L2, the minor capsid protein of papillomavirus

Abstract

The capsid protein L2 plays major roles in both papillomavirus assembly and the infectious process. While L1 forms the majority of the capsid and can self-assemble into empty virus-like particles (VLPs), L2 is a minor capsid component and lacks the capacity to form VLPs. However, L2 co-assembles with L1 into VLPs, enhancing their assembly. L2 also facilitates encapsidation of the ~8kbp circular and nucleosome-bound viral genome during assembly of the non-enveloped T=7d virions in the nucleus of terminally differentiated epithelial cells, although, like L1, L2 is not detectably expressed in infected basal cells. With respect to infection, L2 is not required for particles to bind to and enter cells. However L2 must be cleaved by furin for endosome escape. L2 then travels with the viral genome to the nucleus, wherein it accumulates at ND-10 domains. Here, we provide an overview of the biology of L2.

Introduction

Human papillomaviruses (HPV) are the etiologic agent of 5% of all lethal cases of cancer worldwide, and even benign infections can cause considerable morbidity for patients and expense to the healthcare system. While the oncogenic HPV genotypes are the best studied, many fundamental observations have first been made with animal papillomaviruses (zur Hausen, 2002). Productive infection by papillomaviruses occurs in many vertebrates and exhibits a strict host tropism and a requirement for epithelial differentiation. Animal papillomaviruses, notably cottontail rabbit papillomavirus (CRPV), bovine papillomavirus (BPV), rabbit oral papillomavirus (ROPV) have been invaluable models for virologic, and especially vaccine studies. These animal papillomavirus models remain important because of many technical challenges associated with the oncogenic HPV genotypes; HPV cannot be grown in cultured cancer cell lines, obtaining virions of oncogenic HPV genotypes from clinical lesions is not feasible, infection is not lytic (and thus fails to produce a readily measurable phenotype), and finally infection of animals with HPV does not produce lesions due to strict host tropism. Over the years, many of these hurdles have been circumvented through the use of organotypic raft culture to produce native virions (Conway et al., 2009b; Meyers et al., 1997), or the codon optimization of the capsid genes enabling robust L1 and L2 expression for the formation of pseudovirions (PsV) carrying a reporter construct (Buck et al., 2004) that can be used to challenge mice or even primates (Roberts et al., 2007; Roberts et al., 2011). While much of the seminal work was performed with animal papillomavirus, fortunately there appears to be a strong conservation of both the domains and functions of L2 across the Papillomaviridae.

At present, there are over a hundred known HPV genotypes that are categorized into 5 genera (alpha, beta, gamma, mu and nu) based upon their genomic sequence (Bernard et al., 2010; de Villiers et al., 2004). They can also be subdivided based on their tropism for either cutaneous or mucosal epithelium respectively, as well as their association with cancer or solely benign lesions. Cutaneous HPV types are typically “low-risk” as their infections are either asymptomatic or result in self-limited and benign tumors, as seen for the gamma types such as HPV1 and HPV2 that cause warts. The beta types, notably HPV5 and HPV8, are typically asymptomatic in healthy individuals, but have been associated with non-melanoma squamous cell carcinoma (SCC) in patients with epidermodysplasia verruciformis (a hereditary condition resulting from germline mutations in EVER1 or EVER2), or upon immune suppression due to HIV co-infection or drug treatment upon solid organ transplantation. Most mucosal HPV types are also benign, such as HPV6 or HPV11, but a dozen or so “high risk” or “oncogenic” genotypes are carcinogenic. While the majority of oncogenic HPV infections are self-limited and subject to immune clearance, persistent infections are associated with a dramatically increased risk for the development of cervical, other anogenital and/or oropharanygeal cancers. HPV types 16 and 18 account for 50 and 20% of all cervical cancer cases respectively. Other high/intermediate risk types such as HPV 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68 constitute the remaining 30% HPV-associated malignancies and are associated with a slower onset of dysplasia (de Martel et al., 2012). Importantly, HPV16 is even more predominant (~90%) in other anogenital and oropharyngeal malignancies (Buck and Trus, 2012; zur Hausen, 2002). Given its disproportionate impact on human health, most studies of L2 biology are concentrated on HPV16.

The circular and double-stranded papillomavirus genome encodes six ‘early’ proteins (E1, E2, E4, E5, E6, and E7) and two ‘late’ proteins (L1 and L2), subdivided based on their spatial-temporal expression pattern during the virus life cycle. The ‘early’ proteins can be further subdivided into two regulatory genes involved in replication and transcription (E1 and E2), three oncogenes (E5, E6 and E7), and E4 which contributes to virion production and actually exhibits an expression pattern closer to the late proteins. The late proteins L1 and L2 are structural components of the viral capsid (Doorbar et al., 2012). Expression of L1 and L2 is not detected in infected basal epithelial cells, but they are both detected in the nuclei of the terminally differentiated cells in the uppermost layers of the squamous epithelium, appearing first even later (i.e. higher in the epithelium) than E4. When produced in a variety of recombinant expression systems, L1 can self-assemble, to form empty virus-like particles (VLPs) that are the basis of the licensed HPV vaccines (Kirnbauer et al., 1992). L2 does not form VLPs but can be incorporated when co-expressed with L1. In this review, we will be focusing on the biology of the minor capsid protein, L2, which plays a key role both during virion assembly and the infectious process.

The L2 protein

L2 is just under 500 amino acids in length, which corresponds to an estimated molecular mass of approximately 55KDa. However, L2 typically exhibits an apparent molecular weight of 64-78kDa by SDS-PAGE analysis (Doorbar and Gallimore, 1987; Jin et al., 1989; Komly et al., 1986; Rippe and Meinke, 1989). The reason for this phenomenon is unclear as there are no known post-translational modifications of L2 (with the exception of modification by SUMO of lysine 35 of HPV16 L2 (Marusic et al., 2010)), and a similar size is observed for L2 produced in bacteria. Rose et al reported that L2 in native HPV11 virions exhibited a doublet, but the size of L2 was not impacted by glycosylase treatment, suggesting that glycosylation was not a factor, and the lower molecular weight form might instead represent proteolytic cleavage (Rose et al., 1990). L2 has several key functional roles and numerous interacting partners (Table 2). The main domain sequences used by L2 have been mapped through deletion and/or mutagenesis studies (Figure 1). Much of our discussion and analysis here will be based on HPV16 L2’s primary sequence and these mapped domains because very little information exists onits higher order structure.

