Caroline Petitdemange*,#, Christopher Maucourant*, Nadine Tarantino, Juliana Rey, and Vincent Vieillard
Sorbonne Université, UPMC, Inserm U1135, CNRS ERL 8255, Centre d’Immunologie et des Maladies Infectieuses (CIMI-Paris), Paris, France
*These authors contributed equally to this paper.
#Present address: CNRS, Institut Gustave Roussy, Unité Physiologie et Pathologie Moléculaires
des Rétrovirus Endogènes et Infectieux, Villejuif, France.
Correspondence: Dr Vincent Vieillard
Abstract
Dengue virus (DENV) is the most widespread arbovirus worldwide and is responsible for
major outbreaks. The host’s immune response plays a crucial role in controlling this
infection but might also contribute to the promotion of viral spread and
immunopathology. In response to DENV infection, natural killer (NK) cells preferentially produce cytokines and are cytotoxic in the presence of specific antibodies. Here, we
identified that DENV-2 inhibits glycogen synthase kinase 3 (GSK-3) activity to
subsequently induce MHC class-1-related chain (MIC) A and MIC-B expression and
IL-12 production in monocyte-derived dendritic cells (Mo-DCs), independently of the
STAT-3 pathway. The inhibition of GSK-STAT-3 by DENV-2 or small molecules induced MIC-A/B
expression on Mo-DCs, resulting in autologous NK cells of a specific increase in IFN-γ
and TNF-a production, in the absence of direct cytotoxicity. Together, these findings identified GSK-3 as a regulator of MIC-A/B expression and suggested its role in DENV-2
infection to specifically induce cytokine production by NK cells.
Introduction
Dengue virus (DENV) is one of the most common mosquito-borne viruses ; it is mainly
transmitted by Aedes sp. mosquitoes and is presently a threat in more than 120 countries. The
number of Dengue fever cases reported to the World Health Organization (WHO) has increased steadily from an average of less than a thousand cases per year globally in the 1950s to more
than 3 million cases in 2013 with approximately 13,000 fatalities [1]. Although most primary
DENV infections, which is caused by four closely related viruses, DENV-1 to -4 are
asymptomatic or with minor symptoms, DENV-infected patients develop a self-limited febrile illness with a fever-arthralgia-rash syndrome characterized by an abrupt onset of fever and joint pains [2]. Most patients recover rapidly, while a few develop hemorrhagic fever or shock syndrome. Despite the importance of this disease, no DENV-specific therapies are available, and the sole approved DENV vaccine elicits protection in people with prior DENV exposure
but not in naïve individuals [3].
During a natural DENV infection by an infective mosquito, viral particles are introduced into the subcutaneous space, where resident dendritic cells (DCs) are the first cell type to be probably infected [4]. Although the host rapidly mounts a response to control the virus in the dermis, DENV-infected DCs migrate quickly to the lymph nodes favoring the spread and dissemination to the periphery, in which a high level of infectious particles is detected. This acute phase of infection is accompanied by an early type I interferon (IFN) response, and the recruitment of immune cells, like mast cells and neutrophils to develop inflammatory responses [3]. This process is followed by the activation of inflammatory monocytes and natural killer (NK) cells, in association with the development of the first clinical symptoms [5-7].
NK cells constitute the proverbial first line of defense against a variety of viral infections. Viral immune-surveillance by NK cells is activated by cytokine signals, like 12, 15, IL-18, and type I IFNs, which can be released by infected cells or sentinel DCs and macrophages within infected tissues. Once activated, NK cells play both an antiviral and regulatory role via
a balance of inhibitory and activating signals that enable them to detect and lyse virus-infected target cells while sparing normal cells. Functional NK cell activity occurs when stimulatory signals outweigh MHC class-1 inhibition. Several of these activating receptors have been characterized, including NKG2C, NKG2D, and the natural cytotoxicity receptors (NCRs) ; their ligands are diverse and are mainly up-regulated by cellular stress allowing NK cells to eliminate harmful or unhealthy host cells specifically [8-11].
