Paracrine inhibition of GM-CSF signaling by human cytomegalovirus in monocytes differentiating to dendritic cells

Human CMV (HCMV) is a β-herpesvirus that infects most of the human population during their lifetime. HCMV opportunistically exploits its host immune defects to replicate and spread, so causing severe disease. Transplant recipients, fetuses, and newborns are all at risk of disease or death because of primary infections or virus reactivation. HCMV also encodes a variety of proteins that disrupt TCD4⁺ and TCD8⁺ antiviral responses; the latter are critical for controlling the spread of virus and preventing associated diseases (for a review see Crough and Khanna¹ ). Dendritic cells (DCs) are the sites of primary infection and are essential for capturing the virus Ags needed to mould an appropriate CD8⁺ T-cell repertoire against the virus. Thus, DCs are main targets of virus subversion. DCs infected with HCMV have defects in all stages of Ag presentation, migration to secondary lymphoid organs, and T-cell activation (reviewed in Varani et al² ). We have developed a working hypothesis that these virus-induced changes can be overcome by uninfected DC cross-presenting Ags to CD8⁺ T cells.³  Monocytes are primary targets of HCMV, can harbor latent virus, and are a source of inflammatory DCs in vivo.^4-6  We therefore assumed that DC dysfunction was because of abnormalities arising during the differentiation of their monocyte precursors. DCs derived from infected monocytes in vitro and in patients who received a transplant experiencing an active infection^7,8  all have abnormal phenotypes and function that may contribute to immunosuppression. However, the molecular and cellular mechanisms underlying these defects are poorly documented. We have obtained evidence that the infection of monocytes with HCMV results in defective GM-CSF signaling, with inhibition of STAT5 phosphorylation that leads to dysfunctional derived DCs. These DCs had lost critical features such as the ability to potentiate T_H1 differentiation and proliferation because of secreted factors such as IL-6 that is essential for suppressors of cytokine signaling 3 (SOCS3) synthesis. We have therefore provided new insights into the way HCMV infections cause immunosuppressive damage and the ensuing side effects.

DCs were generated from monocytes, and cell cultures were phenotyped as described in supplemental Methods (available on the Blood Web site; see the Supplemental Materials link at the top of the online article).

The endothelial cell–adapted strain HCMV_{TB40/E GFP} was provided by M. Messerle (Hanover, Germany) and HCMV_VHLE was provided by C. Sinzger (Tubingen, Germany). HCMV BAC_TB40/E GFP (green fluorescent protein) was designed and generated by E. Borst (Hanover, Germany) from a HCMV BAC_TB40/E genome⁹  in which the US1 to US7 genomic fragment was replaced by a GFP cassette under the control of mouse MIEP.

Viruses were amplified on MRC5 fibroblasts and collected by centrifugation at 100 000g for 35 minutes. Virus titers (PFU/mL) were determined on MRC5 cells. Viruses were inactivated by UV irradiation for 20 minutes at 0.17 Amp with the use of a Spectroline device.

Monocytes in differentiation medium were either mock-infected (medium) or infected with live or UV-inactivated virus at a MOI of 5. Alternatively, culture medium was supplemented with human recombinant IL-6 (25-50 ng/mL; Ozyme), anti–IL-6 or anti–IL-6R blocking Abs, or IgG isotype control (15 μg/mL; Gen-Probe Diaclone), and phosphonoacetic acid (PAA; 250 μg/mL; Sigma-Aldrich). CMV-MoDC defines cells produced by the differentiation of infected monocytes and IL-6-MoDC defines cell supplemented with IL-6. In experiments with conditioned medium, the culture medium was a mixture of fresh differentiation medium and the medium to be tested (1/1; vol/vol). When applicable, culture media were cleared of infectious virus by centrifugation (100 000g for 35 minutes).

Cells were treated, stained, and analyzed by flow cytometry on a FACSCalibur cytometer (BD Biosciences; supplemental Methods).

