Tamoxifen protects breast cancer patients from COVID-19: rst evidence from real world data

Background COVID-19 severity is uneven between genders. We hypothesized a role of hormonal therapies in the severity of COVID-19 in breast cancer (BC) patients via the modulation of SARS-CoV-2 susceptibility genes. Patients and Methods We mined the Emilia Romagna region (Italy) registries to compare the rates of hospitalization and mortality for COVID-19 in 2020 amongst 24628 BC patients. Next, we analyzed the modulation of ACE2, TMPRSS2 and NRP1 gene expression and the susceptibility to SARS-CoV-2 infection by tamoxifen, fulvestrant and 17β-estradiol on human ER+ MCF-7 cells in vitro. Results The hospitalization rate observed for 4784 tamoxifen treated BC patients was the lowest (OR, 0.41; 95% CI, 0.18-0.94; p=0.04) among hormonal therapies and no fatalities occurred. A standard mortality rate reduction has been observed also for patients treated with aromatase inhibitors (SMR: 0.73; 95% CI, 0.45-0.90). In vitro experiments showed that fulvestrant, but not tamoxifen, increases ACE2, TMPRSS2 and NRP1 gene expression and susceptibility to SARS-CoV-2 infection and that 17β-estradiol reduces signicantly TMPRSS2 and NRP1 expression. Conclusions Tamoxifen treated BC patients showed a reduced rate of hospitalization and strikingly no fatalities for COVID-19. In vitro experiments conrmed a protective role of tamoxifen while an increased susceptibility to SARS-CoV-2 infection of ER+ cells treated with fulvestrant was observed. test

Introduction SARS-CoV-2 entry into target cells is mediated by the transmembrane protease serine 2 (TMPRSS2) that primes the fusogenic SARS-CoV-2 spike protein following its interaction with ACE2 receptor and viral internalization [1]. TMPRSS2 expression level critically determines SARS-CoV-2 infectivity and production in cells [2] and through pH-dependent endocytosis allows the fusion of the viral envelope with the cell membrane [1]. Endolysosome de-acidi cation can restrict replication of SARS-CoV-2 as acidic conditions are necessary for SARS-CoV-2 to enter into and be released from host cells. Besides ACE2 and TMPRSS2, neuropilin-1 (NRP1) has been recently characterized as a third SARS-CoV-2 susceptibility factor and coreceptor in uencing viral tropism and infectivity [3,4]. Androgens enhance transcriptionally the expression of TMPRSS2 and ACE2 [5,6]. This implies that patients undergoing anti-androgen receptor (AR) treatment may be less susceptible to viral entry due to a systemic downregulation of TMPRSS2 expression as recent data on prostate cancer suggest [7]. The severity of COVID-19 symptoms and mortality is uneven among genders with females being less affected [8,9]. However, the molecular basis of this gender bias is unknown. Interestingly, an epidemiological study on SARS-CoV-1, that similarly to SARS-CoV-2 enters cells via ACE2, showed that mortality rates were lowest in young women compared to men and that the protective effect mediated by gender declines progressively with age suggesting that hormones like estrogens could be involved [10]. Accordingly, ovariectomized female mice developed a more severe form of infection compared to controls and to animals treated with the anti-estrogen receptor (ER) fulvestrant [11]. Hormone receptor (HR) positive breast cancer (BC) patients represent 70-80% of the whole BC population and are principally treated with hormonal therapies including tamoxifen, fulvestrant and aromatase inhibitors (AIs) such as anastrozole, letrozole and exemestane to interfere with estrogen dependent cues according to age and disease stage. Tamoxifen is a selective estrogen receptor (ER) modulator (SERM) acting as an antagonist or agonist depending on the speci c tissue context with several pleiotropic activities [12]. Fulvestrant is a strict selective ER degrading (SERD) antagonist, whereas AIs induce estrogen deprivation in postmenopausal BC women by inhibiting the key aromatization step in the synthesis of estrogens from androgens thereby acting upstream of ER signaling. Given the distinct mechanisms of action of current hormonal therapies, we hypothesized that they may differently affect COVID-19 severity and mortality in BC patients. Hence, we mined real-world data from Emilia Romagna region (Italy) throughout 2020 to evaluate whether hospitalization and mortality for COVID-19 rates differed according to therapy in BC patients. Furthermore, we performed con rmatory in vitro experiments to investigate the molecular mechanisms involved in the therapy-related susceptibility alteration to SARS-CoV-2 infection at the cell level.

