To ensure the recruitment of RNPs to specific areas of the plasma membrane in which GP spikes are concentrated, the production of infectious particles requires a highly organized and well-coordinated process of viral protein-protein interactions during assembly. Since the glycoprotein complex mediates both target cell binding and the subsequent fusion of the viral and host cell membranes during entry, the incorporation of GP into budding particles is essential for facilitating continued infection. The arenavirus glycoprotein is synthesized as an inactive precursor preGP-C that is co-translationally cleaved by signal peptidase into GP-C and the stable signal peptide (SSP) [95,96]. Post-translational maturation cleavage of GP-C by the host cell proprotein convertase S1P (site 1 protease), also known as SKI-1 (subtilisin kexin isozyme-1), then leads to the generation of the distal receptor-binding subunit GP-1 and the transmembrane-spanning fusion competent subunit GP-2 [97,98,99,100]. Since only cleaved subunits are incorporated into budding virions, proteolytic processing of GP-C by S1P is absolutely necessary for the production of infectious particles [98]. Remarkably, the SSP is not only essential for S1P-mediated activation cleavage of GP-C by functioning as a trans-acting maturation factor [101], but remains stably associated with GP, primarily through interaction with a zinc-binding domain present in the cytoplasmic tail of GP-2 [102,103,104]. Together with GP-1 and GP-2, SSP forms the tripartite glycoprotein spike complex on the viral surface, making the mature glycoprotein spikes of arenaviruses unique among enveloped single-stranded RNA viruses [50,105].

In spite of the progress made in understanding the structural details of arenavirus glycoproteins, the intracellular trafficking pathway and the maturation of GP, the mechanism by which the tripartite mature GP complex is incorporated into the lipid bilayer of the nascent virions during virus assembly is still only partially understood. The Z protein has been implicated in this process based on observed intracellular interaction between GP and Z (Figure 5) [72,106]. GP and Z are also capable of interacting in VLPs [106], which coincides with their structural organization in arenavirus particles as observed using cryo-electron microscopy [51]. The N-terminal myristoylation of Z is critical for GP-Z interaction, whereas both the RING domain and the C-terminal late domains are non-essential for GP-Z association [72]. Interestingly, Z directly associates with SSP, even in the absence of other subunits of the GP complex [72]. However, the molecular nature of the SSP-Z interaction still remains to be determined. The identification of the protein-protein interaction motifs within SSP and the Z protein will thus provide further insight into the mechanism underlying the incorporation of GP into nascent virions. In particular, defining the precise role of N-myristoylation of Z in this process will be of great interest, as myristoylation of Z has been linked to various biological functions, including intracellular targeting [71]. Suppression of Z myristoylation - either by mutation of the conserved glycine residue at position 2 (Z-G2A), or by treatment of wild-type Z protein expressing cells with myristic acid analogs - adversely affected the subcellular localization of Z, altering it from a characteristic punctuate cytoplasmic distribution pattern to a more diffuse and predominant perinuclear accumulation [71]. In addition, N-myristoylation has been reported to have a strong impact on protein structure [107]. Further studies are therefore required to elucidate whether the limited association observed between GP and Z-G2A is due to altered intracellular localization or to structural changes of mutant Z that perturb efficient GP- Z interaction.

Association of Z with GP determines directed LASV release from polarized epithelial cells. In dissecting the viral determinants responsible for polarized virus budding, individually expressed Z protein was shown to induce VLP budding from both apical and basolateral surfaces, whereas the addition of GP shifted VLP release to the apical plasma membrane [106]. Accordingly, infectious LASV buds from the apical membranes of polarized kidney cells [106].

It is well established that arenavirus particles assemble at the plasma membrane, therefore virion morphogenesis requires GP-Z co-localization at the site of virus budding. However, neither confocal microscopy [106] nor dual-label immuno-gold staining and electron microscopy [108] produced observable GP-Z association at the plasma membrane. Instead, co-localization between these two proteins was observed in vesicle-like structures in the cytoplasm near the nucleus [106], suggesting the potential existence of an intracellular co-trafficking mechanism to virus assembly sites. Although specific viral assembly pathways remain elusive for arenaviruses, data from our laboratory indicate involvement of late endosomal compartments as a trafficking intermediate for LASV Z protein (S.K. Fehling and T. Strecker, unpublished data). Interestingly, for Marburg virus (MARV), a member of the filovirus family, it has been previously reported that the matrix protein VP40 actively recruits the envelope glycoprotein GP to VP40-enriched late endosomal compartments. Accumulation of MARV GP and VP40 in this compartment is assumed to play an important role in the formation of the filoviral envelope and subsequent release of infectious MARV at the plasma membrane [109,110]. It will be interesting to assess whether distantly related arena- and filoviruses share similarities with respect to their intracellular assembly pathways.