Electrophoresis of cell lysates and protein analysis by immunoblotting was performed with standard procedures (supplemental Methods).

PCRs were performed on reverse-transcribed total RNA as detailed in supplemental Methods.

MRC5 cells were stained with PKH26 (Sigma-Aldrich) and cultured with TNF-α (1 μg/mL) and cycloheximide (25 μg/mL; Sigma-Aldrich) in complete DMEM medium for 24 hours to obtain apoptotic bodies. Apoptotic MRC5 cells were cocultured with DCs (1 DC for 2 MRC5 cells). Cells were harvested and stained with an FITC-conjugated anti–HLA-DR Ab, and double-positive cells were quantified on a FACSCalibur device. The percentage of PKH26⁺/HLA-DR⁺ MoDCs was considered as 100% of phagocytic activity. Alternatively, the Vybrant phagocytosis assay kit (Escherichia coli bioparticles; Invitrogen) was used.

Secreted inflammatory cytokines were detected with a dedicated Ab array from Raybiotech. TNF-α, IL-6, and IL-12p70 were measured with a cytometric bead array assay (CBA; BD Biosciences) and the soluble form of GM-CSF receptor (sGMRα) by ELISA (Gen-Probe Diaclone).

Purified monocytes were treated with Lipofectamine RNAiMax (Invitrogen) and 50nM RNAi was directed against SOCS1 (3′-UCGAAGAGGCAGUCGAAGCUCUCGC-5′) or SOCS3 (3′-AGUAGAUGUAAUAGGCUCUUCUGGG-5′) mRNA (Invitrogen) or control RNAi (Invitrogen). Corresponding protein expressions were checked by Western blot analysis with the use of anti-SOCS3 or anti-SOCS1 Abs (Abcam).

Naive CD4⁺CD45RA⁺ T cells were isolated from PBMCs with the use of a magnetic bead separation system (naive CD4⁺ T-cell isolation kit; Miltenyi Biotec). Allogeneic stimulators were infected with HCMV or were treated with anti–IL-6 or isotypic Abs during differentiation and then with ultrapure lipopolysaccharide (LPS; Invivogen) for 24 hours, washed with culture medium, and cocultured for 5 days in 96-well plates with T cells (95%-98% purity) at a ratio of 1:10 (1 × 10⁴ allogeneic stimulators for 1 × 10⁵ T cells/well in 200 μL). IL-2 (50 U/mL; PeproTech) was then added to each well, and culture was continued for a further 3 days. For T_H2 polarization experiments, IL-4 (5 ng/mL; PeproTech) and anti–IFN-γ Ab (10 μg/mL; BD Biosciences) were added to cocultures. Intracellular cytokines were assayed on day 8 (see supplemental Methods). In CFSE experiments, naive T cells were stained with CFSE (Invitrogen), cocultured for 5 days at a ratio of 1 stimulator to 10 T cells as above, and T-cell proliferation was analyzed.

The project was approved by the ethics committees of the Rangueil Hospital (Toulouse) and Besançon Hospital (Besançon), and all patients who provided samples gave their written informed consent in accordance with the Declaration of Helsinki. PBMCs were isolated from patients' blood samples with the use of a Ficoll density gradient, and CD14⁺ cells were sorted as described in supplemental Methods: “Generation and culture of monocytes derived DCs.” Cells were differentiated to MoDCs for 5 days, and their CD1a and pSTAT5 contents were determined. Secreted IL-6 in blood plasma and MoDC supernatants were assayed with CBA. HCMV DNA in patients' blood was quantified by real-time PCR.¹⁰ 

Values are expressed as the means ± SDs of ≥ 3 independent experiments. Data were analyzed for significance with the Mann-Whitney U test or Student t test, and a P value ≤ .05 was considered significant. All data were analyzed with Prism Version 5 (GraphPad) and FlowJo Version 7.6.5 (TreeStar) software.

