Deletion of the non-adjacent genes UL148 and UL148D impairs human cytomegalovirus-mediated TNF receptor 2 surface upregulation

Human cytomegalovirus (HCMV) is a prototypical β-herpesvirus which frequently causes morbidity and mortality in individuals with immature, suppressed, or senescent immunity. HCMV is sensed by various pattern recognition receptors, leading to the secretion of pro-inflammatory cytokines including tumor necrosis factor alpha (TNFα). TNFα binds to two distinct trimeric receptors: TNF receptor (TNFR) 1 and TNFR2, which differ in regard to their expression profiles, affinities for soluble and membrane-bound TNFα, and down-stream signaling pathways. While both TNF receptors engage NFκB signaling, only the nearly ubiquitously expressed TNFR1 exhibits a death domain that mediates TRADD/FADD-dependent caspase activation. Under steady-state conditions, TNFR2 expression is mainly restricted to immune cells where it predominantly submits pro-survival, proliferation-stimulating, and immune-regulatory signals. Based on the observation that HCMV-infected cells show enhanced binding of TNFα, we explored the interplay between HCMV and TNFR2. As expected, uninfected fibroblasts did not show detectable levels of TNFR2 on the surface. Intriguingly, however, HCMV infection increased TNFR2 surface levels of fibroblasts. Using HCMV variants and BACmid-derived clones either harboring or lacking the ULb’ region, an association between TNFR2 upregulation and the presence of the ULb’ genome region became evident. Applying a comprehensive set of ULb’ gene block and single gene deletion mutants, we observed that HCMV mutants in which the non-adjacent genes UL148 or UL148D had been deleted show an impaired ability to upregulate TNFR2, coinciding with an inverse regulation of TACE/ADAM17.


Introduction
Human cytomegalovirus (HCMV, taxonomic name: Human herpesvirus 5; NCBI Taxonomy ID 10359) is the prototypical member of the betaherpesvirinae. A large proportion of the global human population is latently infected with HCMV as indicated by sero-prevalence rates ranging from 33.0 to 81.0% for developed countries and 59.1 to 95.7% for developing countries, which even further increase when the elderly are assessed (93.8 to 97.7% for developing countries) (1). With only very few exceptions (2), most primary and recurrent HCMV infections in healthy adults progress without overt disease. Conversely, HCMV frequently causes severe and often debilitating or even life-threatening diseases in individuals with immature, impaired, or senescent immunity such as congenitally infected newborns, transplant recipients, and AIDS patients (3). This direct association between clinical manifestations and impaired immunity, as well as several findings documenting how HCMV fundamentally shapes the immune system (4,5), highlight the ongoing immunological battle between HCMV and its host. Accordingly, a comprehensive multi-parametric network analysis of numerous aspects of the immune system among twins of discordant HCMV sero-status showed that HCMV significantly affects 119 of 204 immunological parameters (6). Although the National Academy of Medicine already assigned the highest priority to the development of a HCMV vaccine in the year 2000 (7,8), approved HCMV vaccines have remained unavailable. Besides the lack of defined surrogates of protective immunity, a major obstacle to vaccine development resides in the numerous potent modulators of immunity expressed by HCMV (5,9) that target key aspects of the immune responses such as antigen presentation and T cell recognition (10), antibody responses (11), myeloid cells (12), NK cells (13), cytokine signaling, the interferon (IFN) system (14)(15)(16)(17), and NFkB signaling initiated by tumor necrosis factor (TNF) receptor superfamily (TNFRSF) members (18). Accordingly, HCMV modulates various surface molecules and receptors on infected cells (see for example (19)(20)(21)).
