DYRK1A promotes viral entry of highly pathogenic human coronaviruses in a kinase-independent manner

Identifying host genes essential for Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) has the potential to reveal novel drug targets and further our understanding of Coronavirus Disease 2019 (COVID-19). We previously performed a genome-wide CRISPR/Cas9 screen to identify proviral host factors for highly pathogenic human coronaviruses. Few host factors were required by diverse coronaviruses across multiple cell types, but DYRK1A was one such exception. Although its role in coronavirus infection was previously undescribed, DYRK1A encodes Dual Specificity Tyrosine Phosphorylation Regulated Kinase 1A and is known to regulate cell proliferation and neuronal development. Here, we demonstrate that DYRK1A regulates ACE2 and DPP4 transcription independent of its catalytic kinase function to support SARS-CoV, SARS-CoV-2, and Middle East Respiratory Syndrome Coronavirus (MERS-CoV) entry. We show that DYRK1A promotes DNA accessibility at the ACE2 promoter and a putative distal enhancer, facilitating transcription and gene expression. Finally, we validate that the proviral activity of DYRK1A is conserved across species using cells of nonhuman primate and human origin. In summary, we report that DYRK1A is a novel regulator of ACE2 and DPP4 expression that may dictate susceptibility to multiple highly pathogenic human coronaviruses.


Introduction
Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2), the causative agent of Coronavirus Disease 2019 (COVID- 19), is a beta coronavirus that has launched an ongoing pandemic that continues to threaten public health globally [1,2].Two additional beta coronavirus family members (SARS-CoV and Middle East Respiratory Syndrome Coronavirus (MERS-CoV)) have caused more limited epidemics but with a higher case fatality rate [3][4][5].There are also 4 endemic alpha and beta human coronaviruses that cause the common cold (HCoV-NL63, HCoV-OC43, HCoV-229E, and HCoV-HKU1).Despite zoonotic outbreaks of 3 beta coronaviruses in less than 20 years, our understanding of host factors that support these highly pathogenic human coronaviruses and the immune response to them remain incompletely understood [1,6,7].
We performed a genome-wide CRISPR/Cas9-based inactivation screen in African green monkey kidney Vero-E6 cells and in human lung epithelial Calu-3 cells [23,24].Both Vero-E6 and Calu-3 cells are permissive to SARS-CoV, SARS-CoV-2, and MERS-CoV and endogenously express the cognate receptors ACE2 and DPP4 [23][24][25][26][27]. Our screens revealed several proviral genes shared by SARS-CoV, SARS-CoV-2, and MERS-CoV in both Vero-E6 and Calu-3 cells [23,24].Other genome-wide CRISPR/Cas9 screens for SARS-CoV-2 and other related coronaviruses have since identified additional host dependency factors including proteins involved in viral entry, endocytic trafficking and sorting, cholesterol homeostasis, and autophagy [28][29][30][31][32][33].While reproducibility between screens was excellent when performed on the same cell type, there was otherwise limited overlap across hits in these loss-of-function screens [28][29][30][31][32][33].One notable exception was DYRK1A, which was a top enriched hit in both Vero-E6 and Calu-3 cells [23,24,34,35].DYRK1A was not a hit in cell lines exogenously overexpressing ACE2, such as A549, Huh7.5, or HeLa cells [28][29][30][31][32][33].Here, we sought to elucidate how DYRK1A regulates CoV infection and to determine the cell-type specificity of its function.Because DYRK1A was identified in cells with high endogenous ACE2 expression and was are available under the Human Cell Atlas Data Coordination Platform and NCBI BIOPROJECT Accession code PRJEB31843 at: https://www.ncbi.nlm.nih.gov/bioproject/PRJEB31843 and the European Genome-phenome Archive under accession EGAS00001004344 at: https://egaarchive.org/studies/EGAS00001004344.All mice are available for purchase at the Jackson Laboratories.Viral stocks and plasmids are available BEI or Addgene or can be made available under Material Transfer Agreement.For initiation of the Material Transfer Agreement process, contact the corresponding author (Craig B. Wilen, craig.wilen@yale.edu) also a top hit for a replication competent SARS-CoV-2 pseudovirus, we posited that DYRK1A may function as a novel transcriptional regulator of coronavirus entry [23].As viral entry is the first and rate-limiting step of infection, reducing viral entry can mitigate infection and pathogenesis [36][37][38][39].While SARS-CoV-2 mRNA vaccines have proved highly efficacious at achieving this, the mechanisms that govern receptor expression represent a major gap in our knowledge of coronavirus biology [40].
Despite its roles in many biological processes, the role of DYRK1A in viral pathogenesis remains incompletely explored.DYRK1A can promote human papillomavirus type 16 infection by phosphorylating the viral oncoprotein E7 [75].Similarly, DYRK1A enables transformation by interacting with the oncoprotein E1A from human adenovirus type 5 (HAdV-5) [76].Inhibition of DYRK1A kinase activity or genetic knockdown of DYRK1A also reduces human cytomegalovirus (HCMV) replication [77,78].In the cases of both HCMV and HAdV-5, how DYRK1A performs this function is unclear.DYRK1A also possesses antiviral roles, such as by reducing HIV replication in vitro by down-regulating cyclin L2 and inhibiting long terminal repeat-driven transcription [79,80].
Here, we show that DYRK1A is critical for highly pathogenic human coronavirus infection in nonhuman primate and human cells, but not in mice.We demonstrate that DYRK1A is a novel regulator of coronavirus entry for both SARS-CoV, SARS-CoV-2, and MERS-CoV by promoting ACE2 and DPP4 receptor expression at the mRNA level.We further reveal that DYRK1A performs its proviral role in the nucleus independently of kinase function, suggesting a previously undescribed mechanism of DYRK1A activity in viral infection.

