Novel role of bone morphogenetic protein 9 in innate host responses to HCMV infection

To test if BMP signaling has antiviral activity during acute HCMV infection, we first verified the expression of BMP receptors and their responsiveness to BMP stimulation in human foreskin fibroblasts (HFF-1), which are permissive for HCMV. HFF-1 express transcripts of the BMP type I (ALK1, ALK2, ALK3, ALK6) and type II (BMPR2, ACVR2A, ACVR2B) receptors (Fig. 2A), and endogenous ALK1, ALK2, ALK3, and BMPR2 protein expression was also verified by immunoblotting (Fig. EV1A). For further assays, since BMPs or BMP-like factors are present in fetal bovine serum (FBS) (Kodaira et al, 2006), we serum-starved HFF-1 for 2 h prior to treatment to minimize background levels of stimulation (Fig. 2B). Next, we stimulated HFF-1 with the recombinant BMPs BMP4, BMP6, BMP9, BMP15, and Activin B as representative members of the BMP/Activin family based on their divergent phylogeny, and show that all tested BMPs induced phosphorylation of SMAD1/5/9, and Activin B induced phosphorylation of SMAD2, whereas BMPs and Activin B did not cause phosphorylation of STAT1, which was induced by IFNβ treatment (Fig. 2C). Under these conditions, transcription of Id1 and Id3, two canonical SMAD-responsive genes, was induced in HFF-1 upon BMP and Activin B stimulation, but not after IFNβ treatment (Fig. 2D). This confirms that HFF-1 express functional BMP type I and type II receptor signaling pathways.

Next, we analyzed whether BMP signaling in HFF-1 may induce other target genes than Id1 and Id3, and analyzed the transcript level of interferon-stimulated genes (ISGs) upon BMP stimulation, which was demonstrated previously in other cell types (Eddowes et al, 2019). Importantly, the ISGs Isg15, Irf9, Ifi6, and Stat2 were upregulated upon stimulation with BMPs and Activins, albeit mostly to a lower level compared to IFNβ stimulation (Fig. 2E). These data verify that BMP- and Activin-mediated signaling in HFF-1 is biologically active and leads to transcriptional activation of canonical BMP targets as well as ISGs.

Since stimulation of HFF-1 with BMPs or Activin B induced transcription of ISGs, we assessed whether this results in antiviral activity against HCMV. For this, we either stimulated HFF-1 with BMPs or Activin B or co-stimulated with BMPs/Activin B and IFNβ for 6 h, and then infected them with HCMV (Fig. 3A). Analysis of immediate-early 1 (IE1) protein expression, which is a marker for early HCMV gene expression, revealed that stimulation with BMP15 or Activin B significantly reduced the number of HCMV IE1⁺ cells, independently of added IFNβ (Fig. 3B). BMP4 and BMP6 did not affect the number of HCMV IE1⁺ cells alone (Fig. 3B, left panel) or in combination with IFNβ (Fig. 3B, right panel). Interestingly, while stimulation with BMP9 alone did not affect the number of HCMV IE1⁺ cells (Fig. 3B, left panel), co-stimulation of HFF-1 with BMP9 and IFNβ led to significantly lower IE1⁺ counts compared to IFNβ stimulation alone (Fig. 3B, right panel). This effect was maintained at high and low concentrations of BMP9 (Fig. 3C). Due to this dependence of the antiviral effect of BMP9 on IFNβ, we then focused our study on the investigation of this phenotype, rather than on the BMP15/Activin B phenotype, which we will investigate in the future.

With the potential involvement of BMP9 in the innate immune response, we next investigated whether BMP9 is secreted in response to HCMV infection. For this, we infected HFF-1 with HCMV and collected supernatants in 6 hour increments. To monitor the presence of BMPs in the supernatant of HCMV-infected HFF-1, 293T, which also express all tested BMP receptors (Fig. EV1A, B), were transfected with expression plasmids for a BRE-Luciferase reporter and treated with the supernatants from HCMV-infected HFF-1. The BRE-luciferase reporter responds to BMP stimulation with transcriptional activation of the luciferase gene (Fig. 3D). In addition, supernatants were incubated with a specific neutralizing antibody against BMP9 (Fig. EV1C, D) prior to stimulation of 293T to identify which portion of the induced response can be attributed to BMP9. As shown in Fig. 3D, upon treatment of 293T with HFF-1 supernatant harvested at early and late stages of HCMV infection (0–12 and 42–54 h post infection (hpi)), BRE-Luciferase fold induction was higher than in cells that were treated with supernatant from mock-infected HFF-1. Notably, the majority of the BRE response induced by the supernatants could be neutralized by the addition of the BMP9 antibody (Fig. 3D, dark gray symbols), suggesting that infection of HFF-1 with HCMV induces secretion of BMP9, which may in turn affect the innate immune response to HCMV infection.

