Viruses exploit host metabolic and defence machinery for their own replication. The flaviviruses, which include Dengue (DENV), Yellow Fever (YFV), Japanese Encephalitis (JEV), West Nile (WNV) and Zika (ZIKV) viruses, infect a broad range of hosts, cells and tissues. Flaviviruses are largely transmitted by mosquito bites and humans are usually incidental, dead-end hosts, with the notable exceptions of YFV, DENV and ZIKV. Infection by flaviviruses elicits cellular responses including cell death via necrosis, pyroptosis (involving inflammation) or apoptosis (which avoids inflammation). Flaviviruses exploit these mechanisms and subvert them to prolong viral replication. The different effects induced by DENV, WNV, JEV and ZIKV are reviewed. Host cell surface proteoglycans (PGs) bearing glycosaminoglycan (GAG) polysaccharides — heparan/chondroitin sulfate (HS/CS) — are involved in initial flavivirus attachment and during the expression of non-structural viral proteins play a role in disease aetiology. Recent work has shown that ZIKV-infected cells are protected from cell death by exogenous heparin (a GAG structurally similar to host cell surface HS), raising the possibility of further subtle involvement of HS PGs in flavivirus disease processes. The aim of this review is to synthesize information regarding DENV, WNV, JEV and ZIKV from two areas that are usually treated separately: the response of host cells to infection by flaviviruses and the involvement of cell surface GAGs in response to those infections.
Host cell responses to viral infection
A wide group of viruses have life cycles that involve complex interactions with the metabolic machinery of their hosts. These include the poxviridae, adenoviridae, retroviridae, picornaviridae, flaviviridae and orthomyxoviridae . Several virus types, including those with large genomes, such as poxviruses and herpesviruses, are known to block host apoptotic pathways [2,3]. One early response by host cells to an invading microorganism is to initiate cell death through several mechanisms (see below). Depending on where the invading virus is detected, this could be initiated, through intracellular signalling systems, such as BAX (Bcl-2 regulator protein). Many other host cell responses are, however, available by which cell death can be influenced, enabling viruses to enhance their survival and replication. These include autophagy, apoptosis and pyroptosis; each of which has distinct mechanisms and outcomes (Figure 1):
Autophagy: In autophagy, the cellular constituents are partitioned and fused with lysosomes to allow their degradation and re-cycling . This is a means of extending the longevity of a cell under stress and can be activated by infection. It is a highly conserved surveillance process involving the transport of lipids, proteins and organelles from the cytoplasm into double-membrane vesicles (autophagosomes) and then to lysosomes for degradation (reviewed in ref. ). Autophagy can be activated for DENV (Dengue virus), Japanese Encephalitis virus (JEV) and Zika virus (ZIKV) and potentially also for other flaviviridae, to provide a means by which viral replication can be prolonged and cause pathology .
Apoptosis: Apoptosis is a form of cell death which is programmed and highly regulated, during which the cell undergoes a series of morphological changes, mitochondrial stress and formation of apoptotic bodies, allowing the cell contents to be consumed by phagocytes, avoiding any resulting inflammatory response. Apoptosis can be achieved through the BAX- and BAK (Bcl-2-homologous killer)-mediated intrinsic and TNF-α (tissue necrosis factor alpha)-mediated extrinsic apoptotic pathways.
Pyroptosis: Pyroptosis requires the presence of CASP-1 (caspase cysteine–aspartic protease) in a complex termed the pyroptosome or inflammasome, in each macrophage  and involves activation of the inflammasome within macrophages resulting in the release of cell debris (DNA, ATP, cytokines and other proteins), thereby sustaining the inflammatory response in the surrounding tissue.
Each of these responses is triggered by activation of distinct pathways and each response leads to distinct consequences (Figure 1). From the perspective of the host cell, a pyroptotic response (if it is not overly pronounced, a situation that can lead to a ‘cytokine storm’) can achieve rapid clearance but, in chronic infections, this leads to sustained inflammation that can also cause uninfected cell death . Infection induces many responses, which influence subsequent events, for example, in vitro infection of endothelium by flaviviruses increases adhesion and up-regulates major histocompatibility complex molecules, which can be modulated by cytokines . The ultimate outcome of a viral infection, however, depends not only on which routes are activated but also on the original viral load .
Main cell death modality.