Study population
The cohort of the study included all BC patients resident in Emilia Romagna region and treated in 2020 with one of the following drugs alone (monotherapy): tamoxifen, fulvestrant, AIs (anastrozole, letrozole and exemestane) and anti-HER-2 (trastuzumab or pertuzumab). We retrieved information on whether and when patients were hospitalized for COVID-19 during the same period. Patients treated after the date of hospitalization have been excluded. The study was done using record-linkage processes between multiple databases (FED and AFT -direct and territorial drugs distribution, SDO Hospital discharge cards and REM Mortality detection) of the Emilia Romagna region. The following ATC codes were used to mine the databases for drug delivery to patients: AIs (L02BG03 -ANASTROZOLO, L02BG04 -LETROZOLO, L02BG06 -EXEMESTANE); fulvestrant (L02BA03 -FULVESTRANT); tamoxifen (L02BA01 -TAMOXIFENE) and anti-HER2 (L01XC03 TRASTUZUMAB, L01XC13 PERTUZUMAB) and to retrieve the date of treatments or drug administration. Admission for COVID-19 data were retrieved from hospital discharge cards contained in the regional hospital database (SDO) ltering for the following COVID-19 diagnosis codes: 079.82 SARS-CORONAVIRUS ASSOCIATO, 480.3 POLMONITE DA SARS-CORONAVIRUS ASSOCIATO, V01.82 ESPOSIZIONE A SARS-CORONAVIRUS ASSOCIATO. Deaths from COVID-19 were identi ed within the REM through the search for the following pathologies identi ed as the cause of death: U071, U072 (COVID-19, identi ed and not identi ed), J841 and J849 (Other interstitial lung diseases).

Statistical analysis
Age was summarized as medians and interquartile ranges, while COVID-19 hospitalizations and deaths were summarized by counts and percentages. Logistic regression models were tted on the entire cohort and on both male and female sub-cohorts with COVID-19 hospitalization or COVID-19 death as outcomes. The models were strati ed according to the therapy received in 2020 and age adjusted. Age was considered a potential confounder, since elderly patients are at increased risk for COVID-19 hospitalization and death [13], and were not evenly distributed in different therapy groups. Furthermore, Firth (1993) pro le penalized log likelihood approach was adopted to obtain robust parameter estimates in the presence of rare events such as the present dataset. We obtained odds ratios (OR) and 95% con dence intervals (95% CI) from these models. We used a Wald test to assess the null hypothesis of no difference between the rates of COVID-19 hospitalization/death among the different groups. Additional descriptive analyses were conducted in the COVID-19 hospitalized/died sub-cohorts to explore the time to hospitalization/death from last treatment administration normalized to the drug's half-life (HL). Age standardized hospitalization and mortality ratio (SHR and SMR respectively) were calculated to compare the risk of hospitalization/death for COVID-19 in the cohorts against the general REM population. 95% CI of SHR and SMR were calculated assuming a Poisson distribution of the observed numbers. All statistical analyses were performed using R version 4.0.0 (R Foundation for Statistical Computing) or GraphPad Prism; package logistf version 1.23 was used to compute the estimates of the odds ratios. Anti-HER-2-treated patients were used as a reference group since their age distribution was not skewed and they were not treated with hormonal therapy. T-test was used to analyze the signi cance of gene expression data. Kruskal-Wallis test was used to compare age distribution across therapies.