Adherent monocytes obtained from the peripheral blood of healthy donors were either mock-infected or infected with HCMV (VHLE; MOI of 5). The resulting MoDCs and CMV-MoDCs were obtained by differentiation in a culture medium containing GM-CSF and IL-13 for 5 days. IL-4 treatment always generated the same phenotype as IL-13 (not shown). The MoDCs had the phenotype of immature DCs bearing CD1a (95%) and CD1c (98%), CD80 (52%) and CD86 (53%), but no CD14 (0%) or CD83 (3%). Only 10% of infected MoDCs expressed CD1a, and only 19% had CD1c, whereas most of them had the maturation marker CD83 (83%) and had increased CD80 (85%) and CD86 (97%; Figure 1A). Identical results were obtained when monocytes were positively selected from PBMCs with the use of CD14 magnetic beads (data not shown). Alternatively, the HCMV inoculum was UV-inactivated before adding it to monocytes to determine whether the phenotype modifications required live virus. UVCMV-MoDCs had the regular MoDC phenotype (not shown), indicating that de novo synthesis of virus proteins was necessary for production of the modified phenotype. Surprisingly, monocytes are poorly permissive to HCMV, but most CMV-MoDCs were modified. We therefore measured the percentage of infected cells with an HCMV BAC construct encoding a GFP. Approximately 50% of CMV-MoDCs contained GFP-labeled fluorescence, whereas 82% of them were positive for CD83 and 6% were positive for CD1a compared with 11% (CD83) and 90% (CD1a) of MoDCs (Figure 1B). These data suggest that infected cells inhibited their uninfected counterparts by secreting soluble factors. We therefore cultured monocytes for 5 days in standard medium (GM-CSF, IL-13) supplemented with supernatants recovered from 5-day cultures of MoDCs, CMV-MoDCs, and UVCMV-MoDC. Most of the cells generated from monocytes that had been exposed to medium conditioned by CMV-MoDCs expressed no CD1a (11% of CD1a⁺ cells), whereas those exposed to supernatants from UVCMV-MoDCs were unchanged (Figure 1C). Thus, monocytes infected with live HCMV secrete soluble factors that alter their differentiation into standard immature DCs.

We checked that monocytes that differentiated in the absence of GM-CSF gave rise to DCs bearing no CD1a, as previously reported¹¹  (Figure 2A). We then investigated whether CMV-MoDC phenotype could be because of, at least in part, a defect in GM-CSF signaling. Because GM-CSF triggers STAT5 activation,¹²  which requires the phosphorylation of tyrosine at position 694, we measured the phosphorylated STAT5 (pSTAT5) in CMV-MoDCs by Western blot analysis. The pSTAT5/total STAT5 ratio in DCs generated in the absence of GM-CSF (ΔGM-MoDCs) was lower than that of standard MoDCs (Figure 2B). This was also true for CMV-MoDCs and monocytes differentiated in the presence of CMV-MoDC supernatants. The altered phenotype of CMV-MoDCs could therefore be because of a blockage of GM-CSF signaling, perhaps by soluble factors secreted by infected monocytes during their differentiation. STAT5 phosphorylation was not inhibited in cells treated with UVCMV, confirming that blocking STAT5 activation requires live CMV. We substantiated the impairment of STAT5 activation in CMV-MoDCs by flow cytometry to detect pSTAT5 in cells infected with BAC-GFP (Figure 2C). This again pointed to HCMV impairing monocyte differentiation via a paracrine effect. Inhibiting virus replication with PAA, which inhibits late viral protein synthesis (supplemental Figure 1), did not reverse the inhibition of STAT5 phosphorylation by HCMV, suggesting that immediate early (IE) or early (E) virus products are involved (Figure 2D).