DYRK1A promotes viral entry for SARS-CoV-2, SARS-CoV, and MERS-CoV
We and others recently identified DYRK1A as a critical host factor for SARS-CoV-2, MERS-CoV, and chimeric HKU5 (bat coronavirus) expressing the SARS-CoV spike (HKU5-SARS-CoV-S) in a genome-wide CRISPR/Cas9 inactivation screen in Vero-E6 cells [23,34].In 2 additional independent genome-wide CRISPR screens in Calu-3 human immortalized lung cancer cells, DYRK1A was also identified as a top proviral gene for SARS-CoV-2 [24,34].By comparing the top 10,000 enriched genes ranked by z-score for SARS-CoV-2 across Vero-E6 and Calu-3 cells, we identified DYRK1A as second most strongly enriched gene after only ACE2 [23,24] (Fig 1A).In a separate screen, DYRK1A was the third most strongly enriched hit after ACE2 [34].We generated single-cell knockout (KO) clones of DYRK1A in Vero-E6 cells using CRISPR/Cas9 to validate screening results and clarify the proviral mechanism underlying DYRK1A activity in coronavirus infection (Fig 1B).We challenged clonal KO cells with an infectious clone of SARS--CoV-2 encoding the fluorescent reporter mNeonGreen (icSARS-CoV-2-mNG) and quantified viral infection by microscopy [81].Cells deficient in DYRK1A were less susceptible to SARS--CoV-2 infection (Fig 1C).Consistent with this finding and our screen, loss of DYRK1A conferred a reduction in virus-induced cell death by SARS-CoV-2, HKU5-SARS-CoV-S, and MERS-CoV (Fig 1D).We next tested whether DYRK1A acts at viral entry using a pseudotype assay, where coronavirus spike proteins are expressed on a replication-deficient vesicular stomatitis virus (VSV) encoding a luciferase reporter.This pseudovirus enables a single round of spike-dependent viral entry.Cells lacking DYRK1A exhibited a significant defect in CoV pseudovirus entry with a magnitude similar to ACE2 and CTSL KO (Figs 1E and S1).These data highlight DYRK1A as a proviral host factor for highly pathogenic human coronaviruses, including SARS-CoV-2, which regulates spike-mediated viral entry in Vero-E6 cells.

DYRK1A regulates expression of the receptors ACE2 and DPP4 at the transcript level
Because coronavirus entry is receptor dependent, we next compared the expression of the receptors ACE2 and DPP4 in wild-type (WT) cells and DYRK1A KO cells.In DYRK1A KO cells, ACE2 expression is notably reduced at the protein level by western blot (Fig 2A), but endogenous DPP4 expression was below the limit of detection even in WT cells (Fig 2D , left).To determine whether DYRK1A regulates ACE2 or DPP4 expression at the mRNA level, we performed quantitative reverse transcription PCR (RTAU : PleasenotethatRT À qPCRhasbeenfullyspell -qPCR) to quantify mRNA abundance (Fig 2B).Loss of DYRK1A causes a significant reduction in mRNA transcript levels of both ACE2 and DPP4, suggesting that DYRK1A may transcriptionally regulate the ACE2 and DPP4 loci or that DYRK1A may modulate the mRNA stability of these transcripts (Fig 2B).We then overexpressed ACE2 or DPP4 in DYRK1A KO cells by stable overexpression to clarify whether post-entry or protease-dependent entry may also be regulated by DYRK1A (Fig 2C and 2D).Overexpression of ACE2 and DPP4 both rescued the entry defect conferred by loss of DYRK1A for SARS-CoV-2 and MERS-CoV pseudoviruses, respectively (Fig 2C and 2D).Taken together, these data indicate that DYRK1A promotes viral entry by supporting the expression of ACE2 and DPP4 at the mRNA level.

