As obligate intracellular parasites, viruses have evolved diverse mechanisms to enter into, and exit from, host cells. A requirement that is shared by all animal viruses is the use of cellular receptors for entry into cells to initiate viral infection. Receptors can function in viral attachment to the cell surface and can also mediate viral internalization and penetration of the cell-limiting membrane. Receptors are often grouped into 'primary receptors' and 'secondary receptors' or 'co-receptors' depending either on their function in the viral entry process or historical precedence. For example, HIV-1 uses CD4 as its primary receptor and a chemokine receptor — either CCR5 or CXCR4 — as its co-receptor (Table 1). The interactions between many viral glycoproteins and attachment receptors are highly specific, such as the HIV-1–CD4 interaction, whereas other interactions are less specific, such as the interaction between viral glycoproteins and cell-surface heparan sulphate proteoglycans1,2. Viruses can also bind to cells using host cell-derived molecules that are incorporated into the viral envelope, such as adhesion molecules3.

Table 1 Viruses that spread from cell to cell

For the purpose of discussing animal virus entry into cells after attachment, viruses can be classified into two broad groups: enveloped and non-enveloped. To my knowledge, direct cell-to-cell spread has only been described for enveloped viruses, and so non-enveloped viruses will not be discussed further here. This lacuna may reflect an inability of non-enveloped viruses to spread directly from cell to cell (perhaps owing to their mode of viral exit) or may simply reflect the fact that this has not been studied. Enveloped viruses penetrate the host cell by fusing their lipid envelope with a host cell membrane. Fusion takes place through a pH-independent (for example, HIV-1) or a pH-dependent (for example, influenza A virus) mechanism (Fig. 1). Many viruses that enter cells through pH-independent routes fuse their membranes directly with the cell plasma membrane, whereas viruses that require a pH change to penetrate the lipid membrane of the cell are first internalized by an endocytic mechanism, before fusing with an endosomal membrane2,4,5. Entry of enveloped viruses at the plasma membrane might require binding to one or more receptors that trigger signalling into the cell. Such signalling might activate viral endocytosis (reviewed in Ref. 5), but also has important implications for viruses that fuse at the plasma membrane. For example, signalling that is triggered by the initial engagement of a low number of receptors by viral glycoproteins may lead to the active recruitment of additional viral receptors to the site of virus attachment. Thus, initial interactions of HIV-1 with its cell surface receptors results in cytoskeletal rearrangement that recruits additional receptors to the virus–cell interface to enable efficient membrane fusion and viral penetration6,7,8,9. After entry, the virus uncoats and replicates to produce new infectious particles; in a fluid-phase dissemination model these particles are released from the cell into the extracellular medium (Fig. 1).

Figure 1: The generally accepted paradigm of virus infection.
figure 1

After fluid-phase diffusion, a cell-free, enveloped virus particle binds to the plasma membrane of an attachment receptor-bearing target cell (step 1). The virion diffuses laterally across the plasma membrane, recruiting a critical mass of internalization and/or entry receptors (step 2). For viruses that fuse directly at neutral pH at the plasma membrane, entry-receptor engagement is sufficient to drive virus–cell membrane fusion leading to entry of the virion into the cytoplasm of the cell (step 3a). For viruses that rely on internalization into the cell prior to membrane fusion, virion endocytosis is the next step (step 3b). Some viruses fuse from within vesicular compartments at neutral pH, whereas others require a subsequent reduction in pH that occurs as the endosome matures to trigger membrane fusion (step 3c). Entry of the virion replicative machinery into the cytoplasm leads to a sequence of events that, for a productive infection, allows the synthesis and assembly of new viral components, which may occur in the cytoplasm, nucleus or other intracellular organelle (step 4). Some viruses assemble and bud through intracellular membranes and leave the cell through exocytosis (step 5a), whereas others bud through the plasma membrane (step 5b). New virus particles are released into the extracellular medium to randomly diffuse until they encounter a new target cell (step 6).

Virus particles contain all the necessary information to initiate a new productive infection, and all described modes of virus spread require the partial or complete assembly of virus particles. However, whether virions are released to diffuse and infect new host cells, or spread directly between adjacent cells, depends on the virus and the target cell or tissue. There are advantages and disadvantages of cell-free and cell-to-cell modes of virus spread. Cell-free virus spread enables host-to-host transmission when a virus needs to exit the infected host, disperse and remain infectious for a sufficient period of time to infect a new host. For example, if the virus has a fragile lipid envelope, the virus particle could be transmitted between hosts in a dispersed aqueous environment (for example, influenza A in a saliva droplet) or could resist drying out and spread through desquamated skin scales (for example, smallpox particles). Once inside a host, another advantage of cell-free spread is that viruses transit rapidly through the blood, lymph and cerebrospinal fluid to infect distant tissues.

