key: cord-274528-mr81o9cu
authors: Li, Fei; Han, Ming; Dai, Pengfei; Xu, Wei; He, Juan; Tao, Xiaoting; Wu, Yang; Tong, Xinyuan; Xia, Xinyi; Guo, Wangxin; Zhou, Yunjiao; Li, Yunguang; Zhu, Yiqin; Zhang, Xiaoyu; Liu, Zhuang; Aji, Rebiguli; Cai, Xia; Li, Yutang; Qu, Di; Chen, Yu; Jiang, Shibo; Wang, Qiao; Ji, Hongbin; Xie, Youhua; Sun, Yihua; Lu, Lu; Gao, Dong
title: Distinct mechanisms for TMPRSS2 expression explain organ-specific inhibition of SARS-CoV-2 infection by enzalutamide
date: 2020-09-12
journal: bioRxiv
DOI: 10.1101/2020.09.11.293035
sha: 
doc_id: 274528
cord_uid: mr81o9cu

The coronavirus disease 2019 (COVID-19) pandemic, caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has rapidly become a global public health threat due to the lack of effective drugs or vaccines against SARS-CoV-2. The efficacy of several repurposed drugs has been evaluated in clinical trials. Among these drugs, a relatively new antiandrogen agent, enzalutamide, was proposed because it reduces the expression of transmembrane serine protease 2 (TMPRSS2), a key component mediating SARS-CoV-2-driven entry into host cells, in prostate cancer cells. However, definitive evidence for the therapeutic efficacy of enzalutamide in COVID-19 is lacking. Here, we evaluated the antiviral efficacy of enzalutamide in prostate cancer cells, lung cancer cells, human lung organoids and SARS-CoV-2-infected Ad-ACE2-transduced Tmprss2 knockout (Tmprss2-KO) and wild-type (WT) mice. TMPRSS2 knockout significantly inhibited SARS-CoV-2 infection in vivo. Enzalutamide effectively inhibited SARS-CoV-2 infection in human prostate cancer cells (LNCaP) but not in human lung cancer cells or patient-derived lung organoids. Although Tmprss2 knockout effectively blocked SARS-CoV-2 infection in ACE2-transduced mice, enzalutamide showed no antiviral activity due to the AR independence of TMPRSS2 expression in mouse and human lung epithelial cells. Moreover, we observed distinct AR binding patterns between prostate cells and lung cells and a lack of direct binding of AR to TMPRSS2 in human lung cells. Thus, our findings do not support the postulated protective role of enzalutamide in treating COVID-19.

Coronavirus disease 2019 (COVID-19), which is caused by the novel coronavirus severe acute respiratory 2 syndrome coronavirus 2 (SARS-CoV-2), has emerged as a new worldwide pandemic. COVID-19 has led 3 to nearly 28,000,000 confirmed global cases and 910,000 deaths as of September 11, 2020 1 . SARS-CoV-4 2 is a serious worldwide threat due to its high infectivity 2,3 . The current pandemic and the potential for 5 future pandemics have exposed the urgent need for the rapid development of efficient countermeasures. 6

Therefore, repurposing clinically proven drugs has been postulated as a promising strategy for developing 7 treatments for SARS-CoV-2 infection. 8

Transmembrane serine protease 2 (TMPRSS2) has been reported with essential role in mediating 9

viruses including SARS-CoV-2-driven entry into host cells 4-7 . The spike glycoprotein (S) of SARS-CoV-2 10 and its receptor, angiotensin-converting enzyme 2 (ACE2) have been demonstrated with the function in 11

mediating the attachment of SARS-CoV-2 to host cells 8, 9 . Then, the priming of SARS-CoV-2 S protein is 12 processed by TMPRSS2 5 . Moreover, besides SARS-CoV-2, both SARS-CoV, another type of 13 coronaviruses and H1N1, an influenza virus, also employ TMPRSS2 for viral entry 6,10,11 . Since the 14 conserved role of TMPRSS2 in provoking coronaviruses and influenza viruses-driven entry into host cells 15

has been highlighted, modulating TMPRSS2 expression or its protease activity is postulated to be a 16 potential method for antiviral intervention 12-15 17

Enzalutamide is a potent inhibitor of the androgen receptor (AR) and has been approved for the 18 treatment of castration-resistant prostate cancer (CRPC) patients 16, 17 . Mechanistically, enzalutamide binds 19

to AR, reduces the efficiency of its translocation from the cytoplasm to the nucleus, and impairs the AR-20 mediated signaling pathway 17 . Given the modulation of TMPRSS2 expression by AR in prostate cells 18 , 21

several clinical trials have been initiated to assess the therapeutic efficacy of enzalutamide in COVID- 19 22 patients (ClinicalTrials.gov identifiers; NCT04475601 and NCT04456049). However, it remains elusive 23

whether AR indeed controls TMPRSS2 expression in different organs, especially lung. Thus, an 24 investigation of enzalutamide in the treatment of SARS-CoV-2 infection is urgently needed. Herein, we 25 evaluated the antiviral efficacy of enzalutamide in human lung organoids (LuOs) and human ACE2 26 recombinant adenovirus (Ad-ACE2)-transduced Tmprss2 knockout (Tmprss2-KO) and wild-type (WT) 27 mice. With these powerful approaches, we comprehensively defined the antiviral effect of enzalutamide. 28