Expression analysis for ACE2, TMPRSS2 and NRP1 in human cells MCF7 cells (HTB-22, ATCC) were cultured in RPMI-1640 or high glucose DMEM and MDA-MB-231 (HTB-26, ATCC) were cultured in L-15 medium in standard conditions with or without 10nM 17β-estradiol (E2) as indicated. Cells were seeded into six well plates at a density of 1×10 5 cells/well for 24 hours in duplicate. Monolayers were treated for 48 hours with 10nM or 1uM fulvestrant, 4-hydroxytamoxifen (4-OHT) or vehicle with or without 1nM E2. 10nM and 1uM were chosen since are the steady-state plasmatic concentrations of fulvestrant and tamoxifen in patients, respectively [14,15], RNA was extracted using miRNeasy (Qiagen) and retro transcribed using iScript cDNA synthesis kit (BioRad). Real-time PCR was performed using SYBR Selected Master mix and 7500 Real-Time PCR System (Thermo Scienti c) following manufacturer's instructions. Primers used in RT-PCR are listed in the following . Speci cally, the N gene of SARS-CoV-2 was adopted to quantify the amount of virus. In parallel, the extraction and ampli cation steps were controlled by ampli cation of a spike in control for the albumin gene cloned harbored in a plasmid and added to the samples as 1.1pg DNA [17]. Furthermore, total RNA was extracted from MCF-7 cell pellets (PureLink™ RNA Mini Kit -Thermo Fisher Scienti c) and normalized amounts of total RNA were subjected to reverse transcription and Real-Time PCR assay as described above. The endogenous albumin transcript was adopted as extraction and normalization control. Finally, MCF-7 supernatants were also evaluated for the presence of infectious viral particles by plaque assays on Vero E6 cells (ATCC® CRL-1586), as previously described [18]. N gene expression for 4-OHT, fulvestrant and vehicle-treated cells was normalized to albumin gene expression.
The reported fold changes were calculated between mean values.

Analysis of hospitalization and death for COVID-19 in breast cancer patients
In order to test the hypothesis that hormonal therapies may in uence COVID-19 susceptibility and severity of symptoms, we mined real-world clinical data from the BC population of Emilia Romagna region assessing the frequency of hospitalization and death for COVID-19 in patients treated with tamoxifen, AIs, fulvestrant and anti-HER2 therapies (trastuzumab and pertuzumab) the latter used as a control group as non-hormonal therapy treated BC patients.  Table 2).
Then we compared the COVID-19 hospitalization and mortality rates for female patients treated with the different hormone therapies with those expected from the regional rates. In tamoxifen female treated patients we observed only 14 hospital admissions for COVID-19 out of the 21 expected according to the regional risk matched by age suggesting that the risk of hospitalization in tamoxifen female treated patients was lower (SHR: 0.67; 95% CI, 0.61-1.32; Table 3). Conversely, none of the 4580 tamoxifen treated female BC patients died for COVID-19 while 5 were expected according to the regional mortality risk. Importantly, AIs did not show a decreased hospitalization risk (SHR: 0.98; 95% CI, 0.81-1.12; Table 3) but a signi cantly lower mortality risk (SMR: 0.73; 95% CI, 0.45-0.90; Table 3) compared to the region. Fulvestrant-treated female BC patients showed a higher hospitalization rate for COVID-19 as compared to the region with 10 hospital admissions on 6 expected according to the regional risk adjusted by age (SHR: 1.67; 95% CI, 0.00-2.00; Table 3). Mortality by COVID-19 in the fulvestrant treated female cohort instead was as expected, with 2 fatalities observed based on the regional risk (SMR: 0.89; 95% CI, 0.00-1.92; Table 3).
Interestingly, also anti-HER2 female treated patients showed an increased risk of hospitalization (SHR: 2.50; 95% CI, 0.00-2.72) for COVID-19 with 8 hospital admissions on 3 expected according to the regional risk adjusted by age.  Next, we analyzed the distribution of hospitalization occurrence for COVID-19 normalized to drug half-life (HL), calculating the number of HL equivalents between the date of last treatment or drug administration and the hospitalization, to estimate the duration of protection or of susceptibility for each drug. Strikingly, 80% of hospitalizations were observed within the rst HL since the last administration of fulvestrant ( Figure 1) suggesting that a correlation between the plasmatic concentration of fulvestrant in BC patients and an increased hospitalization risk may exist. Conversely, the COVID-19 hospitalization rates increased over time after treatment discontinuation with tamoxifen and AIs showing respectively 50% and 85% of admissions occurring after the equivalent of 4 HLs (<6.25% putative drug blood concentration) ( Figure 1) suggesting a fade of a protective action as drug concentration lowers.