The proviral role of DYRK1A is kinase independent in nature
To confirm the proviral phenotype and to elucidate the mechanism of DYRK1A proviral activity, we reintroduced DYRK1A into our single-cell KO clones (Fig 1B) by lentiviral  https://doi.org/10.1371/journal.pbio.3002097.g002transduction (Fig 3).We transduced the following constructs: WT DYRK1A, kinase-null DYRK1A (encoding K188R or Y321F inactivating point mutations), and nuclear localization mutant DYRK1A (encoding a nuclear export signal and disruption of the bipartite nuclear localization motif) (Fig 3A).DYRK1A requires ATP for catalytic activity.ATP binds with Lys 188 and a K188R substitution blocks DYRK1A phosphorylation activity [66,67,82].Disruption of Y 321 autophosphorylation by mutating tyrosine to phenylalanine (Y321F) renders DYRK1A catalytically inactive by preventing kinase maturation [66,[83][84][85].Because sustained DYRK1A overexpression can cause cell cycle exit [86][87][88], we generated these constructs under the control of a doxycycline inducible Tet-on promoter.We first confirmed that these addbacks could rescue DYRK1A and ACE2 expression (Fig 3B).Consistent with previous literature, expression of DYRK1A-K188R is weaker relative to WT and DYRK1A-Y321F constructs, likely due to protein destabilization (Fig 3B) [82].Complementation with WT DYRK1A restored ACE2 expression (Fig 3B).Unexpectedly, loss of kinase activity (K188R and Y321F) enabled at least partial rescue of ACE2 expression by western blot, whereas loss of nuclear localization did not elicit detectable ACE2 protein expression, despite partial retention of DYRK1A in the nucleus (Fig 3B and 3C).To confirm DYRK1A complementation could rescue infection with authentic virus, we challenged these cell lines with the icSARS-CoV-2-mNG reporter virus or WT SARS-CoV-2 WA/01/2020 to assess viral replication by mNG production and by plaque assay, respectively [81] ( Figs 3D, 3E, and S2A).Consistent with protein expression data, WT and kinase-dead constructs significantly rescued infection in DYRK1A KO cells (Figs 3D, 3E, and S2A).The nuclear localization mutant DYRK1A also restored infection, possibly due to residual DYRK1A in the nucleus (Figs 3C, 3E, and S2A).Next, we tested whether full or partial restoration of DYRK1A and ACE2 expression could rescue viral entry using pseudovirus.In all cases where DYRK1A was reintroduced, DYRK1A KO clones exhibited full or partial rescue of SARS-CoV-2 and MERS-CoV viral entry, including catalytic and localization mutants (Fig 3F and 3G).Because genetic approaches indicated that DYRK1A kinase activity is dispensable for SARS-CoV-2 infection, we then validated these findings using an orthogonal pharmacologic approach.We inhibited DYRK1A with 4 type I ATP-competitive kinase inhibitors: harmine, INDY, DYR219, or DYR533 (S2B Fig) [89][90][91][92][93].In contrast to a cathepsin L inhibitor, DYRK1A inhibitors did not mitigate virus-induced cell death at tolerated concentrations (S2B Fig).These data suggest that existing DYRK1A active site inhibitors are ineffective against SARS-CoV-2, which is consistent with kinase-independent regulation of infection (S2B Fig) .Collectively, these data support that nuclear DYRK1A regulate receptor expression in a kinase-independent manner.

DYRK1A drives ACE2 and DPP4 expression by altering chromatin accessibility
Because DYRK1A can positively regulate transcription and gene expression [63,69,70,72,73], we next profiled global gene expression (RNA-Seq) and chromatin accessibility (ATAC-Seq) to assess the mechanism of DYRK1A-mediated regulation of ACE2 and DPP4 (Figs 4 and S3).Loss of DYRK1A resulted in up-or down-regulation of approximately 2,000 genes.Analysis of differentially expressed genes (DEGs) confirmed that loss of DYRK1A confers down-regulation of ACE2 and DPP4, as well as CTSL, which encodes the protease for spike cleavage and viral entry in Vero-E6 cells [25] ( Figs 4B, 4C, and S3A).Importantly, complementation with both DYRK1A-WT and DYRK1A-Y321F rescue expression of these genes (Fig 4B and 4D).There was no correlation between DYRK1A-dependent gene regulation and CRISPR genes (ranked by z-score) with the exception of ACE2, DPP4, and CTSL [23] (Fig 4C).Gene ontology analysis also revealed little overlap between biological pathways regulated by DYRK1A and those involved in SARS-CoV-2 infection (S3B Fig) [23].Because RNA-Seq revealed that DYRK1A promotes increased mRNA levels for these genes, we next asked whether DYRK1A altered chromatin accessibility at these sites, thereby altering transcription.Loss of DYRK1A confers generally more open chromatin states across the genome (S3C Fig) .In contrast, absence of DYRK1A resulted in reduced accessibility near the ACE2 transcriptional start site (TSS), a putative proximal enhancer, and a putative distal enhancer situated within BMX (Figs 4F and S3C) [39].Chromatin accessibility was restored at these sites upon reintroduction of DYRK1A (Figs 4F and S3C).A second putative distal enhancer located within ASB11 does  not seem to be altered by DYRK1A (p > 0.05), indicating that DYRK1A-mediated regulation of ACE2 may be context or site specific (Fig 4F).Interestingly, three sites were identified where loss of DYRK1A led to significantly increased (p < 0.0005) chromatin accessibility within 5 kb of the ACE2 TSS, suggesting that DYRK1A may also close off chromatin via repressive activity at the ACE2 locus (S3C Fig) .In comparison to SMARCA4, another ACE2 regulator we have identified, DYRK1A-mediated DNA accessibility significantly overlaps with SMARCA4 at approximately 1/3 of sites (p < 0.00001 by Fisher's exact t test), suggesting that some pathways may be coregulated by the two (S4D Fig) .However, DYRK1A-mediated DNA accessibility (correlation coefficient approximately 0.33) and gene expression output (correlation coefficient approximately 0.08) overall poorly correlate with those regulated by SMARCA4 (S4C and S4E Fig) [39].Moreover, SMARCA4 promotes open chromatin states at both putative distal enhancers (BMX and ASB11) and the putative ACE2 proximal promoter [39].Together, these data suggest that DYRK1A regulates the majority of DNA accessibility and gene expression independently of SMARCA4.Therefore, our data support that DYRK1A is a critical regulator of ACE2 chromatin accessibility and transcription by a novel mechanism.Although enhancer regions and other regulatory elements are not well defined for DPP4 and CTSL, we observed increased chromatin accessibility at sites proximal and distal to the respective TSS for these genes, which was not the case for SMARCA4 at these sites (Fig 4G and 4H).Unlike in the case of ACE2, putative insulator regions (i.e., sites where loss of DYRK1A led to more open chromatin) were not identified for DPP4 or CTSL.These data highlight that DYRK1A-albeit not a canonical epigenetic modifying enzyme or transcription factor-can dramatically alter chromatin accessibility to drive transcription of a proviral gene expression axis that promotes viral entry.