Moreover, we showed the potential mechanism of enzalutamide with its different antiviral activity in the 29 human prostate and lung. 30

Results 31

To elucidate whether TMPRSS2 is crucial for SARS-CoV-2-driven entry into host cells, we employed 33 previously established Tmprss2-KO mouse model (Extended Data Fig. 1a ). In line with previous 34 findings 10,19 , under physiological conditions, Tmprss2 knockout exhibited little effect on multiple organs 35

including lungs (Fig. 1a) . In order to identify Tmprss2 positive cells in multiple organs, we next crossed 36

Tmprss2-KO mice with Rosa26-EYFP mice expressing a CAG-driven YFP Cre-reporter (T2Y), when 37 exposed to tamoxifen, this mouse model can be utilized to trace cells of the Tmprss2-positive lineage via 38 detection of YFP expression (Extended Data Fig. 1b ). With this model, we further investigated the 39 existence of Tmprss2-positive cells in multiple organs. Notably, in addition to prostate, other essential 40 organs, including lung, kidney and liver, which are permissive for SARS-CoV-2 infection in human, were 41 characterized with Tmprss2-postive epithelial cells ( Fig. 1b and Extended Data Fig. 1c ). The broad 1 distribution of Tmprss2-positive cells might indicate the universal function of Tmprss2 in mediating SARS-2

CoV-2-driven entry in multiple organs. 3

To confirm the role of Tmprss2 in SARS-CoV-2 infection, we employed a previously reported Ad-ACE2 4 transduction method to overcome the natural resistance of mice to SARS-CoV-2 infection 20 . Briefly, we 5 first transduced 10-to 18-week-old WT mice and Tmprss2-KO mice with 2.5×10 9 PFU of FLAG-tagged 6

Ad-ACE2 adenovirus. Consistent with previous findings, predominant ACE2 expression was observed in 7 the alveolar epithelium, as indicated by FLAG staining (Extended Data Fig. 1d ). Five days post Ad-ACE2 8 transduction, mice were challenged with 1×10 5 PFU of SARS-CoV-2. Notably, compared to WT mice 9

challenged with SARS-CoV-2, Tmprss2-KO mice exhibited extremely less severe lung pathology, as 10 indicated by the lack of robust inflammatory responses (Fig. 1d, e) . We also quantified the percentage of 11 lung cells with SARS-CoV-2 infection. Based on the quantification of more than one million cells in 5 mice 12

per group via S protein staining, the percentage of S protein-positive lung cells was significantly lower in 13

Tmprss2-KO mice than in WT mice ( Fig. 1f, g) . Collectively, these findings suggested that the lack of 14 TMPRSS2 had marked effects on SARS-CoV-2 infection, highlighting the important role of TMPRSS2 in 15 mediating SARS-CoV-2-driven entry into host cells. 16

Since TMPRSS2 expression is modulated by AR in prostate cells, which promoted us to identify whether 18 AR inhibition can prevent SARS-CoV-2 infection through reducing TMPRSS2 expression. We first 19

surveyed TMPRSS2 and AR expression across a panel of well-characterized prostate cancer cell lines 20

and two previously established organoid lines MSKPCa1 and MSKPCa3 21 . Notably, both qRT-PCR and 21

western blotting demonstrated high AR and TMPRSS2 expression in LNCaP and VCaP cells (Extended 22

Data Fig. 2a, b) . Consistent with previous findings 22,23 , a marked reduction in TMPRSS2 protein and mRNA 23 expression was induced by AR inhibition using enzalutamide treatment and was validated in both LNCaP 24

and VCaP cells (Extended Data Fig. 2c , d, e and f). 25

For sensitive and convenient detection of SARS-CoV-2-driven entry into host cells, we employed a 26

pseudovirus system by incorporating SARS-CoV-2 S protein and luciferase into pseudoviral particles 27 through cotransfection of pNL4-3.luc.RE and PCDNA3.1 encoding the SARS-CoV-2 S protein. Thus, this 28 system allowed the sensitive detection of SARS-CoV-2 pseudotype entry by measuring luciferase activity. 29

The constructed pseudovirus was named SARS-CoV-2-S. We first asked whether LNCaP and VCaP were 30

susceptible to SARS-CoV-2-S-driven entry. Since undetectable ACE2 expression in LNCaP and VCaP 31 cells was identified, we observed lack of robust SARS-CoV-2-S-driven entry into these cells, as expected 32

(Extended Data Fig. 3b, d) . To enable the permissiveness of LNCaP and VCaP cells, we next transduced 33

Ad-ACE2 into these cells (Extended Data Fig. 3a, c) . 34

Given that enzalutamide treatment can reduce TMPRSS2 expression in prostate cells 17,23 , we next 35 sought to ascertain whether enzalutamide can prevent SARS-CoV-2 from infecting prostate cells through 36 downregulation of TMPRSS2 expression. We first investigated the therapeutic efficacy of enzalutamide in 37

blocking SARS-CoV-2-S-driven entry into LNCaP cells. In line with previous results generated from other 38 TMPRSS2 positive cells 5,24 , camostat mesylate, a clinically proven inhibitor for serine protease including 39 TMPRSS2, significantly attenuated the infection of SARS-CoV-2-S, as indicated by the reduction in 40 luciferase activity, suggesting that TMPRSS2 is also an important factor for facilitating SARS-CoV-2-driven 41 entry into LNCaP cells (Fig. 2a) . Remarkably, enzalutamide also significantly blocked SARS-CoV-2-S 42 infection, which even exhibited much higher treatment efficacy than camostat mesylate (Fig. 2a) . In 1 addition, immunofluorescence staining for luciferase also demonstrated the consistent results that 2 enzalutamide significantly reduced the percentage of cells with SARS-CoV-2-S-driven entry (Fig. 2b, c) . 3

