Endogenous IFITMs boost SARS-coronavirus 1 and 2 replication whereas overexpression inhibits infection by relocalizing ACE2

Summary Opposing effects of interferon-induced transmembrane proteins (IFITMs 1, 2 and 3) on SARS-CoV-2 infection have been reported. The reasons for this are unclear and the role of IFITMs in infection of other human coronaviruses (hCoVs) remains poorly understood. Here, we demonstrate that endogenous expression of IFITM2 and/or IFITM3 is critical for efficient replication of SARS-CoV-1, SARS-CoV-2 and hCoV-OC43 but has little effect on MERS-, NL63-and 229E-hCoVs. In contrast, overexpression of IFITMs inhibits all these hCoVs, and the corresponding spike-containing pseudo-particles, except OC43, which is enhanced by IFITM3. We further demonstrate that overexpression of IFITMs impairs cell surface expression of ACE2 representing the entry receptor of SARS-CoVs and hCoV-NL63 but not hCoV-OC43. Our results explain the inhibitory effects of artificial IFITM overexpression on ACE2-tropic SARS-CoVs and show that three hCoVs, including major causative agents of severe respiratory disease, hijack IFITMs for efficient infection of human cells.


Three of five human b-coronaviruses hijack IFITMs for efficient infection
IFITMs are critical for efficient replication of the two causative agents of SARS Overexpression of IFITMs inhibits SARS-CoVs by suppressing ACE2 cell surface expression Inhibitory membrane modulation but not S interaction is conserved between men and mice INTRODUCTION As a consequence of millions of years of an evolutionary arms race with pathogens, humans have evolved elaborate antiviral defense mechanisms. Many human genes encode structurally and functionally diverse antiviral effectors, referred to as 'restriction factors' (RFs). Some of them are constitutively expressed in cells to prevent infection but most are induced on innate immune activation, e.g., by interferons (IFNs). RFs typically target conserved viral features or cellular dependency factors and are thus active against a broad spectrum of viral pathogens. [1][2][3] However, viruses demonstrate a striking ability to adapt to their hosts and evade or counteract antiviral mechanisms. 4,5 Accumulating evidence shows that some of the most successful viral pathogens may even exploit otherwise antiviral cellular factors for efficient infection and replication. 6 One group of broad-spectrum antiviral factors is a member of the IFN-induced transmembrane (IFITM) family. [7][8][9] IFITMs are typically small (125-133 amino acids) transmembrane proteins, which have been reported to inhibit entry of retro-, orthomyxo-, paramyxo-, flavi-, filo-, rhabdo-, influenza A and coronaviruses (CoVs). 7,8,10,11 Humans encode five IFITM proteins and at least three of them (1, 2, and 3) can display antiviral activity. It is believed that IFITMs do not restrict viral entry by targeting specific viral compounds but by modulating membrane composition, fluidity and curvature. [12][13][14] Viral and cellular membranes differ in their lipid composition. 15,16 Viral membranes require strong curvature because viruses are much smaller than their target cells and strong bending is required for virion-cell fusion. Thus, it has been suggested that IFITMs increase membrane ''stiffness'' to prevent efficient virion-cell fusion. 9,[12][13][14]17 Although IFITMs are best known for their broad antiviral activity, accumulating evidence suggests that some viruses are able to exploit them for infection. An elegant study by Zhao et al. showed that the common cold coronavirus (ccCoV) OC43 utilizes IFITM3 for efficient entry into human hepatoma cell lines. 18 In contrast, inhibitory effects of IFITM overexpression have been reported for the three highly pathogenic human coronaviruses SARS-CoV-1, SARS-CoV-2 and MERS-CoV. 19- 21 We previously confirmed that