In the case of TNF, the receptors, and down-stream signaling cascades, the mutual interplay between cytomegaloviruses and their hosts is particularly multifaceted going far beyond simple blockade. HCMV infection initially results in NFkB activation (see e.g. (22)(23)(24)), in part through the host-encoded casein kinase II (25). Accordingly, NFkB signaling is a key mediator in the transition of monocytes to an activated pro-inflammatory state upon HCMV infection (26). However, in cells in which HCMV productively replicates, IkBa degradation induced by exogenously added TNFa is significantly impaired, indicating the existence of viral inhibitors of NFkB signaling downstream of TNFR1 (27,28). Accordingly, NFkB inhibitory activities have been shown for several HCMVencoded gene products including pIE2-pp86 (24), cmvIL-10 (29), pUL44 (30), miR-US5-1 and miR-UL112-3p (31,32), and pUL26 (33). Despite these HCMV-encoded antagonists of NFkB signaling, HCMV also benefits from certain NFkB activity (see. e.g. (34)). Accordingly, it is well known that the major immediate early promoter (MIEP) contains functional kB sites (35-37) and that TNFR1-and TNFR2-dependent signaling is capable to enhance the MIEP activity (38). Furthermore, viral genes expressed later during the replication cycle also benefit in terms of abundance from NFkB activity (34). In accordance with the notion that HCMV takes advantage from NFkB activity, the virus encodes proteins that activate NFkB activity such as pUL76 (39). HCMV even encodes transmembrane glycoproteins with clear sequence similarity to TNFRSF members such as the HVEM mimetic pUL144 (40), which binds BTLA (41,42) and activates NFkB signaling through TRAF6 (43) and TAK supported by the ubiquitin ligase TRIM23 (44). We and others have shown that HCMV expresses the protein pUL138 that enhances surface levels of TNFR1 and thereby sensitizes pUL138-expressing cells to activate NFkB signaling at TNFa concentrations which are too low to activate cells devoid of pUL138 (27,45,46). Accordingly, in HCMV retinitis patients, increased levels of TNFR1 have been observed (47). Higher levels of soluble forms of TNFR1 have also been described during HCMV pneumonitis (48). Intriguingly, the latter work also described increased levels of TNFR2. Based on our observation that HCMV-infected cells exhibit an increased capacity to bind TNFa and aforementioned clinical data suggesting increased levels of soluble TNFR2, we interrogated here if and how HCMV affects TNFR2.
The HSV-1 strain F was kindly provided by David Johnson (Portland, USA). Vaccinia virus (VACV) strain Western Reserve (WR) was originally provided by Bernard Moss (National Institutes of Health, Bethesda, MD).

Cytokines and inhibitors
The activation of NFkB signalling was induced by treatment with TNFa (R&D Systems). For the detection of IkBa degradation, cells were incubated with 20 ng/ml TNFa for 30 min. TNFainduced gene expression was analyzed after 3 h of treatment. Ganciclovir (50 mM; Sigma) was used to prevent herpesviral DNA replication and decrease late gene expression. GCV was added at 2 h post-infection after removal of the virus-containing infection solution.

Generation of recombinant HCMV mutants
Recombinant HCMV mutants were generated according to previously published procedures (52, 53) using AD169-BAC2 or AD169-BAC20. For the construction of the HCMV deletion mutants, a PCR fragment was generated (see Supplementary  Table S1 for primer sequences) using the plasmid pSLFRTKn (54) as the template DNA. PCR fragments containing a kanamycin resistance gene were inserted into the parental BAC by homologous recombination in E. coli. The inserted cassette replaces the target sequence which was defined by flanking sequences in the primers. This cassette is flanked by frt-sites which can be used to remove the kanamycin resistance gene by flp-mediated recombination. The removal of the cassette results in a single remaining frt-site. For the generation of AD169varSDgpt-UL148D-HA, an AD169-BAC20based mutant expressing C-terminally hemagglutinin [HA]tagged UL148D, the gpt sequence of the BAC cassette was replaced by a kanamycin cassette. Subsequently, the kanamycin cassette was removed, generating an frt-site that was used to introduce the UL148D-HA expression cassette, an frt-site-flanked fragment encompassing the cellular EF1 promoter in front of the UL148D-HA coding sequence. Correct mutagenesis of recombinant HCMV BACs was confirmed by southernblot and PCR analysis (data not shown). Recombinant HCMVs were reconstituted from HCMV BAC DNA by Superfect (Qiagen) or FugeneHD (Promega) transfection into permissive MRC-5 cells by following the instructions of the manufacturer. UL148D-HA expression by AD169varSDgpt-UL148D-HA was confirmed by immunoblot analysis (data not shown).