Cytometric analysis showed that there was more GM-CSF receptors on the cell surface of CMV-MoDCs than on MoDCs at day 2 of differentiation (supplemental Figure 2A). Hence, impaired surface expression of the GM-CSF receptor could not cause the defective STAT5 signaling in CMV-MoDCs.

The secretion of a sGMRα by monocytes can be stimulated by an inflammatory stimulation.¹³  We therefore measured the sGMRα in the culture media of MoDCs, CMV-MoDCs, and UVCMV-MoDCs by ELISA. The medium from CMV-MoDCs contained much more sGMRα than that from MoDCs. Inactivation of CMV by UV irradiation resulted in the concentration of sGMR being similar to that of controls, again indicating that live virus is required for the modified phenotype of CMV-MoDCs (supplemental Figure 2B). However, the concentration of sGMRα in the supernatants of MoDCs was as high as those of CMV-MoDCs in some experiments, although the phenotype of the cells was unchanged (data not shown), indicating that other factors were involved.

We assayed the inflammatory cytokines in culture supernatants of MoDCs and CMV-MoDCs with the use of an Ab array to investigate the soluble factors responsible for the defective STAT5 activation. Infected monocytes secreted higher concentrations of IL-6 and IL-8 than did MoDCs after 2 days in culture (Figure 3A). Infected monocytes also secreted more MIP-1β, sTNFRII, and MCP-1, although to a lesser extent. The IL-6 in the culture medium of cells allowed to differentiate for 1-5 days was also assayed by CBA. IL-6 was undetectable in supernatant from MoDCs throughout the 5 days and increased slightly in supernatant from UVCMV-MoDCs from day 1 (1.8 ng/mL) to day 4 (2.9 ng/mL). In contrast, the IL-6 in the supernatant from CMV-MoDCs increased dramatically from 5.7 ng/mL (day 1) to 37 ng/mL (day 5). Because IL-6 influences monocyte differentiation and DC function,^14,15  we examined its influence on the CMV-MoDC phenotype and function. Monocytes differentiated in standard medium supplemented with IL-6 (IL-6-MoDC) had no CD1a or CD1c on their surfaces (Figure 3B). STAT5 phosphorylation was inhibited in cells cultured in medium containing IL-6 concentrations similar to those in the supernatants of CMV-MoDCs (Figure 3C). We monitored STAT5 phosphorylation in CMV-MoDCs differentiated in medium containing anti–IL-6 Ab to confirm that IL-6 influenced the acquisition of the CMV-MoDC phenotype (Figure 3D). STAT5 was phosphorylated in cells when IL-6 was blocked, indicating that IL-6 is important in the paracrine inhibition of STAT5 signaling in CMV-MoDCs.

SOCS plays a key role in inhibiting the Jak/Stat pathways.¹⁶  We therefore monitored SOCS1 and SOCS3 synthesis in MoDCs, CMV-MoDCs, and IL6-MoDCs. The concentrations of mRNAs encoding SOCS1 and SOCS3 were higher in CMV-MoDCs and in IL-6-MoDCs than in MoDCs (Figure 4A). We then blocked IL-6 signaling with Abs directed against the IL-6 receptor to determine how IL-6 inhibited the GM-CSF–mediated activation of STAT5 in CMV-MoDCs. Secreted IL-6 stimulated the synthesis of SOCS3 but not SOCS1 (Figure 4B). Moreover, treating CMV-MoDCs with SOCS3 small interfering RNA restored the phosphorylation of STAT5, whereas siSOCS1 did not (Figure 4C-D; supplemental Figure 3). These data strongly suggest that the IL-6–induced synthesis of SOCS3 is at least partially responsible for blocking GM-CSF–mediated STAT5 signaling in CMV-MoDCs.