DYRK1A is a conserved proviral factor for SARS-CoV-2 in human lung epithelial cells
We then sought to determine whether the proviral role of DYRK1A observed in Vero-E6 cells was conserved in human cells.We previously performed a subpool screen in human lung epithelial Calu-3 cells that identified DYRK1A as a positive regulator of SARS-CoV-2 infection [23].To validate the results of that screen, we generated polyclonal KOs in Calu-3 cells using guides targeting DYRK1A or a nontargeting guide control (NTG).In Calu-3 cells, loss of DYRK1A causes a significant reduction in SARS-CoV-2 infection as quantified by tissue culture infectious dose (TCID50) assay (Fig 5A).Next, we assessed whether DYRK1A was coexpressed with ACE2 and DPP4 in healthy human lung tissue.Using existing RNA-Seq datasets, we identified that DYRK1A is indeed coexpressed with ACE2 and DPP4 (Figs 5B, 5C, and S5) [94,95].Importantly, DYRK1A expression correlates with ACE2 in epithelial cells such as AT2 cells, ciliated cells, and basal cells, which are major cellular targets of SARS-CoV-2 in vivo ( Figs 5B, 5C, and S5) [9,10,[94][95][96].We recently showed that the mSWI/SNF complex regulates ACE2 chromatin accessibility and expression in human cells but not mice; therefore, we tested the species specificity for DYRK1A regulation of Ace2 mRNA levels in mice [39].Since loss of DYRK1A is embryonically lethal [97], we crossed an existing DYRK1A conditional deletion mouse (Dyrk1a F/F ) to a tamoxifen-inducible Cre recombinase under control of the globally expressed ubiquitin c gene (Ubc CreERT2).We treated mice with tamoxifen for 5 days to conditionally ablate DYRK1A and then assessed Ace2 mRNA levels.Loss of DYRK1A did not alter Ace2 expression in the lung or distal small intestine (S6 Fig) .Overall, these findings suggest that DYRK1A is a critical host factor for SARS-CoV-2 infection that is conserved in primate cells.
Using single-cell KO clones and addbacks, we report that nuclear DYRK1A promotes coronavirus entry by positively regulating ACE2 and DPP4 transcription via a kinase-independent mechanism.WT and kinase-null (DYRK1A-Y321F) complementation in vitro offered significant but partial rescue of live SARS-CoV-2 infection (approximately 70% to 80% of WT).Addbacks of another catalytically inactive mutant, DYRK1A-K188R, also exhibited partial but detectable rescue of SARS-CoV-2 infectivity.Of note, DYRK1A-K188R was not as effective at restoring icSARS-CoV-2-mNG infection relative to DYRK1A-Y321F.While this may result from minor residual kinase activity in DYRK1A-Y321F due to noncanonical autophosphorylation at Thr 111 , we anticipate that this is likely not the case due to equal rescue of viral entry by DYRK1A-Y321F and DYRK1A-K188R in both spike-dependent pseudovirus entry experiments and plaque assays [66,82,85].Instead, we suggest that this difference is explained by lower expression of DYRK1A-K188R due to previously described instability of the DYR-K1A-K188R mutant [82].Notably, DYRK1A-NES also partially restored viral entry, likely due to incomplete displacement of DYRK1A to the cytosol.Effective vaccines and direct acting antivirals including nirmatrelvir/ritonavir, remdesivir, and molnupiravir mitigate COVID-19, but novel therapeutic targets against current and future coronaviruses are needed to combat drug resistance and emerging viruses [99][100][101][102][103][104][105].Although small molecule inhibitors of DYRK1A exist, these drugs are constrained by their limited selectivity and toxicity despite ongoing improvements in their design [106,107].Moreover, these inhibitors structurally target the catalytic function of DYRK1A, which we have now shown to be dispensable for coronavirus entry.Therefore, consideration of DYRK1A as a therapeutic target will require new drug classes that are both tolerable and restrict DYRK1A activity independent of its catalytic function.Small molecules such as proteolysis targeting chimeras (PRO-TACs) that could degrade DYRK1A may offer such an approach [108,109].To date, no DYRK1A-specific PROTACs have been generated, but selective degraders of DYRK1A such as CaNDY exist [110].It was recently shown that SNX-544, an inhibitor of Hsp90, can greatly reduce SARS-CoV-2 infection in Vero-E6 and Calu-3 cells [111].Inhibition of Hsp90 leads to destabilization and degradation of DYRK1A, suggesting that indirect mechanisms of DYRK1A depletion may exist that offer therapeutic benefit for SARS-CoV-2 [110].
Our current study further shows that DYRK1A promotes not only expression of ACE2 and DPP4, but also CTSL.DYRK1A presence increases accessibility at recently identified putative ACE2 regulatory elements (TSS, proximal enhancer, and a distal enhancer) and at sites that may regulate DPP4 and CTSL [39].As receptor and protease expression are critical for coronavirus entry, these findings suggest that DYRK1A operates as a critical modulator of chromatin accessibility and transcription to drive a proviral gene program that promotes viral entry.However, deletion of DYRK1A in Calu-3 cells-which lowly express CTSL and instead employ TMPRSS2 for spike proteolytic processing-still conferred increased resistance to SARS-CoV-2 infection.Nonetheless, we cannot preclude the possibility that DYRK1A may also regulate proteases or other aspects of viral entry besides receptor expression.
Because SMARCA4, KDM6A, and HMGB1 are known ACE2 regulators, we asked whether DYRK1A may cooperate with one or more of these proteins, serving as a scaffold to promote ACE2 or DPP4 gene transcription [23,39,115].Both HMGB1 and SMARCA4 alter chromatin accessibility at the ACE2 locus, but how KDM6A modulates ACE2 expression has yet to be defined [23,39].DYRK1A-HMGB1 interactions have not been described, but DYRK1A can interact with SMARCA4 (the catalytic subunit of SWI/SNF) and SMARCB1 (a core subunit of SWI/SNF that stabilizes it and targets it to enhancers), resulting in phosphorylation of SMARCB1 [63,114,116].Whether phosphorylation of SMARCB1 is critical for SWI/SNF activity is unclear, but SMARCA4 can rearrange nucleosomes at the ACE2 promoter and 3 putative enhancers, driving accessibility and transcription of ACE2 [39,117].Overall, the DEGs and chromatin regions regulated by DYRK1A poorly correlate with those altered by HMGB1 or SMARCA4, although some pathways are shared between DYRK1A and SMARCA4 (S4 Fig) [23,39].Furthermore, HMGB1 and SMARCA4 do not alter DPP4 expression or promote MERS-CoV infection [23].Together, these findings suggest that DYRK1A likely functions independently of HMGB1 and SWI/SNF to promote CoV infection.Still, we cannot exclude the possibility that SMARCA4 and DYRK1A may coordinate to coregulate some sites (i.e., ACE2) but not others (i.e., DPP4).Unlike SMARCA4 and HMGB1, the histone methyltransferase KDM6A regulates both SARS-CoV-2 and MERS-CoV infection [23,39,115].KDM6A complexes with KMT2D, a histone demethylase, and the H3K27 acetyltransferase P300 and can activate enhancers [118].DYRK1A can directly interact with KMT2D and P300, hyperphosphorylating P300 to drive H3K27 deposition and enhancer accessibility [44,112].Because kinase activity is dispensable for DYRK1A-driven ACE2 and DPP4 expression, we suspect that DYRK1A also operates independently of KDM6A and KMT2D via a novel kinase-independent mechanism but cannot rule out that DYRK1A may act as a scaffold for these proteins.Clarifying the proteins that coordinate with DYRK1A, including transcription factor(s) that target DYRK1A to these sites, represent an important future direction.
ACE2 and DPP4 perform physiologic functions independent of their roles as coronavirus receptors.ACE2 encodes a dipeptidyl carboxypeptidase that cleaves angiotensin I, maintaining homeostasis in the renin-angiotensin system [119].Located on chromosome Xp22, ACE2 is an X-linked, dosage-sensitive gene known to be transcriptionally regulated by EP300, SP-1, CEBP, GATA3, and HNF1/4 [119,120].Recent efforts have aimed to identify novel ACE2 modifiers, uncovering PIAS1, SMAD4, BAMBI, KDM6A, and GATA6 [115,121].DPP4, which is situated on chromosome 2q24, encodes a serine exopeptidase that regulates glucose homeostasis [122,123].The DPP4 promoter contains binding sites for NF-kB, SP-1, EGFR, NF-1, STAT1, and HNF1 [124][125][126].Until now, only hepatocyte nuclear factor (HNF) family proteins have been identified as shared transcriptional regulators for these genes.Here, we demonstrate for the first time that DYRK1A is a novel regulator for both genes.However, the impact of DYRK1A on the renin-angiotensin system and glucose homeostasis have yet to be thoroughly explored and warrant additional investigation.DYRK1A is also highly dosage sensitive, and whether fine-tuning of DYRK1A expression could alter these critical biological pathways is unknown.
In Down syndrome (trisomy 21), DYRK1A is overexpressed 1.5-fold due to its position on chromosome 21, and dysregulation of DYRK1A dosage directly contributes to neurological defects and disease [45][46][47][48][49][50][51][52].Strikingly, Down syndrome significantly increases (5 to 10×) the risk of COVID-19 infection, hospitalization, and death relative to the general population [54][55][56][57][58][59].While individuals with Down syndrome possess many comorbidities that could partially explain their predisposition to COVID-19 severity, together, they incompletely explain such an elevated risk.A recent network analysis revealed that TMPRSS2-which is also located on chromosome 21 and is subsequently overexpressed in Down syndrome-may contribute to COVID-19 severity by promoting increased viral entry [55,57].Unlike TMPRSS2, DYRK1A does not directly interact with SARS-CoV-2, but we now show that DYRK1A positively regulates ACE2 expression and viral entry.Although ACE2 is not globally overexpressed in Down syndrome, few tissues have detectable ACE2 expression at baseline and transcriptomic datasets for airway epithelial cells derived from individuals with Down syndrome are lacking [55].
Here, we show that DYRK1A deficiency (loss-of-function) leads to a significant reduction in ACE2 expression, but whether DYRK1A overexpression (gain-of-function) in Down syndrome is sufficient to up-regulate ACE2 in humans and subsequently contribute to COVID-19 severity remains to be seen and warrants additional investigation.
Our findings support that DYRK1A can promote DNA accessibility independently of its kinase activity, facilitating ACE2 and DPP4 expression for viral entry.We reveal that the proviral function of DYRK1A dictates susceptibility to diverse highly pathogenic human coronaviruses and is conserved across 2 mammalian species.While chemically targeting DYRK1A remains challenging and existing drugs will not suffice as viable therapeutics, our work highlights the value of considering genetic regulators or interactors in informed therapeutic target design.