We and others have reported the early protective role of NK cells after natural DENV infection and pointed out its role in regulating immune responses. It was observed that NK cell activation is strongly increased following DENV infection [4,5,12,13], and is associated with mild clinical disease [14]. An extensive phenotypic study also revealed profound modifications in their repertoire, including down-regulation of NKp30 and NKp46 [12] ; whereas, a transient
increase of adaptive NKG2C+ NK cells was concomitantly found in patients infected by
DENV-2 [1DENV-2], as described in some other infections [15,16]. Activated NK cells can inhibit viral infection by killing virus-infected cells and the secretion of gamma interferon (IFN-γ). In DENV-2, a high level of IFN-γ production was observed, without stimulation of cytotoxicity [12], in accordance with the up-regulation of major histocompatibility (MHC) class-1 molecules that occur during DENV and other flavivirus infections [17]. However, it is important to note that CD16 on NK cells has been shown to mediate ADCC activity of antibodies induced by DENV infection [18].
High levels of IFN-γ in DENV infection are correlated with milder disease and higher survival rates in patients with dengue hemorrhagic fever [19]. Consistently, immunocompetent C57BL/6J mice infected by DENV caused a peak of IFN-γ in early post-infection [20], and the depletion of NK cells in DENV-infected mice caused a significant increase in the viral titer in association with a decrease of IFN-γ, and a specific IFN-stimulated gene signature [20].
To figure out the mechanisms that regulate NK-cell responses in DENV-2 infection, we used a co-culture system of activated NK cells with autologous monocyte-derived dendritic cells
infected by DENV-2. Herein, we have shown that DENV-2 induced the surface expression of MHC class-1-related chain (MIC)-A and MIC-B, two ligands of NKG2D, via a glycogen synthase kinase-3 (GSK3)-dependent pathway to favor cytokine production by NK cells.
Results
Infection of Mo-DCs by DENV-2 induces the specific production of cytokines by
autologous NK in the absence of cytotoxicity
Stimulation of fresh isolated CD14+ monocytes from healthy donors results in differentiation
into monocyte-derived dendritic cells (Mo-DCs) after six days of IL-4 and GM-CSF (Supporting Information 1A), as described [21,22]. At 48 hours after infection by DENV-2, up to 28.9% of Mo-DCs were infected (Supporting Information 1B). Except for the DC-SIGN receptor, most of the co-stimulatory molecules were up-regulated in both bystander and infected Mo-DCs (Supporting Information 1C), consistent with previous studies [23,24].
In line with ex vivo data observed with NK cells from DENV-2 infected patients [12], NK-cell degranulation, measured by CD107a expression, was not enhanced in NK NK-cells co-cultured with autologous infected Mo-DCs, no matter the effector/target cell ratio (Fig. 1A), the timeline post-infection (Supporting Information 2A) or post-treatment (five hours vs overnight) (data not shown). In contrast, infected Mo-DCs induced significant production of IFN-γ (p = 0.0003) and TNF-α (p = 0.0031) by autologous NK cells, when compared to uninfected Mo-DCs (Fig. 1B ; Supporting Information 2B). Furthermore, in the presence of highly infected Vero cells, a strong increase in cytokine production was observed whereas, degranulation remained close to the level observed in non-infected target cells (Supporting Information 2C), suggesting that the functional modulations of NK cells by DENV-2 could be not specific of Mo-DCs. Furthermore, the cytokine production by NK cells was inhibited in the presence of baicalin (Fig. 1B), a flavonoid that blocks infection of DENV, when cells are treated prior to infection [25]. More importantly, the production of IFN-γ and TNF-α by autologous NK cells was strongly inhibited
in experiments carried out with trans-well chambers, without modulation on degranulation (Fig. 1C ; Supporting Information 3). Altogether, these data suggest that the production of cytokines by NK cells is associated with cell-to-cell contacts with DENV-2-infected cells.