Tamoxifen treatment protects human cells while fulvestrant makes them more vulnerable to SARS-CoV-2 infection in vitro
In order to gain insights into the mechanism of protection from SARS-CoV-2 infection mediated by tamoxifen and of increased susceptibility by fulvestrant, we performed in vitro experiments on human cells. Speci cally, breast cancer cell lines expressing or not ER, MCF7 and MDA-MB-231 cells respectively, were selected as an in vitro model. First, cell susceptibility to SARS-CoV-2 was tested. Once assessed low levels of viral replication in both cell lines, we determined basal levels of ACE2, TMPRSS2 and NRP1 expression and whether 4-OH-tamoxifen (4-OHT), an active metabolite of tamoxifen, and fulvestrant affect the expression of those genes. In MCF7 cells, treatment with fulvestrant at 10nM and 1uM concentration signi cantly increased the expression of ACE2 (fold change>1.8; p=0.025), NRP1 (fold change>4.7; p=0.015) and TMPRSS2 (fold change>9.0; p=0.007) at 48 hours post treatment while 4-OH-T did not (Figure 2a). This effect correlates with ER expression since fulvestrant cannot modulate the expression of those genes in triple negative MDA-MB-231 cells (data not shown). Importantly, we tested whether 17β-estradiol (E2) can modulate ACE2, NRP1 and TMPRSS2 expression in MCF7 cells at physiological concentrations corresponding to pre and postmenopausal ranges (30-400 pg/ml = 0.11-1.47nM; <15 pg/ml = <0.055nM), respectively, and found both NRP1 and TMPRSS2 genes repressed at premenopausal concentrations while not signi cantly affected below 0.05nM E2 (Figure 2b). This suggests that the decline in E2 concentration during aging may promote their expression in tissues. Importantly, E2 treatment does not alter the expression of ACE2 while fulvestrant treatment does, suggesting that fulvestrant may exert its action also through estrogen-independent ER mechanisms.

Discussion
Hospitalization for COVID-19 is motivated by breathing di culties, extreme weakness, loss of appetite, diarrhea, dizziness, confusion or a sudden change in mental state especially in older adults. The worst cases manifest with serious respiratory, cardiovascular insu ciency necessitating intensive care or other life threatening complications that lead to death or a severely compromised health status especially in elderly patients. Recent studies have shown that severe COVID-19 symptomatology is related to an overwhelming production of cytokines in a process called cytokine release syndrome (CRS) where, especially IL-6, plays a central role [20]. Our study, based on real world clinical and epidemiological data, assessed the rate of hospitalization and mortality for COVID-19 in BC patients and provides evidence that tamoxifen and, to some extent AIs, can protect from COVID-19 mortality. Interestingly, fulvestrant and AIs did not provide the same protection despite they both interfere with estrogen signaling albeit with different mechanisms. In particular, fulvestrant leads to ER degradation in cells thereby preventing both estrogen dependent and independent ER signaling while AIs hinder estrogen production from androgens by inhibiting CYP19 aromatase enzymatic activity. This difference is mirrored also at the molecular level where fulvestrant treatment in ER expressing MCF7 cells is able to signi cantly increase the expression of 3 genes involved in SARS-CoV-2 susceptibility (ACE2, NRP1, TMPRSS2) while E2 treatment can only repress NRP1 and TMPRSS2 but not ACE2, the rst driver of SARS-CoV-2 infectivity. AIs e ciently decrease E2 below 5pM concentration in 85% of postmenopausal patients [21]. However, at this very low E2 concentration as well as at postmenopausal levels (<50pM) neither NRP1 nor TMPRSS2 genes are signi cantly repressed in MCF7 cells suggesting that, if the same holds true in patients, AIs treatment should not increase further the expression of those genes in post menopausal patients where AIs are principally used. On the other hand such a low estrogen concentration in women post menopause and treated with AIs could divert farther the equilibrium from estrogen dependent to independent ER signaling resulting in ACE2 repression. Conversely, typical premenopausal E2 concentrations (>100pM) signi cantly downregulated NRP1 and TMPRSS2 in MCF7 cells suggesting that women in their fertile age may express lower levels of both genes in tissues expressing ERs and possibly being less susceptible as compared to women at the postmenopausal stage. Fulvestrant and AIs discrepancies suggest that