Human hepatocytes are highly susceptible to infection by common cold coronaviruses
To establish cellular models for S-mediated pseudo-particle (pp) infection and genuine hCoVs, we first generated single-cycle VSV particles encoding GFP instead of the open reading frame for the viral glycoprotein (VSV(GFP)DG) pseudo-typed with the S proteins of all seven human CoVs ( Figure 1A) and infected a panel of seven cell lines commonly used to study coronavirus infection. The human hepatoma-derived Huh-7 cell line was efficiently infected by Vesicular stomatitis virus pseudoparticles (VSVpp) containing the hCoV-229E, hCoV-NL63, MERS, CoV-1 and CoV-2 S proteins or the VSV-G control ( Figure 1B). Huh-7 cells are commonly used as model for primary human hepatocytes and show significant similarities in gene expression profiles. 24,25 The Huh7.5 derivate, which has a defect in the retinoic-acid-inducible gene-I (RIG-I) pathway, 26 was even more susceptible to S-mediated VSVpp infection. Infection of other cell lines was more variable and often less efficient. The epithelial Caco-2 cell line derived from a colon carcinoma, promoted efficient infection by hCoV-NL63, MERS and CoV-2 S, whereas A549-ACE2 and A549-ACE2-TMPRSS2 cells mediated infection by CoV-1 and CoV-2 S proteins ( Figure 1B). In comparison, parental A549 cells and Calu-3 cells showed low susceptibility to S-mediated VSVpp infection. Altogether, VSVpp provided a suitable system for S-mediated infection for five of the seven hCoVs ( Figure 1B). However, VSV-based pseudo-particles carrying the hCoV-OC43 and-HKU1 S proteins were generally poorly infectious, possibly because of inefficient virion incorporation of these spike proteins. To further expand the panel, we generated lentiviral pseudo-particles (HIVpp) carrying the hCoV-OC43 and-HKU1 S proteins by transfection of HEK293T cells with vectors expressing the S proteins together with a pNL1_HIV-1_NL4-3-Denv-fluc construct. 22 This allowed significant infection mediated by the hCoV-OC43 S (Figure 1C). In contrast, the titers of HIVpp containing hCoV-HKU1 S were too low for meaningful analyses. Notably, it has been previously reported that propagation of hCoV-HKU1 in immortalized cells is highly challenging. 27 For genuine viruses, we initially focused on the seasonal coronaviruses because we had already established Calu-3 cells as suitable system for infection by highly pathogenic CoVs. 22,28 In agreement with the results obtained using S-containing pseudo-particles, Huh7 cells supported replication of three of the four seasonal viruses: hCoV-OC43 (Organ Culture 43), which belongs to the b-coronaviruses (just like the highly pathogenic SARS-and MERS-CoVs; Figure 1A), and the two a-coronaviruses hCoV-NL63 (Netherland 63) and À229E (Figures 1D and 1E). Altogether, our results showed that Huh7 cells are suitable to study S-mediated infection by six of the seven human coronaviruses.

IFITM overexpression inhibits hCoV-229E and-NL63 but promotes hCoV-OC43 infection
To determine the impact of IFITM overexpression on S-mediated pseudo-particle and genuine ccCoV infection, we transfected Huh7 cells with expression constructs for the three IFITM proteins. Western blot analysis verified that transfection with the respective expression vectors resulted in high levels of IFITM1, 2 and 3 expression (Figure 2A) Figure 2B). In contrast, IFITM2 and IFITM3 increased OC43 S-mediated HIVpp infection by $3-fold ( Figure 2C). Notably, we have previously shown that HIVpp containing the SARS-CoV-2 S are inhibited by IFITM overexpression. 22 Thus, the enhancing effect was specific for OC43 S-mediated infection and not dependent on the carrier particle.