Table 2.

Published list of known protein interactions with PV L2.

Stage of PV Virus life cycle	Protein that interacts with L2	Interacting region on L2	Does the protein recognize a specific motif/consensus sequence on L2?	Purpose of interaction	Reference
Prior to infectious cell entry	Furin	9-12	R-X-K/R-R	Cleavage of L2	Richards et al., 2006
	Cyclophilin B (CyPB)	90-110	N/A	Assist in capsid conformational change for secondary receptor uptake	Bienkowska-Haba et al., 2009
	Annexin A2 Heterotetramer	108-120	Neutralizing epitope region 108-120	Putative secondary receptor	Woodham et al., 2012
Vesicular trafficking & Endosome escape	Cyclophilin B (CyPB)	90-110	N/A	Assist in capsid disassembly in the late endosome	Bienkowska-Haba et al., 2012
	Gamma-secretase (GS)	Unknown	Unknown	Facilitates L2/vDNA endosomal escape	Karanam et al. 200
	Sortin Nexin-17 (SNX17)	245-257	F/YxNPxF/Y		Bergant Marusic et al., 2012 Bergant and Banks et al., 2013
Nuclear transport	Syntaxin 18 (BPV1 only)	43-47	DQ/KILQ/K	Transport of L2 toward nucleus	Bossis et al.2005 Laniosz et al., 2007a
	Heat Shock cognate protein 70 (Hsc70)	Unknown	N/A		Florin et al., 2004
	Beta-Actin	25-45	N/A		Yang et al.,2003b
	Dynein motor proteins (DYNLT1 & 3)	456-460, 457-461 (HPV 16 L2)	R/KR/KXXR/K		Florin et al., 2006 Schneider et al., 2011
	Karyopherins (Kapβ2 and Kapβ3)	1-9 or 454-462 (NLS signals at the amino and carboxyl termini respectively)	N/A		Bordeux et al.,2006; Darshan et al.,2004; Klucevsek et al., 2006; Fay et al., 2004; Sun et al., 1995
Viral gene regulation, morphogenesis and assembly	ND10 domain and binding to Daxx	390-420	N/A	Recruits other viral proteins and viral DNA to initiate viral gene transcription or assembly?	Becker et al., 2003 Florin et al.,2002b
	PV E2	1-50 and 301-400	N/A	L2 brings E2 to the ND10 domains (possibly to initiate viral gene transcription or replication)	Okoye et al., 2005 Heino et al., 2000
	PV L1	412-455	PXXP	Viral capsid assembly	Finnen et al.,2003 Lowe et al.,2005
	TBX2 and TBX3	L2 C-terminus region	N/A	L2 brings TBX2/3 to the viral genome’s LCR where TBX2/3 then represses of viral gene transcription	Roanowski et al., 2013
Other known interaction protein partners but with undefined function	PATZ TIP60 TIN-AG-RP PLINP	Unknown	Unknown	Unknown	Gronemann et al., 2002
	SUMO (small ubiquitin-related modifiers)	34-37	PVKE	Unknown	Marusic et al.,2010

Figure 1. Diagram of known neutralizing epitope regions and protein interaction domains of HPV16 L2. Regions that were discovered using other PV L2 types are indicated.

Diagram was adapted and updated from (Karanam et al., 2009)). See Table 1 for a full list of neutralizing antibodies.

Structural Biology: L2 in the context of the papillomavirus capsid

The basic architecture of the papillomavirus virion is a non-enveloped T=7d icosahedral capsid with a diameter of 55-60nm (Figure 2A). Each virion contains 360 L1 proteins. L1 first stably assembles into star-shaped units of five called capsomers (sometimes termed ‘pentamers’) that have a central donut-like cavity, and the capsid is formed by the association of 72 such capsomeres via interlocking arms and disulphide bridging. The capsid also contains an ill-defined number of L2 proteins; although up to 72 L2 proteins have been estimated in a single capsid (1:5 L2 to L1 proteins ratio) (Buck et al., 2008; Buck and Trus, 2012). Smaller amounts have also been described in preparations of native virions purified from bovine and human warts, e.g. 8% (Rippe and Meinke, 1989), 3% (Trus et al., 1997) or 2-5% (Favre et al., 1975). These lower percentages may reflect a naturally lower occupancy, proteolytic degradation and/or the presence of empty particles in the preparations from warts, as the latter have been associated with lower fractions of L2.

Figure 2. Arrangement of L1 and L2 in HPV16 capsids.

3D reconstructions derived by comparing cryo-electron micrograph images of HPV16 capsids made of L1 + L2 versus L1 only (Buck et al., 2008; Buck and Trus, 2012) (A). Interior of L1+L2 capsid without DNA and histones (B). Arrangement of L2 density (red) areas superimposed on the interior view of L1 only capsid (in blue) (C). Clustal analysis of L1 binding domain of L2 showing several conserved proline (PxxP) motifs (D). Clustal Analysis was performed with the UCSF Chimera package. Images were provided by the courtesy of both Christopher Buck and Benes Trus, NCI.