However, cell-free spread has disadvantages, as viral particles are exposed to various biophysical, kinetic and immunological barriers. Biophysical barriers include mucous membranes and intrinsic virion lability that results in progressive loss of infectivity over time. Random fluid-phase diffusion has a kinetic penalty, as it takes time for the virus to encounter a target cell and engage attachment and entry receptors. This is particularly disadvantageous for viruses that have low avidity for their receptors, those that bind receptors expressed at low copy number and/or those that must engage more than one receptor to enter the cell (for example, HIV-1 (Ref. 1); Table 1). Immunological barriers to free virions during acute infection include innate humoral (natural antibodies, complement, chemokines and defensins) and cellular (macrophages) defences. During subsequent rounds of infection, free viruses are faced with adaptive immune responses, which comprise clonally activated lymphocytes and affinity-matured antibodies.

Viruses that spread directly from infected to uninfected cells can avoid many of the obstacles to infection described above for free virus particles. Virus transmission from an infected to an uninfected host by a cell-associated virus might be more efficient than free-virus spread. For example, infected cells entering a new host could adhere to, and cross by transmigration, a mucosal epithelial barrier that would otherwise be impermeable to free virions. Once the initial infection has occurred, the cell-to-cell mode of viral spread enables direct infection of target cells by adjacent infected cells, which eliminates the rate-limiting step of fluid-phase diffusion. If cell–cell contact recruits viral receptors to the intercellular interface, then receptor engagement and viral entry can be efficient. A final advantage of cell-to-cell spread could be that the movement of a virus between cells shields the virus both sterically, and kinetically in terms of exposure time, from the innate and adaptive immune effector mechanisms described above.

This Review focuses on the diverse mechanisms that are used by viruses to spread from cell to cell and relates these mechanisms, when possible, to pathogenesis. Although much of this Review reports on in vitro data, I will consider in vivo results if available. The article concludes by relating some general concepts of direct cell-to-cell viral dissemination to possibilities for therapeutic intervention that might tackle this mode of spread and suggesting some future directions for study.

Virus cell-to-cell spread

Viruses that can move directly from cell to cell are listed with their receptors and virally encoded receptor ligands (viral glycoproteins) in Table 1. Some general principles determine whether a virus can move directly from cell to cell or is limited to cell-free dissemination. For those viruses that lyse the host cell to release infectious progeny, such as the non-enveloped adenoviruses, picornaviruses and reoviruses, release of free virions is likely to be the only mode of spread between cells. Viruses that exit at the plasma membrane, either by budding (for example, paramyxoviruses, retroviruses and rhabdoviruses) or by a form of exocytosis (for example, hepadnaviruses and herpesviruses), can spread directly from cell to cell. Some viruses that remove their cellular receptors during exit (for example, the removal of host cell sialic acid by influenza virus neuraminidase) might be efficiently released from the infected cell and spread preferentially by fluid phase rather than cell-to-cell movement. Conversely, cell-associated viruses, such as certain forms of poxvirus, might be trapped during exit by interactions at cell surfaces, thereby promoting cell-to-cell spread. Whether a virus can exit one cell and immediately enter a new target cell will depend on the proximity of the cells within the tissue, whether the infected cell has in-built polarity (which might determine the route of viral exit; Box 1) and whether the target cell expresses viral receptors at the intercellular interface. Viral assembly in polarized epithelial cells might target virus release towards the apical surface, away from junctions with other cells in the tissue, thereby favouring free virion spread. Some viruses have evolved mechanisms to subvert this process by targeting viral fusion proteins (paramyxoviruses) or viral assembly (herpesviruses) to basolateral surfaces of the cell (Box 1), whereas others (retroviruses) can induce polarization of immune cell types with no fixed polarity, allowing directed assembly and spread of the virus from the infected cell towards an uninfected, receptor-expressing target cell (Box 2).

The main mechanisms that are used by viruses to spread from cell to cell are summarized in Fig. 2. Conceptually the simplest mechanism is the fusion of infected and uninfected cells, which can occur in vitro for herpesviruses, paramyxoviruses and retroviruses. Cell–cell fusion can result in a single, giant cell (syncytium), and therefore might not be considered to be true cell-to-cell spread. However, some viruses, such as paramyxoviruses and herpesviruses, can partially fuse cells (so that only distinct patches of plasma membrane contain pores connecting cellular cytoplasms) to allow cell-to-cell spread without the induction of syncytium formation. Other more sophisticated mechanisms of cell-to-cell virus spread also exist. For example, viruses can direct viral assembly to tight intercellular junctions (herpesviruses), transfer virions or subviral particles across neural synapses (herpesviruses, paramyxoviruses and rhabdoviruses) and manifest various types of cytoskeletal subversion that take advantage of actin tails (poxviruses), filopodia (African swine fever virus (ASFV) and retroviruses), membrane nanotubes (retroviruses) and virological synapses (retroviruses), all of which are thought to enable direct cell-to-cell spread of viruses.