In summary, the collaboration between GP and Z is critical for the production of infectious particles, although the signals and underlying mechanisms for incorporation of glycoprotein spikes into Z-driven virions during assembly remain to be identified.

Due to the bi-segmented nature of the arenavirus genome, a highly coordinated and selective packaging mechanism has to ensure that infectious progeny virions contain at least one of each genome segment during assembly. Arenavirus replication takes place in the cytoplasm of infected cells. Cell cultures infected with LASV and other arenaviruses typically reveal distinctive intracytoplasmic electron-dense viral inclusions that are suggested to be important sites for RNA replication (Figure 6) [111,112]. Thus, the RNP complex not only has to be recruited from these cytoplasmic locations to virus budding sites at the plasma membrane, but also requires tethering to budding membranes in order to be incorporated into nascent virions. Early studies using chemical crosslinking have shown that LCMV Z is associated with the viral nucleocapsid in purified particles [53], suggesting that Z is also involved in these processes. Since the RNP complex is composed of NP, L, and genomic RNA, each of these three components could, in principle, harbor the interaction sites that are important for Z-driven RNP packaging. For many other negative-stranded RNA viruses, such as EBOV, interaction between the matrix protein and the nucleoprotein is crucial for nucleocapsid incorporation into nascent virions [113]. Indeed, initial evidence for direct Z-NP interaction in the absence of other viral proteins has already been demonstrated for LASV (Figure 5) [54]. Follow-up studies showed that the intracellular collaboration of NP and Z leads to the incorporation of NP into LASV Z-induced VLPs [106]. Similar findings have been described for NW JUNV and TCRV, as well as for OW LCMV and MOPV [114,115,116,117,118], suggesting that NP-Z association is conserved within the Arenaviridae family. Collectively, these data indicate that interaction of NP with Z is important for RNP incorporation during virus assembly. Although recent studies have greatly improved our knowledge of the molecular basis for NP-Z interaction, precise details still remain to be determined. Structural studies of LASV NP and functional analysis of NPs from LASV, LCMV, and TCRV have shown that NP consists of two domains [115,116,119,120]. The N-terminal domain is involved in RNA binding and oligomerization [116,121,122], while the C-terminal domain of NP facilitates type I IFN antagonism [115,119,120,123], oligomerization [121], and association with Z [114,115,116]. Interaction between NP and Z requires an intact RING domain of Z, as it has been shown that mutations of the RING structure drastically affected Z-NP interactions and the ability of JUNV Z to incorporate TCRV NP into budding VLPs [118]. However, future studies are necessary to elucidate whether the functional or structural integrity of the RING domain is of critical importance for the reported observations regarding the role of the RING domain in Z-mediated incorporation of NP into VLPs. Also, mutation of residue L79 to alanine within JUNV Z protein severely impaired intracellular Z-NP interactions and the incorporation of NP into VLPs, resulting in decreased VLP infectivity [118]. This observation has been extended to LASV, since mutation of the corresponding amino acid L71 within LASV Z protein similarly affected the infectivity of recombinant VLPs, supporting the role for Z in the packaging of RNPs into mature infectious virions [89]. Interestingly, for MOPV it has been implied that the host cell protein Alix/AIP1 (ALG-2-interacting protein 1) directly participates in Z-NP interaction, presumably by bridging both proteins [124].

Although the interaction between NP and Z plays a central role in genome packaging, additional mechanism(s) for RNP incorporation may exist. Recent studies have demonstrated that the Z protein also binds to the polymerase L, leading to the formation of Z-L heterodimers [85,91]. While the association between Z and L has been shown to play an important role in regulating viral genome replication and transcription, a recent study by Kranzusch and colleagues suggests that Z-L interaction plays a direct role in ensuring the packaging of the viral polymerase as part of the functional RNP complex [91]. This is particularly important for arenaviruses, since the viral genome per se is not infectious, making polymerase packaging crucial to the production of infectious particles. Interaction with Z renders the polymerase catalytically inactive. Importantly, interaction with Z does not prevent L from binding to viral RNA, as demonstrated by the ability of the Z-L complex to bind to viral promotor sequences [91]. Instead, interaction with Z prevents the early steps of RNA synthesis initiation, therefore keeping the polymerase locked and bound to viral promotor sequences, and hence associated with the viral RNPs. In addition, in vitro studies have demonstrated that Z is able to inhibit on-going RNA synthesis, indicating that Z can also contact L during active replication and/or transcription of the viral genome [91]. Whether Z ensures polymerase packaging by interaction of L in the context of a functionally active RNP or whether Z forms a heterodimeric complex with L, which subsequently associates with RNPs through interaction with the viral RNA, remains to be determined.