Abnormal GM-CSF activity had been reported to decrease the phagocytic activity of neutrophils¹⁷  and MoDCs.¹⁸  We have previously shown that the phagocytosis of infected dead cells by MoDCs is crucial for Ag cross-presentation to CD8⁺ T cells.¹⁹  We have now investigated the phagocytic function of CMV-MoDCs by adding PKH26-labeled dead cells to MoDCs, CMV-MoDCs, ΔGM-MoDCs, IL-6–MoDCs, (SN-MoDC)–MoDCs, and (SNCMV-MoDC)–MoDCs. The phagocytic activity of CMV-MoDCs and ΔGM-MoDCs was only 20% that of MoDCs, suggesting that GM-CSF signaling also plays a role in this process (Figure 5A). The CMV-induced IL-6–mediated inhibition of STAT5 accounted at least in part for this defect in CMV-MoDCs because the phagocytic capacity of IL-6–MoDCs was ∼ 50% that of MoDCs. However, although adding conditioned medium from CMV-MoDCs to MoDCs mimicked the effect of IL-6 (50% reduction; Figure 5B), neutralization of IL-6 with anti–IL-6 Abs did not restore phagocytosis. Thus, IL-6 may not be the only secreted factor involved in this effect. Assays of phagocytosis which used E coli bioparticles gave similar results (data not shown). Thus, the defective phagocytosis of dead cells by CMV-MoDCs probably involves both autocrine and paracrine inhibition of STAT5 signaling, some of which involves IL-6.

DCs and their environment help determine the programming of T cells toward T_H1 or T_H2 subsets. We have examined how the polarization of naive CD4⁺ T cells is influenced by CMV-MoDCs. One of the mechanisms by which DCs drive T_H1 differentiation involves IL-12p70. It is generally agreed that the secretion of IFN-γ by differentiated T cells correlates with the amount of IL-12 secreted by stimulated MoDCs²  in mixed lymphocyte reactions (MLRs),²⁰  whereas IL-4 production promotes T_H2 differentiation. We measured the amount of IL-12p70 secreted by MoDCs, CMV-MoDCs, and IL-6–MoDCs that had been stimulated with LPS, a known activator of IL-12 secretion. CMV-MoDCs were much less responsive (67 pg/mL) than MoDCs (380 pg/mL), whereas IL-6–MoDCs produced almost no detectable IL-12p70 (Figure 6A). No DC population secreted any IL-12p70 without LPS stimulation. The production of IL-12p70 by stimulated MoDCs was inhibited by adding CMV-MoDC supernatants. However, neutralization of IL-6 did not restore IL-12p70 production, suggesting that IL-6 is not the only factor involved (Figure 6B). We used MLR experiments to study T_H polarization by DCs. Unactivated and LPS-activated MoDCs, CMV-MoDCs, and IL-6-MoDCs were washed and cocultured for 8 days with allogeneic naive T cells. The T_H1 and T_H2 responses were assessed by flow cytometry by measuring the IL-4 and IFN-γ produced by T cells. MoDCs promoted the production of IFN-γ in 19.6% and LPS-stimulated MoDCs in 33.0% of T cells (Figure 6C), correlating the ability of MoDCs to potentiate the T_H1 response with IL-12p70 secretion. In contrast, LPS-stimulated CMV-MoDCs induced IFN-γ secretion in very few T cells (6.56%). Few T cells secreted IFN-γ when incubated with LPS-unstimulated CMV-MoDCs, suggesting that CMV-MoDCs are intrinsically unable to prime a T_H1 response (data not shown). IL-6–MoDCs treated with LPS also showed a reduced capacity to prime T_H1 (Figure 6C), suggesting that IL-6 alters the ability of MoDCs to polarize T cells toward T_H1. T cells stimulated with MoDCs that had been treated with CMV-MoDC conditioned medium (SN-CMV-MoDCs) and LPS formed fewer IFN-γ–producing T cells (18%) than did those stimulated with supernatants from MoDCs (28.0%). The rate of T_H1 formation was restored (33.9%) by blocking the IL-6 in these supernatants (Figure 6D). The IL-6 secreted by CMV-MoDCs during differentiation therefore prevents them from stimulating T cells to form T_H1. None of the DC populations induced detectable IL-4– (Figure 6) or IL-17– (data not shown) secreting cells after stimulation with LPS. This indicates that CMV-MoDCs do not stimulate naive T cells to form either T_H2 or T_H17, although they were able to differentiate toward T_H2 when stimulated with MoDCs treated with IL-4 and anti–IFN-γ Abs (supplemental Figure 4). Besides impaired T_H1 commitment, incubation of naive T cells with LPS-stimulated CMV-MoDCs or MoDCs conditioned with SN-CMV-MoDCs blocked their proliferation (Figure 6E). Thus, the infection of monocytes with CMV causes profound changes in the ability of CMV-MoDCs to promote T_H1 proliferation and activation, partly because of secreted IL-6.