Mo-DC infected by DENV-2 over-expressed NKG2D ligands, MIC-A, and MIC-B
After infection by DENV-2, expression of MHC class-1 molecules on Mo-DCs remained high (Fig. 2A ; Supporting Information 4A), as described previously for different flaviviruses, and unlike many other viruses that down-regulate MHC class-1 expression on infected cells [17]. Concomitantly, HLA-E expression was significantly increased in DENV-2-infected cells, as described [26,27]. Moreover, the frequency and surface-expression of the MHC class-1-related chain (MIC)-A and MIC-B, two ligands of NKG2D, were significantly induced in DENV-2-infected Mo-DCs (p = 0.0015), but not in bystander Mo-DCs, when compared to non-DENV-2-infected cells (Fig. 2A, 2B). Individually, MIC-A and MIC-B were similarly over-expressed after DENV-2 exposure, both at the cell-surface protein level by flow cytometry and RNA level by qPCR (Supporting Information 4B and 4C). Importantly, the pre-treatment of Mo-DCs with baicalin significantly decreased MIC-A/B expression (p = 0.0005), close to the baseline, without any effect on MHC class-1 molecules (Fig. 2C), suggesting a direct effect of DENV-2 on MIC-A/B cell-surface expression. Mo-DCs infected by DENV-2 were indistinguishable from those of the non-infected controls in terms of cell-surface expression of other tested ligands for the major NK receptors studied ; including ULBP family proteins for NKG2D, and ligands for NKp30 and NKp46 (Fig. 2A, 2B ; Supporting Information 4D).
MIC-A/B trigger NK cells to produce cytokines in the presence of DENV-2-infected Mo-DCs
As DENV-2-infected Mo-DCs expressed MIC-A/B, we wanted to determine whether these ligands trigger NK cells to produce cytokines. As shown in Fig. 3A, IFN-γ and TNF-α
production is significantly decreased in the presence of neutralizing anti-MIC-A/B Abs (p < 0.0001 for IFN-γ and p = 0.007 for TNF-α) to reach a level close to the baseline, only observed with NK cells alone or co-cultured with non-infected Mo-DCs. In contrast, the production of cytokines was not modulated in the presence of the pan W614/A anti-MHC class-1 blocking mAb (Fig. 3A). Consistently, the production of IFN-γ was strongly decreased in the presence of neutralizing NKG2D mAbs (Supporting Information 5). Neither MICA/B nor anti-MHC class-1 mAbs modified the level of degranulation (data not shown). These data suggest that cytokine production by NK cells is mediated by cell-to-cell contacts via an over-expression of MIC-A/B at the surface of DENV-2-infected Mo-DCs, whereas the absence of degranulation is linked to the levels of MHC class-1 and/or HLA-E expression.
Knock-out, small-interfering RNA, or neutralizing antibodies targeting IL-12 subunits revealed a critical role for IL-12 promoted by Mo-DCs in IFN-γ production by NK cells [28], and more particularly after DENV infection [20]. Fig. 3B shows that the production of IFN-γ and TNF-α by NK cells was significantly decreased after co-culture with DENV-2-infected Mo-DCs in the presence of apilimod (formerly STA-5326), a specific inhibitor of IL-12 synthesis [29]. Furthermore, the inhibitory effect of apilimod was enhanced in the presence of anti-MIC-A/B to bring the level of cytokines close to the baseline, as observed in the presence of non-infected Mo-DCs (Fig. 3B). Furthermore, Fig. 3C shows that IL-12 p35 (10.3 ± 1.0-fold-increased) and p40 (13.7 ± 0.5-1.0-fold-increased) gene expression was significantly increased in Mo-DCs infected by DENV-2, compared to non-infected cells. In addition, this increase of IL-12 expression was strongly inhibited after pre-treatment by apillimod in DENV-2-infected Mo-DCs.