The matrix proteins of several different RNA virus families, such as bornaviruses, orthomyxoviruses, rhabdoviruses and retroviruses, possess RNA binding capacities, which may have implications for the incorporation mechanism of RNPs into the membrane envelope during budding [92,125,126,127]. The arenavirus Z protein largely binds to nucleocaspids through its interaction with NP, and since the arenavirus NP and L protein are larger and more positively charged than the NP and L of other negative-strand viruses, it is likely that they function as the major RNA-binding proteins. The Z protein central RING domain is a globular type of zinc-binding domains that more frequently binds other proteins, in contrast to zinc-finger domains which are known for intercalating into nucleic acids [128]. Though, the potential RNA binding activities of Z remain to be established.

Similarly, knowledge of the mechanism underlying intracellular RNP delivery to viral budding zones remains elusive. The functional relationships between Z and NP or Z and L suggest a potential co-trafficking process of the RNP complex via Z trafficking pathways. Consistent with this assumption, the co-expression of Z leads to a redistribution of NP to Z-positive dot-like structures, while solitary expressed NP is evenly distributed in the cytoplasm. Z-NP co-localization was observed either at the plasma membrane, as shown for TCRV [116,118], or within cytoplasmic patches, as documented for LASV, LCMV and MOPV [54,106,114,115]. In contrast, GP seems not to be involved in RNP trafficking since neither a direct interaction nor intracellular co-localization between GP and NP has been detected, which is consistent with the observation that NP is not incorporated into VLPs generated by solitary expression of GP [106]. Co-localization of both proteins in distinct perinuclear clusters was only observed in cells additionally expressing Z protein [106]. Taken together, these observations led to the idea that Z serves as a ‘bridge’ between GP and NP, and therefore links the viral envelope and viral RNP complexes, making Z a key mediator in virus assembly processes. Cryo-electron microscopy of viral particles confirmed that Z is tightly associated with GP and NP, forming a dense layer between both proteins [51]. In a heterologous VLP system using chimeric JUNV/TCRV, the co-expression of TCRV NP greatly enhanced the incorporation of JUNV GP into JUNV Z-induced VLPs, indicating that Z regulates incorporation of GP into nascent virions through interaction with NP [118]. However, when viral protein recruitment was examined in the homologous context, a significant increase in GP incorporation into Z VLPs in the presence of NP was observed neither for NW JUNV and TCRV nor OW LASV [106,117].

Altogether, Z protein is a key regulator of virus assembly. Through its interaction with GP, NP and L, the Z protein ensures the incorporation of viral RNP complexes into budding particles that contain glycoprotein spikes, leading to the release of infectious virions.

Arenaviruses exit their host cells by budding from the plasma membrane (Figure 7A). The final steps prior to virus release involve the envelopment of the viral nucleocapsid by the host cell-derived lipid membrane that harbours mature GP spikes, followed by a subsequent membrane fission event during which the nascent virion is separated from the host cell membrane. Z protein is the key player in these processes. On the one hand, the Z protein mediates all of the essential virion assembly events that ensure the packaging of all viral components required for infectivity, while on the other hand, Z protein alone facilitates virus-host protein-protein interactions that are necessary but also sufficient to create particles (Figure 7B). This assumption is based on the observation that solitary expression of Z protein leads to the formation and release of lipid-enveloped, morphologically regular VLPs at the plasma membrane [52,54,129]. This self-budding activity has been linked to the late domains (Figure 4B) [52,130]. The functional integrity of the late domains is of critical importance for the LASV and LCMV Z-driven release of enveloped particles, since disruption of these motifs abrogated efficient VLP production [52,130].