the former may increase the susceptibility to SARS-CoV-2 infection by removing also an estrogenindependent repression mediated by ER. In addition, fulvestrant treatment exerts a stronger inductive effect as compared to the absence of E2 in MCF7 cells in culture. Interestingly, ER signaling outcome is in uenced by the expression of co-regulators (cofactors and corepressors) at the single cell level and by the activation signals it perceives either via estrogens or estrogen-independent cues [22,23]. It has been shown that growth factor receptor signaling such as epidermal growth factor 1, insulin-like growth factor receptors [24] and epidermal growth factor receptor 2 (HER2) [25] can lead to: ER phosphorylation through the action of downstream kinases [26], estrogen insensitivity and to hormonal therapy resistance in breast cancer [27]. ERs are in equilibrium between estrogen responsive and non-responsive forms in tissues and speci c external cues and conditions such as the presence of estrogens and growth factors, drugs interfering with their signaling and speci c pathological conditions (i.e. diabetes, obesity) can alter this equilibrium. Interestingly, the COVID-19 mortality risk is especially high in patients with type 1 diabetes (T1D) that are characterized by extremely low physiological concentrations of insulin and reduced IGF-1 [28]. It is conceivable that those patients with low IGF1 plasmatic levels may be at increased risk for COVID-19 complications and mortality due to a reduced repression of COVID-19 susceptibility genes by poor estrogen-independent ER signaling. Coherently with a view where receptor tyrosine kinases (RTKs) and ER crosstalk to prevent ACE2, NRP1 and TMPRSS2 expression, we observed that anti-HER2 treated BC patients, who are likely impaired in HER2 signaling at systemic level, and not only in breast cancer tissue, showed an increased risk of COVID-19 hospitalization and mortality. In this subset of BC patients the activity of anti-HER2 therapies may similarly prevent ER phosphorylation, activation and consequently induce SARS-CoV-2 susceptibility genes. Fulvestrant leads to ER degradation uncovering the net effect of estrogen dependent and independent signaling pathways in the repression of SARS-CoV-2 susceptibility genes. This action cannot be recapitulated by AIs or by tamoxifen alone since they can affect only the estrogen dependent pathway (Figure 3). Moreover, the observation that 80% of COVID-19 hospitalizations occurred within the duration of one drug HL equivalent after last treatment in fulvestrant-treated patients further suggests that high fulvestrant plasmatic concentration is linked to higher risk of hospitalization for COVID-19. Epidemiological data are supported by in vitro experiments showing that 4-OHT decreased SARS-CoV-2 infectivity in MCF7 cells, while fulvestrant increased the susceptibility to SARS-CoV-2 infection and, consequently, the production of viral particles. Those results deserve special attention from a clinical standpoint because fulvestrant may expose already fragile individuals to unwanted increased health risks at the time of COVID-19 pandemics. Moreover our results are coherent with other studies showing that tamoxifen and toremifene [29][30][31], and related SERMs like raloxifene and bazedoxifene [30], reduce SARS-CoV-2 infectivity in monkey Vero E6 cells. Interestingly, while increased susceptibility mediated by fulvestrant correlates with the transcriptional upregulation of SARS-CoV-2 susceptibility genes, 4-OHT treatment does not change their expression in MCF7 cells suggesting that tamoxifen protects most likely through the interference with other steps of SARS-CoV-2 life cycle. Recently, two articles have shown that the SIGMA-1 receptor located in the endoplasmic reticulum plays an important role in SARS-CoV-2 replication using a comparative viral-human proteinprotein interaction map [29]. Knockout and knockdown of SIGMAR1 gene caused robust reductions in SARS-CoV-2 production identifying SIGMAR1 as a key therapeutic target for SARS-CoV-2 replication [32]. Drug repurposing screens to select agents with anti-SARS-CoV-2 activity identi ed many sigma receptors (Sig-Rs) ligands [31]. Tamoxifen is a Sig-R ligand [34,35] and was retrieved together with related molecules (i.e. toremifene) in screens for antiviral compounds against RNA viruses (i.e. HCV, SARS-CoV, MERS-CoV, EBOV) suggesting it may block the early steps of the viral replication cycle [36][37][38][39]. In addition, tamoxifen alters endosomal