With BMP9 enhancing the antiviral activity of HFF-1 against HCMV infection when co-applied with IFNβ (Fig. 3B,C), we next evaluated whether BMP9 stimulation enhances IFNβ-mediated signaling, or vice versa (Fig. 4A). For this, we analyzed transcript levels of the BMP9-associated receptors ALK1 (encoded by Acvrl1) and BMPR2 (encoded by Bmpr2) (Jiang et al, 2021; Salmon et al, 2020), as well as the SMAD-responsive gene Id1, in cells stimulated with BMP9 alone or in combination with IFNβ. No significant change in expression was detected when both stimuli were added (Fig. 4B), indicating that IFNβ stimulation did not enhance BMP9-mediated signaling. Notably, while expression levels of some ISGs, namely Isg15, Irf7, and Ifi6, were not affected upon co-stimulation with BMP9 and IFNβ compared to IFNβ stimulation alone (Fig. 4C, upper panels), transcript levels of the ISGs Irf9, Stat2, and Irf1 were significantly higher in the co-stimulation setting (Fig. 4C, lower panels). Congruently, expression of genes encoding negative regulators of BMP signaling (Smurf1 (Zhu et al, 1999)) or IFNAR signaling (Usp18 and Smurf1, (Arimoto et al, 2017; Yuan et al, 2012) were significantly lower in co-stimulated HFF-1 than in cells stimulated with IFNβ alone (Fig. 4D). This indicates that BMP9 modulates transcription of genes related to the antiviral immune response, especially transcripts associated with IFNAR signaling. We next applied inhibitors of canonical BMP- or IFNAR-mediated signaling, DMH1 and Ruxolitinib, respectively, to assess whether Stat2 and Irf9 transcription depended on BMP9-mediated signaling (Fig. 4E). While DMH1, but not Ruxolitinib, negated BMP9-mediated induction of Id3 (Figs. 4E and EV1E), transcription of Stat2 and Irf9 upon BMP9 stimulation was independent of canonical BMP signaling, but dependent on intact IFNAR signaling. The same result was obtained in the BMP9/IFNβ co-stimulation context (Fig. 4E). Collectively, these data suggest that BMP9-induced enhancement of ISG expression is mediated by IFNAR signaling.

Phosphorylation of the transcription factors STAT1 and STAT2 is a hallmark for the activation of IFNAR signaling. As Ruxolitinib treatment prevented BMP9-mediated induction of ISG transcription (Fig. 4E), we hypothesized that BMP9-mediated signaling may directly affect IFNAR signaling. To investigate this, we stimulated HFF-1 with either BMP9, IFNβ or in combination with decreasing concentrations of IFNβ and analyzed STAT1 phosphorylation 1 and 8 h later (Fig. 5A). While BMP9 alone did not lead to detectable levels of STAT1 phosphorylation, co-stimulated HFF-1 displayed stronger phosphorylation of STAT1 compared to HFF-1 stimulated with IFNβ alone at both time points (Fig. 5A, quantification in graph). Substituting IFNβ with IFNα, another type I IFN, led to comparable results, while BMP9 did not affect responses mediated by IFNγ, a type II IFN (Fig. 5B). Congruently, transcription of Irf9, Stat2, and Irf1 was enhanced upon addition of BMP9 to IFNα or IFNβ stimulated cells, but not when applied with IFNγ (Fig. 5C). Transcription of the SMAD-responsive gene Id3 was not affected by addition of any of the IFNs (Fig. 5C). These results reveal that BMP9 modulates type I IFN-, but not type II IFN-mediated signaling, likely by increasing STAT1 phosphorylation.

HCMV US18 and US20 are transmembrane proteins harboring seven transmembrane domains (Fig. EV2A), and have been associated, via a proteomic screen, with downregulation of the type I BMP receptors ALK1 and ALK2 during HCMV infection (Fielding et al, 2017). This is of particular interest as ALK1 is one of the main type I BMP receptors associated with BMP9-mediated signaling (Salmon et al, 2020). Thus, we investigated whether the expression of HCMV US18 and US20 specifcally inhibits BMP9-mediated signaling without affecting type I IFN signaling pathways. For this, we transfected 293T with the reporter plasmids BRE-Luciferase or MX1-Luciferase, responsive to BMP or IFNβ stimulation, respectively, and stimulated them with BMP9, IFNβ or both (Fig. 6A). To verify our reporters, we included the M27 protein of murine CMV, a well-characterized inhibitor of IFNAR signaling (Zimmermann et al, 2005), and expression of all proteins of interest was verified (Fig. 6C). Notably, expression of HCMV US18 or US20 inhibited BMP9-mediated induction of the BRE-Luciferase reporter, but not of the MX1-Luciferase reporter (Fig. 6B), and co-expression of US18 and US20 inhibited BMP9 signaling even stronger (Fig. 6D). Moreover, we could exclude modulatory effects of US18 and US20 expression on pattern recognition receptor (PRR) signaling (cGAS-STING, RIG-I or downstream of the PRR-activated transcription factor IRF3) (Fig. 6E), suggesting specific downregulation of BMP9-induced, but not of PRR- or IFNAR-induced signaling.

To verify these observations in HFF-1, we generated HFF-1 with doxycycline-inducible expression of V5-tagged US18 and/or HA-tagged US20 (Fig. 7A). Expression of HCMV US20 in HFF-1 showed a double-band likely corresponding to N-linked glycosylation since PNGase F treatment resulted in a single band (Fig. 7B). We next assessed whether expression of US18-V5 and/or US20-HA in HFF-1 can blunt the response to BMP9 stimulation. Indeed, in the presence of US18/US20, phosphorylation of SMAD1/5/9 was not detectable (Fig. 7C), and Id3 transcription was significantly reduced after BMP9 stimulation (Fig. 7D, upper graph), in line with the observation that both downregulate cell surface expression of BMPRs (Fielding et al, 2017). Strikingly, expression of either US18-V5 or US20-HA was sufficient to abrogate BMP9-mediated enhancement of Stat2 transcription, while they did not affect IFNβ-induced activation of Stat2 transcription (Fig. 7D, lower graph). These data show that expression of HCMV US18 and US20 in HFF-1 specifically inhibits BMP9-mediated signaling, as well as BMP9-mediated enhancement of IFNAR signaling, while they do not affect IFNAR signaling.