In agreement with the results obtained using 229E S VSVpp, overexpression of all three IFITMs inhibited RNA production ( Figure 2D) and infectious virus yields ( Figure 2E) in Huh7 cells infected with genuine hCoV-229E. The inhibitory effects of IFITM overexpression was confirmed using a replication-competent version of hCoV-229E expressing a luciferase reporter gene ( Figure 2F). 29 Overexpression of IFITMs had even stronger inhibitory effects on hCoV-NL63: IFITM1 and IFITM2 reduced viral RNA levels by $90% and IFITM3 blocked infection almost entirely ( Figure 2G). In contrast, IFITM overexpression moderately increased intracellular RNA levels ( Figure 2H) and viral RNA production ( Figure 2I) of hCoV-OC43. The

Effect of endogenous IFITMs on ccCoV infection
To determine whether IFNs induce expression of IFITMs and an antiviral state in Huh7 cells, we pre-treated these cells with IFN-a, IFN-b and IFN-g and analyzed them by western blot. Similar to our previous results iScience Article obtained using Calu-3 cells, 22 type I and II IFNs induced efficient expression of all three IFITMs in Huh7 cells ( Figure 3A). As expected, IFN treatment efficiently inhibited hCoV-229E, hCoV-NL63 and CoV-2 S-mediated VSVpp infection ( Figure 3B), as well as OC43 S-mediated HIVpp infection ( Figure 3C). In comparison, IFNs also inhibited genuine hCoV-229E and-NL63 but enhanced replication of hCoV-OC43 (Figure 3D). IFN-a and IFN-g were most effective and resulted in up to 35-fold higher levels of vRNA production in hCoV-OC43 infected Huh7 cells. Altogether, all S-containing pseudo-particles showed similar susceptibility to IFN inhibition and did not recapitulate the enhancing effect of IFNs on genuine hCoV-OC43 replication. These results suggest that IFN sensitivity of VSVpp and HIVpp is determined by the viral backbone rather than the pseudo-type S-proteins they carry.
To examine the effect of endogenous IFITM expression on S-mediated infection, we performed siRNA knock-down (KD) studies in Huh7 cells. siRNA treatment reduced the levels of IFITM mRNA expression by >90% ( Figure S1A) but had only modest enhancing effects on infection by hCoV-229E, hCoV-NL63, MERS, CoV-1 and CoV-2 S VSVpp ( Figure S1B). In comparison, silencing of IFITM3 markedly reduced OC43 S-mediated infection of Huh7 and Huh7.5 cells (Figures S1C and S1D). To examine the impact of IFITM depletion on genuine hCoVs, we performed siRNA KD analyses in untreated and IFN-b-treated Huh7 cells. Treatment with siRNAs generally decreased the levels of targeted IFITM mRNAs by about one order of magnitude ( Figure S1E). Silencing of endogenous IFITM expression had little effect on viral RNA production ( Figure 3E) and infectious virus yields ( Figure 3F) in hCoV-229E infected Huh-7 cells. In accordance with this, IFITM KD had little impact on infection by the hCoV-229E luciferase reporter virus ( Figure 3G). Similarly, individual silencing of the three IFITM proteins did not significantly affect hCoV-NL63 replication in Huh7 ( Figure 3H) and Calu-3 cells ( Figure 3I). In contrast, silencing of IFITM3 reduced the levels of viral RNA production in hCoV-OC43 infected Huh7 cells by 70-80% in both untreated and IFN-b-treated Huh7 cells ( Figure 3J). Altogether, these results indicate that IFITM overexpression clearly inhibits hCoV-229E and hCoV-NL63, while endogenous expression had little effect. In contrast, IFITM3 considerably enhanced hCoV-OC43 replication under all experimental conditions used.