Figure 2: Mechanisms of virus cell-to-cell spread.
figure 2

a | Cell–cell plasma-membrane fusion followed by movement of infectious viral material (shown as viral cores) into the uninfected target cell. This type of cell-to-cell spread could result in syncytium formation or might be restricted to localized 'microfusion' events that allow the fused cells to maintain structural independence. Herpesviruses, paramyxoviruses and retroviruses can spread by cell–cell plasma-membrane fusion. b | Passage of virions across a tight junction. The virus exits basolaterally from an infected cell and is trapped between the infected and uninfected cell membranes at the tight junctions. Using viral entry receptors that are located on the target cell within the tight junctions, virions fuse with and penetrate the uninfected target cell. Herpesviruses can move across tight junctions. c | Movement of virions across a neural synapse. Virions, either mature or incomplete (naked cores are shown), assemble in either the postsynaptic or presynaptic cell depending on the virus, and either bud through the membrane into the synaptic space or are released from synaptic vesicles into the cleft. Virions then either fuse directly with the opposing synaptic cell or are endocytosed. Rhabdoviruses, herpesviruses and paramyxoviruses move across neural synapses. d | Viral induction of actin- or tubulin-containing structures. Poxviruses project from infected to uninfected cells using actin tails (left). Infected cells produce cross-linked bundles of branched actin directly underneath the virus particle. The nucleation and growth of actin outwards then projects the virus particle out of the cell towards a neighbouring uninfected cell. Herpesviruses can induce actin- and tubulin-containing structures that project virions towards adjacent cells (middle). Asfarvirus-induced actin-containing filopodia project African swine fever virus virions towards adjacent cells (right). e | Viral subversion of actin-containing structures. Filopodial bridges can join retrovirally infected and uninfected fibroblastic cells in culture. Murine leukaemia virus-infected cells stably anchor filopodia that project from uninfected cells, which allows virions to travel from an infected to an uninfected cell in an actin-myosin-dependent manner. Infection of the target cell requires expression of viral receptors, which is suggestive of virus–cell membrane fusion. Only retroviruses have so far been shown to use this method of spread. f | Membrane nanotube subversion. Nanotubes can form between immune cells, including myeloid and lymphocytic cells. Nanotubes appear to be continuous with the plasma membrane, contain actin but not tubulin and can join T cells at distances of up to 300 μm. Protein and organelle transport occurs along nanotubes, with speeds that are commensurate with actin motor-mediated movement. HIV-1 virions that bud from an infected cell can move along nanotubes and infect distant target cells in a receptor-dependent manner, which is suggestive of virus–cell membrane fusion. g | Virological synapses. Immune cells are not constitutively polarized, but contain machinery that allows them to polarize their secretory apparatus towards a second cell that is involved in an immunological synapse. This machinery can be subverted by retroviruses so that an infected cell can polarize viral budding towards the receptor-expressing target cell in a structure called a virological synapse. Virions bud from the infected cell into a synaptic cleft, from which they fuse with the target-cell plasma membrane.

Most analyses of virus cell-to-cell spread have been carried out in tissue culture systems, and therefore the in vivo relevance and importance to pathogenesis of this mode of viral spread needs to be addressed. Some viruses induce clinical symptoms that are in part defined by cell-to-cell spread, such as the syncytium formation that is induced in vivo by paramyxoviruses, and the neurotropic spread of rhabdoviruses and herpesviruses. For several viruses, evidence for cell-to-cell spread has been obtained from animal models: in the CNS few, if any, free virions of rhabdoviruses, herpesviruses and paramyxoviruses are observed, but the viruses nevertheless transit efficiently between neurons and other cell types. In other systems, mutated viruses that are inefficient at cell-to-cell spread in vitro are attenuated in vivo. Thus, mutations in the poxvirus vaccinia virus (VV) that eliminate actin-tail formation in vitro also dramatically attenuate viral replication in vivo10,11, and envelope glycoprotein-inactivated mutants of the herpesvirus pseudorabies virus (PRV), which cannot spread from cell to cell in vitro, are unable to disseminate transneuronally in vivo12,13. Some viruses, including the retrovirus human T-lymphotrophic virus-1 (HTLV-1), seem to disseminate within and between hosts entirely by cell-to-cell transfer, as infectious virions are rarely released from the cells14.

Cell–cell fusion

Fusion of cell plasma membranes can result in the formation of syncytia. In principle, any virus that expresses its fusion machinery on the surface of an infected cell that can contact another cell, such as the basolateral surfaces of polarized epithelial cells, and can undergo pH-independent fusion at the plasma membrane, can mediate cell–cell fusion. Viruses that replicate in the nucleus (for example, herpesviruses) could benefit from the presence of additional nuclei within a syncytium.