Next, we evaluated the role of US18 and US20 during HCMV infection. For this, we generated recombinant viruses with either V5-tagged US18 or HA-tagged US20 since antibodies against US18 and US20 are not available. Interestingly, US18-V5 and US20-HA proteins were detected in viral particles by immunoblotting (Fig. EV2B), indicating that US18 and US20 are components of the virions. Moreover, US18-V5 and US20-HA were already detected 3 h post infection (hpi) in HFF-1 (Fig. EV2C), too early for de novo expressed viral proteins, suggesting that US18 and US20 were delivered into cells. De novo expression of US18-V5 and US20-HA was detectable from 24 hpi (Fig. EV2D), which is in line with previous studies (Cavaletto et al, 2015; Fielding et al, 2014). Interestingly, the expression of US18 and US20 during HCMV infection overlapped with the time windows when we detected BMP9 secretion of HCMV-infected cells (Fig. 3D), indicating a concurrence of the host-generated BMP9-mediated enhancement of the antiviral immune response with the antagonistic virus-encoded inhibitors of BMP signaling.

Next, we evaluated the impact of US18 and US20 expression on BMP9-induced signaling during HCMV infection. For this, we generated HCMV recombinants lacking US18, US20 or both proteins by introducing stop codons in the respective genomic locus, and designated them as HCMV US18stop, US20stop, and US18/20stop. Since US18, US19, and US20 are transcribed as a tri-cistronic mRNA (Guo and Huang, 1993), this approach allowed minimal editing of the viral genome and ensured that only the proteins of interest were not expressed. With these viruses, we infected HFF-1 with a high or low multiplicity of infection for 3 h or 48 h, respectively, then stimulated the infected cells with BMP9 or mock-treated them for either 2 h for immunoblotting or 6 h for RT-qPCR analysis (Fig. 8A). At early stages of infection (3 hpi), HFF-1 infected with either HCMV US18stop, HCMV US20stop, or HCMV US18/20stop displayed a strong phosphorylation of SMAD1/5/9 already in the absence of BMP9 stimulation, potentially due to endogenously secreted BMP9 (see Fig. 3D). Notably, this was not visible in HCMV WT-infected cells (Fig. 8B). This suggests that incoming US18 and US20 (likely present in the virion, see Fig. EV2B, C) are involved in downregulation of BMP signaling. Interestingly, US18/US20-mediated inhibition of BMP signaling could be overcome in the presence of an excess of exogenous BMP9, visible by detectable levels of SMAD1/5/9 phosphorylation in HCMV WT-infected cells (Fig. 8B). In line with these results, HFF-1 infected with HCMV WT induced Id3 and Stat2 transcription in response to BMP9 stimulation, while HFF-1 infected with HCMV US18stop, HCMV US20stop, or HCMV US18/20stop induced transcription of Id3 and Stat2 independently of BMP9 (Fig. 8B).

At late stages of infection (48 hpi), when US18 and US20 are de novo expressed (see Fig. EV2D), BMP9 stimulation led to higher SMAD1/5/9 phosphorylation levels in HFF-1 infected with HCMV lacking US18/20 compared to HCMV WT, indicating that also de novo expressed US18/US20 downregulate BMP signaling at this time point of infection (Fig. 8C). Congruently, HCMV WT-infected HFF-1 did not respond to BMP9 stimulation with changes in Id3 and Stat2 transcript levels, while HFF-1 infected with HCMV US18stop, HCMV US20stop, or HCMV US18/20stop readily responded to BMP9 stimulation with an increase in Id3 and Stat2 transcripts (Fig. 8C).

In summary, our data shows for the first time that US18 and US20 downmodulate BMP9-mediated signaling at early and late stages of HCMV infection.

Since BMP9 enhances the antiviral immune response and HCMV US18/US20 downmodulate BMP9-mediated signaling, we raised the hypothesis that HCMV lacking US18 and US20 expression could be more sensitive to IFNβ treatment than WT HCMV. To test this, we infected HFF-1 with either HCMV WT or HCMV US18/20stop for 3 h, followed by stimulation with either BMP9, low-dose IFNβ, or co-stimulation with BMP9 and low-dose IFNβ (Fig. 9A). In line with Fig. 8B, HFF-1 infected with HCMV US18/20stop induced transcription of Id3 independent of BMP9, and correlating with data from Figs. 4 and  7, stimulation with IFNβ did not affect Id3 transcript levels in any condition tested (Fig. 9B). Transcript levels of the ISGs Irf9 and Stat2 in HFF-1 infected with HCMV WT were significantly higher in the co-stimulation context compared to IFNβ stimulation alone (Fig. 9C), resembling the phenotype observed in Fig. 4. Strikingly, this enhanced transcription of Irf9 and Stat2 was associated with lower transcript levels of the viral IE1 and UL44 transcripts in HCMV WT-infected cells (Fig. 9D). In comparison, IFNβ stimulation of HFF-1 infected with HCMV US18/20stop induced higher transcript levels of Irf9 and Stat2 compared to HCMV WT-infected cells (Fig. 9C). This indicates that the lack of US18 and US20 at early stages of HCMV infection, which in turn leads to activated BMP-mediated signaling (Fig. 8B), promotes stronger immune signaling in response to IFNβ stimulation. Congruently, HFF-1 infected with HCMV US18/20stop displayed significantly lower transcript levels of viral IE1 and UL44 upon IFNβ stimulation compared to HFF-1 infected with HCMV WT (Fig. 9D). To rule out that the differences in viral transcript levels are caused by different infection efficiencies of WT and US18/20stop viruses, we compared the expression level of the tegument protein UL82/pp71, which is a component of the viral particle and delivered into infected cells, 3 h post infection between the two viruses. As shown in Fig. EV3, UL82 expression was comparable among HCMV WT and HCMV US18/20stop.

In summary, our data show that an HCMV mutant lacking US18 and US20 expression is more sensitive to IFNβ treatment at early stages of infection, consistent with an important involvement of BMP9 in the antiviral immune response to HCMV infection.