Since pseudovirus system was limited to investigations on SARS-CoV-2-S-driven entry into host cells, we 4

next assessed whether enzalutamide interferes with authentic SARS-CoV-2-driven entry and the 5 subsequent steps of the viral replication cycle. Consistent with findings from the pseudovirus system, 6 enzalutamide efficiently exerted antiviral activity against SARS-CoV-2 in LNCaP cells, as demonstrated by 7 the significantly reduced viral titers of SARS-CoV-2 in both culture medium supernatant and cellular 8 contents (Fig. 2d, e) . Moreover, we evaluated whether enzalutamide can prevent SARS-CoV-2-S-driven 9 entry into VCaP cells. Recapitulating results in LNCaP cells, enzalutamide treatment also significantly 10 blocked SARS-CoV-2-S-driven entry into VCaP cells (Extended Data Fig. 3e ). Taken together, utilizing 11 both the pseudovirus system and authentic SARS-CoV-2, we demonstrated that enzalutamide efficiently 12

prevented SARS-CoV-2-driven entry into prostate cells by inhibiting AR to reduce TMPRSS2 expression. 13

Since enzalutamide efficiently inhibited infection of human prostate cells with SARS-CoV-2, we next sought 15

to evaluate its therapeutic efficacy in human lung cells. To this end, we surveyed single-cell RNA-16

sequencing data from healthy human lungs 25 , consistent with previous results 26 , TMPRSS2-positive cells 17

were broadly distributed in various cell types, potentially indicating an important role of TMPRSS2 in 18 mediating SARS-CoV-2 infection in multiple lung cell types (Extended Data Fig. 4a , b and d). In particular, 19

high expression levels of both TMPRSS2 and ACE2 were identified in SFTPC-positive alveolar type II (ATII) 20 cells, potentially indicating TMPRSS2-dependent entry of SARS-CoV-2 into these cells (Extended Data 21 Fig. 4a , c, e and f). We further assessed AR expression and found that similar to TMPRSS2, AR was also 22 highly expressed with a wide distribution (Extended Data Fig. 4a Since human lungs were characterized with both AR and TMPRSS2 expression, we next sought to 29 determine whether AR can modulate TMPRSS2 expression in the lungs. We firstly established human lung 30 organoids (LuOs) derived from adjacent normal lung tissues with similar culture protocol as previously 31

reported 27 (Fig. 3a) . To verify whether LuOs are an appropriate model in which to evaluate the therapeutic 32 efficacy of enzalutamide, we performed immunofluorescence staining for AR and TMPRSS2 in LuOs. By 33 staining of serial sections, we identified both AR expression and TMPRSS2 expression in LuOs, which 34 also contained AR/TMPRSS2 double-positive cells (Fig. 3b) . We next employed LuOs to explore whether 35 enzalutamide could manipulate TMPRSS2 expression. Distinct from above findings in prostate LNCaP 36 cells, enzalutamide treatment did not significantly reduce TMPRSS2 expression, validated in three LuOs 37 lines (Fig. 3c, d and e) . To ensure the permissiveness of LuOs for SARS-CoV-2-S-driven entry, we also 38 transduced these organoids with Ad-ACE2 (Fig. 3f) . Twenty-four hours post Ad-ACE2 transduction, LuOs 39

were pretreated with 10 μM camostat mesylate or 10 μM enzalutamide for 48 hours before virus infection 40 (Fig. 3g ). Camostat mesylate but not enzalutamide inhibited infection of LuOs with SARS-CoV-2-S, 41

confirming that enzalutamide could not protect lung cells against SARS-CoV-2 infection (Fig. 3h , i and j).

Moreover, we also evaluated whether enzalutamide blocked authentic SARS-CoV-2-driven entry and viral 43 replication. Consistent with the results obtained with the SARS-CoV-2 pseudovirus, enzalutamide did not 1 exhibit antiviral activity against authentic SARS-CoV-2 (Fig. 3k ). 2

Given that lack of treatment efficacy of enzalutamide in blocking SARS-CoV-2-driven entry was 3 characterized in human lung organoids, we also employed multiple lung cancer cell lines to validate 4

whether these results could be recapitulated. Among these cell lines, three of eight, namely, H1437, H2126 5

and A549 cells were AR-positive, confirming the wide distribution of AR expression across multiple lung 6 cell types (Extended Data Fig. 5a ). Since only H1437 and H2126 cells exhibited detectable TMPRSS2 7 expression, we treated these two cell lines with the AR ligand dihydrotestosterone (DHT) and the AR 8

inhibitor enzalutamide to assess changes in TMPRSS2 expression (Extended Data Fig. 5b ). Notably, 9

unlike in LNCaP cells, in which DHT stimulated and enzalutamide reduced TMPRSS2 expression, no 10 obvious changes in TMPRSS2 expression were observed in H2126 and H1437 cells treated with these 11 two agents (Extended Data Fig. 5c , d, e and f). In addition, we also performed immunofluorescence for 12 TMPRSS2 to validate these results. Distinct from results in LNCaP cells that DHT stimulated TMPRSS2 13 expression and enzalutamide reduced TMPRSS2 expression respectively, no obvious changes in 14