Cloning of expression vectors and transient transfection
For the cloning of the UL148 and UL148D expression constructs, primers containing restriction sites and C-terminal HA epitope tag were used to amplify the respective gene product from AD169-BAC2 DNA. PCR fragments were cloned into pIRES2-EGFP by use of the introduced restriction sites. Plasmids were confirmed by the sequence determination of the inserted fragment. HeLa cells were transfected with pIRES2-EGFP plasmids expressing UL148-HA or UL148D-HA using FugeneHD (Promega) by following the manufacturer's instructions.

Flow cytometry
For flow cytometry, cells were detached using Alfazyme (PAA Laboratories), washed, and incubated in 50 ml of 3% (v/v) FCS in phosphate-buffered saline (PBS) with labeled antibody (phycoerythrin [PE]-conjugated) anti-TNFR1, anti-TNFR2, or the respective PE-conjugated isotype control (R&D Systems) for 45 min on ice in the dark. For the quantification of TACE/ADAM17 surface densities, cells were incubated with an antibody detecting the TACE ectodomain (MAB9301, R&D Systems) followed by the incubation with a PE-conjugated secondary antibody (Biolegend). After antibody staining, cells were washed three times and measured in the PE channel of a FACS Canto II (Becton Dickinson). For detection of TNFa binding, cells were incubated with TNFa-Biotin and Streptavidin-FITC or let unstained. Cells were washed three times and measured in the FACS Canto II FITC channel. Histograms were generated by the use of FlowJo (Tree Star).

Immunoblot analysis
Immunoblotting was performed according to standard procedures. Briefly, cells were washed in PBS and lysed in lysis buffer ( ). Equal amounts of protein were subjected to SDS-PAGE and transferred to nitrocellulose membranes. Immunoblot analysis was performed using mouse monoclonal antibody (MAb) anti-b-actin (Sigma), mouse MAb recognizing HCMV pUL83/pp65 (3A12; Abcam), and rabbit polyclonal anti-IkBa (sc-371; Santa Cruz). The proteins were visualized using peroxidase-coupled secondary antibodies (Dianova) and the ECL chemiluminescence system (Cell Signaling).

Northern blotting and semi-quantitative RT-PCR analysis of specific transcripts
Total RNA was extracted from cells using the RNeasy minikit (Qiagen). Total RNA was subjected to morpholinepropanesulfonic acid (MOPS) gel electrophoresis and transferred to nylon membranes using a TurboBlotter (Schleicher and Schuell). Probes were prepared by PCR with gene-specific primers (see Supplementary Table S2 for primer sequences) and digoxigeninlabeled dUTP (Roche) for the detection of the indicated transcripts. Hybridization and detection were performed as described by Roche manuals. For semi-quantitative RT-PCR, total RNA was digested with DNase I and used as the template for one-step RT-PCR (Qiagen) with gene-specific primers (see Supplementary Table S2 for primer sequences).

Statistical analysis
Statistical significance was determined using an one-way ANOVA test as indicated in the figure legends. A p value of <0.05 was considered statistically significant. *, p value <0.05. **, p value <0.01. ***, p value <0.001. Calculations of p values were performed using GraphPad Prism.

Results
HCMV-AD169 was the first HCMV strain that was sequenced and annotated and that was, and still is, used for numerous studies. The AD169varATCC genome differs from clinical isolates by its lack of the ULb′ region. However, HCMV-AD169 variants were identified that retained large parts of the ULb′ region. AD169 variants with long (AD169varL) and short (AD169varS) UL genome regions differ concerning 11.8 kb of coding capacity within the ULb′ region. We used the two representative BACderived clones AD169-BAC2 and AD169-BAC20, which genetically correspond to AD169varL and AD169varS, respectively (49). For convenience, we named the AD169-BAC2-and AD169-BAC20derived viruses AD169varL and AD169varS, respectively.