IFITM2 and IFITM3 are critical for efficient replication of SARS-CoV-1 We have shown that efficient SARS-CoV-2 replication in Calu-3 cells is highly dependent on IFITM2 and (to a lesser extent) on IFITM3, while silencing of IFITM1 had only modest effects. 22,23 Thus, we asked whether endogenous IFITM2 expression also affects infection of human lung cells by the two most virulent human coronaviruses, SARS-CoV-1 and MERS-CoV. SARS-CoV-1 is closely related to SARS-CoV-2 and infected over 8,000 people with a case fatality rate of about 10% between 2002 and 2004. 30 MERS-CoV is only distantly related to SARS-CoV-2 and was first reported in 2012. Since then it has infected over 2,500 humans resulting in 900 deaths. It has previously been established that Calu-3 cells express all three IFITM proteins and are susceptible to infection by highly pathogenic human CoVs. 22,28 Our siRNA silencing approach reduces IFITM2 expression by about 90%, 22  IFITM2 and IFITM3 are highly homologous 7 and it has been reported that both localize in early and late endosomes, 31 iScience Article Potential interactions between hCoV spike proteins and IFITMs Our previous results from proximity ligation and mammalian-membrane two-hybrid assays as well as coimmunoprecipitation analyses, strongly suggested that IFITM-dependent SARS-CoV-2 infection requires specific interactions of IFITMs with the viral S protein. 22 To examine this further, we performed comprehensive co-immunoprecipitation studies and found that the hCoV-229E,-NL63,-OC43 and SARS-CoV-2 S-proteins co-purified with all three IFITM proteins ( Figure 5A). In comparison, the MERS-CoV S did not co-precipitate IFITM1 and IFITM2 and the SARS-CoV-1 S failed to pull down IFITM1. Notably, the S proteins of all CoVs exploiting IFITMs, i.e., hCoV-OC43 (IFITM3), SARS-CoV-1 (IFITM2 and IFITM3) and SARS-CoV-2 (IFITM2), interacted with the IFITM proteins that were most critical for efficient infection. To determine potential effects on virion binding to the cells, Calu-3 cells were transfected with IFITM-targeting or non-targeting siRNAs, treated with IFN-b or left untreated, and exposed to SARS-CoV-2 (MOI of 2.5) for 2 h on ice or at 37 C. Western blot analyses showed that different levels of IFITM expression had no significant effects on the amounts of cell-associated SARS-CoV-2 nucleocapsid (N) ( Figure 5B). Altogether, the results suggest that interactions between CoV S proteins and IFITMs may be required but not sufficient for effective utilization of IFITMs as infection cofactors.
It has recently been reported that IFITM3 knockout (KO) mice demonstrate increased susceptibility to SARS-CoV-2 induced disease. 33 A protective effect of mouse IFITM3 in vivo may seem at odds with a role of IFITMs as critical cofactors for efficient SARS-CoV-2 infection. To address this, we examined whether the broad inhibitory effect but not the specific interaction between the S protein in IFITMs may be conserved between humans and mice. We found that the SARS-CoV-2 S co-immunoprecipitated human but not mouse IFITM2 and IFITM3 proteins ( Figure 5C). In comparison, overexpression of both human and mouse IFITMs inhibited S-mediated infection by VSVpp (Figures 2B and 5D). Thus, conservation of the broad inhibitory function and lack of specific interactions with the viral spike may explain protective effects of IFITM3 in mouse models.

