Langerhans-type dendritic cells (LC) are the only type of innate immune cells found in the epithelial layer of human oral mucosae, where they provide an initial line of defense against invading microorganisms (Barrett et al., 1996; Allam et al., 2003, 2006, 2008; Cutler and Jotwani, 2006; Novak et al., 2008). Acquisition of human cytomegalovirus (CMV), a source of serious disease in newborns, transplant recipients, and AIDS patients, normally occurs early in life by contact between contaminated bodily fluids, such as urine and saliva, and host oronasal mucosae. Limited replication within mucosal cells at the time of infection is followed by viral spread to the bone marrow, where latency is established (Pass, 2001; Britt, 2007; Mocarski et al., 2007). Intermittent bouts of reactivation then release virus back in the urine and saliva, thus promoting viral spread to new hosts.

Being present at densities of 20–80 cells/mm² in the outermost layers of the oral mucosa (Cruchley et al., 1989, 1994; Lasisi et al., 2013; Upadhyay et al., 2013; Omine et al., 2015), LC are among the first cell types to encounter CMV during entry, together with epithelial cells. Their differentiation from CD34⁺ hematopoietic progenitors (Iijima et al., 2007), a well-known site of CMV latency (Minton et al., 1994; Mendelson et al., 1996; Zhuravskaya et al., 1997; Slobedman and Mocarski, 1999; Khaiboullina et al., 2004), was also shown to be associated with viral genome maintenance (Reeves et al., 2005a,b; Reeves and Sinclair, 2010), and reactivation (Reeves et al., 2005a,b; Reeves and Sinclair, 2010; Huang et al., 2012).

Although mucosal LC are conventionally considered immature, their actual maturation status is likely to be more heterogeneous, as the oral mucosa is normally colonized by a variety of microorganisms whose identity and pathogenicity depend on the health and hygiene of the host. Local contacts with infiltrating CD4⁺ and CD8⁺ lymphocytes expressing the CD40 ligand (CD40L), with bacterial products such as lipopolysaccharide (LPS) or with pro-inflammatory cytokines such as interleukin-1β (IL-1β), tumor necrosis factor α (TNF-α), and prostaglandin E2 (PGE2) can indeed stimulate mucosal LC maturation and migration (Yavuzyilmaz et al., 1995; Heasman et al., 1998; Orima et al., 1999; Loro et al., 2001; Orozco et al., 2006).

We previously showed that infection initiates in only a small fraction of in vitro-differentiated immature LC (iLC), as compared to mature LC (mLC) obtained by iLC stimulation with granulocyte-macrophage colony-stimulating factor (GM-CSF), fetal bovine serum (FBS), LPS and CD40L (Lauron et al., 2014; Coronel et al., 2015). Despite this and quite surprisingly, iLC produced higher amounts of viral progeny than mLC (Coronel et al., 2015), suggesting that CMV replication in these cells undergoes at least two blocks: one acting early and restricting infection onset to only specific cells within the population, and one acting later, reducing viral yields in mLC. Exposure of iLC differentiated from latently infected progenitors to the mLC cocktail also failed to enhance viral DNA replication and progeny production, suggesting that signaling by FBS, CD40L, and LPS in combination does not trigger robust CMV reactivation (Coronel et al., 2015).

In this work, we sought to determine: (1) if select stimuli can increase the proportion of LC that initiate infection by relieving the initial block; (2) if these stimuli also increase viral yields; (3) if the magnitude of progeny production can be predicted from the number of viral antigen⁺ cells detected at infection onset; (4) if populations containing large amounts of infected cells are also the most efficient at producing progeny; (5) if mLC's low yields are due to restrictions in progeny production or release; (6) which specific component(s) of the mLC cocktail is responsible for increasing the proportion of infected cells and for reducing viral yields, and (7) if similar restrictions also apply to progeny produced following viral reactivation from latency.

We show that iLC stimulation with the mLC cocktail, with pro-inflammatory cytokines or with CD40L, but not with FBS or LPS, increases the proportion of infected cells. Populations exposed to pro-inflammatory cytokines or CD40L also produced yields similar to those of iLC, while treatment with FBS, LPS, or the mLC cocktail reduced progeny amounts to different extents. A linear correlation was observed between the number of antigen⁺ cells present at infection onset and the amount of progeny produced at peak times, with the notable exception of mLC cultures. As the proportion of infected cells changed over time with different kinetics in each population, an inverse correlation was detected between the amount of progeny produced per cell and the total number of antigen⁺ cells present at late times post-infection, suggesting that conditions which support prolonged viral protein expression may reduce the ability of each infected cell to produce virus.