The primary human foreskin fibroblast (HFF-1; SCRC-1041), MRC-5 (CCL-171) and human embryonic kidney 293T (293T; CRL‐3216) cell lines were obtained from ATCC. 293T and MRC-5 were maintained in Dulbecco’s modified Eagle’s medium (DMEM; high glucose) supplemented with 8% fetal calf serum (FCS) and 1% penicillin/streptomycin (P/S). HFF-1 were maintained in DMEM (high glucose) supplemented with 15% FCS, 1% P/S, and 1% non‐essential amino acids (NEAA). Cells were cultured at 37 °C and 7.5% CO₂. To generate HFF-1 with doxycycline-inducible expression of US18-V5 and/or US20-HA, lentiviral transduction was performed. For this, 293T (730,000 cells per well, six-well format) were transfected with 600 ng psPAX2, 600 ng pCMV-VSV-G and 800 ng of either pW-TH3 empty vector (EV1), pW-YC1 empty vector (EV2), pW-TH3 US18-V5, or pW-YC1 US20-HA complexed with Lipofectamine 2000. Sixteen hours post transfection, medium was changed to lentivirus harvest medium (DMEM h.gl. supplemented with 20% FCS, 1% P/S, and 10 mM HEPES). Forty-eight hours post transfection, lentivirus was harvested, diluted 1:2 with HFF-1 medium, and polybrene was added to a final concentration of 4 µg/ml. HFF-1 were seeded the day before transduction in a six-well format with 250,000 cells per well. For transduction, HFF-1 medium was replaced by medium containing lentivirus, and cells were infected by centrifugal enhancement at 684 × g and 30 °C for 90 min. Three hours post infection, medium was replenished with fresh HFF-1 medium. Successfully transduced cells were selected with 2 µg/ml Puromycin (HFF-1 transduced with pW-TH3 vectors) or 150 µg/ml Hygromycin (HFF-1 transduced with pW-YC1 vectors). Selection of cells with puromycin lasted over ~1.5 weeks with at least three passages, and selection with hygromycin over ~2.5 weeks with at least five passages. Untransduced control cells were treated with the antibiotics side-by-side to evaluate when the selection reached its endpoint. Occasionally, antibiotics were added to the medium when cells were passaged. Experiments were always performed with cells that were not treated with antibiotics for at least one passage prior the experiment.

The wild-type HCMV TB40-BAC4 (hereinafter designated as HCMV WT) was characterized previously (Sinzger et al, 2008) and kindly provided by Martin Messerle (Institute of Virology, Hannover Medical School, Germany). HCMV BACs were reconstituted after transfection of MRC-5 with purified BAC DNA. The reconstituted virus was propagated in HFF-1 and virus was purified on a 10% Nycodenz cushion. The resulting virus pellets were resuspended in virus standard buffer (50 mM Tris–HCl pH 7.8, 12 mM KCl, 5 mM EDTA) and stored at −70 °C. Infectious titer was determined by standard plaque assay and IE1 labeling in HFF-1. Manipulation of the HCMV genome was carried out by en passant mutagenesis (Tischer et al, 2010). pEP-KanS (Tischer et al, 2010) served as the template for PCR. For construction of the recombinant HCMV US18stop, a linear PCR product was generated using primers US18stopFOR: 5’- CGACGCCTACCTTAGACCGACAGCGGTCGTAAGCGGCAGCTAAGGCGACACCGCCTCCGTCTCCGAACACCATGAGTCGCCAGGATGACGACGATAAGTAGGG-3’ and US18stopREV: 5’-GCGGCACGATGGTGACCGTCGGCGACTCATGGTGTTCGGAGACGGAGGCGGTGTCGCCTTAGCTGCCGCTTACGACCGCTGCAACCAATTAACCAATTCTGATTAG-3’ to replace the start codon (ATG) with a stop codon (TAA, bold) at nucleotide positions 11,150 to 11,152 (accession #EF999921). HCMV-specific sequences are underlined. For construction of the recombinant HCMV US20stop, a linear PCR product was generated using primers US20stopFOR: 5’- AGAGAAGGGTAGGTGCGCCGCAGCGGCTTTGTGCCGAGACGGTCGCCACCTAACAGGCGCAGGAGGCTAACGCAGGATGACGACGATAAGTAGGG-3’ and US20stopREV: 5’- CCATGCGGGAGAGCAGCAGCGCGTTAGCCTCCTGCGCCTGTTAGGTGGCGACCGTCTCGGCACAAAGCCGCTGCAACCAATTAACCAATTCTGATTAG-3’ to replace the start codon (ATG) with a stop codon (TAA, bold) at nucleotide positions 12,821 to 12,823 (accession #EF999921). HCMV-specific sequences are underlined. For construction of the HCMV US18/20stop virus, the HCMV US18stop BAC was further manipulated and the US20stop mutation was incorporated by en passant mutagenesis. For construction of the recombinant HCMV US18-V5, a linear PCR product was generated using primers US18-V5-FOR: 5’- CGGCGGACGCATCGACGGCGTGAGTCTCCTCAGCTTGTTGGGTGGCGGAGGTTCGGGCAAACCGATTCCGAACCCGCTGCTGGGCCTGGATAGCACCTAAAAAGATGGGGGAGATGAAGGATGACGACGATAAGTAGGG-3’ and US18-V5-REV: 5’- TCTAAGCGAGCGGGAGCGTGTCATCTCCCCCATCTTTTTAGGTGCTATCCAGGCCCAGCAGCGGGTTCGGAATCGGTTTGCCCGAACCTCCGCCACCCAACAAGCTGAGGAGACTCACAACCAATTAACCAATTCTGATTAG-3’ to introduce a C-terminal V5-tag with a GGGS linker (bold) in between nucleotide position 10,330 and 10,331 (accession #EF999921). HCMV-specific sequences are underlined. For the construction of the recombinant HCMV US20-HA, a linear PCR product was generated using primers US20-HA-FOR: 5’- CAACGGGACTTTGACAGCCGCCAGTACGACGGGGAAGTCCGGTGGCGGAGGTTCGTATCCGTATGATGTGCCGGATTATGCGTAATGCCTATAAAACCGCGCAGGATGACGACGATAAGTAGGG-3’ and US20stopREV: 5’- CGCGGCTGCTGTGAAAACGGGCGCGGTTTTATAGGCATTACGCATAATCCGGCACATCATACGGATACGAACCTCCGCCACCGGACTTCCCCGTCGTACTGGCAACCAATTAACCAATTCTGATTAG-3’ to introduce a C-terminal HA-tag with a GGGS linker (bold) in between nucleotide position 12,061 and 12,062 (accession #EF999921). HCMV-specific sequences are underlined. Genome regions of the recombinant HCMV BAC US18stop, US20stop and US18/20stop were amplified and sequenced to verify the correct replacement of respective start codons, and no further addition of errors were introduced compared to the HCMV TB40/E WT BAC. Recombinant HCMV BACs were reconstituted and purified as described above.