Based on previous findings regarding the upregulation of TNFR1 surface levels by the HCMV-encoded protein pUL138, we tested how HCMV affects the binding of biotinylated TNFa to infected fibroblasts. Upon ligand binding, TNFR1 and TNFR2 rapidly internalize (55,56). Making use of this mechanism as a specificity control, cells were incubated for 1 h at 37°C which resulted in the expected decrease in the signal intensity, most likely through the internalization of TNFa-bound receptor complexes ( Figures 1A, B). Cells infected with the ULb'-positive AD169varL (described in (27)) showed a TNFa binding which clearly exceeded the binding of cells infected with AD169varS which lacks the ULb' region ( Figures 1A, B). The increased TNFa binding was only partially dependent on pUL138 as indicated by cells infected with a UL138 deletion mutant (DUL138). We concluded that at least one additional gene product of the ULb' region enhances the binding of TNFa to HCMV-infected fibroblasts.
To study the signaling of TNFa in HCMV-infected cells, we first assessed the TNFa-mediated proteasomal degradation of the NFkB inhibitory protein IkBa (also known as NFKBIA), which represents an essential step of the canonical NFkB signaling. On protein level, IkBa degradation is a hallmark of canonical NFkB signaling, while on mRNA level, a subsequent ikba induction as negative feedback regulation indicates NFkB-dependent gene expression. In accordance with the interpretation that the ULb' region encodes gene products that increase TNFa-induced NFkB activation, we observed TNFa-induced IkBa degradation at 48 and 72 hpi in cells infected with the ULb'-positive AD169varL, but not in cells infected with the ULb'-negative AD169varS ( Figure 1C).
While the IkBa protein needs to be degraded to enable canonical NFkB signaling, ikba transcription is strongly induced, which returns the system to its initial state. Therefore, we studied the upregulation of the well-known NFkB-responsive host gene ikba. In accordance with aforementioned publications documenting an inhibition of TNFa-induced signaling in HCMV-infected cells (28), we observed that HCMV impaired the TNFa-induced upregulation of ikba transcripts ( Figure 1D). This effect occurred irrespective of UL138 and the ULb' region. Given that HCMV also harbors functional kB sites in its genome, for example in the MIEP, we tested if TNFa treatment may increase the transcription of such viral genes. A clear upregulation of ie1 and ie2 mRNA amounts in TNFa-treated cells infected with an ULb'-positive HCMV, but not in cells infected with an HCMV variant lacking the ULb' region, indicated that ULb' gene product(s) indeed enhance(s) the TNFainduced signaling upstream of viral promoters containing functional kB sites. For a more general assessment, we inspected the HCMV genome for additional kB sites in the vicinity of promoter regions. Among others (data not shown), we found such sites adjacent to the genes US9 and UL139. Corroborating the notion that TNFa enhances the expression of HCMV-encoded genes harboring kB sites in an ULb' gene region-dependent manner, we observed an upregulation of US9 and UL139 mRNA amounts by TNFa in cells infected with the ULb'-positive AD169varL but not in case of the ULb'-negative AD169varS ( Figure 1D). Since deletions of UL138 or UL136 did not abrogate this effect, we concluded that the ULb' region codes for yet unidentified positive regulators of TNFa binding which enhance the activity of kB-containing HCMV promoters including the MIEP.
The signaling pathways downstream of TNFR1 and TNFR2 are only partially overlapping (Figure 2A). One fundamental difference is the absence of a death domain in TNFR2 through which TNFR1 induces TRADD/FADD-dependent caspase activation. We inspected publicly available data sets (58) regarding HCMVinduced changes in transcript abundance and ribosome footprints of all components of the TNFR1 ( Figure 2B) and TNFR2 ( Figure 2C) downstream signaling listed in the TNF signaling pathway -Homo sapiens (human) of the KEGG database ("hsa04668" (57)). Since we were interested in both the transcript abundance of these factors and HCMV-induced alterations, we plotted RPKM as well as fold changes. Intriguingly, we observed that various mRNAs coding for mediators of TNFR2 signaling are continuously transcribed and appear to be translated (as indicated by the ribosome footprints) after HCMV infection, and that HCMV even enhances the abundance and ribosome occupancy of mRNAs coding for factors inherently involved in TNFR2 signaling such as TRAFs, and IKKs ( Figure 2C). Conversely, TRADD, FADD, and Caspase 8, which are involved in TNFR1 signaling, show an opposite regulation ( Figure 2B).