The most extensively studied flavivirus in this regard is DENV, but information concerning virus–host (vertebrate and insect vector) interactions of other flaviviruses, such as WNV (West Nile virus), JEV and ZIKV, is also emerging. There are fewer data available regarding interactions between many of the flaviviruses and insect tissues or cells. Responses in insects are addressed first and are followed by a survey of the interactions of DENV, WNV, JEV and ZIKV and their consequences for mammalian cells.
Responses to DENV, WNV, JEV and ZIKV in mosquitoes
Viruses are introduced into the mosquitoes through a blood meal from their animal host and replicate in the midgut within a few minutes, then spread through the insect tissue in the haemolymph. Viruses are then passed to another animal host during feeding through transmission of infected saliva [10,11]. Life-long infection of Aedes and Culex mosquitoes apparently occurs without pathological effects and viral interference, in which infection with one virus inhibits infection with another, is also observed . For example, Aedes aegypti infected by DENV-2 and -3 strains simultaneously show higher levels of RNA of the DENV-2 form, but the origins of this phenomenon are not well understood. Viral interference and viral persistence involve the JAK–STAT (Janus kinase–signal transducer and activation of transcription proteins) and Toll-mediated pathways of the immune system of the insect. Several cellular systems are implicated in flavivirus replication including, for DENV, EF1α (translation elongation factor) and La protein [13–15]. The immune response in insects is distinct from that in vertebrates. In mosquitoes, this involves three main signalling pathways:
Innate immune system (involving the production of antimicrobial peptides, mainly against Gram-negative bacteria).
The Toll-mediated pathway (against viruses, Gram-positive bacteria or fungi and also involves the production of some peptides).
Signal transduction through the JAK–STAT pathway (against viruses, particularly in Aedes).
The innate response involves the activation of two proteolytic cascades following infection, leading to blood clotting and melanization, resulting in the release of reactive oxygen species (ROS). Phagocytosis of bacteria and encapsulation of larger parasites by blood cells can also occur. For virus infection, the Toll and JAK–STAT pathways are the most important and have been studied most extensively for DENV in A. aegypti [16–20]. The flavivirus genome comprises a single open-reading frame, which encodes a total of 10 viral proteins (Figure 2). Several molecular determinants have been identified, which contribute to the infection of mosquito cells by the virus . These comprise the molecular hinge region of the E-protein, for infection of A. aegypti and the FG loop of domain III, a co-receptor-binding region [22,23].
The structure of the flavivirus genome.
Mammalian cellular responses to DENV, WNV, JEV and ZIKV infection
There are four serologically distinct strains of DENV (DENV 1–4). Dengue pathogenesis is often explained in terms of antibody enhancement, cytokine release and promotion of both apoptosis and pyroptosis in response to viral NS1 (flavivirus non-structural protein) protein . Dengue induces apoptosis in hepatocytes, both in vitro and in vivo [25–29], and up-regulated IL-8 (interleukin-8) and RANTES [a chemokine; regulation on activation, normal T-cell expressed and secreted, also known as CCL5 chemokine (C–C motif) ligand 5] have been observed to accompany apoptosis .
The DENV non-structural, NS2, NS3, NS5, capsid (C) and E proteins (flavivirus envelope protein) are all known to trigger the extrinsic apoptotic pathway in a variety of cell types, which include endothelial cells, hepatocytes or immune cells, and these processes involve the proinflammatory cytokines . The small membrane (M) protein from all four serotypes of DENV, on the other hand, induces apoptosis in hepatocytes and neurones . HMGB1 [high mobility group box-1 protein (amphoterin)] has also been identified as being released during non-apoptotic cell death  which, outside the cell, acts as a cytokine  and has recently also been identified in the tissue of severe (fatal) dengue victims .
Autophagy has been observed in epithelial cells, involving the non-structural viral proteins, NS4 , which viruses can exploit to boost energy production by the host cell to assist viral replication [36,37]. DENV induces autophagy and protects against other stress agents [35,36], and this also serves as a means of protection against apoptosis . DENV can induce autophagy by several routes, involving endoplasmic reticulum (ER) stress and altered signalling followed by production of ROS . The inhibition of ER stress signals limits the ability of DENV to induce autophagy, while increased autophagy ultimately protects from apoptosis and increases the potential of DENV to replicate.
