (Lee et al., 2021).
Given our computational findings, we considered cis-regulatory elements within the ACE2 locus which may be proximate mediators of this immune-response regulatory mechanism. After identifying six such intronic regions, we prioritized and then experimentally deleted three putative intronic enhancers, containing predicted STAT1 and IRF1 binding motifs. Deletion of these three elements (Regions 1, 4, and 5) separately led to a consistent downregulation of ACE2 transcripts. The downregulation of ACE2 upon deletion, relative to mock-transfected controls, suggests that these enhancers may contribute to basal expression both individually but possibly collectively. This latter possibility could not be examined because experimental deletion in tandem of enhancers spanning separate introns would likely generate ACE2 loss-of-function contexts as well as complex gene/regulatory element interactions at the locus. Interestingly, performing individual deletion of each enhancer in the context of immune stimulation, we did observe a significant attenuation of the downregulation caused by our deletions, while immune stimulation in wild-type cells caused moderate changes to ACE2 expression ( Figure S3). These results corroborate our previous findings that ll OPEN ACCESS iScience 25, 104614, July 15, 2022 9 iScience Article SARS-CoV-2 infection does not significantly upregulate transcript levels of ACE2 in Calu-3 cells in spite of significant increases in type I and type III IFNs, as well as upregulation of known interferon-stimulated genes (e.g., IFIT1, IRF7, and OAS2) . The observed attenuation effect may suggest an upperthreshold or ''saturation'' of ACE2 expression from baseline, such that immune signaling does not lead to a substantial increase. However, following deletion of these putative enhancers, proximate regulators of immune signaling (e.g., STAT1 and IRF1) acting elsewhere in the ACE2 locus (e.g., at the promoter level (Ziegler et al., 2020)) may be able to compensate for the loss/reduction in enhancer activity. Further experimentation (e.g., using a viral-infection model system) into this complex regulatory system may elucidate the role that these enhancers play in upregulating ACE2 during an immune response.
An understanding of the mechanisms regulating ACE2 expression during viral infection is important from a disease-pathology point of view, given that this may inhibit the protective effects of ACE2 activity. In addition, understanding the regulatory mechanisms controlling ACE2 expression prior to viral exposure may be of equal importance from the perspective of disease prevention, given that baseline levels of ACE2 in highexposure tissues (e.g., lung) may modify viral susceptibility (Devaux et al., 2020).
It has been suggested, though not conclusively shown, that chronic smokers are at elevated risk to both SARS-CoV-2 infection as well as severe disease (Alqahtani et al., 2020;Cattaruzza et al., 2020;Patanavanich and Glantz, 2020). This follows with previous studies of other coronaviruses, e.g., MERS-CoV, for which epidemiological evidence does suggest smoking status as a key risk factor (Arcavi and Benowitz, 2004;Park et al., 2018). In terms of increased susceptibility, it may be that smokers have elevated baseline ACE2 expression in lung tissues, increasing the likelihood that SARS-CoV-2 may bind their target receptors (Qiao et al., 2020). In the second part of our study, we analyzed microarray expression data from two independent datasets consisting of current smokers, non-smokers, and patients with COPD. Considering the expression of ACE2, we also observed previously reported increases in baseline expression within smokers (Qiao et al., 2020;Smith et al., 2020). With this, as well as genes showing similar transcriptional behaviors, we identified enrichments for oxidative stress-response pathways, including transcriptional regulators such as NRF2 and the AP-1 complex. These signals are indicative of another potential regulatory mechanism acting on the ACE2 locus, and are expected given the chronic oxidative stresses experienced by habitual smokers and patients with COPD (Pierrou et al., 2007).
Looking again at the ACE2 locus, we found enrichment in open-chromatin regions (putative enhancers) for DNA sequences bearing similarity to known FOSL2:JUN binding motifs, further suggesting the regulatory effects of oxidative-response signaling at this locus. We therefore performed another set of targeted deletion experiments of intronic enhancers most likely to behave as antioxidant-response elements (AREs) (e.g., contain NRF2 motifs, AP-1 ChIP-seq signals, etc.). Performing these deletions in the context of exogenous oxidative stress yielded a substantial decrease in ACE2 expression for the first element tested (Region 1), with a fold-change decrease below that observed in wild-type cells following treatment.