tra cking and increases the pH of endolysosomes [40] thereby hindering protease mediated membrane fusion events. Besides host cell-speci c effects, tamoxifen treatment systemically doubles the sex hormone-binding globulin (SHBG) concentration in BC patients [41]. SHBG is produced by the liver and binds to testosterone, dihydrotestosterone (DHT) and E2 in the blood regulating sex hormones availability to cells. Tamoxifen may therefore exert its anti-androgen effect in vivo also by altering sex hormones bioavailability in females. Our results suggest a model where tamoxifen tackles SARS-CoV-2 infection at different stages of its life cycle and in multiple ways at once ( Figure 3). Sex hormones balance is known to in uence key cell functions. Our experiments indeed highlight the role of E2 at premenopausal concentrations in restraining the expression of TMPRSS2 and NRP1 in human cells suggesting that E2 protection from severe COVID-19 in fertile females exists, in part may be due to this regulation and progressively diminishes with menopause onset. In addition, E2 concentration critically affects pro versus anti-in ammatory responses in humans and animals. High E2 concentrations (periovular to pregnancy levels) produce immune suppressive actions while the opposite occur for low E2 concentrations with the production of proin ammatory cytokines (IL-1, IL-6, TNF-α, IL-4, IL-10 and IFN) through NF-kB regulation [42,43]. Acute respiratory distress syndrome (ARDS), a severe complication in COVID-19, has been investigated in experimental studies and in agreement with our model a protective role of E2 in preventing systemic and pulmonary release of cytokines upon acute lung injury [44], viral infection with H1N1 [45] and SARS-CoV-1 [11] has been documented. Interestingly, estrogens are known repressors of IL-6 and are used to treat osteoporosis in women [46]. Similarly, E2 treatment inhibits IL-6 expression in MCF7 cells and tamoxifen treatment does not interfere with this repression. If the same regulation takes place in patients, the potative protective effect of E2 in restraining IL-6 production from cells should not be hindered by tamoxifen therapy. Therefore, premenopausal BC patients treated with tamoxifen may be characterized by a triple protection mediated by estrogens, by estrogen independent mechanisms (ER phosphorylation and activation) and last by the pleiotropic antiviral effects of tamoxifen.
The conclusions of our study on the outcome of COVID-19 hospitalization and mortality are supported by observations on a very large sample size of BC patients and are also compared with the whole population of Emilia Romagna region across 2020. In addition those data are coherent with in vitro experiments on the expression of SARS-CoV-2 susceptibility genes and virus infectivity. We emphasize that we tested the effects of 4-OHT and fulvestrant at their therapeutic concentrations and demonstrated that they can signi cantly in uence the susceptibility of human ER expressing MCF7 cells to SARS-CoV-2 infection. Our results envision a therapeutic use of tamoxifen in infected people to restrain the severity of the symptoms, COVID-19 mortality and the vicious circle induced by SARS-CoV-2 infection on ACE2 and TMPRSS2 expression [47]. Potentially also a prophylactic use could reduce the risk of infection and spread of the disease. A shortcoming of the study is that we cannot distinguish whether a reduced rate of hospitalization and mortality for COVID-19 in tamoxifen treated patients is due to a decreased rate of infection, whether it is due to the mitigation of symptoms or both. However, in a recently published letter Competing interests GM has competing interests with Novartis, BMS, Roche, P zer, ARIAD, and MSD that are not related to the present study. All the other authors have no con icts of interest to declare.

Ethical Approval
The study has been approved by the local ethical committee (CEROM; protocol number IRST 100.51 ACT4COVID).

Data and materials availability
All data and materials used in the analysis are available upon request. Materials transfer agreements (MTAs) will be issued if required.

Figures
Page 19/21 Figure 1 Hospitalization rates expressed as percentage of all COVID-19 hospitalizations occurred since the last drug administration. In order to compare different treatments, the time from the last drug administration is expressed as drug half live equivalents. In particular, 50 days HL for fulvestrant14, 7 days HL for tamoxifen15, 2 days HL for AIs19 were considered.