Viral stocks
Viral stocks were generated in Vero-E6 or Vero-E6-ACE2-TMPRSS2 cells seeded at approximately 80% confluency inoculated with HKU5-SARS-CoV-1-S (NR-48814), SARS-CoV-2 isolate USA-WA1/2020 (NR-52281), or MERS-CoV (icMERS-CoV EMC/2012) (NR-48813) from BEI resources at an MOI of approximately 0.01 for 3 days to generate a P1 stock.Vero-E6 or Vero-E6-ACE2-TMPRSS2 cells were inoculated with the P1 stock and incubated for 3 days or until 50% to 70% cytopathic effect was observed to generate a P2 stock.To generate icSARS-CoV-2-mNG stocks, lyophilized icSARS-CoV-2-mNG (World Reference Center for Emerging Viruses and Arboviruses, Galveston, TX) was resuspended in 500 μl deionized water and diluted 100-fold in medium.Diluted virus was added to 10 7 Vero-E6 cells grown in T175 (Corning) for 3 days.All viral stock harvests were clarified by centrifugation (500 × g/5 minutes) and filtered through a 0.45-μm filter (Millipore Sigma), aliquoted, and stored at −80˚C.All viral stocks were tittered by at least 2 independent plaque assays or TCID50 assays.Final viral stocks generated at Yale University were sequenced to confirm no mutations were generated during viral stock propagation.All work with infectious virus was performed in a Biosafety Level 3 facility in accordance to regulations and approval from the Yale University Biosafety Committee and Yale University Environmental Health and Safety.