TMPRSS2 expression were observed in H2126 and H1437 cells with these two treatments (Extended 15

Data Fig. 5g ). To enable these lung cells to be susceptible to SARS-CoV-2-S-driven entry, we also 16

transduced Ad-ACE2 into these cells (Extended Data Fig. 5h ). We next compared camostat mesylate and human lung organoids and lung cancer cells, we demonstrated that TMPRSS2 expression was 26 independent of AR expression in human lung epithelial cells, thus AR inhibition using enzalutamide did not 27 reduce TMPRSS2 expression to block SARS-CoV-2-driven entry into human lung epithelial cells. 28

In order to identify whether AR was not capable of modulating TMPRSS2 expression utilizing in vivo mouse 31 models, we next treated WT mice and castrated mice with enzalutamide for 7 days. Notably, in castrated 32 mice, enzalutamide treatment impaired the function of AR by blocking its nuclear translocation in prostate 33 cells (Extended Data Fig. 6a ). As observed in human prostate cells, reduced Tmprss2 mRNA levels were 34 identified in prostate epithelial cells in enzalutamide-treated mice and castrated WT mice (Fig. 4a) . No 35 significant changes in Tmprss2 mRNA levels in response to enzalutamide treatment and castration were 36 observed in the lungs of male mice (Fig. 4b ). In addition, consistent results were obtained in the lungs of 37 female mice treated with enzalutamide ( Fig. 4c) . Finally, consistent with findings in human prostates and 38

lungs, in vivo experimentation in mice also demonstrated the organ-specific role of AR in regulating 39

TMPRSS2 expression. 40

To demonstrate enzalutamide treatment efficacy in preventing SARS-CoV-2 driven entry into lung cells 41

utilizing in vivo models, we also employed Ad-ACE2 transduced mouse models (Extended Data Fig. 6b ).

Briefly, we firstly treated 10-12-week-old wild type C57BL/6 mice with or without enzalutamide treatment 1 by daily intragastric gavage. We next transduced control and enzalutamide-treated mice with 2.5×10 9 PFU 2 of Ad-ACE2 adenovirus. Five days post Ad-ACE2 transduction, mice were challenged with 1×10 5 PFU of 3 SARS-CoV-2. Mouse lungs were collected for pathological analysis and viral load determination 3 days 4 post SARS-CoV-2 challenge. Notably, histopathological analysis revealed similar levels of inflammatory 5 infiltration in control and enzalutamide-treated mice (Fig. 4d , e). In addition, the viral titers did not differ 6 significantly between control and enzalutamide-treated mouse lungs (Fig. 4f) . Moreover, the percentage 7 of lung cells infected with SARS-CoV-2 in control mice did not differ significantly from that in enzalutamide-8 treated mice (Fig. 4g, h) . Taken together, utilizing in vivo mouse models, we obtained consistent results 9

indicating that enzalutamide did not inhibit SARS-CoV-2-driven entry into lung cells and subsequent viral 10

replication. 11

Given the discrepancy between prostate cells and lung cells in the changes in TMPRSS2 expression in 13 response to enzalutamide treatment, we next sought to elucidate whether such discrepancy was attributed 14

to distinct AR binding pattern. We first performed chromatin immunoprecipitation with sequencing (ChIP-15 seq) on AR in prostate cells LNCaP and assay for transposase-accessible chromatin using sequencing 16

(ATAC-seq) in both prostate cells LNCaP and lung cells A549, H1437 and H2126. Based on AR ChIP-seq 17

in LNCaP cells, we compared chromatin accessibility among these four cell lines of AR binding sites. 18

Notably, distinct from extensive chromatin accessibility of these sites in LNCaP cells as expected, the other 19 three lung cell lines were characterized with much less open chromatin (Fig. 5a) . In principle, transcription 20 factors modulate transcriptional regulation through binding to regulatory elements of target genes, which 21

tightly associates with chromatin accessibility 28 . The chromatin accessibility disparity might indicate distinct 22

AR binding pattern between prostate cells and lung cells. To further characterize AR binding pattern in 23 prostate cells and lung cells respectively, we categorized these AR binding sites into two main groups: Compatible with AR binding in TMPRSS2, two extra AR binding sites were verified with extensive 33 chromatin accessibility in LNCaP cell but not in other lung cells (Fig. 5e ). In addition, unlike in prostate 34

cells, ChIP-qPCR demonstrated the lack of robust AR binding in the upstream region of TMPRSS2 locus 35

in lung cells (Fig. 5f ). These results indicated lack of AR binding in TMPRSS2 in lung cells, which coincided 36

with the above findings that AR inhibition utilizing enzalutamide did not reduce TMPRSS2 expression to 37

inhibit SARS-CoV-2-driven entry. Furthermore, Gene Set Enrichment Analysis (GSEA) revealed that 38

androgen response genes were significantly enriched in AR-positive prostate cells (LNCaP, VCaP and 39 22RV1) when compared with AR-positive lung cells (A549, H1437 and H2126) (Fig. 5g) . In accordance 40

with GSEA results, a significantly higher sum of z-scores for androgen responsive genes was observed in 41 AR-positive prostate cells than that in AR-positive lung cells (Fig. 5h ).