Overexpression of IFITM proteins prevents ACE2 cell surface expression
Increasing evidence supports that endogenously expressed IFITMs promote infection by SARS-CoVs, whereas overexpression inhibits it. 20,21 To elucidate the mechanisms underlying these opposing effects, we analyzed whether IFITMs may interact with ACE2 and/or affect the subcellular localization of this primary receptor of SARS-CoV-1, SARS-CoV-2 and hCoV-NL63. To assess potential interactions, we co-transfected HeLa cells with constructs expressing ACE2 and the three IFITM proteins and performed proximity ligation assays (PLA). The results revealed high numbers of PLA foci for ACE2 and all three IFITMs in the cytoplasm rather than the cell surface ( Figure 6A). These results showed that ACE2 and IFITM proteins are in close proximity and may interact in living cells. To examine potential effects on subcellular localization, we cotransfected HeLa with vectors expressing ACE2 and either IFITM1, IFITM2 or IFITM3 and analyzed them by confocal microscopy. As expected, ACE2 was mainly detected at the surface of HeLa cells transfected with the control vector ( Figure 6B). In striking contrast, the proportion of ACE2 localized at the cell surface dropped from $80% to $20% on IFITM overexpression ( Figures 6B and 6C). Similar results were obtained in A549-TMPRSS2-ACE2 transfected with a control vector or constructs expressing IFITM1, IFITM2 or IFITM3 ( Figure S3A). Flow cytometric analyses of permeabilized and non-permeabilized cells confirmed that overexpression of IFITM1, IFITM2 or IFITM3 significantly reduced the levels of ACE2 exposed at the surface of Hela-ACE2 and A549-TMPRSS2-ACE2 cells but not the total expression levels ( Figures S3B and S3C). The effect of IFITM overexpression was specific for ACE2 because it had no significant impact on cell surface expression of CD4 and CD46, the primary receptors of HIV and measles virus, respectively ( Figures 6D-6F).
The above-mentioned results raised the possibility that silencing of endogenous IFITM expression may enhance ACE2 cell surface expression. Unlike IFITM overexpression, however, siRNA silencing of endogenous IFITM had no significant effect on ACE2 ( Figure 7A). In agreement with published data, 34 IFN-b moderately increased the levels of ACE2. Silencing of IFITMs, however, neither affected ACE2 expression nor the proportion accessible at the cell surface ( Figure 7B). To challenge our hypothesis that IFITM overexpression inhibits SARS-CoV-2 by reducing ACE2 surface expression, we silenced expression of this receptor by siRNA treatment of Calu-3 cells. Reduced ACE2 expression strongly inhibited replication of the original SARS-CoV-2 Hu-1 strain and the Delta variant of concern in Calu-3 cells ( Figure 7C)  Pseudo-typed viral particles have proven highly useful to examine the activity of neutralizing antibodies and other agents targeting the spike proteins of human coronaviruses including SARS-CoV-2. However, IFITMs are thought to exert their antiviral activity by affecting the composition and fluidity of cellular membranes instead of specifically engaging spike or other viral proteins. [12][13][14] Previous studies reporting inhibitory effects of IFITMs used MLV-GFP pseudo-types carrying CoV-1 S, 21 or lentiviral particles carrying S proteins of different CoVs including SARS-CoV-2 and MERS-CoV. 19,20 Our present and earlier 22 results obtained using S-containing HIV-and VSV-based pseudo-particles agree with these previous findings. However, shapes, lipid composition and curvatures of S-containing pseudo-virions differ drastically from genuine CoV particles and more closely resemble those of the carrier virus. In addition to the S protein, infectious CoV particles also contain the membrane (M) and envelope (E) proteins as essential components of their lipid bilayer. 35 The M protein is highly abundant in CoV particles and may bind to target cells using heparan sulfate proteoglycans as initial attachment factors. 36 Thus, although single-round pseudo-viruses may reflect key aspects of S-mediated virus attachment and its inhibition, it is conceivable that they do not represent faithful models for attachment and fusion of genuine hCoV particles. Consequently, our data indicate that viral pseudo-particles carrying SARS-CoV-1 or SARS-CoV-2 S-proteins do not reproduce the ability of the native viruses to hijack IFITMs for efficient infection. Potential reasons for this are that the pseudo-virions may be particularly susceptible to the membrane modulating effects of IFITMs and/ or that the ability of SARS-CoVs to exploit IFITMs for infection requires specific features of the genuine viral particles.