Paramyxoviruses. Paramyxoviruses, such as measles virus (MV) and respiratory syncytial virus (RSV), induce classical multinucleate syncytia in vitr o15,16 and in viv o17 in all cell types that they infect. However, MV tagged with green fluorescent protein (GFP) can also spread without forming syncytia by inducing partial cell–cell fusion of neighbouring glial-cell processes18. All strains of MV use the cellular receptor human signalling lymphocyte activation molecule (SLAM), whereas laboratory-adapted strains use CD46, and the viral ligands are the haemagglutinin and fusion glycoproteins19. Although virus budding mainly takes place from the apical surface of polarized cells (Box 1), the haemagglutinin and fusion glycoproteins are trafficked separately to the basolateral surface20. It is thought that apical budding from respiratory epithelia facilitates transmission between hosts through release of the virus in aerosol droplets, whereas fusion between cells at the basolateral surfaces allows systemic dissemination of MV21,22.

Herpesviruses. Alphaherpesviruses induce syncytium formation in vitro and in viv o23,24, and express a range of glycoproteins that have a role in cell–cell fusion. Cell–cell fusion by herpes simplex virus-1 (HSV-1) requires the coordinated action of glycoproteins gB, gD and gH–gL on the virus-infected cell, with a further role for the gD receptor on the target cell25,26. Syncytial and non-syncytial cell-to-cell viral spread can be distinguished by the selective use of the gD receptor nectin 1 (Ref. 27). However, syncytium formation by clinical isolates of HSV-1 is rarely observed in vitro, and seems instead to be linked to tissue-culture adaptation. By contrast, all clinical isolates of varicella zoster virus (VZV) form syncytia in culture24. This difference might relate to mechanisms of viral dissemination: HSV-1 spreads readily in peripheral tissues by releasing free virions, whereas VZV mainly relies on cell-to-cell spread24. Cytomegalovirus (CMV) also disseminates by cell-to-cell spread in vivo: little free virus is detected and viruses seem to move directly between leukocytes, endothelial cells and fibroblasts28,29. In common with MV, CMV-induced cell–cell fusion does not necessarily result in syncytia. Foci of CMV-infected cells in the liver did not show evidence of syncytia30 and electron-microscopy analysis revealed limited areas of membrane fusion between CMV-infected endothelial cells and leukocytes29.

Retroviruses. HIV-1 and HTLV-1 form syncytia in tissue culture. Although CCR5 and CXR4 co-receptor using HIV-1 can induce syncytia in T-cell and macrophage cultures, the relevance of this property for virus spread in vivo is unclear. Syncytia are rarely observed in ex vivo samples from HIV-1-infected individuals31, and the few available reports describe fusion between myeloid-derived cells, mainly in the brain32,33 and adenoid34. Although there is no clear evidence that syncytium formation can enhance HIV-1 dissemination in vivo, it nevertheless contributes to cell death in tissue culture and in vivo by apoptosis and may contribute to HIV-1-associated dementia33.

Movement across neural synapses

Rhabdoviruses. The rhabdovirus rabies virus (RV) infects neurons that express the viral receptors nicotinic acetylcholine receptor, neuronal cell-adhesion molecule and p75 neurotrophin receptor35,36. Viral surface glycoprotein G binds to cellular receptors that define cellular tropism in a pH-dependent entry process in which fusion takes place in acidified endosomal vesicles35. The first electron-microscopy images of polarized budding and the synaptic transmission of RV in mice and skunks were published more than 30 years ago37,38,39. Virions budding from the axon terminal of postsynaptic cells into the synaptic cleft were taken up directly into the presynaptic cell by a form of endocytosis. Virus exit from other parts of the cell was rarely observed. RV also replicates in myocytes at the site of inoculation38, with neuronal infection subsequently taking place at nerve endings across neuromuscular junctions36. Although microscopic evidence is lacking, transfer of viruses across such neuromuscular junctions might also occur by direct cell-to-cell spread40,41.

Herpesviruses. In common with other neurotropic viruses, herpesviruses enter sensory neurons at peripheral sites, such as the skin and mucous membranes, before being transported by retrograde traffic towards the cell body, where they replicate and become latent. The virus subsequently reactivates and disseminates through anterograde axonal transport and synaptic transfer42. Initial replication in respiratory epithelia is followed by infection of neurons using the essential entry glycoproteins gB, gD and gH–gL43 (Fig. 3). gB and gE–gI are important for neurotropism in vivo and in vitro, whereas gD is not important for PRV43. Because gD is essential for free PRV spread, but not for transneuronal spread, this glycoprotein differentiates between these two processes, which confirms the importance of cell-to-cell herpesvirus spread in the nervous system13. Despite a substantial body of work, herpesvirus assembly within, and spread between, neurons remains incompletely understood, and there is controversy regarding the form of the virus that is trafficked along axons. Envelope glycoproteins and tegumented HSV capsids can be independently transported from the cell nucleus to the axon terminus, and only assemble at the synapse42,43,44,45. Alternatively, virus particles may acquire glycoprotein envelopes in cell bodies and be transported and released at the axon termini, as proposed for PRV46,47. The trafficking signals present in gE–gI that target these glycoproteins to the basolateral surfaces of epithelial cells (discussed above; Box 1) might also target virus assembly to axon termini13,48, but the mechanism that underlies such targeting is unclear, particularly because axon termini are considered to be the trafficking equivalent of an apical surface on epithelial cells49 (Box 1).