SARS-CoV-2 plaque assays
Vero-E6 cells were seeded at 4 × 10 5 cells/well in 12-well plates (Corning) and were incubated overnight.The next day, media were removed and replaced with 100 μl of 10-fold serial dilutions of virus.Plates were incubated at 37˚C for 1 hour, with gently rocking every 15 minutes to promote viral adherence.Wells were then covered with 1 mL overlay media (DMEM, 2% FBS, 0.6% Avicel RC-581) and incubated for 48 hours at 37˚C.At 2 dpi, plates were fixed with 10% formaldehyde (Ricca Chemical) for 30 minutes and then stained with crystal violet solution (0.5% crystal violet (Sigma-Aldrich) in 20% ethanol) for 30 minutes.Crystal violet was aspirated and wells were rinsed with deionized water to visualize plaques.

SARS-CoV-2 infections of polyclonal Calu-3 cells by TCID50
Polyclonal KOs of DYRK1A were generated in Calu-3 cells using 2 independent guides (sgRNA#1 and sgRNA#2).Calu-3 cells were seeded into 24-well plates at a density of 2 × 10 5 cells/well.Two days later, cells were infected with SARS-CoV-2 post-seeding at an MOI of approximately 0.05.Viral inocula was incubated with cells for 30 minutes at 37˚C.Unbound virus was aspirated and cells were washed once with 1× PBS.Cells were incubated for 24 hours and then subjected to mechanical lysis by freeze thaw.Infectious viral particles were tittered by serial dilution and incubated on 96-well plates coated with Vero-E6 cells.Each dilution was applied to 8 wells.Cytopathic effect was determined visually at 3 dpi, and TCID50/mL was calculated using the dilution factor required to produce CPE in half of the wells.Viral titers were assessed and compared against the NTG.

SARS-CoV-2 fluorescent reporter virus assay
Cells were plated at 2,500 cells/well in a 384-well plate (Greiner) and adhered at 37˚C for approximately 5 hours.icSARS-CoV-2-mNG was added at an MOI of approximately 1.0.Infection frequency was measured by mNeonGreen expression at 1 or 2 dpi by high-content imaging (Cytation 5, BioTek) configured with bright field and GFP cubes.Total cell numbers were determined from bright field images using Gen5 software.Object analysis was used to quantify the number of mNeonGreen-positive cells.Percent infection was calculated as the ratio between mNeonGreen+ cells and total cells.