Viral late domains are highly conserved motifs and have been found within viral matrix proteins of numerous enveloped RNA viruses, including rhabdoviruses, filoviruses, and paramyxoviruses as well as within the Gag proteins of various retroviruses (reviewed in [80,131]). Through interaction with individual proteins of the cellular ESCRT complexes, they facilitate the recruitment of the VPS machinery to virus budding sites. The VPS pathway normally functions in the downregulation of cell surface receptors and their sequestering into MVBs, which subsequently deliver their intraluminal cargo for lysosomal degradation. Notably, the sorting of protein cargo into the lumen of the MVB occurs by a vesicular budding event that is catalyzed by the coordinated action of the ESCRT complexes [132,133]. The ESCRT machinery comprises the four multiprotein complexes, ESCRT-0, ESCRT-I, ESCRT-II, ESCRT-III, as well as several ESCRT-associated factors that include Alix/AIP1 and the AAA-ATPase Vps4. ESCRT-0 plays a role in recruitment of ubiquitinated cargos and their concentration at the endosomal membrane. ESCRT-I and II cooperate to induce membrane bud formation and confine cargo, while the ESCRT-III protein complex causes membrane scission that results in the release of vesicles into MVBs [134]. Finally, Vps4 plays an essential role in the completion of vesicle formation by disassembling ESCRT-III components from the target membrane. Because both processes are directed away from the cytosol, ESCRT-driven budding of vesicles into the lumen of the MVB topologically resembles the budding of enveloped particles from the plasma membrane (reviewed in [135,136,137]). It is therefore not surprising that viruses recruit the cellular ESCRT machinery to organize budding of viral progeny, mediated through their late domains. Viral proteins that encode P[T/S]AP late domains gain access to the ESCRT pathway through interaction with the ESCRT-I component Tsg101 (tumor susceptibility gene 101) [138,139,140], while the YxxL motif mediates binding to Alix/AIP1, which is an ESCRT-I and III binding partner [141,142,143]. The PPxY late domain motif has been linked to interactions with various members of the family of mammalian Nedd4 (Neuronal precursor cell-expressed developmentally downregulated 4) E3 ubiquitin ligases (hereafter called Nedd4-like proteins) [144,145,146,147,148,149,150].

Tsg101 has been reported to be involved in LCMV and LASV budding. Perturbing Tsg101 function by RNA interference resulted in decreased levels of both LASV and LCMV Z-mediated budding [129,130]. However, the mechanism by which Tsg101 is recruited to virus budding sites is not fully understood. Particularly in the case of LCMV it will be of interest to elucidate the molecular relationship between Z protein and Tsg101, since Z lacks the classical Tsg101-binding motif PTAP. Direct interaction between Tsg101 and Z protein has not yet been established and it remains to be determined whether the late domains within Z play a role in this process. To investigate Z-Tsg101 interaction, wild-type LASV Z protein or Z mutants with individually disrupted PTAP and PPPY late domains were tested for binding to Tsg101 in a co-immunoprecipitation assay. Wild-type Z protein is efficiently co-precipitated with Tsg101 following recombinant expression in human embryonic kidney 293T cells (Figure 8A and B). Mutation of the PTAP motif (ATAA) resulted in a decreased ability of Z to interact with Tsg101, demonstrating that - in the case of LASV - the PTAP motif contributes to, but is not the sole determinant of Z-Tsg101 binding. Remarkably, the PPPY motif appears to be very important for interaction with Tsg101. Mutation of PPPY to PPPA (which has been shown to be sufficient to block VLP release just as effectively as a PPPY deletion mutant [52]) completely eliminated the association between LASV Z and Tsg101 (Figure 8B). Consistent with these observations, the PPPY motif plays a dominant role in Z-mediated budding, whereas the effects of mutating the PTAP late domain were much less severe [52]. A very similar observation has been made for the interaction of Tsg101 with the MARV matrix protein VP40, which was reported to be dependent on the presence of a functional PPPY motif [151]. In contrast, mutations of the PPPY motif within the Gag protein of human T-cell leukemia virus type 1 (HTLV-1) had no effect on its ability to bind Tsg101 [152], suggesting that various viruses have evolved different mechanisms for recruiting Tsg101.