Figure 3: Two modes of herpesvirus cell-to-cell spread.
figure 3

a | Cell-to-cell spread of herpes simplex virus-1 (HSV-1) in nervous tissue, showing movement in, and between, axons and other cell types. Arrows show the direction of movement of viral material. Viral glycoproteins gB and gH–gL are always required for neuronal cell-to-cell spread, whereas gD is required for HSV-1, but not for pseudorabies virus, inter-neuron spread. Viral material travels from the cell body in an anterograde movement towards axon termini or to varicosities that can be induced by gD. Once at an axon terminal or varicosity, infectious virions are transferred across the synaptic space into an uninfected cell: either another neuron in a neural circuit or another cell type, such as a glial cell. b | HSV-1 virions arrayed at intercellular tight junctions. An HSV-1-infected cell is shown, with budding virions lined up along epithelial-cell tight junctions (upper panel). The lower panel shows a magnified version of the epithelial-cell tight junction to show the aligned viruses. Part b reproduced, with permission, from Ref. 62 © 2001 American Society of Microbiology.

Herpesvirus neuronal spread has also been observed at sporadic regions along the axon. Using wild-type and gD-deleted PRV-infected rat optic nerves, infected non-neuronal cells, such as glia, were observed at intervals along axons of infected neurons50. This was confirmed in cultures in which HSV- or PRV-infected neurons infected epithelial cells, or other neurons, through direct axonal contacts51. HSV-1 infection of cultured fetal dorsal root ganglia52 or swine PRV infection of porcine trigeminal ganglial neurons53 revealed virus assembly in varicosities. These varicosities form synaptic contacts and seem to be sites of cell-to-cell virus transmission to surrounding neuronal and non-neuronal cells (Fig. 3a). The authors speculate that recruitment of the herpesvirus receptor nectin 1 by gD might be sufficient to induce the formation of varicosities53.

Paramyxoviruses. Rare infection of the nervous system by MV causes a range of syndromes, including fatal subacute sclerosing panencephalitis (SSPE). SSPE is characterized by the absence of cell-free MV and the presence of high titres of neutralizing antibody in the cerebrospinal fluid, which suggests that cell-to-cell spread might be the dominant mode of viral dissemination and persistence54. When infecting nervous tissue, MV switches receptor use from SLAM (and/or another unknown receptor) to a different receptor (possibly neurokinin 1) that is expressed on neurons55. This receptor switch is accompanied by loss of the viral haemagglutinin protein, which is required for free virus infection but dispensable for subsequent cell-to-cell spread between neurons55. Immature budding structures that were observed in neurons led two groups to propose that MV moves from cell to cell rather than by release of free virions56,57. Mechanistically, this cell-to-cell movement was proposed to occur by vesicular release of infectious viral material, or immature virions, from the presynaptic cell followed by uptake by the postsynaptic cell57.

Movement across tight junctions

Tight junctions, which are mainly found between adherent epithelial cells, seal cells together to prevent small-molecule and particle exchange (Box 1). They are formed by interactions between transmembrane adhesion proteins, the best characterized of which is claudin. Although tight junctions prevent pathogens from passing freely across epithelial barriers, they seem to have been subverted by some viruses to spread directly between cells. The alphaherpesviruses are the best-studied family that uses tight junctions, not only to spread by cell–cell fusion, but also to travel between cells in a neutralizing antibody-resistant manner in the absence of syncytium formation58. The glycoproteins that are required for this process are gB, gD, gH–gL and gE–gI58. Viruses that lacked gE or gI failed to cause disease in rodents and produced smaller lesions in the corneal epithelium than wild-type viruses59. Moreover, growth of the mutant viruses in culture resulted in smaller plaques, and reduced cell-to-cell spread as inferred from minimal viral replication in the presence of neutralizing antibody59. Subsequent studies confirmed this finding, extended it to VZV and PRV24,58,60,61,62 and unequivocally showed that gE–gI are involved in trafficking virions through the trans-Golgi network to the basolateral surfaces of polarized cells62,63 (Fig. 3b). The mechanism of diverted viral trafficking from apical to basolateral cell membranes (Box 1) relies on gE–gI accumulation at intercellular junctions, potentially mediating interactions with as-yet-unidentified cellular receptors64.