Since we demonstrated lack of AR binding in TMPRSS2 in lung cells, utilizing freshly dissociated lung 1 cells from 43 normal human lung tissue samples, we next sought to validate whether the correlation 2 between the expression of AR and TMPRSS2 coordinated these findings. Concordant with above findings, 3 no significant correlation relationship was identified between AR and TMPRSS2 expression (Fig. 5i, j, k) . 4

We also analyzed normal lung tissues and normal prostate tissues from TCGA datasets. A significant and 5 positive correlation between the mRNA expression of AR and a both-open gene PARP14 was observed in 6 both lung and prostate tissues, in keeping with AR binding in this gene (Extended Data Fig. 7a, d) . The 7 mRNA levels of both TMPRSS2 and FKBP5, which were characterized with specific AR binding in prostate 8 cells, significantly correlated with AR mRNA levels in prostate tissues but not in lung tissues (Extended 9

Data Fig. 7b, c, e, f) . These findings established a distinct AR binding pattern between the prostate and 10 the lungs, providing clinical evidence that TMPRSS2 expression is not responsive to AR inhibition in lungs. 11

Collectively, these results revealed a distinct AR binding pattern between human prostate and lung cells.

This finding not only offers a mechanistic explanation for the inability of AR to modulate TMPRSS2 13 expression but also suggests that enzalutamide is not a promising drug for blocking SARS-CoV-2-driven 14 entry into host cells. 15

TMPRSS2 has been demonstrated with a pivotal role in promoting SARS-CoV-2-driven entry into host 17 cells through facilitating S protein priming via its serine protease activity 4-7,29 . These previous findings 18

suggest that the modulation of TMPRSS2 expression may provide a novel strategy to treat SARS-CoV-2 19

infection by blocking viral entry into host cells. It is well known that TMPRSS2 expression is regulated by 20 AR in prostate epithelial cells. Enzalutamide, an AR inhibitor approved for use in CRPC patients, can 21 reduce TMPRSS2 expression in prostate cancer cells. Thus, enzalutamide has been proposed as a 22

promising repurposed drug to inhibit SARS-CoV-2 infection and subsequent replication, which even 23

provoked the initiation of two clinical trials. Here, we further confirmed the indispensable role of TMPRSS2 24

in SARS-CoV-2 infection using human ACE2-transduced Tmprss2-KO mice (Fig. 1) . Consistently, 25 enzalutamide significantly decreased TMPRSS2 expression and inhibited SARS-CoV-2 infection in human 26 prostate cancer cells (Fig. 2) . However, we did not observe any antiviral activity of enzalutamide against 27 SARS-CoV-2 in the lungs of Ad-ACE2-transduced WT mice or human lung organoids. These results 28

suggested that enzalutamide may have antiviral activity in the prostate in male COVID-19 patients, but 29 also indicated that enzalutamide may have no clinical efficacy in treating COVID-19 patients with lung 30

infection. 31 in human lung cancer cells. Surprisingly, in AR/TMPRSS2 double-positive H2126 and H1437 lung cancer 38 cells, neither AR inhibition using enzalutamide nor AR stimulation using DHT resulted in a significant 39 change in TMPRSS2 expression, implying that AR cannot regulate TMPRSS2 expression in human lung 40 cancer cells. These findings seemed inconsistent with those of previous studies indicating that androgen 41 exposure enhanced TMPRSS2 expression in another lung cell line, A549 30 . This discrepancy might be due 42 to 100nM testosterone, since such higher concentration of testosterone to treat cells might result in 43 misleading findings, which could not reflect the physiological function of AR. In addition, notably, TMPRSS2 1 mRNA was hard to detect under physiological conditions in A549 cells, which exhibit high AR expression 2 (Extended Data Fig. 5b) . Given that enzalutamide did not downregulate TMPRSS2 expression in these 3 cells, we further demonstrated that enzalutamide failed to inhibit the entry driven by SARS-CoV-2-S, as 4 expected. Since lung cancer cell lines harbor many genetic alterations, which might lead to disparities in 5 findings with respect to normal lung cells, we employed early-passage benign human LuOs for further 6 study. Compatible with findings in lung cancer cells, enzalutamide had no treatment efficacy in preventing 7

authentic SARS-CoV-2 and SARS-CoV-2-S pseudovirus in benign human lung organoids. Moreover, we 8 also employed Ad-ACE2-transduced mouse models and demonstrated that enzalutamide lacked antiviral 9

activity against SARS-CoV-2 in vivo. 10

A previous study demonstrated that androgen-deprivation therapy (ADT) significantly reduced the risk 11

of SARS-CoV-2 infection in prostate cancer patients 31 . However, a subsequent study demonstrated that 12 the lethality rate of SARS-CoV-2 in metastatic prostate cancer patients with ADT was not lower than that 13

in other cohorts of infected Italian male patients 32 , which did not suggest that ADT exhibited antiviral activity 14

against SARS-CoV-2 in patients with metastatic prostate cancer. The inconsistent findings from these two 15 studies might be attributed to the different populations selected for investigation. However, to date, no 16 concordant and definitive clinical evidence indicates that ADT, including enzalutamide treatment, 17

significantly inhibits SARS-CoV-2 infection. 18

Utilizing multiple models, including human lung cancer cells, human lung organoids and Ad-ACE2-19

transduced WT mice, we demonstrated that enzalutamide failed to inhibit SARS-CoV-2 infection, which 20 was attributed to the lack of AR-driven modulation of TMPRSS2 expression in lung epithelial cells. To 21 elucidate the mechanisms underlying the disparity of AR-driven regulation between prostates and lungs, 22

we further performed AR ChIP-seq and ATAC-seq in AR-positive lung cells and AR-positive prostate cells.