The broad antiviral effect of IFITM as a modulator of membrane rigidity does not explain why endogenous and artificially overexpressed IFITMs have opposing effects on genuine SARS-CoV-2 infection. 19,22 Here, we demonstrate that no less than three of the five human b coronaviruses, hCoV-OC43, SARS-CoV-1 and SARS-CoV-2, require IFITMs for efficient replication. SARS-CoV-1 and SARS-CoV-2 are closely related and both utilize ACE2 as major receptor, just like hCoV-NL63. [37][38][39] We discovered that artificial overexpression of IFITM1, 2 and 3 prevents ACE2 expression at the cell surface. Our ongoing studies suggest that lower expression levels at least in part explain why endogenously expressed IFITMs do not compromise ACE2 cell surface expression. Because ACE2 is essential for infection by SARS-CoV-1, SARS-CoV-2 and hCoV-NL63, this effect provides a plausible explanation for previously reported inhibitory activities of IFITM overexpression. Our results agree with the recent finding that ACE2 is required for inhibitory effects of IFITM3 on SARS-CoV-2 S-mediated infection. 40 It also explains the diverse effects of IFITM overexpression on common cold coronaviruses: inhibition of hCoV-NL63 is particularly effective because this ACE2tropic virus should be affected by both changes in membrane fluidity and down-modulation of ACE2. In contrast, IFITM3 overexpression enhances hCoV-OC43 infection because this virus exploits IFITM3 for infection and does not use ACE2 as entry receptor. Thus, the entry receptors of OC43 remain available even under conditions of IFITM overexpression allowing robust enhancement by IFITM3. Finally, iScience Article hCoV-229E exploits neither ACE2 nor IFITMs for infection and showed moderate sensitivity to inhibition by IFITM overexpression that is most likely because of IFITM-mediated changes in cellular membrane rigidity.
Although knock-down of endogenous IFITM expression had little, if any, effect on hCoV-229E, hCoV-NL63 and MERS-CoV it strongly reduced replication of SARS-CoV-1, SARS-CoV-2 and hCoV-OC43 even in the absence of IFN treatment. This shows that low basal levels of IFITM expression are sufficient to boost infection by several hCoVs, whereas marked inhibition of genuine viruses requires overexpression. The exact mechanisms underlying IFITM-dependent enhancement of infection remain to be determined. However, results of co-immunopreciptation and PLA assays, together with the inhibitory effect of anti-IFITM2 antibodies 22,41 suggest that SARS-CoV-2 S proteins may interact with the N-terminal region of IFITMs to promote virion attachment or entry. We found that IFITM expression did not increase the levels of cell-associated SARS-CoV-2 nucleocapsid (N) protein after brief (2 h) exposure of Calu-3 cells to viral particles ( Figure 5B). However, our previous studies showed that depletion of IFITM2 significantly reduces the intracellular levels of viral RNA already 6 h after virus exposure. 22 Thus, endogenous IFITM2 expression promotes an early step of SARS-CoV-2 infection but does not seem to have a major impact on the overall levels of virus attachment. Our finding that IFITMs are in close proximity to spike and ACE2 raises the possibility that both may cooperate to promote SARS-CoV entry, although further studies are required to challenge this hypothesis. iScience Article SARS-CoV-1 and SARS-CoV-2 are highly similar to one another but only distantly related to hCoV-OC43. The finding that all three hCoVs hijack IFITMs for efficient infection, suggests that utilization of IFITMs as entry cofactors may be more common among coronaviruses than previously anticipated and has possibly been independently acquired several times during their evolution. In line with this possibility the underlying mechanisms appear to differ. The OC43 S exploits IFITM3 irrespectively of the carrier particle. 18 In contrast, IFITMs specifically enhance infection of genuine SARS-CoVs but not of the corresponding S-containing HIV-or VSV-based particles. 22,23 It has been shown that the C-terminal domain of IFITM3 is essential for enhancement of hCoV-OC43 infection, 18 whereas the N-proximal region of IFITM2 seems critical for enhancement of SARS-CoV-2 infection. Further analyses of the interactions between IFITMs, S proteins, and ACE2 at sites of viral fusion are required to elucidate how SARS-CoVs can use these otherwise antiviral proteins for efficient entry into their host cells.