Recruitment of actin-containing structures

Poxviruses. VV is proposed to spread between cells by several mechanisms. For example, the intracellular, mature form of the virus can be released through cell lysis, leading to direct infection of adjacent cells, or release of the extracellular enveloped form and immediate reinfection of neighbouring cells65. However, by far the most efficient mode of cell-to-cell spread, which is resistant to neutralizing antibody, is actin-mediated movement of cell-associated enveloped particles using so-called actin tails (Fig. 4a). Mutations in VV that reduce the production of enveloped viral forms (either free extracellular or cell-associated) or viral gene products that mediate actin-tail formation result in virus particles that are inefficient at cell-to-cell spread65. How VV induces actin tails is described in several recent reviews65,66,67,68 and so will not be discussed here. The mechanism by which virus that is attached to actin tails enters new target cells is unknown, but the same mechanism that allows the entry of extracellular enveloped virus is probably used: through loss of the outer envelope, either in a receptor-mediated manner at the cell surface69 or through disruption within an endosomal compartment, followed by fusion of the inner membrane with a cellular-limiting membrane68,70,71.

Figure 4: Interaction of viruses with actin-containing structures.
figure 4

a | Actin tails induced in vaccinia virus (VV)-infected cells. VV particles can be seen at the tips of actin tails, some of which are interacting with a neighbouring cell. The inset is a higher magnification image of an actin tail with VV at the tip. Actin is labelled green, VV protein A36R is labelled red and colocalization is shown in yellow. The scale bar represents 10 μm. b | Actin tails propel VV into adjacent cells. Thin section electron micrograph of a VV particle at the tip of an actin tail that has entered an adjacent cell. Arrows point to the actin tail and PM indicates the plasma membrane. The scale bar represents 0.5 μm. c | African swine fever virus (ASFV)-induced filopodia in an infected macrophage. Virions labelled in green at the tips of virus-induced actin-containing filopodia are labelled in red. The scale bar represents 8 μm. d | ASFV interacting with a neighbouring cell. Virions labelled in red are seen associating with actin fibres labelled in green within an adjacent cell. The scale bar represents 4 μm. e | Murine leukaemia virus (MLV) trafficking along filopodia to infect an adjacent cell. Actin-containing fliopodia (MLV receptor mCAT1 labelled in red) that extrude from an uninfected fibroblast are anchored to an infected neighbouring cell (plasma membrane labelled green with palmitylated YFP) via MLV envelope glycoprotein-receptor interactions. Virions travel in a retrograde manner from the infected to the uninfected cell. The scale bar represents 5 μm. f | HIV-1 infecting a CD4+ T cell across a nanotube. An HIV-1-infected T cell (bottom) is connected to an uninfected CD4+ T cell (top; CD4 is labelled in green) by a nanotube. HIV-1 Gag is labelled red and Env is labelled blue. Gag and Env can be seen trafficking along the nanotube and within the previously uninfected target cell. The scale bar represents 10 μm. Part a reproduced, with permission, from Ref. 140 © (1999) American Society for Microbiology. Part b reproduced, with permission, from Nature Ref. 78 © (1995) Macmillan Publishers Ltd. All rights reserved. Parts c and d reproduced, with permission, from Ref. 77 © (2006) Blackwell Publishing. Part e reproduced, with permission, from Nature Cell Biology Ref. 81 © (2007) Macmillan Publishers Ltd. All rights reserved. Part f courtesy of C. Jolly, University of Oxford, UK.

VV can also increase cellular motility72 by expressing the F11L protein, which blocks the RhoA signalling cascade that regulates cellular adhesion and motility73. The rapid, random movement of cells on which cell-associated virions are attached would, at least in principle, increase the number of cells infected within a tissue through direct cell–cell contact.

Herpesviruses. HSV-1 (Refs 74, 75) and PRV76 promote the formation of extended, actin-containing protrusions that contact neighbouring cells and that traffic virions in a neutralizing antibody-resistant manner. In PRV, the conserved US3 protein kinase induces the protrusions76. Interestingly, inhibition of the Rho-associated kinase ROCK complemented US3-mediated projection formation in cells infected with a US3-deleted virus76. This is the same pathway that is inhibited by VV F11L to enable cell motility and form long cellular projections that contact distant cells73. Thus, two different families of viruses might have evolved distinct strategies to target the same cellular signalling pathway to increase cell-to-cell spread.

Asfarviruses. ASFV, the sole member of the asfarvirus family, can induce the formation of actin-containing filopodia-like protrusions at the surface of infected cells77 (Fig. 4b). These ASFV-induced structures are distinct from those produced by poxviruses, in that the linear actin fibres are bundled into sheaths, contrasting with the extensively branched and cross-linked arrays of actin observed in VV78,79. After reaching the plasma membrane, ASFV induces localized actin nucleation in a process that requires as-yet-unidentified viral and/or host proteins77.