We observed an increased TNFa binding and an induction of HCMV genes harboring kB sites in cells infected with ULb'-positive HCMVs together with continuous or even enhanced expression of various mediators of TNF receptor signaling. Therefore, we considered it more likely that the increased TNFa binding results from the upregulation of cellular TNFRSF members rather than virus-encoded decoy receptors. Although TNFRSF ligands and receptors form a complex and often converging network, TNFa is only recognized by TNFR1 and TNFR2 (59). Since a relevant fraction of the upregulation of TNFa binding and most of the TNF-dependent upregulation of HCMV genes harboring kB sites was observed irrespective of the TNFR1 modulator pUL138, we analyzed surface levels of TNFR2. Contradicting a common belief that TNFR2 expression is restricted to immune cells, we observed a strong upregulation of TNFR2 on the surface of HCMV-infected MRC-5 cells ( Figure 3A). This TNFR2 upregulation was not a general phenomenon elicited by DNA viruses because it was not observed when the cells were infected with herpes simplex virus (HSV)-1 or vaccinia virus (VACV) (Figures 3A, B). Treatment with ganciclovir (GCV), that inhibits the viral genome replication and drastically reduces the expression of viral late genes, which is functionally linked to genome amplification, resulted in intermediate TNFR2 surface levels ( Figures 3C, D), indicating that at least one of the causative HCMV-encoded gene products is expressed in part with early (or even earlier kinetics) and in part with GCV-sensitive late expression profiles.
In the absence of HCMV infection, U937 and THP-1 cells, as expected, showed TNFR2 on the surface, whereas MRC-5, HeLa, and retinal pigment epithelial (RPE)-1 cells did not ( Figure 4A). Intriguingly, however, uninfected MRC-5 and to a lesser extend RPE-1 cells expressed tnfr2 mRNA ( Figure 4B), suggesting that TNFR2 surface levels in MRC-5 cells are negatively regulated, at least in part, by post-transcriptional mechanisms that are altered upon HCMV infection. Accordingly, RT-PCR analysis using graded mRNA template dilutions showed that HCMV does not upregulate the tnfr2 mRNA in MRC-5 cells (data not shown), indicating that the increase in TNFR2 surface levels of infected MRC-5 fibroblasts cannot be explained by an HCMV-induced transcriptional upregulation. Since tnfr2 mRNA levels were increased upon HCMV infection in the data set published by Tirosh et al., who had infected human foreskin fibroblasts (HFF), while we did not observe an upregulation in MRC-5 cells, we tested another HCMVpermissive fibroblast cell line. In BJ-5ta, HCMV induced tnfr2 transcription, indicating that our analysis was capable to recognize transcriptional differences, but that MRC-5 cells are special with regard to their baseline tnfr2 transcription ( Figure 4C, see a potential explanation in the discussion). Despite these transcriptional differences in uninfected cells, HCMV encounters tnfr2 transcripts in all assessed fibroblasts, either constitutively expressed (MRC-5) or conditionally induced upon infection (BJ-5ta). Accordingly, TNFR2 surface levels are upregulated by HCMV in MRC-5 (e.g., Figure 3) and in BJ-5ta cells (see below).
To test whether the HCMV-induced TNFR2 upregulation occurs virus dose-dependently, we applied graded TB40-HCMV doses to RPE-1 cells, which express relatively low levels of tnfr2 mRNA amounts ( Figure 4B). Indeed, we observed a strong and multiplicity of infection (MOI)-dependent upregulation of TNFR1 and TNFR2 surface levels ( Figure 4D).