Inhibition of GSK-3 enhances MICA/B expression in DENV-2-infected Mo-DCs
In order to characterize the mechanism behind the induction of MIC-A/B induction after DENV-2 infection, we explored the role of glycogen synthase kinase-3 (GSK-3) and signal
transducer and activator of transcription 3 (STAT-3) activities, previously associated with MIC-A/B expression in cancer [30,31]. Fig. 4A shows that exposure to 5.15-DPP, a specific inhibitor of STAT-3, had no effect on MIC-A/B expression, whereas treatment with inhibitors of GSK-3 activity, LiCl, and SB415286, significantly increased MIC-A/B expression, both in frequency and MFI, close to the level observed after DENV-2 infection (Fig. 4A). These data are confirmed by qPCR, both in MIC-A and MIC-B (Supporting Information 4C). Altogether, these data indicated that MIC-A/B up-regulation is mediated through the inhibition of GSK-3 (tyrosine 216) phosphorylation, independently of the phosphorylation of STAT-3. Inhibitors of GSK3 and STAT-3 activity have no effect on MHC class-1 expression, both in frequency and MFI (Supporting Information 4A). To determine the significance of these findings, we examined the overall functional capacity of NK cells in the presence of GSK3 inhibitors. After treatment with LiCl and SB415286, the baseline level of degranulation was unchanged (Fig. 4B), while IFN-γ and TNF-α production by NK cells was significantly increased, at levels observed in the presence of DENV-2-infected Mo-DCs (Fig. 4C). In contrast, in the presence of the inhibitor of STAT-3, the functional activity (degranulation and cytokines production) remained close to the baseline (Fig. 4B, 4C). To confirm the role of GSK-3 activity in DENV-2 infection, the total and phosphorylated forms of GSK-3 were quantified in lysates of Mo-DCs infected by DENV-2. Although the total amount of proteins was not modified, the quantity of phosphorylated GSK-3 in DENV-2-infected Mo-DCs was decreased, similar to the levels observed after 48 hours of treatment with specific inhibitors, when compared to non-infected and LPS-activated Mo-DCs (Fig. 4D).
Discussion
In this study, we used an in vitro model of co-culture to analyze the cellular and molecular events required for the control of DENV-2 infection by NK cells. We demonstrated that
specifically, and did not mediate degranulation, in accordance with data previously observed with ex vivo NK cells of infected patients [12-14] ; this is possibly explained by the upregulation of MHC class-1 molecules, and HLA-E in DENV-2 infection, confirmed herein and previously
described [26,27,32]. However, we cannot exclude that the cytotoxic functions of NK cells are
modulated differently in the function of the DENV serotype. We have also shown that IFN-γ
production by NK cells is profoundly decreased after treatment of DENV-2-infected Mo-DCs with apilimod, which selectively and potently inhibits IL-12 [31]. Consistently, neutralizing IFN-γ is sufficient to abolish the NK cell response against DENV in an optimized humanized
mouse model [33], the infection of IL-12p40−/− and IL-18−/− mice by DENV induces a profound
decrease of IFN-γ production ; in association with an increase in disease severity displayed by higher viral loads and enhanced lethality [20]. Similar data were reported in patients infected by DENV-3 [34], suggesting that the role of IFN-γ can be extended to different serotypes of DENV. IFN-γ is known to be a key mediator in the up-regulation of MHC class-1 molecules [35], a phenomenon also reported after infection by different flaviviruses, including DENV, unlike many other viruses [17]. Together, these data suggest that the production of IFN-γ is a
crucial defense mechanism to fight infected cells, or alternatively, that the escape mechanism
mediated by DENV is mainly against NK cells, with little regard for avoiding T cell detection.
The importance of cell-to-cell cross-talk for cytokine production by NK cells during 2 infection prompted us to analyze, in detail, the mechanism by which NK cells control DENV-2 infection in an in vitro co-culture system. We demonstrated that blocking HLA-class-1 molecules has no effect on cytokine production by NK cells. Concerning the activating receptors, the ligands for NKp30 and NKp46 were weakly expressed on DENV-2-infected cells and thus are not likely to trigger NK cell functional activity. Consistently, Costa et al.,[33] showed that cell-to-cell contacts between Mo-DCs and autologous NK cells are important for NK-cell activation and IFN-γ production while blocking ligands for natural cytotoxicity receptors ; NKp30, NKp44, and NKp46 had a minimal impact in humanized mice. The E
protein of DENV interacts directly with NKp44 [36], expressed predominantly on activated NK cells. Whether this direct interaction between the virus and a specific NK receptor could modulate cell functions remains to be determined.