It has been reported that knock-out of IFITM3 in transgenic mice expressing human ACE2 is associated with more severe disease and early death compared to control mice. 33 A protective effect of IFITM3 against severe disease in SARS-CoV-2 infected mice may seem at odds with our finding that IFITMs are important cofactors of SARS-CoV-2 infection. However, as antiviral restriction factors, IFITMs are under high pressure for positive selection and are thus highly variable. 42 We found that mouse IFITMs do not interact with the SARS-CoV-2 S protein in co-immunoprecipitation assays but display antiviral activity. This agrees with findings that overexpression of mouse IFITMs inhibits and silencing of endogenous mouse IFITM3 promotes SARS-CoV-2 infection. 19 Altogether, these results suggest that the broad antiviral activity of IFITMs but not the hijacking by SARS-CoV-2 S proteins is conserved between humans and mice.
IFITMs are expressed in human lung, gut, heart and brain cells and strongly induced in SARS-CoV-2 infected individuals. 22,43,44 High levels of expression at the primary targets for viral transmission and dissemination make IFITMs particularly suitable to be used as entry cofactors. We and others have shown that IFITM-targeting antibodies may inhibit SARS-CoV-2 infection. 22,41 In addition, recent data identified IFITMs as SARS-CoV-2 dependency factors. 45 Thus, increasing evidence suggests that IFITMs may be suitable targets for therapeutic approaches against SARS coronaviruses and further studies on their role in viral dissemination and pathogenesis are highly warranted.

Limitations of the study
We demonstrate that three human coronaviruses including the two responsible for severe respiratory disease utilize IFITMs for efficient replication. Our results further demonstrate that down-modulation of cell surface ACE2 explains why artificial IFITM overexpression inhibits ACE2-tropic hCoVs. However, the exact mechanism(s) underlying the enhancing effects of IFITMs on SARS-CoV-1, SARS-CoV-2 and hCoV-OC43 infection remain to be defined. It will also be interesting to determine whether isoforms of IFITM2 that differ in their antiretroviral activity, 46 also differ in their ability to promote SARS-CoV-1 and SARS-CoV-2 infection. In addition, further studies are required to clarify to relevance of IFITMs for transmission, tissue tropism and pathogenesis of SARS-CoVs in vivo. Finally, our results suggest that IFITMs may directly interact with the spike proteins of human coronaviruses and perhaps also ACE2. Detailed functional and structural characterization of these interactions might allow us to develop antivirals targeting IFITMs.

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

DECLARATION OF INTERESTS
The authors declare no competing interests.

Expression constructs
Expression plasmids encoding spike proteins of hCoV-229E and hCoV-NL63, pCG_hCoV-229E-spike C-V5-tag_BFP and pCG_hCoV-NL63-spike C-V5-tag_BFP, were cloned as described. 49 Expression plasmids encoding for IFITM1, IFITM2 and IFITM3 (pCG_IFITM1, pCG_IFITM2, pCG-IFITM3, pCG_IFITM1-IRES_eGFP, pCG_IFITM2-IRES_eGFP and pCG_IFITM3-IRES_BFP) were cloned as described. 22 Expression plasmids encoding for flag-tagged IFITM1, IFITM2 and IFITM3 (pTwist_EF1a_ 3Xflag_IFITM1, pTwist_EF1a_3Xflag_ IFITM2 and pTwist_EF1a_3Xflag_IFITM3) were PCR amplified and subcloned in pTwist based backbones using flanking restriction sites SpeI and NheI. The ORF of ACE2 was extracted with XbaI and MluI and then inserted into the XbaI-MluI site of pCG. For hCoV-229E propagation, Huh-7 cells were inoculated with a MOI of 0.1 in Dulbecco modified Eagle medium supplemented with 2% FCS. Cells were incubated at 33 C, and washed with phosphate-buffered saline and further cultured in fresh medium at 1 day post-infection. On day 4, supernatant was harvested, aliquoted, and stored at À80 C. HCoV-NL63 was propagated as described for hCoV-229E but using LLC-MK2 cells. HCoV-OC43 was propagated as described for hCoV-229E but using HCT-8 cells and harvesting viral stocks on day 6 and 8. BetaCoV/Netherlands/01/NL/2020 or B.1.617.2 (Delta) were propagated on Vero E6 infected at an MOI of 0.003 in serum-free medium as previously described. 