Unlike in prostate cells, the lack of specific AR binding at TMPRSS2 locus in lung cells, as demonstrated 24

by AR ChIP-seq, were consistent with the finding that TMPRSS2 expression was independent of AR 25 expression in human lung epithelial cells. These findings indicated mechanisms explaining that the lack of 26 antiviral activity of enzalutamide against SARS-CoV-2 is due to a lack of direct AR binding at the TMPRSS2 27 locus in lung epithelial cells. 28 However, our study had limitations. Microenvironmental components, including immune cells, nerve 29 cells and stromal cells, are involved in viral infection and subsequent replication 33-35 . Although we 30

employed Ad-ACE2-transduced in vivo mouse models, our models did not consider the human lung 31 microenvironment. Since stromal cells in multiple human organs, including the lungs, are also 32 characterized by AR expression, we cannot exclude the possibility that enzalutamide might display antiviral 33 activity by altering the expression of some essential cytokines or chemokines in stromal cells. 34

It was also noting that SARS-CoV-2 could still infect the lungs of Tmprss2-KO mice with lower effectivity. 35

Besides, when transduced with Ad-ACE2, both TMPRSS2-negative prostate cells PC3 (data not shown) 36

and lung cells H23 were permissive for robust SARS-CoV-2-driven entry (Extended Data Fig. 5k ). These 37 findings implied that besides TMPRSS2, other factors may also play a crucial role in promoting SARS-38

CoV-2 infection. Further studies to identify these factors and their precise functions in mediating SARS-39

CoV-2 infection will be really necessary. 40

Finally, we took advantage of multiple models of human prostate and lung cells, patients-derived benign 41 lung organoids and Ad-ACE2-transduced Tmprss2-KO and WT mice to comprehensively confirm the pivotal function of TMPRSS2 in SARS-CoV-2 infection. Our findings validated that enzalutamide 1 significantly inhibits SARS-CoV-2 infection in AR and TMPRSS2 double positive prostate cancer cells, 2 identified that enzalutamide does not exhibit antiviral activity in human lung cancer cells and patients-3 derived benign lung organoids in vitro and in the lungs of Ad-ACE2-transduced WT mice in vivo, and 4 demonstrated the distinct AR binding pattern between prostate and lung epithelial cells. These findings will 5 enhance our understanding of TMPRSS2 in SARS-CoV-2 infection and indicate the potential failure of 6 clinical trials using enzalutamide to treat COVID-19 patients. 7

We thank members of the Core Facility of Microbiology and Parasitology (SHMC) and the Biosafety Level 

The authors have no competing interests to declare. 

Transduction and infection of mice. Mice were anesthetized with Avertin (Sigma-Aldrich, T48402-5G) 2 and transduced intranasally with 2.5×10 9 FFU of Ad5-ACE2 adenovirus in 75μL DMEM (Gibco 3 C11995500BT). Mice were infected intranasally with 1×10 5 PFU of SARS-CoV-2 at the fifth day after Ad-4 ACE2 transduction. Three days post infection, lungs were harvested for virus titer measurement and 5 pathogenicity analysis using qPCR and immunohistology, respectively. 6

Castration and enzalutamide treatment of mice. Enzalutamide (Selleck, S1250) (10 mg/kg; the vehicle 7 contained 1% carboxymethyl cellulose, 0.1% Tween 80, and 5% DMSO) was administered intragastrically 8

to castrated mice daily for 10-30 days as previously described 22 . for 2 minutes and centrifuged at 12,000 rpm for 15 minutes at 4°C. The aqueous phase was transferred 27 into a new tube, and an equal volume of isopropanol was added. The mixture was centrifuged at 13,000 28 rpm for 10 minutes at 4°C. The supernatant was discarded, and the pellet was resuspended in 75% ethanol 29

and centrifuged at 8,000 rpm for 7 minutes at 4°C. The supernatant was then thoroughly removed and 30 discarded. The pellet was resuspended in 50 μL of nuclease-free water. Reverse transcription was further 31 performed with PrimeScriptTM RT Master Mix (TaKaRa, RR036A) with 400 ng of total RNA as input. qRT-32

PCR was conducted with SYBR qPCR Mix (Qiagen, 208054) using the manufacturer's protocol. The 33

primer sequences are listed as followed: Mouse-Actb-F: CATTGCTGACAGGATGCAGAAGG; 5

Mouse-Actb-R: TGCTGGAAGGTGGACAGTGAGG. 6

Western blotting. Cell lysates were prepared in RIPA buffer supplemented with proteinase/phosphatase 7 inhibitors. The protein content was quantified with a BCA (Thermo) assay. Fifteen micrograms of protein 8

were separated via SDS-PAGE and transferred onto a 0.45 mm PVDF membrane (GE). The membrane 9

was blocked for 1 hour at room temperature in TBST buffer containing 5% milk and was incubated 10 overnight at 4°C or for 2 h at room temperature with primary antibodies diluted in TBST buffer containing 11 5% milk. The membrane was then incubated with rabbit HRP-conjugated secondary antibodies (SAB, 12