Studies by Finnen et al on HPV11 revealed that there is an L1 binding site at the carboxy-terminus of L2 (residues 396-439 in HPV11 L2) and Okun et al mapped an analogous site mapped in BPV1 L2, but also observed another L1 interaction domain between 91-246 (Finnen et al., 2003; Okun et al., 2001). The C-terminal L1-binding region of L2 is characterized by several proline residues (PxxP) and similar PxxP motifs have been associated with protein-protein interactions in other systems. Homologous PxxP motifs were observed in a similar region of the C-termini of other L2 thus suggesting that this is a well conserved L1-binding site on PV L2 (Figure 2D). As binding of L2 to L1 was not affected significantly under conditions of high salt, weak detergents, urea and pH, it appears that the main mode of L1:L2 interaction is hydrophobic in nature. However, the L1 residues mediating the interaction with L2 in the capsid remain undefined, although some speculative modeling has been done based on the available X-ray crystallographic structure of L1 VLPs which suggest that L2 might interact with L1 via the N-terminus (Buck et al., 2008; Lowe et al., 2008). Also, purified L1 VLP and L2 are unable to form complexes in vitro without co-expression and this suggests that their interaction must occur prior to capsid assembly or at the capsomer level rather than insertion of L2 into a formed L1 VLP (Finnen et al., 2003; Okun et al., 2001).

The amino terminus of L2 contains two highly conserved cysteine residues (C22 and C28) (see Figure 5) across all PV types which form an intra-molecular disulfide hairpin loop rather than bridging with L1. Studies using HPV16 PsV showed that the point mutation of either or both cysteine residues resulted non-infectious virions but did not affect virus capsid assembly. However, this effect was not observed for BPV1 PsV (Campos and Ozbun, 2009; Gambhira et al., 2009). Interestingly, using PV made from organotypic raft cultures, Conway et al found that mutating these cysteine residues improved infectivity compared to wild type. The authors suggested that this remarkable difference could possibly be attributed to differences in virion preparation producing subtle differences in capsid structure between PV made in cell culture versus differentiating epithelial tissue. In addition, it was suggested that these cysteine residues potentially play a key role in late stage capsid stabilization by modulating the accessibility of L2 on the surface of raft-derived virions (Conway et al., 2009a).

Figure 5. Strong sequence conservation of 17-36aa region (recognized by the RG-1 neutralizing antibody) among different HPV types.

Note the two cysteine residues (C22 and C28) in this region are important for infection and recognition by RG-1 (Campos and Ozbun, 2009; Gambhira et al., 2009). Clustal Analysis was performed with the UCSF Chimera package.

Antibody binding studies suggest that L2 is predominantly hidden below the surface of native virions although its configuration remains unknown. A detailed study using antisera to overlapping peptides covering the entire HPV16 L2 sequence suggested that several regions (residues 32-81, 212-231, 272-291 and 347-381) are accessible on the surface of HPV16 L1/L2 VLPs (Heino et al., 1995). However, in another study using several monoclonal antibodies raised against BPV1 L2 protein, only the region 61-123 was actually exposed on the surface of BPV1 virions (Liu et al., 1997). Likewise, other studies mapping neutralizing epitopes in HPV16 L2 (see below in immunology section) suggest that only the first amino terminal ~120 amino acids are available for binding (Kawana et al., 1998; Rubio et al., 2011; Yang et al., 2003a). However, care must be taken in interpreting the studies with neutralizing antibodies as the HPV virions undergo conformational changes during the infectious process that appear to alter access of L2 to the capsid surface. Further, these conformational shifts might also be triggered upon binding of mature virions to the surface of microtiter plates for ELISA-type studies. In conclusion, it is still controversial which regions of L2 are actually exposed on the capsid surface of fully mature virions in solution. For example, the epitope in HPV16 L2 17-36 recognized by the RG-1 monoclonal antibody is accessible on virions bound to ELISA plates, but not mature virions in solution (Day et al., 2008). Rather, it becomes revealed after binding to the extracellular matrix and furin cleavage of L2 (Kines et al., 2009). However, other studies indicate that L2 residues 13-31 and 100-120 are constitutively exposed on the capsid surface (Richards et al., 2006; Yang et al., 2003a). Thus much of L2 is buried below the capsid surface of mature virions, but certain amino-terminal regions of L2 can be exposed on the cell surface during the early events of viral infection (Buck and Trus, 2012)(see section on viral host cell entry for details).

Unfortunately, no X-ray crystallographic structures are available for capsids containing L2. Several efforts at high resolution three dimensional image reconstruction of cryo-electron micrographs of native BPV1, HPV1 or CRPV virions, HPV1 or HPV16 VLPs have attempted to visualize L2 in the capsid (Baker et al., 1987; Belnap et al., 1996; Buck et al., 2008; Buck and Trus, 2012; Hagensee et al., 1994; Trus et al., 1997). However, most studies failed to visualize L2, possibly reflecting inadequate incorporation into the capsid, degradation, disorder or lack of symmetry. A study by Trus et al detected some protein density at the center of the capsomers at the 5-fold axes of symmetry (pentameric capsomers) of native BPV1 virions, but not the hexameric capsomers, that might be L2. However protein density associated with L2 could not be definitely assigned and the ratio of L1:L2 in BPV1 virions used in this study was ~30:1(Trus et al., 1997). However, in a more recent study comparing the structure of HPV16 L1 only versus L1+L2 particles in which the ratio of L1:L2 was driven close to 5:1, protein density was observed at the base of the axial lumen of all capsomers that was absent in L1 only particles (Figures 2A-C). This is consistent with the notion that much of L2 is buried below the capsid surface, although little of L2 was visualized as the density corresponded with only a small fraction (~12kDa) of its mass (Buck et al., 2008; Buck and Trus, 2012).

L2 facilitates genome encapsidation

L1 and L2 were found to have DNA-binding activity in vitro (Li et al., 1997; Mallon et al., 1987; Schafer et al., 2002), first suggesting that both capsid proteins may aid in viral genome encapsidation by directly binding to DNA without sequence specificity. These putative DNA-binding regions are conserved short sequences of positively charged basic residues found at the carboxy terminus of L1 and in both the amino and carboxy termini of L2. The DNA binding region of L1 of HPV33 is necessary for DNA encapsidation (Schafer et al., 2002). Likewise, L2 (Roden et al., 1996; Zhao et al., 1998) has also been reported to be required for efficient DNA/viral genome encapsidation for BPV1, although the effect may be less pronounced for HPV16, HPV31 (Buck et al., 2004; Holmgren et al., 2005) and HPV33 (Unckell et al., 1997). Deletion of either the N-terminal or the C-terminal DNA binding domain from BPV1 L2 had no impact upon BPV1 genome encapsidation but rendered the virions non-infectious (Roden et al., 2001). However, the impact of deletion of the DNA binding domains at both termini was not tested because these domains can also function as nuclear localization signals (NLS) (Bordeaux et al., 2006; Doorbar et al., 2012; Fay et al., 2004; Klucevsek et al., 2006; Zhou et al., 1994), complicating interpretation of these data.