We investigated the relevance of these observations in vivo with the use of monocytes from patients who received a transplant experiencing CMV viremia (Table 1). The monocytes were differentiated into DCs (MoDCs). The MoDCs from viremic patients formed cell clusters (Figure 7A), a feature indicating cell activation as observed in cultured CMV-MoDCs. The medium recovered after 5 days of differentiation contained no detectable viral genome, suggesting the absence of virus replication. Significantly fewer MoDCs from CMV-positive patients bore CD1a than those from CMV-negative patients (Figure 7B). IL-6 was quantified in the growth medium and blood plasma, and the phosphorylation of STAT5 in MoDCs was analyzed. Patients 7 and 8 had very high virus loads (480 000 and 340 000 copies/mL, respectively); these were correlated with large amounts of IL-6 secreted by their MoDCs and in their plasma (Figure 7C-D) and few CD1a⁺ cells (Figure 7A). Furthermore, the MoDCs derived from all the viremic patients lost their ability to phosphorylate STAT5 on differentiation (Figure 7E). These data reinforce the in vitro findings of the role of IL-6 and the subsequently defective GM-CSF signaling in DCs derived from infected monocytes.

Infection by HCMV is known to disturb both innate and adaptive immune responses. DCs are main virus targets to prevent the stimulation of antiviral T cells. However, this is always incomplete because the virus immune evasion functions are circumvented by the host.³  Because monocytes are the precursors of DCs in vivo and HCMV may infect them to further establish latency, we examined the effects of infection on monocyte differentiation and DC dysfunction.

Monocytes were differentiated into DCs by the coordinated action of GM-CSF and IL-13, which models the phenotype and function of inflammatory and migratory DCs in vivo.²¹  MoDCs infected by HCMV lost their standard CD1a⁺ immature phenotype and acquired the maturation marker CD83, probably because of the bystander effect of TNF-α, a known inducer of DC maturation, secreted by infected cells in the early phase of differentiation. Discrepancies with published data⁷  about both cell maturation and the requirement for live virus could be partly because of the use by others of the laboratory strain AD169 that has a 19-kbp genome deletion not found in clinical strains. CD1a and CD1c bind and present lipid Ags to T cells,²²  in addition to being DC phenotype markers. Their expressions are subnormal in CMV-MoDCs and in infected DCs,²³  but their role in controlling HCMV infection remains to be studied. We have found evidence for a powerful paracrine mechanism that leads to modification of the phenotype of the whole population of CMV-MoDCs. Even though only a fraction of the cells contained the IE/E virus proteins, this viral protein expression seems to be critical for initiating this process. We considered the possibility that there was a failure of GM-CSF signaling in CMV-MoDCs because GM-CSF is needed to generate inflammatory DCs from monocytes in vivo²¹  and because there are functional and phenotypic defects (CD1a⁻) in MoDCs generated in the absence of GM-CSF.¹¹  We showed that STAT5 activation was inhibited in the whole population of CMV-MoDCs, which may block GM-CSF signaling, and that this was because of soluble factors secreted by a fraction of infected monocytes.