Biosafety-relevant experiments with viruses were approved by the labor inspectorate (Staatliches Gewerbeaufsichtsamt) Braunschweig, Lower Saxony, Germany, and controlled by the Occupational Safety Department of TU Braunschweig in accordance with the Occupational Health and Safety Act.

HCMV US18 and US20 were amplified from the HCMV TB40/E BAC (accession #EF999921) and subcloned into pcDNA3.1-V5/His (Invitrogen) via the KpnI/NotI sites to generate pcDNA3.1 US18-V5/His and pcDNA3.1 US20-V5/His, respectively. psPAX2, pCMV-VSV-G, and pW-TH3 (clone 10), which contain a doxycycline-inducible promoter for expression of the respective insert and a puromycin selection cassette, were kindly provided by Thomas Hennig (University of Wuerzburg, Germany). To generate pW-YC1, the puromycin selection cassette in pW-TH3 was replaced by a hygromycin selection cassette previously amplified from pMSCVhygro (Clontech). To generate pW-TH3 US18-V5, US18 was amplified from the HCMV TB40/E BAC (accession #EF999921) with oligos introducing a C-terminal V5-tag with a GGGS linker, and subcloned into the NheI/EcoRI sites of pW-TH3. pW-YC1 US20-HA was generated by amplification of US20 from the HCMV TB40/E BAC (accession #EF999921) with oligos introducing a C-terminal HA-tag with a GGGS linker and subcloning into NheI/EcoRI sites of pW-YC1. Expression constructs for M35-V5/His and M27-V5/His (both in pcDNA3.1-V5/His, Invitrogen) have been described previously (Munks et al, 2006). pEF1α-Renilla, which expresses renilla luciferase under control of the EF1α promotor, pGL3basic MX1-Luciferase, which expresses firefly luciferase under control of the murine MX1 promoter, and pCAGGS Flag-RIG-I N, expressing a constitutively active truncation mutant of RIG-I, were kindly provided by Andreas Pichlmair (Technical University Munich, Germany). The reporter plasmid pGL3basic IFNβ-Luc (IFNβ-Luc), expressing the firefly coding sequence under the control of the murine IFNbeta enhancer, was described previously (Chan et al, 2017). pGL3 BRE Luciferase (BRE-Luc), expressing the firefly coding sequence under control of a synthetic promoter containing SMAD-binding elements (SBEs) was obtained from AddGene (#45126). pEFBOS mCherry-mSTING expressing monomeric Cherry fused to the N-terminus of murine STING, and pIRESneo3 cGAS-GFP (GFP fused to the C-terminus of human cGAS) were kindly provided by Andrea Ablasser (Global Health Institute, Ecole Polytechnique Fédérale de Lausanne, Switzerland). pIRES2-GFP was purchased from Clontech. CMVBL IRF3-5D codes for human IRF3 containing five amino acid substitutions (S396D, S398D, S402D, S404D, S405D) which render it constitutively active and was provided by John Hiscott (Institut Pasteur Cenci Bolognetti Foundation, Rome, Italy). All constructs were verified by sequencing. Primer sequences, as well as sequences of all constructs, are available upon request.