Taken together, there is no evidence for sequence-specific DNA interaction with either capsid protein, but in vitro there is a non-specific ionic interaction between the negatively charged DNA and these short sequences rich in positively charged amino acids (although there significance for virion assembly is unclear). Further, L2 can bring components of the PV virion to specific locations in the nucleus defined as ND-10 and Day et al hypothesized that L2 co-localizes/organizes L1, E2 and viral genome to facilitate encapsidation of the histone-bound genome and virion assembly (Day et al., 1998; Florin et al., 2006), although E2 is not required for packaging reporter constructs into PsV (Buck et al., 2004).

L2 and host cell entry

There are currently two competing models of papillomavirus infection events prior to cell entry (Kines et al., 2009; Surviladze et al., 2012). However both agree that L2 starts predominantly buried within the capsid and its exposure requires a conformational change induced by capsid binding to heparan sulfate proteoglycans that is anchored in the extracellular matrix in vitro or the basement membrane in vivo, or possibly in solution. This binding induces changes in the conformation of the capsid such that the very amino terminus of L2 extrudes to the capsid surface and becomes susceptible to cleavage by the pro-convertase enzyme furin in the extracellular milieu, thus removing one of the putative NLS (Day and Schiller, 2009; Johnson et al., 2009; Kines et al., 2009; Richards et al., 2006; Selinka et al., 2003). Analysis of the N-terminus of all PV types reveals that the consensus furin cleavage motif site (R-X-K/R-R) (Figure 3B) is conserved at around amino acids 9-12 (Richards et al., 2006). Numerous PV types exhibit reduced infectivity upon exposure to furin inhibitiors both in vitro and in vivo, and furin deficient cells are not infected unless complemented with the a furin expression vector or extracellular addition of furin. Point mutation of the furin cleavage site also rendered virions non-infectious (Richards et al., 2006). Thus cleavage of L2 by furin is required for infection and results in further capsid conformational changes. The virus is subsequently internalized via an undefined secondary receptor found on the host cell basal keratinocyte (Figure 3A). Importantly, furin cleavage also exposes a broadly neutralizing epitope (amino acid region 17-36) which was first characterized using the RG-1 monoclonal antibody (Gambhira et al., 2007b).

Figure 3. A model depiction of the early events of PV infection in the cervical epithelium in vivo.

Exposure of the basement membrane by micro trauma allows PV to bind to Heparan Sulfate Proteoglycans (HSPG). Binding of PV to HSPG and subsequent interaction with host cell protein Cyclophilin B causes a conformational change in capsid structure resulting in the exposure of the N-terminus of L2 (Bienkowska-Haba et al., 2009). L2 has a furin cleavage motif and is cleaved by furin (Richards et al., 2006). Subsequently, virions are internalized into the basal cells to deliver the viral genome to their nucleus (Day et al., 2013) (A). Clustal analysis of the furin cleavage motif, R-x-K/R-R (boxed in black) on a variety of HPV and other commonly studied animal PV L2 (B). Clustal Analysis was performed with the UCSF Chimera package.

Richards et al showed L2 with point mutation of L2 preventing furin cleavage did not affect PsV production. These mutant PsV had typical levels of L2 and were able to encapsidate the reporter genome normally. These mutant capsids were also able to bind to the cell surface, enter the cell, traffic through the endosomes and uncoat normally. However, the mutant L2 and reporter genome were unable to leave the endosome even after 24 hours post-infection while wildtype L2 and reporter genome accumulated in the nucleus. Endosomal retention was also observed in infection studies with wildtype PsV incubated with a furin inhibitor. These findings suggest that furin cleavage of L2 is important for the exit of the L2/genome complex from the endosomal compartment (Day and Schiller, 2009; Richards et al., 2006).

There is no clear difference in the initial internalization of L1 VLPs and L1/L2 VLPs visualized by fluorescence microscopy. This suggests that it is either L1 that binds to the secondary unknown entry receptor or these particles are able to internalize but via subtly distinct pathways that are not currently discernable by fluorescence microcopy. Assuming the latter, it has been proposed that furin cleavage exposes on the surface of the capsid a previously buried region of L2 that can bind a currently undefined epithelial cell surface entry receptor, and that it is this L2-receptor interaction facilitates viral entry along the true infectious pathway. Studies have suggested two candidate sites on L2 for binding to a putative entry receptor; one coincident with the RG-1 epitope (13-31), and another at a region on the surface exposed region of L2 (108-120) that is recognized by another neutralizing antibody (Kawana et al., 2001b; Yang et al., 2003a). The 13-31 region binds to cell surfaces and contains several residues critical for infection by HPV16 (Yang et al., 2003a), notably the previously mentioned two cysteines residues that are completely conserved (Campos and Ozbun, 2009; Gambhira et al., 2009).

The L2 108-126 region, which contains a neutralizing epitope, can also bind to epithelial cells and exhibits sequence conservation (Kawana et al., 1998; Kawana et al., 1999). Interestingly, pre-incubation of 108-126 L2 peptide with cultured cells reduced PsV infectivion by 60% compared to control peptides (Woodham et al., 2012). Further, annexin A2 heterotetramers (A2t) on epithelial cell surfaces were recently found to interact with this region and they were proposed as a putative candidate for L2 receptor binding (Woodham et al., 2012). In summary, furin cleavage of L2 is both a critical and universal event in PV infection and it presumably renders L2 able to perform other roles further downstream of the infectious process. Whether L2 is involved in the initial uptake (i.e binding to the secondary receptor) remains a controversial topic and is an active area of both debate and PV research.