Because Nedd4 and Nedd4-like proteins have been implicated as critical host factors for PPxY motif-dependent budding, the requirement of a functional PPPY motif indicates that upstream ubiquitin ligase activity may be necessary before Z protein recruits Tsg101. Nedd4 and Nedd4-like ubiquitin ligases are part of an ubiquitination enzyme cascade that involves an ubiquitin-activating enzyme E1 and an ubiquitin-conjugating enzyme E2. The final step mediated by E3 ubiquitin ligases results in the covalent attachment of ubiquitin to target proteins. Ubiquitination often plays a central role in recruitment of proteins by the ESCRT machinery, which may explain also the link between ubiquitin ligases and virus budding [153,154]. The potential involvement of Nedd4 in Z-mediated budding is demonstrated by the observation that LASV Z protein efficiently binds to this host cell protein (Figure 8C). Co-immunopreciptation studies also show that this interaction requires an intact PPPY motif, since the PPPA mutant failed in its ability to bind to Nedd4, while mutations of the PTAP motif have no effect of Z-Nedd4 interaction (Figure 8C). In addition, several Nedd4-like proteins, including WWP1, WWP2, Itch, Smurf1, and Smurf2, are capable of interacting with Z protein (T. Strecker, unpublished data). These observations indicate that Nedd4 or Nedd4-like proteins may link the Z protein to the VPS pathway, although the exact contribution of the individual ubiquitin ligases in Z-mediated budding remains to be determined. However, previous work has shown that overexpression of a dominant-negative mutant of Nedd4 and depletion of Nedd4 by siRNA does not reduce LASV Z-driven VLP production [129]. Notably, human Nedd4 proteins exist in at least two principal isoforms designated Nedd4-1 and Nedd4-2, which are encoded by different genes [155]. Nedd4-1 (but not Nedd4-2) is recruited by HTLV-1 Gag and subsequently facilitates its ubiquitination [149]. More work will be necessary to elucidate the details of the specific role of individual Nedd4 proteins in arenavirus budding. The observation that LASV Z is physically able to interact with Nedd4 and Nedd4-like proteins raises the question of a potential role of ubiquitination of Z and its functional relevance in arenavirus budding. The Z protein of LASV exhibits multiple ubiquitinated forms in cells that also overexpress ubiquitin, indicating that Z protein is a target for ubiquitination. Interestingly, a similar protein pattern of Z was observed in cells that coexpress various Nedd4-like proteins (T. Strecker, unpublished data). Future studies will be aimed at understanding the precise function of ubiquitination in arenavirus budding.

The molecular details of how the Z-Tsg101 and Z-Nedd4 interactions promote virus budding from the plasma membrane are currently being elucidated in our laboratory. For LASV we propose a model in which Z interacts successively with Nedd4 and Tsg101. Both factors were efficiently incorporated into Z-specific VLPs, supporting the concept of their intracellular interaction (Figure 8D). Interaction with Nedd4 is assumed to lead to ubiquitination of Z and subsequent Tsg101-mediated recruitment of the ESCRT machinery, which then facilitates the budding of virus particles. Tsg101 not only interacts with PTAP late domain motifs, but also recognizes ubiquitin that is attached to proteins [156,157], which could explain the mechanism underlying the participation of Tsg101 in PPxY motif-dependent LCMV Z protein budding. TCRV Z protein buds from cells despite lacking the classical late domain motifs P[T/S]AP and PPxY and does not require Tsg101 activity [158]. Additionally, the potential late domain motifs ASAP and YLCL found in TCRV Z do not play a role in Z self-budding activity [117,158], indicating that TCRV Z employs a budding mechanism distinct from that found in other arenaviruses. The differences in type and number of late domains within various Z proteins are still not fully understood and it is not currently known whether the P[T/S]AP and PPxY motifs act in a redundant or synergistic way in arenaviruses in which both motifs are present. Similarly, further studies are required to address whether the individual late domains act in a cell type-dependent manner. For example, the PSAP motif within the matrix protein of vesicular stomatitis virus was recently reported to contribute to cell type-dependent virion egress [159]. The identification of cell type or host cell-species specific factors will be important to better understand the mechanisms that determine efficient arenavirus budding, and thus viral tropism. In spite of the mechanistic differences between the ways various arenaviruses utilize the cellular VPS pathway for virus budding by recruiting different components of the ESCRT machinery, they all share the requirement of Vps4 for the final release of virus particles from the host cell membrane [129,158]. Although many questions still remain unanswered, current results highlight the important role of Z protein in the final stage of arenavirus budding. In order to better understand the sequence of events that finally lead to the release of infectious arenavirus particles, one of the major future tasks is to identify the individual ESCRT components that are necessary and also sufficient to mediate the budding process.