Populations matured using the mLC cocktail remained the only ones to display large(r) amounts of infected cells but extremely low yields, a phenotype possibly ascribable to the opposing effects of exposure to CD40L, which increased the proportion of infected cells and viral yields, and to LPS, which reduced both parameters. All LC populations contained approximately twice as much cell-associated than cell-free virus, suggesting that defects in progeny release are unlikely to be the source of mLC's low yields. By contrast, the number of cell-associated particles per infected cell ranged from ~0.2 in mLC to ~6 in FBS cultures, implying the existence of a severe block to progeny production in mLC populations. With the notable exception of pro-inflammatory cytokines, none of the tested stimuli increased the extent of viral reactivation beyond that observed in iLC cultures, indicating that the intracellular environments that restrict lytic viral replication also constrain viral reactivation. Together, these data highlight the remarkable functional plasticity of LC: instead of being strongly polarized toward either a fully permissive or a fully resistant phenotype, LC responses to CMV infection and reactivation follow a continuum whereby the strength and extent of the restrictions imposed vary depending on the nature and combination of the stimuli received.

Horizontal CMV transmission is thought to occur by contact between virions and host oronasal mucosae, whose outer layers are exclusively populated by epithelial cells and LC. These cell types are thus likely to play important roles in CMV acquisition and intra- as well as inter-host viral spread.

Both the phenotype and functions of mucosal LC can be strongly modulated by contact with soluble factors, other immune cells, and pathogenic or commensal microorganisms. LC responses to incoming CMV virions are thus likely to vary depending on their degree and mode of activation. In this work, we sought to assess the impact of different activation stimuli on the ability of in vitro-generated LC to support CMV infection and reactivation.

Consistent with our previous data (Hertel et al., 2003; Lauron et al., 2014; Coronel et al., 2015), we found that IE expression in iLC populations is very limited, and ~20-fold lower than what expected for high MOI infections of fully permissive cells such as HFF (Coronel et al., 2015). These findings provide further support to our hypothesis that a block restricting transcription or translation of the UL122/123 genes, encoding the IE1/IE2 proteins, exists in a large proportion of iLC cultures (Lauron et al., 2014). The fact that exposure to select activation stimuli increases the number of infected cells (Figure 1) suggests that some signaling pathways can partially relieve this block. We think that these signals are most likely to act directly on viral gene transcription, rather than on virion entry or intracellular trafficking, as iLC contain more viral particles than mLC at early times post-infection (Lauron et al., 2014), similar numbers of viral genomes reach the nucleus of both cell types (Lauron et al., 2014), and prolonged exposure to the inoculum does not substantially increase the proportion of IE⁺ cells (not shown). The antiviral activities of nuclear domain 10 bodies (ND10) are also unlikely to be the main source of this block, as expression of specific ND10 components such as PML and hDaxx was increased, rather than reduced, in cells exposed to IL-6 (Hubackova et al., 2012), IL-1β (Heuser et al., 1998; Ha et al., 2006), or CD40L (Salomoni et al., 2006). The enhancement of viral gene transcription ensuing signaling by activation stimuli may occur, for instance, if the abundance, availability or activation status of transcription factors such as the major immediate-early promoter activators NF-kB, CREB, and Sp1 (Hunninghake et al., 1989; Sambucetti et al., 1989; Lang et al., 1992) is altered. Alternatively, signaling may affect the degree of chromatinization of viral genomes, or change their epigenetic landscape, tilting the balance in favor of transcription activation by promoting histone acetylation and/or demethylation. Clearly, additional studies are needed to define the exact mechanisms acting in each population, keeping in mind that these are likely to differ, either qualitatively or quantitatively, in cultures exposed to each activation mix.