L2 and vesicular trafficking of papillomavirus

For successful infection, PV needs to enter the host cell, cross the cytoplasm and transport the viral genome to the nucleus. With respect to entry, this is actually a two-step process involving viral entry and subsequently, viral movement into the appropriate organelles. PV apparently can utilize several different cellular pathways for internalization, although one particular pathway might be dominant and the virus is then able to exploit alternative pathways should the typical pathway be blocked. It should be noted that the typical entry route may vary between the cell line/type and the particular PV genotype studied (See (Horvath et al., 2010; Sapp and Bienkowska-Haba, 2009) for a review). As mentioned earlier, there has been no discernable difference in uptake mechanisms in epithelial cells with respect to L1-VLPs and L1/L2-VLPs with current imaging techniques. However, subtle differences were noted when studying PV uptake using immune cells. For example, L1-VLPs enter cells via either clathrin-mediated or caveolae-dependent mechanisms (Bousarghin et al., 2005; Yan et al., 2004), whereas L1/L2 VLPs were taken up by clathrin- and caveolae-independent mechanisms, suggesting that L2 may impact the uptake pathway. Interestingly while Langerhans cells that are exposed to L1 VLPs functionally matured and L1/L2 VLPs lacked this effect. Fahey et al suggest that L2 may play a role in immune escape by re-routing uptake (Fahey et al., 2009). However, inclusion of L2 in VLPs did not impact their activation of bone marrow-derived dendritic cells (Lenz et al, 2001).

In the case of infection of epithelial cells, regardless of the mode of internalization, consistent findings suggest that following cell entry, papillomavirus enters the early endosome and transitions to the late endosome in an acidification-dependent process. Interestingly, the virions have recently been found to pass via the trans-golgi network, wherein L1 is retained. Entry to the golgi required furin cleavage of L2, and golgi trafficking mediated by Rab 9a and Rab7b (Day et al., 2013). Other organelles have also been implicated in the trafficking process, notably caveosomes and the endoplasmic reticulum for BPV1, HPV31 and HPV16 (Bossis et al., 2005; Dabydeen and Meneses, 2009; Laniosz et al., 2009; Laniosz et al., 2008; Laniosz et al., 2007a). Virions that gain entry to the cytoplasm by an unknown and possibly L2-dependent mechanism have also been visualized by transmission electron microscopy, but L1 does not make it to the nucleus (Ishii et al., 2010; Yang et al., 2003a). Numerous host cell proteins have also been identified as interacting partners with L2 and have been suggested to facilitate entry and trafficking by various mechanisms to the nucleus. These details will be discussed in the next few sections.

The first example is the chaperon protein cyclophilin B (CyPB). Cyclophilins are peptidyl-prolyl cis/trans isomerases, and it has been shown in other viruses such as HIV whereby CyPA binds via the HIV capsid protein motif 85-PXXXGPXXP-93, which is a similar proline motif those of to L1 interacting motif on L2 (Bienkowska-Haba et al., 2009). In a similar fashion, CyPB binds to L2 amino acid region 90-110 (now termed the CyPB binding site) and it was thought initially that CyPB-binding merely facilitates the cell-surface exposure of L2 N-termini for furin cleavage and aid subsequent cell internalization (Figure 3A). The authors of the study however noted that if CyPB was inhibited, non-infectious HPV16 PsV internalization could still occur. In addition, PsVs with mutated CyPB motifs could also internalize into the cell independently of CyPB but were still sensitive to CyPB inhibitors (Bienkowska-Haba et al., 2009). This suggests that CyPB plays 2 distinct roles in HPV infection. Recently, the second role was elucidated in a study whereby it was shown that CyPB aids in the uncoating of HPV capsid in the late endosome. Specifically, CyPB causes the dissociation of the viral capsid proteins L1 and L2 complexed with the viral DNA into a separate endoyctic compartments (Bienkowska-Haba et al., 2012).

Another host protein that has been recently found to be important for virion trafficking is the cytosolic adaptor protein Sortin Nexin 17 (SNX17). SNX17 is an important player in endosomal recycling and can be found in early endosomes and recycling tubules. Key substrates of SNX17 are typically transmembrane cargo proteins with a NPxY motif at their cytosolic tails. Interestingly, studies by the Banks group showed that numerous PV types have a highly conserved NPxY motif at around amino acid region 245-257 of L2 that binds to SNX17. They also showed that knockdown of SNX17 dramatically reduced HPV PsV infectivity. Conversely, over-expression of SNX17 increased infectivity in a dosage-dependent manner. Mutation of the NPxY region reduced viral infectivity suggesting that there is a critical SNX17-L2 interaction. This interaction may prevent premature lysosomal degradation of the virus as NPxY mutant virions experience premature lysosomal degradation and the L2-vDNA complex does not escape the endosome to travel to the nucleus (Bergant and Banks, 2013; Bergant Marusic et al., 2012).

Lastly, another possible candidate put forward as aiding in PV transport is the tSNARE syntaxin 18 protein (Bossis et al., 2005; Laniosz et al., 2007b). Syntaxin 18 binds to L2, and L2 residues 41-45 are critical for this interaction. Mutation of L2 41-45 in the context of BPV1 eliminated infectivity without compromising encapsidation (Bossis et al., 2005). Syntaxin 18 is an ER resident protein and overexpression of a dominant negative inhibitor blocked both ER trafficking and BPV1 PsV infection. This suggests a distinct route of trafficking for BPV1.

L2 and vesicular escape during infection

Following virion dissociation, the viral DNA (vDNA) must escape the vesicular compartment to be able to travel to the host cell nucleus. Studies using labeled PsV genome and L2 antibodies have suggested that L2 in complex with HPV vDNA somehow exits the late endosome to carry out this process. The information on the mechanism of HPV endosomal escape is an area of active investigation. Kamper et al showed that HPV33 L2 C-terminus has a 23 amino acid region that contains adjacent hydrophobic and basic clusters of amino acids. The authors showed that full length L2 with this region or this region fused to GFP was able to integrate into cellular membranes. The deletion or mutation of this region also abrogated viral infectivity and consequently the viral genome together with L2 was retained in the late endosome suggesting that this region could potentially act as a membrane destabilizing peptide for L2/vDNA endosomal escape (Kamper et al., 2006).