Rabbit polyclonal anti-STAT1 (#9172), rabbit monoclonal anti-phospho-STAT1 (#7649, clone D4A7), rabbit monoclonal anti-SMAD1 (#6944, clone D59D7), rabbit monoclonal anti-phospho-SMAD1/5/9 (#13820, clone D5B10), rabbit monoclonal anti-SMAD2 (#5339, clone D43B4), rabbit monoclonal anti-phospho-SMAD2 (#18338, clone E8F3R), rabbit polyclonal anti-p38 (#9212), rabbit monoclonal anti-phospho-p38 (#4631, clone 12F8), rabbit monoclonal anti-p44/42 (#4695, clone 137F5), rabbit monoclonal anti-phospho-p44/42 (#4377, clone 197G2), rabbit monoclonal anti-HA (#3724, clone C29F4), and rabbit monoclonal anti-GAPDH (#2118, clone 14C10) were obtained from Cell Signaling. Mouse monoclonal anti‐β-actin (A5441, clone AC‐15) and rabbit polyclonal anti-Calnexin (#C4731) were obtained from Sigma‐Aldrich. Mouse monoclonal anti‐V5 (BLD-680605, clone 7/4) and rabbit polyclonal anti-V5 (BLD-903801) were obtained from BioLegend. Mouse monoclonal anti‐ICP36 (anti-UL44) (#MBS530793, clone M612460) was purchased from MyBioSource. Mouse monoclonal anti-UL35 was described previously (Fabits et al, 2020). Rabbit polyclonal anti-UL82/pp71 (Tavalai et al, 2008) was kindly provided by Thomas Stamminger (Institute of Virology, Ulm University Medical Center, Ulm, Germany). Mouse monoclonal anti-IE1 (clone 63-27, originally described in (Andreoni et al, 1989) was a kind gift from Jens von Einem (Institute of Virology, Ulm University Medical Center, Ulm, Germany). Mouse monoclonal anti-ALK1 (#MAB370, clone 117702), anti-activin RIA/ALK2 (#MAB637, clone 71412) and goat polyclonal anti-BMPR-IA/ALK3 (#AF346) were purchased from Novus Biologicals. Mouse monoclonal anti-BMPR2 (#ab130206, clone 1F12) was purchased from Abcam. Alexa Fluor^®‐conjugated secondary antibodies were purchased from Invitrogen. The transfection reagents Lipofectamine 2000, FuGENE HD and jetPEI were purchased from Life Technologies, Promega and Polyplus, respectively. Polybrene was obtained from Santa Cruz Biotechnology and Doxycycline was obtained from Sigma-Aldrich. OptiMEM and SYTO^TM 59 red fluorescent nucleic acid stain were purchased from Thermo Fisher Scientific. Protease inhibitors (#4693116001) and phosphatase inhibitors (#4906837001) were purchased from Roche. Recombinant human BMP4 (#314-BP-010), human BMP6 (#507-BP-020), human BMP9 (#3209-BP-010), human BMP15 (#5096-BM-005), human Activin B (#659-AB-005), and the human/mouse anti-BMP9 antibody (#AF3209) were purchased from R&D Systems. Effective concentrations for each BMP and Activin used in this study differ and are based on the information stated on the respective datasheet. Recombinant human IFNβ was purchased from PeproTech (#300-02BC), recombinant IFNα2 (#592702) and recombinant IFNγ (#570202) were obtained from BioLegend. PNGase F (#P0704S) was obtained from New England Biolabs, and DMH1 (#S7146) and Ruxolitinib (#S1378) from Selleckchem.

Levels of BMP type I and type II receptor transcripts were determined using 100,000 HFF-1 or 293T, respectively. Cell lysis, RNA extraction, and RT-PCR were performed as described below. For experiments with ligand stimulations, HFF-1 were seeded the day before the experiment (100,000 cells per well in a 24-well format). Two hours prior stimulation, medium was replaced by HFF-1 growth medium lacking FCS (hereinafter referred to serum starvation medium) in order to induce serum starvation of the cells, followed by the addition of the stimuli BMP4 (final concentration 18 nM), BMP6 (final concentration 18 nM), BMP9 (final concentration 3 nM), BMP15 (final concentration 18 nM), Activin B (final concentration 4 nM), IFNα2 (final concentration 5 ng/ml), IFNβ (final concentration 5 ng/ml), or IFNγ (final concentration 5 ng/ml). After 6 h, cells were lysed in RL buffer and processed as described below. When using DMH1 or Ruxolitinib to inhibit BMP- and IFNAR-mediated signaling, respectively, the inhibitors were added to the serum starvation medium to a final concentration of 10 µg/ml after the first hour of the 2 h serum starvation step. The ligand stimulation for 6 h was performed as described above, cell lysis and processing of samples as described below. For infection experiments, HFF-1 were seeded the day before the experiment (100,000 cells per well in a 24-well format) and infected by centrifugal enhancement (684 × g at 30 °C for 45 min) the next day with HCMV WT, HCMV US18stop, HCMV US20stop, or HCMV US18/20stop (MOI 0.5 or 4). After centrifugation, cells were incubated at 37 °C for 30 min. Then, medium was exchanged to either serum starvation medium or normal growth medium for samples infected with MOI 4 or MOI 0.5, respectively. At 3 h post infection, samples infected with an MOI of 4 were stimulated with BMP9 (final concentration 3 nM) or IFNβ (final concentration 1 or 5 ng/ml) for 6 h. For samples infected with an MOI 0.5 medium was exchanged to serum starvation medium at 46 h post infection, followed by the ligand stimulation with BMP9 (final concentration 3 nM) or IFNβ (final concentration 5 ng/ml) at 48 h post infection for 6 h. Cells were lysed in RL buffer and RNA was purified using the RNA isolation kit (IST innuscreen, former AnalytikJena, #845-KS-2040250), following the manufacturer’s protocol. After RNA extraction, 1500 ng of RNA per sample was used for further processing. DNase treatment and cDNA synthesis was performed with the iScript^TM gDNA Clear cDNA Synthesis kit (BioRad, #1725035) following the manufacturer’s protocol. Generated cDNA was diluted 1:5 before performing PCR to obtain 100 µl of cDNA. Then, RT‐PCR was performed with 4 µl of cDNA per sample and the GoTaq^® qPCR Master Mix (Promega, #A6001) on a LightCycler 96 (Roche). GAPDH was used for normalization. The following oligo sequences were used: Gapdh_FOR: 5’-GAAGGTGAAGGTCGGAGTC; Gapdh_REV: 5’-GAAGATGGTGATGGGATTTC; Alk1_FOR: 5’-GCGACTTCAAGAGCCGCAATGT; Alk1_REV: 5’-TAATCGCTGCCCTGTGAGTGCA; Alk2_FOR: 5’-GACGTGGAGTATGGCACTATCG; Alk2_REV: 5’-CACTCCAACAGTGTAATCTGGCG; Alk3_FOR: 5’-CTTTACCACTGAAGAAGCCAGCT; Alk3_REV: 5’-AGAGCTGAGTCCAGGAACCTGT; Alk6_FOR: 5’-CTGTGGTCACTTCTGGTTGCCT; Alk6_REV: 5’-TCAATGGAGGCAGTGTAGGGTG; Bmpr2_FOR: 5’- AGAGACCCAAGTTCCCAGAAGC; Bmpr2_REV: 5’-CCTTTCCTCAGCACACTGTGCA; Acvr2a_FOR: 5’-GCCAGCATCCATCTCTTGAAGAC; Acvr2a_REV: 5’-GATAACCTGGCTTCTGCGTCGT; Acvr2B_FOR: 5’-CGCTTTGGCTGTGTCTGGAAG; Acvr2B_REV: 5’-CAGGTTCTCGTGCTTCATGCCA; Id1_FOR: 5’-GTTGGAGCTGAACTCGGAATCC; Id1_REV: 5’-ACACAAGATGCGATCGTCCGCA; Id3_FOR: 5’-CAGCTTAGCCAGGTGGAAATCC; Id3_REV: 5’-GTCGTTGGAGATGACAAGTTCCG; Isg15_FOR: 5’-CATGGGCTGGGACCTGA; Isg15_REV: 5’-GCCGATCTTCTGGGTGATCT; Irf1_FOR: 5’-GAGGAGGTGAAAGACCAGAGCA; Irf1_REV: 5’-TAGCATCTCGGCTGGACTTCGA; Irf7_FOR: 5’-CCACGCTATACCATCTACCTGG; Irf7_REV: 5’-GCTGCTATCCAGGGAAGACACA; Irf9_FOR: 5’-CCACCGAAGTTCCAGGTAACAC; Irf9_REV: 5’-AGTCTGCTCCAGCAAGTATCGG; Ifi6_FOR: 5’-TGATGAGCTGGTCTGCGATCCT; Ifi6_REV: 5’- GTAGCCCATCAGGGCACCAATA; Stat2_FOR: 5’-CAGGTCACAGAGTTGCTACAGC; Stat2_REV: 5’-CGGTGAACTTGCTGCCAGTCTT; Usp18_FOR: 5’-TGGACAGACCTGCTGCCTTAAC; Usp18_REV: 5’-CTGTCCTGCATCTTCTCCAGCA; Smurf1_FOR: 5’-AGTCCTCAGACACGAACTGCG; Smurf1_REV: 5’-GTCGCATCTTCATTATCTGGCGG.