Generation of DYRK1A clonal Vero-E6 knockout and complemented cells
Vero-E6 DYRK1A KO cells were generated by lipofectamine transfection of Cas9-ribonucleoproteins (RNPs) CRISPR guide RNAs (gRNA) were synthesized by IDT (sequence: TCAGCAACCTCTAACCAACC).gRNAs were complexed in a 1:1 molar ratio with tracrRNA in nuclease-free duplex buffer by heating at 95˚C for 5 minutes and then cooled to room temperature.Duplexes were combined with Alt-R Cas9 enzyme at room temperature for 5 minutes to form RNPs in Opti-MEM (Gibco) with 200 μl total volume.Complexes were mixed with RNAiMAX transfection reagent (Invitrogen) in Opti-MEM at room temperature for 20 minutes before transfection.Transfection was performed with 3.2 × 10 5 Vero-E6 cells in suspension in a 12-well plate (Corning).Cells were incubated for 48 hours then stained with 1:500 Zombie Aqua (BioLegend) in 1× PBS (Gibco) prior to flow cytometry-based sorting on live cells.Single cells were sorted into 96-well plates (Corning).Clones were screened by increased resistance to rcVSV-SARS2-S virus-induced cell death, western blot, and were confirmed by Sanger sequencing.DYRK1A KO clones were complemented by lentiviral transduction of pCW57.1-purovector containing full-length DYRK1A, kinase-dead DYRK1A (K188R or Y321F), or nuclear localization defective DYRK1A (NES) with an N-terminal 3× FLAG tag.Two days posttransduction, puromycin was added and cells were selected for 3 days to select for stably expressing addbacks.Stable DYRK1A expression was induced by the addition of 20 or 200 μg/ml doxycycline hyclate (Sigma-Aldrich) dissolved in DMSO (Sigma-Aldrich) for 24 or 72 hours.Expression of DYRK1A in complemented cells was confirmed by western blot for 3× FLAG and DYRK1A.

Pseudovirus production
VSV-based pseudotype viruses were generated as previously described in 10-cm dishes [127].Briefly, HEK293T cells were transfected with pCAGGS or pCDNA3.1 vectors expressing the CoV spike (S) glycoprotein by calcium phosphate transfection and then inoculated with a replication-deficient VSV encoding Renilla luciferase in place of the G glycoprotein.After 1 hour at 37˚C, unbound inoculum was removed and cells were washed with 4× PBS.Fresh media were added with anti-VSV-G clone I1 (8G5F11) (Absolute Antibody) to neutralize residual VSV-G virus [128].After 24 hours, supernatant was harvested as viral stock and centrifuged at 3,000 rpm for 10 minutes to clarify cellular debris.VSVpp-SARS2-S stocks were concentrated using Amicon Ultra 100 kD filter columns (Millipore Sigma) at 3,000 rpm.VSVpp-SARS1-S and VSVpp-MERS-S stocks were not concentrated.All stocks were aliquoted and stored at −80˚C.Plasmids encoding codon-optimized sequences of SARS-CoV-S and MERS-SΔCT were previously described [12,129].Vector pCAGSS containing the SARS-CoV-2, Wuhan-Hu-1 S Glycoprotein Gene (NR-52310), was produced under HHSN272201400008C and obtained through BEI Resources, NIAID, NIH.

Pseudovirus entry assay
Approximately 4AU : PleasenotethatasperPLOSstyle; numeralsarenotallowedatthebeginningofasentenc × 10 5 Vero-E6 cells were seeded in 100 μl volume of each well of a blackwalled clear bottom 96-well plate (Corning) and incubated at 37˚C for approximately 5 hours to allow cells to adhere.VSV pseudovirus at 1:10 final concentration volume/volume for unconcentrated virus (VSVpp-SARS1-S or VSVpp-MERS-S) or 1:20 for concentrated virus (VSVpp-SARS2-S).Virus was incubated with cells for 24 hours, and cells were subsequently lysed with Renilla Luciferase Assay System (Promega) according to manufacturer's instructions.Luciferase activity was measured using a microplate reader (BioTek Synergy).Pseudovirus entry was normalized to VSV-G within each condition, and percent entry was calculated relative to WT Vero-E6 cells after normalization.

SARS-CoV-2 in vitro DYRK1A inhibition assays
Harmine and INDY were purchased from Cayman Chemical, and DYR219 and DYR533 were synthesized in-house.Compounds were resuspended at a stock concentration of 40 to 50 mM in DMSO.Drugs were diluted 2-fold in DMSO and spotted into 384-well black skirted plates (Corning) in 20 nL at 1,000× drug stock using the Labcyte ECHO dispenser at the Yale Center for Molecular Drug Discovery.Approximately 1AU : PleasenotethatasperPLOSstyle; numeralsarenotall .25 × 10 3 Vero-E6 cells were plated in total volume of 20 μl.Two days later, cells were infected with an MOI of approximately 1 SARS--CoV-2 isolate USA-WA1/2020 in 5 μl media.Cells were incubated for 3 days before assessing viability and virus-induced cell death by CellTiter-Glo according to manufacturer's protocol (Promega).Luminescence was quantified using a plate reader (Cytation 5, BioTek).For each cell line, viability was determined in SARS-CoV-2-infected cells relative to uninfected cells.