To identify the genetic determinant(s) of the HCMV-mediated TNFR2 upregulation, we compared a set of HCMV strains that differ, among other aspects, in the degree to which the ULb' region is still present in the genome ( Figure 5A). While TB40-HCMV, which harbors the entire ULb' region, and AD169varL, which harbors most of the ULb' region, upregulated TNFR2 surface densities, the ULb'-negative AD169varS failed to do so ( Figures 5B, C). Interestingly, Towne-HCMV, which lacks parts of the ULb' region, showed an intermediate TNFR2 upregulation phenotype ( Figures 5B, C), suggesting that more than one ULb' gene product influences the regulation of TNFR2. Since MRC-5 and BJ-5ta fibroblasts differ with regard to the baseline tnfr2 transcription ( Figures 4C, 5D), we also assessed the level of tnfr2 transcripts and TNFR2 protein surface levels in BJ-5ta cells infected either with AD169varL or AD169varS (Figures 5D, E). Irrespective if cells express tnfr2 constitutively (MRC-5) or conditionally upon HCMV infection (BJ-5ta), ULb'-positive HCMVs (AD169varL) strongly enhanced TNFR2 surface levels, while ULb'-negative HCMVs (AD169varS) did not.
To identify genes that are essential for the TNFR2 upregulation, we generated and tested a comprehensive panel of gene block and single gene deletion HCMV mutants. Starting with bigger deletions, we confirmed that indeed at least two gene products influence the upregulation of TNFR2, as indicated by the intermediate phenotypes of the two non-overlapping deletions DUL148-133 and UL148A-UL150 ( Figure 6A). By iterative cycles of gene (block) deletion in HCMV by BACmid mutagenesis and functional TNFR2 analysis, we further narrowed down the responsible gene loci. While the gene region spanning from UL139 to UL133 was dispensable for TNFR2 upregulation, the regions UL148 to UL146 and UL148A to UL148D were not ( Figure 6A). Using a set of single gene deletions of the affected canonical genes, we found that deletion of UL148 impaired the HCMV-mediated upregulation of TNFR2 ( Figures 6B, C). We assume that the slight effect observed with the UL147A deletion may result from a deregulation of the adjacent UL148 gene (Figures 6B, C). The observation that the DUL148A-UL148D mutant as well as the DUL149-UL150 mutant showed an effect ( Figure 6A) suggested that the gene UL148D, which was affected in both gene block mutants, may be involved in TNFR2 upregulation. This assumption turned out to be true as indicated by the loss of function observed with a specific UL148D HCMV deletion mutant (Figures 6B, C). Again, we assume that partial effects observed by deletions of UL148B and UL148C may due to do indirect effects on adjacent genes. Taken together, UL148, located close to the UL region, and UL148D, located close to the RS region, are required for the HCMV-mediated upregulation of TNFR2 in fibroblasts. Since Towne-HCMV harbors the UL148 gene but lacks UL148D, this argumentation is in agreement with the intermediate TNFR2 upregulation phenotype observed for Towne-HCMV ( Figures 5B, C). Our data document that HCMV encodes at least two non-adjacent genes in the ULb' region that are indispensable for a post-transcriptional upregulation of TNFR2 surface levels.  (61), which may reveal other pUL148 targets. Only a non-significant trend towards TNFR2 downregulation in the absence of UL148 was observed (0.82-fold; p~0.22), maybe due to the presence of pUL148D. Intriguingly, the supplementary data shows an increase of disintegrin and metalloproteinase domain-containing protein 17 (ADAM17; DUL148/wt fold ratio: 4.05; supplementary data of (61)), which is also known as tumor necrosis factor-a-converting enzyme (TACE). TACE induces TNFR2 shedding (62,63). Based on our TNFR2 findings and aforementioned published work, we tested if the upregulation of TNFR2 levels by pUL148 and pUL148D may be explained through decreased levels of TACE. Compared to the parental AD169varL, HCMV mutants lacking the capacity to express pUL148 or pUL148D replicated to similar levels in fibroblasts ( Figure 7A). With regard to TACE, however, AD169varL significantly decreased surface levels compared to AD169varS, and DUL148 and DUL148D showed intermediate phenotypes ( Figure 7B). Conversely, ectopic expression of UL148 or UL148D in HeLa cells significantly decreased TACE levels ( Figure 7C), and the insertion of the UL148D gene into the ULb'negative AD169varS restored the intermediate phenotype of TACE downregulation (varSDgpt-UL148D-HA, Figure 7D), suggesting that the herein described TNFR2 upregulation mediated by pUL148 and pUL148D is caused, at least in part, by their capacity to downregulate TACE.