West Nile virus
While the strategy adopted by WNV appears to be less versatile than that used by DENV, being unable to affect autophagy [39,40], WNV is nevertheless capable of initiating apoptosis in the central nervous system causing neuropathology, or necrosis when the viral load is high . West Nile virus activates Toll-like receptor 3 and increases the permeability of the blood–brain barrier through TNF-α . Distinct strains of WNV affect pathways of ER stress to enable the viral load to be increased. The capsid protein interacts with p53 in vivo  to activate the intrinsic apoptotic pathway and blocks apoptosis via PI31-Akt (phosphatidylinositol 3-kinase/protein kinase B signalling pathway) , while the NS3 viral protein is involved in extrinsic pathway activation in neuroblastoma and cervical cancer cells. Inflammation follows activation of the CASP-9 apoptotic pathway , and cell death through the CASP-3 pathway contributes to pathogenesis . Furthermore, WNV has the ability to influence microRNA , a capacity which may also extend to other flaviviridae. WNV induces ER stress to degrade ATF6 (one of the major unfolded protein response pathways whereby the ER responds to a high load of viral proteins by up-regulating protein folding machinery) rapidly, causing phosphorylation of a second unfolded protein response pathway and CHOP (cyclic AMP response element-binding transcription factor homologous protein)-dependent premature cell death. The latter has been proposed as a potential host defence mechanism that is capable of limiting viral replication and which explains neuronal loss in WNV .
Japanese Encephalitis virus
Neutrophil chemotaxis can be induced by JEV , which has the ability to influence both the intrinsic and extrinsic apoptotic pathways. The expression of two chemokines, Cxc11 and Cxc12, rapidly increases macrophage numbers and attracts polymorphonucleate leucocytes, key players in the innate immune response. These also serve, during early stages of infection by WNV, as reservoirs of infection, later contributing to clearance . Oxidative stress and programmed cell death can be induced by JEV infection [51,52]. In neural cells, JEV induces apoptosis by stimulating an up-stream stress response. The NS3 viral protein of JEV, in contrast with both DENV and WNV, induces apoptosis via the intrinsic pathway and JEV can also replicate in the absence of CASP-3, to induce CASP-8 and CASP-9 through a mitochondrion-dependent pathway [53,54]. Inhibition of CASP, however, does not block viral production, and JEV appears to depend on mitochondrial apoptosis [55,56] for pathogenesis. In addition, like WNV, JEV activates CHOP  and also employs autophagy [58,59] to mediate the proinflammatory cytokine response in neural cells.
In general, cell death triggered by viruses can be either immune-mediated or induced via cell autonomous injury once viral infection has altered the expression of pro- and anti-apoptotic proteins. Several studies have reported apoptosis as a mechanism of cell death in ZIKV-infected NPCs (neural progenitor cells) [60,61]. However, apoptosis has also been found in adjacent foetal neural cells without evidence of infection and a similar phenomenon of bystander apoptosis has been reported for human immunodeficiency viruses , DENV  and WNV . The proposed mechanism of bystander apoptosis still remains to be elucidated; however, it is likely that cytotoxic factors released by infected NPCs might damage uninfected NPCs . Furthermore, other non-neural cells, such as astrocytes, microglia and lymphocytes, could become activated in the brain and release proinflammatory cytokines that damage uninfected neurones. The complete picture of the immune response to ZIKV is rather complex and warrants further investigation. Similarly to DENV, however, ZIKV can induce autophagy that, during infection of foetal NPCs, can lead to defective neurogenesis . In addition to caspase-dependent cell death, caspase-independent cell death has also been reported to cause cell death in infected cells such as epithelial cells, primary skin fibroblasts and astrocytes  in vitro. This mode of cell death was explained by extensive vacuolization of the ER, which is the major intracellular site of ZIKV replication . The accumulation of ZIKV vacuoles triggers cell collapse defined as paraptosis, involving cytoplasmic vacuolization, independent of caspase involvement .
Infection is clearly a complex and multi-faceted event, but one class of cell surface molecules that are present in some form on both mammalian and insect cells and which are coming under increasing scrutiny, are the GAG polysaccharide components of proteoglycans (PGs).