RNA-Seq
WT Vero-E6 cells, DYRK1A KO#1 + vector, and DYRK1A KO#1 + complements were seeded at 3 × 10 5 cells/well in a 6-well plate and were treated with doxycycline for 72 hours to enable rescue of ACE2 and DPP4.Samples were performed in biological duplicate and harvested by scraping.Total cellular RNA was extracted using the Direct-zol RNA MiniPrep Plus (Zymo Research), and libraries were prepared with rRNA depletion by the Yale Center for Genome Analysis.RNA-Seq libraries were sequenced on Illumina NovaSeq 6000 with a goal of at least 25 × 10 6 reads per sample.Reads were aligned to reference genome chlSab2, NCBI annotation release 100 using STAR aligner v2.7.3a with parameters-winAnchorMultimapNmax 200 -out-FilterMultimapNmax 100 -quantMode GeneCounts [130].Differential gene expression was obtained using the R package DESeq2 v1.32 [131].Heatmaps were generated using R package [132].
Analysis of existing single-cell RNA-Seq (scRNA-Seq) datasets from human lung tissue Human lung scRNA-Seq datasets from healthy human lungs were accessed using the European Genome-phenome Archive (Accession EGAS00001004344) or the Human Cell Atlas Data Coordination Platform and NCBI BIOPROJECT (Accession code PRJEB31843) [94,95].

L
; DYRK1A, Dual Specificity Tyrosine Phosphorylation Regulated Kinase 1A; hpi, hours postinfection; Rluc, Renilla luciferase; SARS-CoV-2, Severe Acute Respiratory Syndrome Coronavirus 2; VSVpp, VSV pseudovirus; WT, wild-type.(PDF) S2 Fig. DYRK1A active site inhibitors do not block SARS-CoV-2 infection at tolerated concentrations.(A) WT Vero-E6 cells, DYRK1A KO cells, and cells overexpressing DYRK1A after DOX induction for 72 hours were infected with SARS-CoV-2 WA/01/2020 at an MOI of approximately 0.1 for 1, 24, or 48 hours.Plaque assays were performed in biological duplicate.Shown is 1 representative replicate.(B) WT Vero-E6 cells were treated with the positive control protease inhibitor calpain inhibitor III or a potent DYRK1A inhibitor (harmine, INDY, DYR219, and DYR533) for 48 hours.Cells were then infected with SARS-CoV-2 (MOI approximately 1), and cell viability was assessed 72 hpi via CellTiter-Glo.% Viability was calculated relative to uninfected or untreated controls.The half-maximal inhibitory concentration (IC50) and half-maximal cytotoxic concentration (CC50) were calculated for each drug using nonlinear regression dose response curves.Each experiment was performed at least 2 independent times.Shown are the means of 2-3 technical replicates.Data underlying this figure can be found in S1 Data.DOX, doxycycline; DYRK1A, Dual Specificity Tyrosine Phosphorylation Regulated Kinase 1A; hpi, hours postinfection; KO, knockout; SARS-CoV-2, Severe Acute Respiratory Syndrome Coronavirus 2; WT, wild-type.(PDF) S3 Fig. DYRK1A regulates global gene expression and chromatin accessibility.(A) RNA--Seq volcano plots depicting DEGs in cells where DYRK1A is absent or reintroduced.(B) Enrichr pathway analysis for biological function enriched in the presence of DYRK1A.Gene set criteria included p < 0.05, L2FC<0 for KO+vec vs. WT Vero-E6 and L2FC>1.5 for both KO+WT and KO+Y321F vs. KO+vec.(C) ATAC-Seq volcano plots depicting DEGs in cells where DYRK1A is absent or reintroduced.All experiments were performed in biological duplicate.Data underlying this figure can be found under GEO Accession GSE213999.DEG, differentially expressed gene; DYRK1A, Dual Specificity Tyrosine Phosphorylation Regulated Kinase 1A; KO, knockout; L2FC, log2 fold-change; WT, wild-type.(PDF) S4 Fig. DYRK1A and SMARCA4 share limited similarities in regulating chromatin accessibility and gene expression.(A) Principal component analysis of ATAC-Seq experiments performed in DYRK1A KO+vec, DYRK1A KO+WT, SMARCA4 KO+vec, and SMARCA4 KO +WT cells generated in a parental Vero-E6 background.Each experiment was performed in biological duplicate (replicates 1 and 2).DYRK1A and SMARCA4 loss share some molecular impacts, suggesting that some pathways may be coregulated (PC1), whereas others may be independently regulated (PC2).(B) Correlation heatmap comparing all sites from ATAC-Seq experiments in DYRK1A KO+vec, DYRK1A KO+WT, SMARCA4 KO+vec, and SMARCA4 KO+WT cells.(C) Correlation heatmap comparing chromatin accessibility by ATAC-Seq in DYRK1A or SMARCA4 complemented cells, identifying a correlation coefficient of 0.33 supporting approximately 33% of clusters may be correlated by DYRK1A and SMARCA4.(D) Venn diagram highlighting shared peaks gained by DYRK1A and SMARCA4 complementation.(E) Correlation heatmap comparing changes in RNA abundance in DYRK1A or SMARCA4 complemented cells, identifying a correlation coefficient of 0.08 supporting <10% of the top up-regulated/down-regulated genes are shared between DYRK1A and SMARCA4.(F) Gene set enrichment analysis from RNA-Seq experiments showing shared pathway regulation by DYRK1A and SMARCA4.Data underlying this figure can be found under GEO