Discussion
Here, we show that HCMV actively stimulates the surface disposition of TNFR2 on infected fibroblasts. This seeding observation was not restricted to our laboratory. After we knew what to look for, we recognized that supplementary data sets accompanying publications of our colleagues support our observation regarding the HCMV-induced TNFR2 upregulation.
Although not mentioned at all in the article, the associated supplementary data by Weekes et al. also indicates a significant upregulation of TNFR2 at the plasma membrane (see Supplementary Table 2 of (21)). The fact that the ULb'-positive HCMV strain Merlin was used and primary human foreskin fibroblasts were infected while we used AD169varL and TB40-HCMV infections of MRC-5 and RPE-1 cells, respectively, confirms and generalizes the HCMV-induced TNFR2 upregulation. Intriguingly, we found that this upregulation is strictly associated with the presence of the ULb' gene region, which becomes increasingly established as central hub for the regulation of the members of the TNF receptor superfamily and their ligands. A surprising side observation was the abundant expression of tnfr2 mRNAs in MRC-5 cells, which did not result in detectable TNFR2 surface levels in the absence of HCMV infection. The promoter of the human tnfr2 gene comprises various enhancer elements that respond to numerous transcription factors such as STATs, IRFs, and AP-1, which may explain the expression (see e.g. (64,65)). In addition to immune cells such as regulatory T cells, mesenchymal stem cells (MSCs) express TNFR2 and require TNFR2 for their immune suppressive capacities (66,67). Intriguingly, Zhang et al. recently noticed that MRC-5 cells share several properties with human umbilical cord-derived MSCs (hUC-MSCs) including the promotion of Tregs and the induction of the immunomodulatory molecule IDO in response to IFNg and TNFa (68).
Using a comprehensive panel of HCMV ULb' mutants, we found that deletion of UL148 and UL148D impaired the HCMVmediated upregulation of TNFR2. Intermediate TNFR2 levels were observed upon deletion of either UL148 or UL148D, whereas the combined loss of both genes resulted in an additive loss of TNFR2 surface levels. The fact that HCMV devotes at least two genes to upregulate TNFR2 indicates an important biologic relevance. Both pUL148 and pUL148D are expressed with early/late kinetics (21) which is in line with our observation that GCV treatment decreased HCMV-induced TNFR2 upregulation. So far, functional characterization of the UL148D gene region mainly focused on the role of hcmv-miR-UL148D, which targets the ERN1, ACVR1B, IEX1, and CCL5 mRNAs blocking apoptosis (69-73). In their global assessment of RNA-induced silencing complex (RISC)-associated mRNAs in Normal human dermal fibroblasts (NHDF) infected either with the ULb'-negative AD169varATCC or the ULb'-positive HCMV TR, Pavelin et al. did not observe a relevant change of the TNFRSF1B mRNA in RISC complexes (see supplementary data of (74)), arguing against HCMV-encoded or induced miRNAs directly targeting tnfr2 mRNA. Although pUL148D is expressed (21) and is not disrupted in so far sequenced HCMV genomes (75), it awaits further in-depth molecular characterizations. In clear contrast to pUL148D, different groups investigated functions of pUL148. The Kamil laboratory showed that pUL148 is an ER-resident transmembrane glycoprotein which alters the tropism of HCMV (76). The pUL148 interacts with SEL1L, activates the unfolded protein response (77), reorganizes the ER (78), and influences ERassociated degradation (ERAD) (79). TNFR2-mediated signaling induces TRAF2 degradation by a mechanism relying on the ubiquitin E2 ligase Ubc6 (80), which is also important for the ubiquitination of target proteins in ERAD (81), and ER stress leads to increased TNFR2 levels based on an association with progranulin (82). This raises the intriguing question if the role of pUL148 in the ER and the TNFR2 upregulation are mechanistically coupled.