#L3012) for 1 h in 5% milk at RT. The primary antibodies included anti-β-Actin (Sigma-Aldrich, A3854, 13 1:5,000), anti-AR (Abcam, ab108341, 1:2,000), anti-TMPRSS2 (Abcam, ab92323 Transduction of cells with Ad-ACE2. Cells were seeded in 6-well plates before transduction. The next 37 day, Ad-ACE2 was transduced into cells at a multiplicity of infection (MOI) of 100 with polybrene. The 38

culture medium supernatant was replaced with fresh medium 12 hours post transduction. 39

Pseudovirus production. SARS-CoV-2 pseudovirus was produced by cotransfection of 293T cells with pNL4-3.luc.RE and PCDNA3.1 encoding the SARS-CoV-2 S protein using Vigofect transfection reagent 1 (Vigorous Biotechnology, T001). One hour before transfection, the medium was replaced with fresh DMEM 2 (GIBCO, C11995500BT). Further transfection was performed according to the manufacturer's protocol. 3

The supernatants were harvested at 48 hours post transfection, filtered through a 0.45 μm cell strainer, 4

and split into 1.5 mL tubes for storage at -80°C. 5

Pseudovirus infection assay. Cells transduced with or without Ad-ACE2 were seeded in 96-well plates 6

at initial count of between 15, 000 and 20, 000 cells per well and treated with different agents (10 μM 7 camostat mesylate (Selleck, S2874) or 10 μM enzalutamide (Selleck, S1250)). Two days post seeding and 8 treatment, cells were incubated with pseudovirus for 12 hours. The culture medium supernatant was then 9

replaced with fresh medium. Two days post virus infection, the culture medium supernatant was removed, ChIP-seq library preparation. ChIP-seq was performed as previously described 36 . In brief, 10 million cells 27

were fixed with 1% formaldehyde at room temperature for 10 minutes with rotation. Then, 125 mM 28 glycine was added to quench the formaldehyde at room temperature for 5 minutes. After preclearing using 20 µL of protein G beads (Invitrogen, 10003D), 3 µL of anti-AR antibody was 35 added (Abcam, ab108341) for immunoprecipitation overnight. To bind the anti-AR antibody, 60 μL of 36 protein G beads was added and incubated with rotation for two hours at 4°C. The beads were washed 37

twice each with Low Salt Wash Buffer, High Salt Wash Buffer and LiCl Wash Buffer and resuspended 38

in 100 μL of freshly prepared DNA Elution Buffer (50 mM NaHCO3 and 1% SDS). The ChIP sample 39 beads were placed on a magnet, and the supernatant was collected into a new tube. The above elution 40 step was repeated with another 100 μL volume of elution buffer. The samples were then digested with 41 10 μL of Proteinase K (Invitrogen, 25530049) with incubation at 67°C for 4 hours. DNA was purified 1 with DNA Clean & Concentrator TM -5 (Zymo Research, D4004). One nanogram of eluted DNA was used 2 as input for library construction with a TruePrep DNA Library Prep Kit V2 for Illumina (Vazyme, TD503). 3

Libraries were sequenced with the Illumina NovaSeq sequencing system (PE 2×150 bp reads) at Berry 4

Genomics. 5

ChIP-qPCR. Eluted DNA (0.5 μL) was used as the template for ChIP-qPCR with SYBR qPCR Mix (Qiagen, 6 208054) following the manufacturer's protocol. The primer sequences used for ChIP-qPCR are listed as 7 follows: 8

AR_ChIP_TMPRSS2_ARE_F: TGGTCCTGGATGATAAAAAAGTTT; 9

AR_ChIP_TMPRSS2_ARE_R: GACATACGCCCCCACAACAGA. 10 ChIP-seq data processing and analysis. Raw fastq files were first trimmed to remove adaptors using 11

TrimGalore-0.5.0 with the following parameter settings: -q 25 --phred33 --length 35 -e 0.1 --stringency 4. 12

Trimmed fastq files were then mapped to hg19 genome utilizing Bowtie2 37 . Sambamba_v0.6.6 was 13 conducted to remove duplicates 38 . For IGV visualization, deepTools was then performed using function 14 bamCoverage to generate normalized CPM .bw files 39 . For peak calling, MACS2 was utilized with -q 0.05 15 parameter setting. DeepTools was further applied for heatmap visualization with the function of 16 computeMatrix and plotHeatmap. 17 ATAC-seq library preparation. To reduce the amount of contaminating mitochondrial DNA, we 18 performed a previously reported optimized ATAC-seq protocol 40 . In brief, 50,000 cells were collected 19

and washed once with PBS. Cells were then lysed in 50 μL of ice-cold lysis buffer (10 mM Tris-HCl, pH 20 7.4; 10 mM NaCl; 3 mM MgCl2; 0.1% NP-40; 0.1% Tween 20; and 0.01% digitonin) for 3 minutes on 21

ice. Immediately after lysis, nuclei were washed with 1 mL of wash buffer (10 mM Tris-HCl, pH 7.4; 10 22 mM NaCl; 3 mM MgCl2; and 0.1% Tween 20) and then centrifuged at 500g for 10 minutes at 4°C. To 23 prepare sequencing libraries, a TruePrep DNA Library Prep Kit V2 for Illumina (Vazyme, TD501) was 24 utilized for the following steps. 25 ATAC-seq data processing and analysis. The approach used or ATAC-seq data processing was quite 26 similar to that used for ChIP-seq data processing. However, the peak calling step differed due to the lack 27 of input control files. In brief, after raw reads were trimmed with TrimGalore-0.5.0, Bowtie2 was used for 28 mapping the reads to the hg19 genome 37 . SAMtools was further utilized for bam file sorting and indexing. 29

