Incorporation of RG1 epitope into HPV16L1-VLP does not compromise L1-specific immunity.

The candidate pan-Human Papillomavirus (HPV) vaccine RG1-VLP are HPV16 major capsid protein L1 virus-like-particles (VLP) comprising a type-common epitope of HPV16 minor capsid protein L2 (RG1; aa17-36). Vaccinations have previously demonstrated efficacy against genital high-risk (hr), low-risk (lr) and cutaneous HPV. To compare RG1-VLP to licensed vaccines, rabbits (n = 3) were immunized thrice with 1 µg, 5 µg, 25 µg, or 125 µg of RG1-VLP or a 1/4 dose of Cervarix®. 5 µg of RG1-VLP or 16L1-VLP (Cervarix) induced comparable HPV16 capsid-reactive and neutralizing antibodies titers (62,500/12,500-62,500 or 1000/10,000). 25 µg RG1-VLP induced robust cross-neutralization titers (50-1000) against hrHPV18/31/33/45/52/58/26/70. To mimic reduced immunization schedules in adolescents, mice (n = 10) were immunized twice with RG1-VLP (5 µg) plus 18L1-VLP (5 µg). HPV16 neutralization (titers of 10,000) similar to Cervarix and Gardasil and cross-protection against hrHPV58 vaginal challenge was observed. RG1-VLP vaccination induces hrHPV16 neutralization comparable to similar doses of licensed vaccines, plus cross-neutralization to heterologous hrHPV even when combined with HPV18L1-VLP.

Production and characterization of a novel HPV anti-L2 monoclonal antibody panel.

The major capsid protein of HPV, L1, assembles into pentamers that form a T = 7 icosahedral particle, but the location of the co-assembled minor capsid protein, L2, remains controversial. Several researchers have developed useful monoclonal antibodies targeting L2, but most react with linear epitopes toward the N-terminus. As a means to better define the virus capsid and better assess the localization and exposure of L2 epitopes in the context of assembled HPV, we have developed a panel of 30 monoclonal antibodies (mAbs) which target the N-terminus of L2 amino acids 11-200, previously defined as a broadly protective immunogen. Select mAbs were processed with enzymes and anti-L2 Fabs were generated. These new mAb/Fab probes will be beneficial in future studies to unravel the placement of L2 and to help better define the role of L2 in the HPV lifecycle and the nature of the broadly protective epitopes.

Aberrant expression of homeobox gene HOXA7 is associated with müllerian-like differentiation of epithelial ovarian tumors and the generation of a specific autologous antibody response.

Autologous serum antibodies to molecules that are aberrantly expressed in tumors represent potential biomarkers for early diagnosis of cancer. In this study, we identified the homeobox gene HOXA7 as encoding an antigen in epithelial tumors of the ovary. These tumors are thought to arise from the simple epithelium lining the ovarian surface, but they often resemble the specialized epithelia derived from the müllerian ducts. Expression of HOXA7 was detected in ovarian tumors exhibiting müllerian-like features and correlated with the generation of anti-HOXA7 antibodies by patients. In contrast, it was observed that healthy women lack anti-HOXA7 antibodies (P < 0.0001) and that HOXA7 expression is absent from normal ovarian surface epithelium. Interestingly, HOXA7 expression was detected in the müllerian-like epithelium lining inclusion cysts in normal ovaries and in the müllerian duct-derived epithelium of normal fallopian tubes. Furthermore, ectopic expression of HOXA7 enhanced the epithelial phenotype of immortalized ovarian surface epithelial cells, as indicated by the appearance of cobblestone morphology, induction of E-cadherin expression, and down-regulation of vimentin. These findings associate aberrant HOXA7 expression with the müllerian-like differentiation of epithelial ovarian tumors and suggest diagnostic utility of serum antibodies to HOXA7.

Positively charged termini of the L2 minor capsid protein are necessary for papillomavirus infection.

Coexpression of bovine papillomavirus L1 with L2 mutants lacking either eight N-terminal or nine C-terminal amino acids that encode positively charged domains resulted in wild-type levels of viral genome encapsidation. Despite wild-type binding to the cell surface, the resulting virions were noninfectious. An L2 mutant encoding a scrambled version of the nine C-terminal residues restored infectivity, in contrast to an L2 mutant encoding a scrambled version of the N-terminal residues.

L1 interaction domains of papillomavirus l2 necessary for viral genome encapsidation.

BPHE-1 cells, which harbor 50 to 200 viral episomes, encapsidate viral genome and generate infectious bovine papillomavirus type 1 (BPV1) upon coexpression of capsid proteins L1 and L2 of BPV1, but not coexpression of BPV1 L1 and human papillomavirus type 16 (HPV16) L2. BPV1 L2 bound in vitro via its C-terminal 85 residues to purified L1 capsomers, but not with intact L1 virus-like particles in vitro. However, when the efficiency of BPV1 L1 coimmunoprecipitation with a series of BPV1 L2 deletion mutants was examined in vivo, the results suggested that residues 129 to 246 and 384 to 460 contain independent L1 interaction domains. An L2 mutant lacking the C-terminal L1 interaction domain was impaired for encapsidation of the viral genome. Coexpression of BPV1 L1 and a chimeric L2 protein composed of HPV16 L2 residues 1 to 98 fused to BPV1 L2 residues 99 to 469 generated infectious virions. However, inefficient encapsidation was seen when L1 was coexpressed with either BPV1 L2 with residues 91 to 246 deleted or with BPV1 L2 with residues 1 to 225 replaced with HPV16 L2. Impaired genome encapsidation did not correlate closely with impairment of the L2 proteins either to localize to promyelocytic leukemia oncogenic domains (PODs) or to induce localization of L1 or E2 to PODs. We conclude that the L1-binding domain located near the C terminus of L2 may bind L1 prior to completion of capsid assembly, and that both L1-binding domains of L2 are required for efficient encapsidation of the viral genome.