PG roles in host–virus interactions
The role of PGs, particularly HSPGs (heparan sulfate proteoglycan), is well documented as a means of attachment of flaviviruses, including DENV [69–72], YFV , JEV  and TBE [75–77]. The principal PGs are the four integral membrane syndecans and the six glycosylphosphatidylinositol anchored glypicans. In the syndecans, the HS chains are carried on the ectodomain, tending to maintain them distal from the cell membrane, whereas in the glypicans, the HS chains tend to be adjacent to the plasma membrane. There are also many other PGs, carrying GAG chains, including so-called part-time PGs, such as CD-44 (a cell surface glycoprotein, also known as HCAM; homing cell adhesion molecule). The PGs bind and regulate a multitude of proteins in the extracellular matrix to control key biological processes including cell proliferation and differentiation, wound repair, host defence mechanisms, as well as responses to stress and inflammation. Other more subtle roles are also now emerging for PGs in viral infection. Recently, it was observed that the addition of exogenous heparin (a GAG polysaccharide with close structural similarity to endogenous HS) to cells infected with ZIKV showed a surprising increase in survival and continued viral replication, shown by high viral load in surviving cells, while control cells had died . While the mechanism behind this phenomenon and the extent to which it reflects in vivo physiology remain unknown, the possibility that heparin may play a role as a survival factor in trophoblasts has been raised , since it was found to influence several anti-apoptotic pathways, including CASP-3. Furthermore, HSPG has been documented as stimulating TNF-α in murine microglia , but the action of heparin seems unlikely to be through binding to TNF-α , which is a master regulator of the inflammation response. It is not known currently whether heparin is reinforcing a mechanism that the virus exploits, in order to prolong cell survival and promote viral persistence, or whether it acts through an unrelated mechanism.
A dependency on the HS biosynthetic machinery, in particular on the NDST1 (N-deacetylase and N-sulfotransferase-1) and EXT1 (Exostosin-1) enzymes, has been found using a functional genomics approach for ZIKV infection of HeLa cells . Viral infection results in activation of Toll-like receptor signalling pathways of the innate immune response, sequestration of growth factors, chemokines and cytokines, and altered leucocyte adhesion and also influenced the behaviour of metalloproteases . The influence of flaviviruses on PG regulation and expression, or the interplay between them, which may exert considerable influence on disease progression, has yet to be explored in detail. Nevertheless, it is noteworthy that the PG, syndecan, is shed in response to stress or damage  and, in the case of the bacterium, Pseudomonas aeruginosa, is exploited to enhance microbial virulence [85,86]. In addition to the role played by HS in DENV attachment, the NS1 viral glycoprotein that is secreted by infected cells and has been implicated in the disease process attaches to cell membranes via HS and chondroitin sulfate (CS) . Thus, GAG components of PGs are involved in several stages of the viral infection process. It will be intriguing to discover the extent to which similar events occur with other flaviviruses. Mechanical stress, which could, under certain strongly inflammatory conditions, be relevant to infection, has also been shown to alter PG expression in endothelial cells. The HSPGs contained different polysaccharide (HS) components and exhibited distinct biochemical properties . Stress induced by NO alters the turnover of matrix components , which, although recorded for rheumatoid arthritis, may also apply more widely. Interestingly, in the case of infection of endothelial cells by another class of virus, herpes virus, both HSPGs and CSPGs were down-regulated .
There are several routes available for cell death resulting in varied degrees of longevity and stress on host cells and hence a variety of mechanisms available for exploitation by the virus. It seems highly likely that the response of cells varies between the causes of the stress to the cells  and between different virus types, the viral load, as well as the cell type and its stage in the cell cycle . The response to inflammation can also be complex, as is the case in kidney where some PGs, such as agrin and glomerular basement membrane-associated HS levels, are decreased, but CD44 is increased . While there are broad similarities between flaviviruses in this regard, particularly between WNV and JEV, which can both induce neural symptoms, it is not yet clear in mechanistic terms, where this propensity originates. A role for GAGs beyond attachment, which itself could explain some facets of tissue tropism of the flaviviruses, in the infection process and cellular responses, is also beginning to emerge. Comparisons between phylogenetically close viruses, such as DENV and ZIKV, will be particularly interesting in this regard. One observation that can be made is that there is little information available concerning PG expression in relevant insect vectors and still less relating the effects of flavivirus infection to PG expression in insects. This represents a significant gap in our knowledge of the life-cycle and infection process of these important viruses.
flavivirus capsid protein
caspase cysteine–aspartic protease
cyclic AMP response element-binding transcription factor homologous protein
flavivirus envelope protein
heparan sulfate proteoglycan
interleukin one beta
Janus kinase–signal transducer and activation of transcription proteins
Japanese Encephalitis virus
flavivirus small membrane protein
neural progenitor cells
flavivirus non-structural protein
reactive oxygen species
tissue necrosis factor alpha
West Nile virus
Yellow Fever virus
The authors acknowledge funding from the Italian Ministry of Health (E.V.), BBSRC (E.A.Y., M.A.S. and S.L.T.), FAPESP and CNPq (M.A.L. and E.A.Y.).
The Authors declare that there are no competing interests associated with the manuscript.