We suggest that this putative ARE, and potentially others which trend in the same direction, act to counter the inhibitory effects of oxidative stress on ACE2 expression, which has been previously observed in a mouse model of hyperoxia (Fang et al., 2019), preventing a more deleterious loss of ACE2 protein following acute exposure. ACE2 plays an important role in mitigating acute oxidative stress (Xia et al., 2011;Zheng et al., 2014), particularly in the context of cardiovascular and lung disease (Rabelo et al., 2011;Shenoy et al., 2011). We further propose that the repression of ACE2 upon acute oxidative stress, when repeated on the order of decades in chronic smokers, may lead to an ''overcompensation'' of baseline ACE2 expression-establishing higher levels of ACE2 protein to protect lung tissues from further damage. This process could be mediated by a number of oxidative stressresponse mechanisms; in particular, our observed enrichments for NRF2-regulated genes co-expressed with ACE2, along with the presence of NRF2 motifs within intronic ACE2 enhancers, follow with the protective role of NRF2 signaling induced in response to cigarette smoke (Ma, 2013). Furthermore, a mousemodel study of cigarette smoke found significant increases in ACE2 activity only after three weeks of exposure (Hung et al., 2016), while additional studies have found dose-response effects with increased treatment time (Gebe et al., 2010;Liu et al., 2021). Human smokers also exhibit a dose-response effect of ACE2 expression with increasing pack-years (Smith et al., 2020). However, we acknowledge the speculative nature of this proposed over-compensating effect and note the importance of additional experimental testing. While the links between smoking status and COVID-19 severity are controversial, the link ll OPEN ACCESS 10 iScience 25, 104614, July 15, 2022 iScience Article between COPD status and COVID-19 severity may be clearer (Leung et al., 2020a(Leung et al., , 2020bOlloquequi, 2020;Zhao et al., 2020). More generally, it has been suggested that the detrimental effects of smoking, most notably attenuation of antiviral innate immune responses (Sopori, 2002), can increase susceptibility to pathogen infection (Arcavi and Benowitz, 2004).
In summary, here, we explored the regulatory mechanisms which may act on the ACE2 locus in the context of both immune stimulation as well as oxidative stress, leading us to identify two putative pathways which may mediate this transcriptional regulation. It is important to note that these pathways are not mutually exclusive; the links between immune signaling and oxidative stress are well established (Rahman et al., 1996), and this is particularly true for ACE2 given its biological role in RAS regulation Wang et al., 2020). We suggest that further experimental testing is warranted to confirm these predicted mechanisms, and furthermore to develop potential strategies taking advantage of this knowledge to modify susceptibility and disease severity of coronavirus infections, particularly SARS-CoV-2.

Limitations of the study
The microarray analysis performed is limited to the detection of transcripts for which corresponding probes exist on the given chip, meaning that splice isoforms and rare variants are unlikely to be detected. This limitation impacts our study by preventing the detection of dACE2 and differentiation between this isoform and the canonical ACE2 within the sample population. Given that our identified intronic elements span multiple different exons in ACE2, it is extremely difficult to generate serial deletions in cis of these enhancers without generating any ACE2 loss-of-function scenarios. Thus, we cannot comment on the potential combinatorial effects of different intronic ACE2 regulatory elements to regulating expression.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

ACKNOWLEDGMENTS
The authors would like to thank Harvard University Dean Christopher Stubbs and Dean Francis Doyle for granting restricted access to the Capellini Laboratory during the COVID-19 pandemic to perform these studies. The authors would like to thank members of the Capellini, Doxey, and Hirota labs for critical insight into this work. AB was the recipient of a fellowship from the Natural Sciences and Engineering Research Council of Canada (NSERC). VIDO receives operational funding for its CL3 facility (Inter-Vac) from the Canada Foundation for Innovation through the Major Science Initiatives. VIDO also receives operational funding from the Government of Saskatchewan through Innovation Saskatchewan and the Ministry of Agriculture. ACD is funded in part by an NSERC Discovery Grant (RGPIN-2019-04266). TDC is funded in part by The