HFF-1 were seeded one day prior to the experiments (100,000 cells per well in a 24-well format). HFF-1 were washed once with serum-free medium and cultured in serum starvation medium 2 h before stimulation. Then, cells were stimulated with BMP4 (18 nM final), BMP6 (18 nM final), BMP9 (3 nM final), BMP15 (18 nM final), Activin B (4 nM final), IFNα2 (0.2, 1, or 5 ng/ml final), IFNβ (0.2, 1, or 5 ng/ml final) or IFNγ (0.2, 1, or 5 ng/ml final). Immunoblot analysis of the BMP receptors (ALK1, ALK2, ALK3, and BMPR2) was conducted in unstimulated HFF-1 and 293 T. Cells were lysed at indicated time points with radioimmunoprecipitation (RIPA) buffer (20 mM Tris–HCl pH 7.5, 1 mM EDTA, 100 mM NaCl, 1% Triton X‐100, 0.5% sodium deoxycholate, 0.1% SDS). Protease and phosphatase inhibitors were added freshly to lysis buffers. Cell lysates and samples were separated by standard SDS–PAGE or Tricine-SDS-PAGE (for HCMV US18/US20 blots) and transferred to PVDF membrane (GE Healthcare) using wet transfer in Towbin blotting buffer (25 mM Tris, 192 mM glycine, 20% (v/v) methanol). Membranes were probed with the indicated primary antibodies and respective secondary HRP‐coupled antibodies diluted in 5% BSA in TBS‐T. Immunoblots were developed using SuperSignal West Pico or SuperSignal West Femto (Thermo Fisher Scientific) chemiluminescence substrates. Membranes were imaged with a ChemoStar ECL Imager (INTAS). Quantifications of immunoblot band intensities were performed with ImageJ. For HFF-1 lines with doxycycline-inducible expression of US18-V5 and/or US20-HA, doxycycline was added to a final concentration of 1 µg/ml 2 h post cell seeding. Eighteen hours post doxycycline treatment, medium was exchanged to serum starvation medium containing 1 µg/ml doxycycline for 2 h, followed by ligand stimulation with BMP9 (3 nM final) or IFNβ (5 ng/ml final). Effects of the inhibitors DMH1 and Ruxolitinib were tested by seeding HFF-1 the day before the experiment (100,000 cells per well in a 24-well format) and exchanging the medium to serum starvation medium containing either DMSO, Ruxolitinib (10 µM final) or DMH1 (10 µM final) for 2 h, followed by ligand stimulation with BMP4 (concentration 18 nM final) or IFNβ (5 ng/ml final) for 2 h. Cells were lysed and processed as described above. To determine the glycosylation status of US18-V5 and US20-HA, cell lysates were treated with PNGase F at 37 °C for 3 h according to the manufacturer’s protocol prior to Tricine-SDS-PAGE. Protein expression in luciferase-based assays was verified by analyzing the corresponding lysates stored in 1× passive lysis buffer (Promega) by Tricine-SDS-PAGE. For immunoblotting of viral particles, 50,000 PFU per virus were diluted from the purified virus stock in RNase-free H₂O, SDS loading buffer was added and incubated at room temperature for 1 h, followed by separation of the proteins by Tricine-SDS-PAGE. For infection experiments, HFF-1 were seeded the day before the experiment (100,000 cells per well in a 24-well format) and infected by centrifugal enhancement (684 × g at 30 °C for 45 min) with HCMV WT, HCMV US18stop, HCMV US20stop or HCMV US18/20stop, HCMV US18-V5 or HCMV US20-HA (MOI 0.5 or 4) the next day. After centrifugation, cells were incubated at 37 °C for 30 min. Then, medium was exchanged to either serum starvation medium or normal growth medium for samples infected with MOI 4 or MOI 0.5, respectively. At 3 h post infection, samples infected with an MOI of 4 were stimulated with BMP9 (3 nM final) or IFNβ (5 ng/ml final) for 2 h. For samples infected with an MOI 0.5, medium was exchanged to serum starvation medium at 46 h post infection, followed by stimulation with BMP9 (3 nM final) or IFNβ (5 ng/ml final) at 48 h post infection for 2 h. Cells were lysed at indicated time points and processed as described above. For analysis of UL82/pp71 expression, HFF-1 were infected by centrifugal enhancement with MOI 4 of HCMV WT or HCMV US18/20stop as described above, incubated for 3 h at 37 °C, and lysed for immunoblotting.