Retroviruses. Retroviral subversion of actin-containing cellular structures for cell-to-cell spread has been reported for murine leukaemia virus (MLV) and HIV-1. An initial report showed that MLV bound preferentially to filopodia, then 'surfed' along the protrusion towards the cell body, where virus–cell fusion occurred80. Subsequently, uninfected cells were shown to produce filopodia that were anchored through endocytosis into infected cells, resulting in stable associations across which the virus could infect receptor-expressing cells81. Filopodia from uninfected cells were associated at the infected cell plasma membrane by Env-receptor clusters (Fig. 4c). Virions moving from the infected cell remained associated with clusters of receptors, and the underlying receptor association with actin allowed retrograde flow towards the uninfected cell81. The role of the virus in these interactions differs from that reported for poxviruses and ASFV, in that infection did not induce actin polymerization and formation of cellular protrusions, but instead anchored existing actin-rich structures into uninfected cells.

Analysis of HIV-1 movement between T cells revealed the use of actin-containing membrane nanotubes82. Membrane nanotubes are a new form of intercellular communication83 that has been observed in cultures of neuronal cells84, immune cells85, myocytes and endothelial cells86. Intercellular exchange of materials, including calcium ions, lipids, proteins and organelles, takes place along nanotubes83. T-cell nanotubes form spontaneously in vitro, frequently following immunological synapse formation, and can be several hundred micrometres in length82. HIV-1 was shown to transfer along nanotubes from an infected to an uninfected T cell at a speed commensurate with actin-mediated transport. Virus then entered target cells in a receptor- and fusion-dependent manner82 (Fig. 4d), which indicates that viruses are transported along the exterior of the nanotube before conventional receptor engagement and entry. As multiple T cells can be connected by linear and branching nanotubes, this mode of HIV-1 spread might be efficient in secondary lymphoid tissue, which is densely packed with susceptible target T cells.

Virological synapses in the immune system

T cell virological synapses. HTLV-1 and HIV-1 can spread directly between immune cells by forming an infectious, or virological, synapse14,87,88. Unlike most other cells in the body, immune cells are, in general, neither permanently fixed within a tissue nor constitutively polarized (Box 1). Thus, exchange of information and material between these cells takes place by the formation of transient adhesive contacts and polarization of receptors and ligands at the intercellular interface. Some of the known molecular interactions that take place at the intercellular interface are shown in Fig. 5. Virological synapses are related to, but distinct from, immunological synapses (Box 2). The observation that the microtubule organizing centre (MTOC) polarizes at HTLV-1 (Refs 89, 90) and HIV-1 (Q.S., unpublished observations) virological synapses in T cells is consistent with the notion that the regulated secretory machinery of the T cell is activated by viral infection89,90,91. The signals that drive this mechanism have not been determined for HIV-1, but HTLV-1-infected T cells express the viral transactivator protein Tax, which accumulates at the MTOC. Engagement of intercellular adhesion molecule 1 (ICAM1) on the infected cell by its cognate ligands initiates MTOC relocation by an as-yet-undefined pathway90, and Tax seems to synergize with this pathway, resulting in efficient MTOC polarization to virological synapses92.

Figure 5: Anatomy of a retroviral virological synapse.
figure 5

A hypothetical virological synapse between a retrovirally infected T cell (left) and a receptor-expressing target cell (right). a | HIV-1 envelope glycoproteins (Env) are expressed on the infected cell plasma membrane and interact with the receptors CD4 and CCR5 or CXCR4 on the target cell132. The adhesion molecules intercellular adhesion molecule 1 (ICAM1) and lymphocyte function-associated antigen 1 (LFA1) engage to stabilize the cellular conjugate102. ICAM1 can signal into a human T-lymphotrophic virus-1 (HTLV-1)-infected cell to polarize the secretory apparatus90. The secretory apparatus, exemplified here by the microtubule organizing centre (MTOC), aligns proximal to the cell–cell contact zone. MTOC polarization is enhanced in HTLV-1-infected cells by the viral Tax protein, which associates with the MTOC92. b | In the HIV-1 virological synapse, viral assembly and budding are polarized towards the synapse, and virus is released into the synaptic cleft before fusing with the target cell plasma membrane132.

APC–T cell virological synapses. The best studied form of immunological synapse is that assembled between antigen-presenting cells (APCs), particularly dendritic cells (DCs), and lymphocytes during lymphocyte activation93. HIV-1 has taken advantage of this interaction to move from DCs into T cells, and this has been proposed as a mechanism of early spread of HIV-1 from mucosal surfaces to secondary lymphoid tissue during mucosal virus transmission94,95. DCs can spread HIV-1 to T cells after DC infection with HIV-1 or after short-term DC 'pulsing', during which the virus seems to be 'archived' prior to DC encounter of the T cell96,97,98. The C-type lectin DC-SIGN (DC-specific intercellular adhesion molecule-3-grabbing non-integrin) could act as a viral attachment receptor to keep HIV-1 at the DC surface99 or could mediate internalization of the virus into an as-yet-undefined compartment in which the virus remains transiently infectious100. Other molecules, such as heparan sulphate proteoglycans101, might have similar roles that could allow the virus to be trapped and presented to T cells during immunological synapse formation. Adhesion molecules could influence viral infectivity and formation of virological synapses: the integrin lymphocyte function-associated antigen 1 (LFA1) has been implicated in the capture of free virions through its cognate adhesion partner ICAM1 (Ref. 3), which seems to be central to virological synapse assembly and stability102. A novel interaction in which the HIV-1 glycoprotein gp120 binds the α4β7 integrin, which is involved in the homing of T cells to gut-associated lymphoid tissue (GALT), could provide insight into the dramatic tropism of HIV-1 for GALT during the acute infection103. The interaction of gp120 with α4β7 is proposed to activate LFA1, which could increase HIV-1 cell-to-cell spread across virological synapses in GALT103.