Bronnimann et al recently described another region in L2 that potentially could also aid endosome escape. They found that L2 has a transmembrane-like domain region at its N-terminus spanning residues 45-67, that includes several highly conserved GxxxG motifs (Figure 4). Mutagenesis of some of these GxxxG motifs resulted in endosomal retention of L2 suggesting the importance of this region in infection. The authors also noted that the predicted structure of this domain is alpha-helical in lipid environments and that these motifs lie on two opposite faces of this helix (Figure 4B). Interestingly, based upon some in vitro studies, these faces could self-associate within biological membranes using the GxxxG motif. Furthermore, they provide evidence that multiple L2 molecules can associate either homo- or heterotypically via these GxxxG motifs to potentially form a higher order structure that could synergistically facilitate endosomal penetration (Bronnimann et al., 2013).

Figure 4. The putative transmembrane domain of L2.

Clustal analysis of the transmembrane-like (TM-like) domain in L2 of fourteen PVs (A). Alpha helical modeling of the TM domains with the colors indicating the conserved GXXXG domains on each face (B) (Bronnimann et al., 2013). Images were provided by the courtesy of Samuel Campos, University of Arizona. Clustal Analysis was performed with the UCSF Chimera package.

These findings do not rule out other host cell protein interactions with this TM domain that could aid endosome escape. Indeed, we have recently shown that gamma-secretase (GS) is required for PV infection. Interestingly, GS inhibitors prevent HPV16 PsV infection and result in retention of L2/genome complex in the late endosome (Karanam et al., 2010). This suggests that GS facilitates egress from the late endosome, an L2-depdenent process, but its key substrate in this process is not known. GS cleaves substrates at their transmembrane domain and the above-mentioned studies suggest that L2 has such a domain indicating a possible interaction. Indeed substrates of GS are also typically cut with a pro-protein convertase at a position less than 30 residues from the transmembrane domain. Interestingly, L2 is cleaved by the proprotein convertase furin at approximately amino acid region 9-12 which is ~30 residues from the putative transmembrane domain. As mentioned, inhibition of furin and GS cleavage both result in endosomal retention of L2 and the viral genome, suggesting these processes may be functionally related.

L2 and the cytoskeleton

Upon endosome escape, the L2/vDNA complex may utilize the cytoskeleton to traverse the cytoplasm. L2 is able to interact with beta-actin at a conserved site found at residues 25-45 (Yang et al., 2003b), but the locomotive mechanism involving beta-actin was not defined. Florin et al describe the binding of L2 with dynein motors via its C-terminal 40 amino acids. The authors demonstrated via immunofluorescence and co-immunoprecipitation that L2 via binding to the motor protein dynein could interact with the actin microtubule network and conduct minus-end-directed transport of L2/vDNA to the nucleus (Florin et al., 2006). In a subsequent study, the same authors further substantiated this using a combination of two-hybrid assays and immunofluorescence to reveal DYNLT1 and DYNLT3 (Dynein light chain 1 and 3) as the main components that work with L2 for microtubule transport (Schneider et al., 2011).

Nuclear entry by L2

Multiple possible interacting partners of L2 have now been identified as aiding traversal of the cytoplasm and vesicular trafficking. Importantly the mechanism of nuclear entry by L2 differs by stage of the virus life cycle, i.e. viral entry into the nucleus during the establishment of initial infection versus during virion production. With respect to the former, it is well established that L2 is required to go to the nucleus with the vDNA to enact initial infection (Day et al., 1998; Kondo et al., 2009). Initially, it was assumed that the L2-vDNA complex entered the nucleus via nuclear pore complexes via the previously mentioned NLS on both the N and C-termini of L2. However, the N-terminus NLS is located upstream of the furin cleavage site and thus might not be present after entering the cell during infection. Similarly, the C-terminus NLS is near the membrane destabilizing peptide and may also be a DNA-binding domain, potentially blocking this NLS. In light of this, Pyeon et al showed in a surprising study that the L2-genome complex enters the nucleus, not via the nuclear pores during interphase, but rather requires entry to mitosis associated with the breakdown of the nuclear envelope (Pyeon et al., 2009). Mamoor et al also described an arginine-rich nuclear retention signal in L2 has been described in HPV16 L2, comprising residues 296-SRRTGIRYSRIGNKQTLRTRS-316 that is required for infection but not virion assembly. It is possible that this region allows for the association of L2 and the genome complex with the nuclear matrix during metaphase. In the same study, a leucine-rich nuclear export signal (NES) was also described at the C-terminus of HPV16 L2, encompassing residues from 462-LPYFFSDVSL-471. This leucine-rich region on L2 along with some deletion mutants of the same region was subsequently fused to GFP and was found to exert nuclear export in a CRM1 (Exportin)-dependent manner. At present, the role of this NES is not clear but it does not appear to be critical for either pseudovirion assembly or infection (Mamoor et al., 2012).

While the roles of the nuclear localization signals (NLS) at both its N- and C-termini are unclear with respect to initiation of infection, these regions can independently interact with karyopherins, specifically kapα2β1, kapβ2 and kapβ3, for entry into the nucleus via the nuclear pore complex after the initial ribosomal synthesis of L2. It is interesting to note BPV1 L2 interacts with kapα2β1 only which suggests that the nuclear import pathways are different for different PV types (Bordeaux et al., 2006; Darshan et al., 2004; Klucevsek et al., 2006; Sun et al., 1995). Heat shock cognate protein 70 (Hsc70) can also assist in transporting newly synthesized L2 back into the nucleus via forming an active complex with L2. Hsc70 is necessary for L2 transport to the nucleus as depletion of Hsc70 resulted in L2 accumulation in the cytoplasm. Apart from nuclear transport and entry, Hsc70 may facilitate viral assembly possibly by maintaining L2 in a specific state of folding that is optimal for interaction with the L1-capsomers. This role was proposed upon finding that Hsc70 could be co-sedimented with L1/L2-VLPs but not L1-VLPs suggesting that Hsc70 is associated with L2 upon its integration into VLPs. Hsc70 could not be detected with PsV suggesting that it can be eliminated by an undefined mechanism during the process of DNA encapsidation (Florin et al., 2004; Florin et al., 2002a).