HFF-1 (5000 cells per well of a 96-well plate) were seeded one day prior the experiment. The next day, medium was exchanged to serum starvation medium for 2 h, followed by stimulation with BMP4 (18 nM final), BMP6 (18 nM final), BMP9 (0.25 or 3 nM final), BMP15 (18 nM final), Activin B (4 nM final) or IFNβ (5 ng/ml final) for 6 h. Cells were then infected with HCMV at an MOI of 0.5 in normal growth medium for 16 h. Cells were washed with TBS and fixed with 4% paraformaldehyde (PFA) for 10 min. Fixed cells were permeabilized with 0.1% Triton X-100 in TBS for 5 min and blocked with TBS containing 5% FCS and 1% BSA for 30 min. Cells were immunolabeled with mouse anti-IE1 antibody, washed with TBS and labeled with secondary anti-mouse AF488 conjugated antibody. Nuclei were stained with the SYTO^TM 59 red fluorescent nucleic acid stain. Red nuclei as a measurement for the total cell number and IE1 positive nuclei were counted with the IncuCyte S3 (Essen BioSciences, Sartorius, Göttingen, Germany) and the ratio of infected cells was calculated.

BRE- and MX1-luciferase assays: 293T (15,000 cells per well, 96-well format) were transiently transfected with 100 ng pGL3 BRE Luciferase or pGL3basic MX1-Luciferase, 1 ng pEF1α-Renilla and 120 ng plasmid of interest complexed with 0.8 µl FuGENE HD (Promega) diluted to 10 µl total volume in OptiMEM. Sixteen hours post transfection, medium was changed to serum-free DMEM to start the serum starvation process for 8 h, followed by the stimulation with BMP4 (18 nM final), BMP6 (18 nM final), BMP9 (3 nM), BMP15 (18 nM final), Activin B (4 nM final), IFNβ (5 ng/ml final) or supernatants from HCMV-infected cells (50 µl per well). To generate supernatants from HCMV-infected cells, HFF-1 were infected by centrifugal enhancement (MOI 0.5, 684 x g at 30 °C for 45 min, followed by 30 min incubation at 37 °C and replacement of virus-containing medium with fresh growth medium) with HCMV WT. For each 6 h increment, medium was exchanged to serum starvation medium for 6 h prior to harvest. Supernatants treated with the α-BMP9 antibody were first incubated with the antibody (final concentration 1 or 5 µg/ml) for 15 min prior to the 293 T stimulation. 16 h post stimulation cells were lysed in 1× passive lysis buffer (Promega).

cGAS-STING luciferase assay: 293T (25,000 cells per well, 96-well format) were transiently transfected with 60 ng pEFBOS mCherry-mSTING, 60 ng pIRESneo3 human cGAS-GFP (for stimulated conditions), 60 ng pIRES2-GFP (for unstimulated conditions), 100 ng pGL3basic IFNβ-Luc, 1 ng pEF1α-Renilla, 120 ng plasmid of interest and 1.2 µl FuGENE HD (Promega) diluted to 10 µl total volume with OptiMEM.

RIG-I N luciferase assay: 293 T (25,000 cells per well, 96-well format) were transiently transfected with 13 ng pCAGGS Flag-RIG-I N (stimulated) or pcDNA3.1 (unstimulated) together with 50 ng pGL3basic IFNβ-Luc, 1 ng pEF1α-Renilla and 130 ng plasmid of interest and 0.66 µl FuGENE HD (Promega) diluted to 10 µl total volume with OptiMEM.

IRF3-5D luciferase assay: 293T (25,000 cells per well, 96-well format) were transiently transfected with 10 µl of FuGENE HD/DNA complexes composed of 100 ng CMVBL IRF3-5D (stimulated) or 100 ng pIRES2-GFP (unstimulated), 100 ng pGL3basic IFNβ-Luc, 1 ng pEF1α-Renilla, 120 ng plasmid of interest and 1 µl FuGENE HD (Promega) diluted to 10 µl total volume with OptiMEM.

For the cGAS-STING, RIG-I N, and IRF3-5D luciferase assays cells were lysed 20 h post transfection in 1× passive lysis buffer (Promega). Luciferase activity was measured with the dual-luciferase assay system (Promega) and an Infinite M Plex plate reader (Tecan). Luciferase fold induction was calculated by dividing Renilla-normalized values from stimulated samples by the corresponding values from unstimulated samples.

Topology predictions for HCMV US18 and US20 were carried out using the online tools PredictProtein (https://predictprotein.org/), CCTOP (http://cctop.ttk.hu/) and DeepTMHMM (https://dtu.biolib.com/DeepTMHMM). Differences between the two datasets were evaluated by Student’s t test (unpaired, two-tailed), in the case of viral transcript levels after log transformation of the datasets, using GraphPad Prism version 5.0 (GraphPad Software, San Diego, CA). P values < 0.05 were considered statistically significant.