Macrophages can spread HIV-1 directly to T cells across a virological synapse-like interface104. Because macrophages are readily infected with HIV-1 and produce and/or store large amounts of the virus over extended periods of time, they might contribute substantially to HIV-1 cell-to-cell dissemination105. Notably, the macrophage–T cell synapse seems to be transient, as adherent T cells become rapidly infected and then spontaneously detach106. This implies that a single infected macrophage can infect high numbers of T cells over its lifespan, thought to be a number of weeks, in vivo.

Perspectives and future directions

The effects of cell-to-cell spread on viral pathogenicity extend beyond the efficiency of spread within and between tissues. One of the early operational definitions of cell-to-cell spread was that of neutralizing antibody resistance, as exhibited by some herpesviruses107, even though other viruses, such as CMV, might be neutralizing antibody sensitive30. Escape from neutralizing antibody inhibition of cell-to-cell viral dissemination has also been reported for poxviruses108 and hepatitis C virus (HCV)109. For retroviruses, it is unclear whether antibodies can effectively neutralize virus spread from cell to cell. Some reports suggest that HIV-1 spread through this route is resistant to neutralization, whereas others imply HIV-1 is in fact neutralization susceptible110,111,112, and therefore more work is clearly required. Ultrastructural analysis of HTLV-1 virological synapses shows that the plasma membranes of the infected and target cells are tightly apposed, which could prevent access of antibodies and complement components to virions in transit between cells113. Whether the kinetic advantage imposed on viruses by cell-to-cell spread can reduce the impact of cell-mediated immunity remains to be seen: it is conceivable that the speed with which viruses assemble and pass directly between cells might reduce the ability of cytotoxic effector mechanisms to eliminate infected cells rapidly enough to prevent virus propagation. HIV-1 infection has been proposed to dysregulate immune responses beyond that expected by simple elimination of CD4+ T cells. For example, the virus might disrupt the function of immunological synapses, as described for HIV-1 infected T cells114, potentially by disrupting signal transduction pathways shared by both virological and immunological synapses115.

Where is this field of virology heading? For herpesviruses, orthomyxoviruses, poxviruses and rhabdoviruses, cell-to-cell spread is an accepted mode of dissemination that is well characterized. For other viruses, such as HIV-1 and HCV, the mechanisms that underlie cell-to-cell spread have yet to be elucidated and its importance for viral transmission, tropism, immune evasion and overall pathogenesis in vivo remains unclear. Progress in understanding how viruses subvert cellular processes to move between cells will add to our knowledge of how receptors used by viruses signal into cells, how the cytoskeleton functions in health and disease8,9,116, and how non-polarized cells can be induced to polarize their secretory apparatuses91. This will be satisfying, as aspects of constitutive cell polarity were initially characterized in the context of viral protein trafficking 30 years ago117,118.

There are obvious parallels between viral and bacterial subversion of the cell cytoskeleton for infection and dissemination66,67. For example, the parallel strands of actin that propel Rickettsia spp. resemble those found in filopodia, and might therefore be analogous to the assemblies that are induced by ASFV77 and herpesviruses76. Comparative analyses of the different types of interaction between viruses and prokaryotes and their host cells will enrich not only our understanding of microbial pathogenesis, but may also shed light on molecular mechanisms that underlie normal cellular processes.

Viruses are subject to intrinsic mechanisms of intracellular antiviral restriction. These are exemplified by the family of cytidine deaminases of which the prototype, APOBEC3G, interferes with retroviruses119 and hepadnaviruses120, the TRIM family of inter-species retroviral restriction factors121,122 and the newly described tetherins123. Some of these factors, such as the TRIMs, are saturable, meaning that if sufficient amounts of the virus can be introduced into a cell the restriction can be overcome. It will be interesting to see whether the enhanced efficiency of cell-to-cell spread, compared with cell-free spread, of viruses can overcome such resistance to infection.

Further characterization of the molecular basis of viral cell-to-cell spread will increase our understanding of viral pathogenesis and could help uncover new avenues of therapeutic interest, either in terms of novel viral targets that are specifically involved in cell-to-cell spread or cellular targets that facilitate this mode of virus dissemination. Genome-wide functional screens are already providing potential new antiviral targets124,125,126,127 that could allow us to reduce or prevent cell-to-cell spread by human pathogenic viruses.