Our first hypothesis was that the soluble GM-CSF receptor sGMRα blocked the binding of GM-CSF to its receptor. The significantly increased sGMRα concentration in some preparations of CMV-MoDCs pointed to its implication in impairing STAT5 signaling. However, individual variations in the concentrations of constitutive sGMRα in monocytes and our inability to neutralize sGMRα in culture supernatants have made it impossible to decide on the effect of sGMRα at this stage. One of the inflammatory cytokines induced by HCMV, IL-6, is suspected to be a key factor in the failure of organ and BM transplantations.²⁴  IL-6 has also been identified in the HCMV secretome of endothelial cells as a mediator of angiogenesis and cell survival²⁵  and in tumor cells as a paracrine factor in tumor progression and vascularization through activation of the STAT3 cascade.²⁶  We find that DCs derived from the differentiation of monocytes supplemented with IL-6 bore no CD1a or CD1c. Moreover, neutralizing IL-6 during the differentiation of infected monocytes restored phosphorylation of STAT5, suggesting that IL-6 blocks GM-CSF receptor signaling. Previous data have shown that IE1 increases the activity of the IL-6 promoter,²⁷  which may explain why UV-inactivated virus barely increased IL-6 secretion and had no effect on STAT5 phosphorylation. Thus, the secretion of IL-6 by UVCMV is initiated by the activation of NF-κB but cannot be maintained because of blockade of IE1 transcription after UV irradiation.

The inhibition of STAT5 phosphorylation in DCs generated from monocytes involves SOCS3 and SOCS1,¹⁸  and SOCS1 can abolish GM-CSF signal transduction in DC precursors²⁸  but not in conditions of viral infection. Furthermore, induction of SOCS3 by IL-6 is crucial for the regulation of the inflammatory response, whereas the IL-6–mediated induction of SOCS3 could inhibit insulin receptor signaling.^16,29  We have now obtained evidence that IL-6 secreted by CMV-MoDCs is crucial for the induction of SOCS3, which in turn leads to the inhibition of STAT5 phosphorylation that may reflect blockage of the GM-CSF receptor signaling by SOCS3 but not SOCS1.

The influence of STAT5 blockage by SOCS on the phagocytic activity of MoDCs was deduced¹⁸  from the inability of cells to internalize dextran. Defects in the phagocytosis of dextran and yeast particles by CMV-MoDCs have been reported,⁷  but they have little physiologic relevance. The sequestration of tumor and viral Ags through phagocytosis of dead cells by DCs is a key step in T-cell activation.^30,31  We examined the capacity of CMV-MoDCs to ingest apoptotic and necrotic bodies prepared from infected fibroblasts because this phagocytosis is a critical step in the cross-presentation of HCMV Ags to CD8⁺ T cells.¹⁹  The inhibition of this process involved soluble factors secreted by CMV-MoDCs, but IL-6 was not the only factor involved. It has been reported that IL-6 can cause monocytes to differentiate to macrophages rather than DCs,¹⁵  but no information was provided about their phagocytic capacities. IL-6 can also reactivate latent HCMV in a model of primary monocytes infected in vitro,³²  which makes our findings also relevant to the differentiation of latently infected monocytes. Giving GM-CSF to patients with carcinoma improved their clearance of apoptotic cells,³³  suggesting a link between the activation of STAT5 and the expression of genes promoting phagocytosis. Because PU.1, a transcription factor induced by pSTAT5, drives such genes,³⁴  we looked for PU.1 mRNA and protein in CMV-MoDCs, but their expressions were similar to those in MoDCs (data not shown). CD36 and αvβ5 integrin could be involved in the uptake of apoptotic cells by DCs,³¹  but we observed no alteration in CD36 in DCs (data not shown). The αvβ5 integrin remains to be explored. Because inhibition of phagocytosis could provide the virus with a new way of escaping the host immune response, it might be worth exploring the capacity of CMV-MoDCs to cross-present viral Ag to CD8⁺ T cells.