Importantly, the cellular ligands of NKG2D, MIC-A, and MIC-B, but not ULBP family proteins, are equally likely to trigger NK cells to produce cytokines sharply, but not to induce cytotoxicity, contrasting with previous studies. Discrepancies with these results may well reflect the difference between the experiments performed by us in an autologous condition with infected Mo-DCs that over-expressed HLA class-1 and HLA-E molecules and studies performed in heterologous conditions with tumor cell lines that for instance down-modulated HLA-class 1 molecules [31,37]. MIC-A and MIC-B are generally not observed at the surface of normal cells but constitutively expressed in stress situations, like cancer and viral/bacterial infections [11]. Interestingly, in a large number of Vietnamese children and adults, MIC-A and MIC-B allelic types have been shown to be involved in DENV pathogenesis [38]. In addition, elevated levels of soluble MIC-B were reported in the sera of infants infected with DENV [39]. After infection of Mo-DCs, MIC-A/B are mainly induced by DENV-2-infected cells, when compared to bystander cells, suggesting that a viral protein could be implicated in the expression of MIC-A/B in infected cells. It should also be pointed out that not all infected cells expressed MIC-A/B, suggesting that other NK-cell activating ligands may be involved in triggering NK-cell functions. In the presence of blocking anti-MIC-A/B mAbs, the production of IFN-γ and TNF-α returned close to the baseline, at levels observed in the presence of non-infected target cells. These data suggest that MIC-A/B ligands are central in the cross-talk between NK cells and Mo-DCs in response to DENV infection, and allows to hypothesize that DCs and NK cells coordinate their response through direct contacts to create a positive feedback loop for DCs-derived IL-12 and NK-derived IFN-γ which drives T cell-mediated immunity in response to DENV-2 infection, as also seen after an influenza infection [40]. In a context of ADE/ADCC, monocytes infected by DENV can produce type I IFN (IFN-α), TNF-α, and
IL-10, but not IL-12 [41]. It is likely that the ligation of FcR during ADE, in addition to IFN-α and IL-10 cytokines detected in ADE-affected monocytes, suppressed IL-12 production and could explain the short-lived IFN-γ production, detectable at 24 hours but not at 48 hours post-infection in ADE condition [41]. These data contrast with our and other data observed in the absence of FcR ligation, in which a high level of IFN-γ production was detected whatever the time post-infection and associated with the production of IL-12 by infected Mo-DCs [42].
The expression of MIC-A/B at the surface of Mo-DCs is certainly a consequence of the virus infection. The modulation process may occur at different stages, including transcription, RNA stabilization, protein stabilization, and cleavage from the cell membrane [43]. Several transcription factors, such as heat shock transcription factor 1, NF-kB, STAT-3, and GSK-3 have been reported to promote the transcription of MIC-A/B [31,44]. However, treatment with 5.15-DPP suggests that the inhibition of GSK-3 activity is certainly independent of STAT-3 in our model. Discrepancy effects of STAT-3 inhibitors on MIC-A and MIC-B expression were previously reported in cancer cells ; thus, Garrido-Tapia et al.,[45] recently published that STAT3 inhibition by Stat21 increases the cell-surface expression of MIC-B, but had no effect on MIC-A in gastric adenocarcinoma cells, whereas Stat21 affects MIC-A and MIC-B expression in colorectal adenocarcinoma cells [37], but with little or no effect on MIC-B and myeloma cells [31]. This suggests that the role of STAT3 inhibition on MIC-A and/or MIC-B expression perhaps could be dependent on the pathology.
Importantly, however, the inhibition of GSK-3 activity induces MIC-A/B expression on Mo-DCs and subsequently induces the production of IFN-γ and TNF-α by autologous NK cells, at