Wang et al. published mass-spectrometry data, for which cells had been infected with HCMV either expressing or lacking UL148 (61). Among the top five hits of differentially regulated proteins, their supplementary data showed disintegrin and metalloproteinase domain-containing protein 17 (ADAM17; DUL148/wt fold ratio: 4.05 (61)), which is also known as tumor necrosis factor-aconverting enzyme (TACE). TACE induces TNFR2 shedding (62,63). For pUL148, we confirmed this down-regulation of TACE. Furthermore, we extended the mechanism of TACE downregulation to pUL148D. Thus, the pUL148/pUL148D-mediated increase in TNFR2 surface levels may be explained, at least in part, through their decreasing effect on TACE. One intriguing question is why HCMV regulates TNFR2. Apparently, it is not a general phenomenon of betaherpesvirinae because HHV-7 seems not to upregulate TNFR2 levels (83). The facts that HCMV does not replicate in mice, and that MCMV decreases surface levels of TNFR1 and TNFR2 (84) rule out in vivo experiments in mice to address immune suppressive aspects of the CMV-induced TNFR2 upregulation. TNFR1 and TNFR2 both recognize TNFa (see e.g (85)). For certain viruses, TNFa mediates its antiviral activity through TNFR1 rather than through TNFR2 (86,87). Thus, TNFR2 upregulation as a sequestration mechanism diverting TNFa from TNFR1 to TNFR2 may appear as a plausible argument at first glance. However, we and others showed that HCMV encodes pUL138 which upregulates TNFR1 surface levels (27,45,46). Therefore, it is highly likely that additional reasons beyond withholding TNFa from TNFR1 drove HCMV to express at least two gene products upregulating TNFR2. In the case of HIV, TNFa signaling via TNFR2 inhibits the viral entry into primary tissue culturedifferentiated macrophages (88,89). For non-mammalian viruses, a role of fish TNFa favoring virus replication has been described (90). In case of the rhabdovirus SVCV, TNFa signals through its receptor TNFR2 to enhance viral replication (91). Thus, TNFR2 can in principle elicit antiviral and proviral signals. However, global CRISPRi and CRISPRn screens for host factors affecting HCMV infections (conducted with HCMV-Merlin in HFFs) neither showed significant advantages nor disadvantages regarding HCMV infection associated with a loss of TNFRSF1B coding capacity (see supplementary data set of (92)). This argues against direct effects of TNFR2 on HCMV replication in fibroblasts. Accordingly, a siRNA screen of factors modulating HCMV replication comprised TNFRSF1B but also did not observe a relevant increase or decrease in virus replication (see Supplementary Tables 1-3 in (93)). The interpretation that HCMV does not require TNFR2 for replication in cell culture systems such as fibroblasts may explain why fibroblast-adapted HCMV strains such as AD169varS have lost the ULb' region comprising the two TNFR2 regulators. If the TNFR2 upregulation would be critical for HCMV replication in fibroblasts, the replacement of the ULb' region should have been counter selected.
In the absence of pathogens and tumors, TNFR2 expression is usually restricted to cells of the immune system. Accordingly, TNFR2-selective agonists have been mostly applied to lymphocytes (94,95). It is not trivial to anticipate which TNFR2 downstream signaling events and gene expression changes occur in HCMV-permissive cells and might be advantageous for HCMV. However, TNFR2 "makes TNF a friend of tumors" (96), e.g., by protecting malignant cells from DNA damage (97). Furthermore, TNFR2 engagement on endothelial progenitor cells leads to increased expression of anti-inflammatory mediators such as IL-10, TGFb, and HLA-G (98). Thus, the upregulation of TNFR2 on HCMV-infected cells appears as another example of immunoregulatory functions that are shared by tumors and viruses. It will be interesting to investigate the exact proviral function(s) resulting from TNFR2 upregulation in the futuremaybe by using herein described HCMV mutants.

Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

Author contributions
VTKL-T, FM, and BK performed research. VTKL-T, FM, HH, and MT analyzed data. VTKL-T, HH, and MT interpreted data. VTKL-T and MT supervised the project. VTKL-T and MT wrote the manuscript. All authors contributed to the article and approved the submitted version.