53 Briefly, the cells were inoculated for 2hat 37 C before the inoculum was removed. The supernatant was harvested 48 h post-infection upon visible cytopathic effect (CPE). To remove the debris, the supernatants were centrifuged for 5minat 1,000 3 g, then aliquoted and stored at À80 C. Infectious virus titre was determined as plaque forming units (PFU). N gene RNA copies were determined by RT-qPCR. BetaCoV/Munich/Bav-Pat1/2020 or rSCV were propagated on Vero E6, rMERS-CoV on Vero B4 cells infected with low passage virus stock solution (approximately 1,000,000 PFU per ml) in serum-free medium. Three days post-inoculation, supernatant was harvested upon visible cytopathic effect (CPE). To remove the debris, the supernatants were centrifuged for 5minat 1,000 3 g and virus particles were purified from cytokines and concentrated using Vivaspin 20 (Sartorius, filtration units with a size exclusion of 100 kDa) according to the manufacturer's instructions. Virus concentrate was resuspended in 2-3 ml PBS, diluted 1:2 in virus preservation medium (0.5% gelatine in OptiPRO serumfree medium) and stored at -80 C. Infectious virus titre was determined as plaque forming units (PFU) and viral RNA concentration was quantified by RT-qPCR. All stocks were sequenced by next generation sequencing methods and the absence of additional mutations was confirmed to occur in less than 20% of the virus-specific reads.

Renilla luciferase assay
To determine replication level of Renilla hCoV-229E, the supernatant of infected Huh7 cells was removed 24h postinfection and the cells lysed in 200ml of Renilla luciferase lysis buffer. Renilla luciferase of hCoV-229E was determined by Renilla Luciferase Assay Kit (Promega) according to the manufacturer's instructions on an Orion microplate luminometer (Berthold).

TCID50 assay
To determine the infectious titer of hCoV-229E, 10,000 Huh-7 cells were seeded 1 day before infection in a 96-well plate. The following day, cells were inoculated with a 10-fold serial dilution of the respective virus stock. 4 days after infection, cytopathic effects were observed by light microscopy and the tissue culture infectious dose 50 (TCID50) was calculated according to Reed-Mü nch. For determining the TCID50 of hCoV-NL63 and hCoV-OC43 virus stocks, LLC-MK2 cells or HCT-8 cells were used and treated as described for hCoV-229E. iScience Article were fixed with 4% PFA for 20 minutes at room temperature, washed with PBS, permeabilized and blocked for 30 minutes at room temperature with perm/block solution containing 0.5% Triton X-100 and 5% FCS in PBS. Afterwards cells were stained 2h with primary antibody (a-ACE2 1:100, a-flag 1:100 and a-CD46 1:100) diluted in PBS in a wet chamber at room temperature. After washing with PBS-Tween 20, slides were incubated with secondary antibodies (Alexa Fluor IgG H+L 1:500) and 500 ng/ml DAPI 2h at 4 C. After washing with PBS-T and water, slides were mounted with Moviol-G. Images were acquired using a LSM 710 system.

QUANTIFICATION AND STATISTICAL ANALYSIS
Statistical analyses were performed using GraphPad PRISM 8 (GraphPad Software). P-values were determined using a two-tailed Student's t test with Welch's correction. Unless otherwise stated, data are shown as the mean of at least three independent experiments GSEM. Significant differences are indicated as: *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001. Statistical parameters are specified in the figure legends.