The bamCoverage function in deepTools was used to generate .bw files with counts per million (CPM) 30 normalization 39 . R package Diffbind was used to identify overlapped peaks between AR ChIP-seq-31 generated peaks in LNCaP cells and ATAC-seq-generated peaks in all four cell lines, respectively. Then, 32

both-open peaks were defined by overlapping the above generated peaks in all four cell lines. To further 33

identify specific prostate-open peaks, we employed the intersect function in bedtools 41 to exclude peaks 34 that emerged in any of the three lung cell lines in LNCaP cells. 35 GSEA analysis. We downloaded gene expression matrices of multiple cancer cell lines from 36 cBioPortal 42,43 (derived from Cancer Cell Line Encyclopedia). We next performed GSEA to determine 37

whether hallmark androgen response genes show significantly differences between AR-positive prostate 38 cancer cells (LNCaP, VCaP and 22RV1) and AR-positive lung cancer cells (A549, H1437 and H2126) 44 .

In addition, we also compared sum of z-scores for hallmark androgen response genes between these two 40 groups.

Coronavirus disease (COVID-19) Weekly Epidemiological Update

Pandemic Preparedness: 3 Developing Vaccines and Therapeutic Antibodies For COVID-19

Aerodynamic analysis of SARS-CoV-2 in two Wuhan hospitals

Evidence that TMPRSS2 activates the severe acute respiratory syndrome coronavirus 8 spike protein for membrane fusion and reduces viral control by the humoral immune response

SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically 11

Efficient activation of the severe acute respiratory syndrome coronavirus spike protein 13 by the transmembrane protease TMPRSS2

A transmembrane serine protease is linked to the severe acute respiratory syndrome 15 coronavirus receptor and activates virus entry

Structure of SARS coronavirus spike receptor-binding domain 17 complexed with receptor

Receptor and viral determinants of SARS-coronavirus adaptation to human ACE2

Tmprss2 is essential for influenza H1N1 virus pathogenesis in mice

TMPRSS2 Contributes to Virus Spread and Immunopathology in the Airways of 23

Murine Models after Coronavirus Infection

Inhibition of SARS-CoV-2 entry through the ACE2/TMPRSS2 pathway: a 25 promising approach for uncovering early COVID-19 drug therapies

TMPRSS2 and COVID-19: 28 Serendipity or Opportunity for Intervention?

A SARS-CoV-2 protein interaction map reveals targets for drug repurposing

Rapid repurposing of drugs for COVID-19

Emerging mechanisms of resistance to androgen receptor 35 inhibitors in prostate cancer

Development of a second-generation antiandrogen for treatment of advanced prostate 37 cancer

COVID-19 and androgen-targeted therapy for prostate cancer patients

Phenotypic analysis of mice lacking the Tmprss2-41 encoded protease

Generation of a Broadly Useful Model for COVID-19 Pathogenesis, Vaccination, and Treatment

Organoid cultures derived from patients with advanced prostate cancer

SOX2 promotes lineage plasticity and antiandrogen resistance in TP53-and RB1-deficient 3 prostate cancer

Acquired resistance to the second-generation androgen receptor antagonist enzalutamide 5 in castration-resistant prostate cancer

TMPRSS2 and TMPRSS4 promote SARS-CoV-2 infection of human small intestinal 7 enterocytes

A cellular census of human lungs identifies novel cell states in health and in asthma

SARS-CoV-2 entry factors are highly expressed in nasal epithelial cells together with 11 innate immune genes

Long-term expanding human airway organoids for disease modeling

Chromatin accessibility and the regulatory epigenome

SARS-CoV-2 Receptor ACE2 Is an Interferon-Stimulated Gene in Human Airway 17

Androgen receptor and androgen-20 dependent gene expression in lung

Androgen-deprivation therapies for prostate cancer and risk of infection by SARS-22

CoV-2: a population-based study (N = 4532)

On the relationship between androgen-deprivation therapy for prostate cancer and risk of 25 infection by SARS-CoV-2

Author Correction: Pathological inflammation in patients with COVID-19: a key 27 role for monocytes and macrophages

Transplantation of ACE2(-) Mesenchymal Stem Cells Improves the Outcome of Patients with 29

COVID-19 Pneumonia

Lung innervation in the eye of a cytokine storm: neuroimmune interactions 31 and COVID-19

ERG orchestrates chromatin interactions to drive prostate cell fate reprogramming

Fast gapped-read alignment with Bowtie 2

Sambamba: fast processing of NGS alignment 37 formats

deepTools2: a next generation web server for deep-sequencing data analysis. Nucleic 39

An improved ATAC-seq protocol reduces background and enables interrogation of 41 frozen tissues

BEDTools: a flexible suite of utilities for comparing genomic features

Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal

The cBio cancer genomics portal: an open platform for exploring multidimensional cancer 3 genomics data

Gene set enrichment analysis: a knowledge-based approach for interpreting 5 genome-wide expression profiles