Similar to other permissive cells, the viral yields of each population (with the exception of mLC) were directly proportional to the amount of infected cells present at infection onset (Figure 1). The calculated number of particles produced by each initially infected cell (0.7 pfu/cell), however, remained substantially lower than that of other cell types (10 or more pfu/cell; Coronel et al., 2015), and could not be increased by LC activation. Also in contrast to other cell types, the efficiency of progeny production per cell was larger in populations containing smaller number of infected cells (Figure 2). This could be due to the fact that the initial pool of progeny-producing cells present in populations with low numbers of IE⁺ cells such as iLC and FBS is not expanded by activation, so that each additional IE⁺ cells contributes little to final yields. Alternatively, activation may lower the ability of each IE⁺ cells to produce particles. In either case, infection progress appears to be restricted in most activated LC populations. These restrictions did not appear to affect progeny release (Figure 3A). Rather, activation with most stimuli led to a sharp decline in the amount of intracellular virus produced by each IE⁺ cell, indicating that one or more steps leading to the production of infectious particles, such as capsid egress from the nucleus and virion assembly in the cytoplasm, may be blocked in these cells (Figure 3B).

Mature LC populations were the most severely affected of all. While containing higher numbers of IE⁺ cells, total yields as well as yields per IE⁺ cell are dramatically reduced. Of the three components of the mLC cocktail, CD40L appeared to exert a positive effect on both IE protein expression and yields, while FBS diminished both, so that the efficiency of progeny production per cell of both treatments were similar to that of iLC (Figure 2). LPS, by contrast, had no effect on the number of infected cells but strongly reduced yields, thus lowering the extent of particle production per cell (Figure 2). Combined, CD40L, and LPS increased the number of IE⁺ cells but not yields, further reducing the efficiency of progeny production (Figure 2). This suggests that the mLC phenotype may arise, in part, from the opposing effects of these two molecules: while CD40L signaling may stimulate transcription or translation of the IE proteins, perhaps by acting on the major immediate early genes promoter, LPS signaling may induce the expression of cellular proteins that inhibit viral progeny production.

Very interestingly, the iLC, FBS, CD40L, LPS, and mLC cocktails appeared to have similar effects on viral progeny production starting from genomes introduced by virions during lytic infection as well as from latent viral DNA (Figure 4). This suggests that the restrictions imposed by each stimulus to viral cycle progression are not strongly affected by the action of the tegument proteins released during entry, and can affect progeny production rates independently of the origin of viral DNA (i.e., whether already present in the nucleus or newly introduced by virions). This is an interesting finding, considering that the number and availability of viral genomes should be higher in directly infected cells. Equally interesting is the fact that iLC cultures can support viral reactivation to a similar extent as mature cells. In keeping with our previous results (Coronel et al., 2015), this suggest that differentiation of latently infected CD34⁺ cells into iLC is sufficient to initiate the process of reactivation without an absolute requirement for activation. This is in agreement with data from Reeves et al. (Reeves and Sinclair, 2013), showing that circulating, immature DC express the UL122/123 transcripts and release low amounts of reactivated virus upon co-culture with human fibroblasts, but are in contrast to other reports showing that LPS treatment of circulating DC (Reeves and Sinclair, 2013) and of monocyte-derived LC (Huang et al., 2012) enhance the frequency of reactivation events. The specific reasons for this discrepancy are unclear, but may be related to the fact that we stimulated iLC with LPS in the presence of GM-CSF, thus potentially triggering different signaling pathways.

The only treatment capable of increasing and/or prolonging the incidence of reactivation above that of iLC was exposure to pro-inflammatory cytokines. Inflammation has been associated with increased CMV reactivation rates in immunocompromised (Fietze et al., 1994; Humar et al., 1999; Reinke et al., 1999) and immunocompetent but critically ill patients (Limaye and Boeckh, 2010; Cook and Trgovcich, 2011), and in naturally or experimentally infected myeloid cells (Stein et al., 1993; Prösch et al., 1995; Hahn et al., 1998; Hargett and Shenk, 2010; Reeves and Compton, 2011; Huang et al., 2012; Reeves and Sinclair, 2013). We believe that these stimulatory effects are due to the up-regulation of UL122/123 genes transcription, as it occurs in other cell types after treatment with TNF-α, IL-1β, IL-6, and PGE2 (Stein et al., 1993; Prösch et al., 1995; Kline et al., 1998; Zhu et al., 2002). To our knowledge, this is the first report showing a stimulatory effect of these cytokines, combined, on CMV reactivation from latency.

LH conceived and designed the experiments; RC, DJ, and LD performed the experiments; LH analyzed the data; JM provided comments and critiques; LH wrote the paper.