We performed MLRs with CMV-MoDCs as stimulators because IL-6 influences T-cell activation and the commitment of naive T cells to form TH₁ or TH₂. Cells were treated with LPS, which induces APCs to secrete IL-12 and is usually a prerequisite for naive cells to produce IFN-γ. Our data showing that HCMV inhibited IL-12p70 secretion by LPS-activated CMV-MoDCs extend previous reports that used infected DCs from adults³⁵  and neonates.³⁶  The defect in IL-12p70 secretion by CMV-MoDCs was not because of IL-6, even though SOCS1, a negative regulator of both IL-12 synthesis^37,38  and T_H1 promotion,³⁹  was induced. Nevertheless, the secretion of IFN-γ by T cells was restored when the IL-6 in supernatants of CMV-MoDCs was neutralized before adding it to T cells, suggesting the dissociation of IL-12p70 secretion from T_H1 commitment as reported in experiments that implicated Notch signaling.⁴⁰  We have demonstrated the capacity of HCMV to impair the commitment of T cells to form T_H1, by a mechanism that may disturb the overall host response. Protein array studies have reported that CMV-MoDCs secreted more MIP-1β and IL-6 than did MoDCs. The chemokine MIP-1β reduces the polarization of T cells to T_H1 when added to LPS-stimulated MoDCs in MLRs.²⁰  Thus, HCMV may interfere with the polarization of naive T cells via MIP-1β and the resulting ERK activation, but this needs further investigation. The negative regulation of GM-CSF receptor signaling by a Src-like adaptor⁴¹  and the disruption of DC metabolism because of SOCS3 blocking M2 pyruvate kinase⁴²  have yet to be investigated. PPARγ activation may be involved in the alterations in DC phenotype and function⁴³ ; this is being investigated because we have shown that HCMV activates PPARγ.⁴⁴ 

The relevance of our in vitro observations is supported by data obtained from naturally infected MoDCs derived from viremic patients whose high virus loads were correlated with the absence of CD1a, confirming previous observations.⁸  The blockage of STAT5 phosphorylation in MoDCs and the high IL-6 contents of both MoDC supernatants and patients' blood plasma further suggest abnormal GM-CSF signaling because of HCMV. Because MoDCs from patients produced no virus suggests that the ability of monocytes infected in vivo to differentiate into standard DCs was modified. HCMV may enhance the host's susceptibility to secondary bacterial infections, because GM-CSF is essential for neutrophil function and proliferation. These findings provide new insights into how HCMV infection impairs the overall innate and adaptive responses, and these should be taken into account when using immunomodulatory therapies that are based on GM-CSF to treat infectious diseases in solid-organ transplant recipients,⁴⁵  as in prophylaxis against sepsis in preterm neonates⁴⁶  and in fighting cancer.⁴⁷ 

The authors thank all the patients who donated blood, S. Chavanas, E. Champagne, and K. Leghmari for critically reading the manuscript, and F. L'faqihi and V. Duplan for technical assistance in flow cytometry. The English text was edited by Dr Owen Parkes.

This work was supported by institutional grants from Inserm. J.C. is supported by a Fondation pour la Recherche Médicale fellowship.

Contribution: J.C. designed and performed experiments, analyzed data, and wrote the manuscript; H.M. designed experiments; B.M. prepared HCMV-BAC; B.R. analyzed data; C.M. performed virologic diagnosis; H.W. collected patients and analyzed data; A.C., C.V., P.R., N.K., L.R., and G.H. collected patients; C.D. designed experiments, analyzed data, and wrote the manuscript; and all the authors approved the final manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Christian Davrinche, Inserm 1043, CPTP Bat B, CHU Purpan, Toulouse, 31024 France; e-mail: christian.davrinche@inserm.fr.