Nuclear activities of L2

There is a motif around amino acid region 390-420 that targets L2 to the ND10 subdomains of the nucleus (also known as PML-oncogenic domain/POD/PML bodies and defined by co-localization with PML) (Becker et al., 2004; Becker et al., 2003; Day et al., 1998; Florin et al., 2002a; Florin et al., 2002b). HPV E1 and E2 can also associate with ND-10 domains during viral DNA synthesis (Swindle et al., 1999). Taken together, it was proposed that L2 may recruit other PV viral components to this area possibly to facilitate structural assembly and/or vegetative replication of the PV genome (Day et al., 2004; Day et al., 1998; Heino et al., 2000; Schafer et al., 2002). Whereas BPV1 L1, E1 and E2 exhibit diffuse nuclear staining when produced individually using recombinant Semliki Forest Virus vectors, both BPV and HPV L2 when overexpressed would cause both L1 and E2, but not E1, to co-localize in a specific nuclear region known as the ND-10 domain (Day et al., 1998). Given the importance of L2 in assembly of BPV virions (and to a smaller extent HPV virions), these findings suggested that L2 might act to organize/co-localize the virion components adjacent to ND-10. However, Buck et al found no requirement for E1 or E2 in the assembly of HPV pseudovirions (Buck et al., 2004), although this production system may not fully mimic vegetative replication and virion assembly occurring in differentiated epithelium.

It has also been suggested that ND-10 localization is actually non-physiological and that the L2 accumulation observed is due to the over-expression system and/or misfolding (Kieback and Muller, 2006). However, Day et al also observed that establishment of a papillomavirus infection and viral transcription is reduced 10-fold in the absence of PML protein (a critical structural component of ND-10), and that upon infection L2 traffics to ND-10 in association with the viral genome (Day et al., 2004). These findings once again point to the importance of the ND-10 localization of L2. However, here it is proposed that the organization of viral components at ND-10 by L2 may facilitate the early burst of transcription and replication events at the initiation of infection.

Studies with HPV PsV also show that they bring their encapsidated reporter DNA vectors to ND-10 and efficiently initiate reporter gene expression and vector replication via SV40 T antigen at low MOIs. While this implies that the interaction between L2 and the encapsidated DNA is not sequence specific, and initiation of infection does not require direct interaction between L2 and E2, it does not rule out a role for L2-E2 interactions in the initiation of infection by authentic PV virions. Indeed, several studies suggest interactions between L2 and E2. It was found that the amino terminal residues 1-50 of L2 are responsible for binding to E2. However, L2 residues 301-400 are also required for the down-regulation of E2-dependent transcriptional activation. L2 does not reduce transcriptional activation by either lowering E2 protein levels via reduction of mRNA levels or by enhancing its proteasomal degradation (Okoye et al., 2005). Heino et al also found that L2 inhibits the transcriptional activation function of E2 transcriptional transactivator (E2TA), but not its capacity to support viral DNA replication. Interesting, L2 can bind to and bring the E2-TR transcriptional repressor form to ND-10, as it binds to two separate domains of E2 (Heino et al., 2000). Taken together, these findings indicate a possible role for L2 in early viral transcription as infection is initiated.

L2 also causes changes in the ND-10 environment; overexpression of HPV33 L2 induces the recruitment of Daxx, a nuclear transcriptional repressor protein that represses transcriptional activators, to ND-10. The interaction of Daxx with L2 was further confirmed via co-immunoprecipitation with L2. Furthermore, overexpression of L2 results in the expulsion from ND-10 and subsequent degradation of the transcriptional activator, SP100. Interestingly, proteasome inhibitors prevent the accumulation of L2 in the nucleus under these conditions and the accumulation of L1 at ND-10 with L2 only occurred after the exit of SP100. This adds to the speculation that the accumulation of L2 and Daxx followed by the loss of SP100 potentially could result in a major reorganization of the ND-10 environment to facilitate the assembly of virions (Becker et al., 2003; Florin et al., 2002b). The domain of HPV 33 L2 (390-420) that was required for ND-10 localization, was also required for the loss of SP100 and accumulation of Daxx. It was also suggested that interaction with Daxx may be also be the driving force for the localization of L2 into ND-10(Becker et al., 2003; Florin et al., 2002b). However, it is important to recognize the caveat that these data were derived using an over-expression system (Becker et al., 2004; Kieback and Muller, 2006).

Given L2’s interaction with E2 and inhibition of E2-dependent transcriptional activity, it seems plausible that L2 has a role also in transcriptional regulation. Indeed, a number of relevant L2-interacting partners have been identified using yeast 2-hybrid approaches. Gornemann et al detected four nuclear body-associated proteins that could interact with the L2 of different HPV types and co-localize in the ND-10 domains in some cases. PATZ is a transcription factor which was found to co-localize in discrete subnuclear domains and to bind with L2. Interestingly, PATZ is a zinc finger family transcriptional regulator and it was postulated that L2-PATZ interactions might play a role in gene regulation and cell differentiation during papilloma formation. Other interesting potential interactions were identified, including TIP60, TIN-Ag-RP and PLINP. However, all these identified interactions require further study to understand their function (Gornemann et al., 2002). In a more recent study using the 2-hybrid approach again, the transcription factors TBX2 and 3 were identified as interacting partners of L2 at its c-terminus using HPV11, 16 and 18. The study also found both TBX2 and TBX3 repress viral transcription from the viral LCR of several HPV types and this repression was enhanced when L2 was over-expressed. In particular, the levels of early genes such as E6 and E7 were reduced in HeLa cells when TBX2 was increased. Interaction of TBX2 with the LCR was confirmed by ChIP analysis, and L2 seems to be a key player in stabilizing this interaction. TBX2/3 co-localization with L2 can also occur either in the ND-10 domains or throughout the nuclei depending on the distribution of L2. It is important to recognize that L2 is usually diffused throughout the nucleus in the upper stratified layers of CIN1/2 lesions (Lin et al., 2009), and it is possible that L2-TBX2/3 interactions here serve to inhibit early viral gene expression and aid the transition to viral assembly during infection (Schneider et al., 2013).

L2 Immunology

Two L1 VLP vaccines (GlaxoSmithKline’s Cervarix® and Merck’s Gardasil®) have been approved by the FDA for the prevention of HPV-associated cervical neoplasia. Cervarix® contains HPV16 and 18 L1 VLPs, whereas Gardasil is a mixture of these two types as well as L1 VLPs of low risk types HPV6 and 11 that cause genital warts. While both vaccines showed high efficacy in preventing HPV infection in many human clinical trials (Dillner et al., 2010; Kaufmann and Nitschmann, 2010; Kreimer et al., 2011; Lehtinen et al., 2012; Romanowski et al., 2009), the protection these vaccines confer is type-restricted. Given this limitation, second generation HPV vaccines are being developed. Indeed, Merck is currently testing a nonavalent L1-VLP HPV vaccine targeting 7 high risk HPV types to broaden protection among the types most commonly detected in cervical cancer. However, complex formulation is expensive which could drive up costs and thus indirectly discourage vaccine usage in low resource settings where a broadly protective HPV vaccine is most needed (Schiller and Lowy, 2012).

Alternatively, the HPV minor capsid protein L2 is an interesting candidate that may address the aforementioned requirements of cost and broad coverage. L2 can be expressed as a single antigen in bacteria and this could potentially bring down production costs. More significantly, vaccination studies with L2 in animal challenge models showed the production of antibodies neutralizing for a wide spectrum of PV types, and similarly broad protection from experimental viral challenges (Campo et al., 1997; Embers et al., 2002; Gambhira et al., 2007a; Jagu et al., 2009; Kawana et al., 2001a; Kawana et al., 2001b; Lin et al., 1992). Passive transfer of L2-specific neutralizing antibodies is sufficient to mediate protection of mice from cutaneous or vaginal challenge with HPV pseudovirions (Day et al., 2010; Gambhira et al., 2007a; Gambhira et al., 2007b; Jagu et al., 2013). While some variation in these neutralizing epitopes is apparent (Seitz et al., 2013), sequence comparison at the N-terminus of L2 reveals high conservation across PV types (Figure 5), including regions that are recognized by neutralizing monoclonal antibodies (Figure 1 and Table 1)(Christensen et al., 1991; Rubio et al., 2011). The reason for sequence conservation in these L2 epitopes which is potentially detrimental to the virus has been debated. Several explanations have been put forward as to why the neutralizing epitopes are conserved in L2, but not L1. Firstly, L2 is hidden in the virus capsid unlike the surface loops of L1 that form the immunodominant neutralizing epitopes of VLPs. This could explain the lack of evolutionary pressure for these epitopes drift in sequence (Roden et al., 2001). Secondly, the conservation of the L2 neutralizing epitopes also suggests an evolutionary constraint that may reflect binding to a cellular entry factor or some other critical role during infection (Kawana et al., 2001b; Woodham et al., 2012; Yang et al., 2003a).

Table 1.

Published HPV16 L2-specific neutralizing and cross-neutralizing antibodies.

Monoclonal antibody designation	Epitope region in HPV16 L2	Cross Neutralizing ability	Isotype	References
RG-1	17-36	Yes	IgG2aκ	(Gambhira et al., 2007b)
WW1	17-36	Yes	IgG1	Wu et al, Manuscript in preparation
MAb24B	58-64	Yes	IgG2b	(Nakao et al., 2012)
MAb13B	64-73	Yes	IgG1
K4L220–38	21-30	Yes	IgG2b	(Rubio et al., 2011)
K18L220–38	22-30	Yes	IgG1
K8L228–42	32-39	16 only	IgG1
K1L264–81	73-79	16 only	IgG1
MAb5/Mab13	108-120	Yes	IgG2a/3	(Kawana et al., 1999)
MAb6	65-81	?
16L2.4B4	108-120	Yes	Undefined	(Woodham et al., 2012)

Unfortunately, L2 provided no discernable benefit in the context of L1/L2 VLPs (Breitburd et al., 1995; Kirnbauer et al., 1996), and the immune response produced to vaccination with L1/L2 VLPs was found to be pre-dominantly an anti-L1 response (Roden et al., 2000). This skewed response has not been fully understood but has been attributed to the overall number of L1 (which have numerous immunogenic epitopes also) compared to L2 in a single virus capsid, and distant spacing of L2 compared to L1, and possibly low occupancy of L2 in L1/L2 VLPs. Further, L2 is buried in the PV capsid and thus possibly hidden from or only exposed transiently to B cells. The challenge ahead for L2-based vaccines is to find methods to expose L2 to the immune system to induce and maintain protective levels of antibodies to these L2 epitopes for many decades. Many methods have been utilized with varying success at the preclinical setting and the reader is directed to (Karanam et al., 2009; Schiller and Lowy, 2012; Wang and Roden, 2013) for a more comprehensive coverage of this topic.

Conclusion

It is clear that L2 has complex and multifunctional roles in the biology of all PVs, notably in virion assembly and early events of infection, and has potential as a vaccine antigen. A recurring theme in our discussion of L2 biology is that while many functional domains and numerous host-cell protein interacting partners are now recognized (Table 2), the detailed mechanism of their action remains controversial or unclear in many cases. The challenges ahead for PV L2-based research are to obtain an atomic structure of L2, and to decipher how host proteins that interact with L2 impact the biology of PV. The PaVE website will be a valuable tool to collate and disseminate these findings.

RESEARCH HIGHLIGHTS.

• L2 is the minor antigen of the non-enveloped T=7d icosahedral Papillomavirus capsid

• L2 is a nuclear protein that can traffic to ND-10 and facilitate genome encapsidation

• L2 is critical for infection and must be cleaved by furin

• L2 is a broadly protective vaccine antigen recognized by neutralizing antibodies