A serologically identified tumor antigen encoded by a homeobox gene promotes growth of ovarian epithelial cells.

Ovarian carcinomas are thought to arise from cells of the ovarian surface epithelium by mechanisms that are poorly understood. Molecules associated with neoplasia are potentially immunogenic, but few ovarian tumor antigens have been identified. Because ovarian carcinomas can elicit humoral responses in patients, we searched for novel tumor antigens by immunoscreening a cDNA expression library with ovarian cancer patient serum. Seven clones corresponding to the homeobox gene HOXB7 were isolated. ELISAs using purified recombinant HOXB7 protein revealed significant serologic reactivity to HOXB7 in 13 of 39 ovarian cancer patients and in only one of 29 healthy women (P < 0.0001). Ovarian carcinomas were found to express HOXB7 at markedly higher levels than normal ovarian surface epithelium, suggesting that immunogenicity of HOXB7 in patients could be associated with its elevated expression in ovarian carcinomas. Overexpression of HOXB7 in immortalized normal ovarian surface epithelial cells dramatically enhanced cellular proliferation. Furthermore, HOXB7 overexpression increased intracellular accumulation and secretion of basic fibroblast growth factor, a potent angiogenic and mitogenic factor. These results reveal HOXB7 as a tumor antigen whose up-regulated expression could play a significant role in promoting growth and development of ovarian carcinomas.

Enhancement of sindbis virus self-replicating RNA vaccine potency by targeting antigen to endosomal/lysosomal compartments.

Self-replicating RNA vaccines (RNA replicons) have emerged as an attractive approach for tumor immunotherapy. RNA replicons do not integrate into host chromosomes, eliminating the concern for oncogenicity associated with a DNA vaccine. In this study, we used human papillomavirus type 16 (HPV-16) E7 as a model antigen and evaluated E7-specific immunity generated by a Sindbis virus self-replicating RNA vector, SIN-rep5. Three different constructs were created to target E7 antigen to different cellular localizations: (1) E7, a cytosolic/nuclear protein; (2) Sig/E7, a secretory protein; (3) Sig/E7/LAMP-1, in which we linked the transmembrane and cytoplasmic regions of the lysosome-associated membrane protein 1 (LAMP-1) to E7 protein to target E7 to the endosomal/lysosomal compartment. We found that the RNA replicon vaccine containing the Sig/E7/LAMP-1 fusion gene generated the highest E7-specific T cell-mediated immune responses and antitumor effects relative to RNA vaccines containing either wild-type E7 or Sig/E7. Our in vitro studies demonstrated that E7 antigen from Sig/E7/LAMP-1 RNA replicon-transfected apoptotic cells can be taken up by bone marrow-derived dendritic cells (DCs) and presented more efficiently through the MHC class I pathway than wild-type E7 RNA replicon-transfected apoptotic cells. Furthermore, our data revealed that CD8(+) T cells, CD4(+) T cells, and NK cells were important for the antitumor effects generated by Sig/E7/LAMP-1 RNA vaccination. These results indicate that targeting antigen to the endosomal/lysosomal compartment via fusion to LAMP-1 may greatly enhance the potency of self-replicating RNA vaccines.

Safety and immunogenicity trial in adult volunteers of a human papillomavirus 16 L1 virus-like particle vaccine.

BACKGROUND: Studies in animal models have shown that systemic immunization with a papillomavirus virus-like particle (VLP) vaccine composed of L1, a major structural viral protein, can confer protection against subsequent experimental challenge with the homologous virus. Here we report results of a double-blind, placebo-controlled, dose-escalation trial to evaluate the safety and immunogenicity of a human papillomavirus (HPV) type 16 (HPV16) L1 VLP vaccine in healthy adults. METHODS: Volunteers were given intramuscular injections with placebo or with 10- or 50-microg doses of HPV16 L1 VLP vaccine given without adjuvant or with alum or MF59 as adjuvants at 0, 1, and 4 months. All vaccine recipients were monitored for clinical signs and symptoms for 7 days after each inoculation. Immune responses were measured by an HPV16 L1 VLP-based enzyme-linked immunosorbent assay (ELISA) and by an HPV16 pseudovirion neutralization assay. The antibody titers were given as the reciprocals of the highest dilution showing positive reactivity in each assay. All statistical tests were two-sided. RESULTS: The prevaccination geometric mean ELISA titer for six seropositive individuals was 202 (range, 40--640). All vaccine formulations were well tolerated, and all subjects receiving vaccine seroconverted. Serum antibody responses at 1 month after the third injection were dose dependent in recipients of vaccine without adjuvant or with MF59 but were similar at both doses when alum was the adjuvant. With the higher dose, the geometric means of serum ELISA antibody titers (95% confidence intervals) to purified VLP 1 month after the third injection were as follows: 10,240 (1499 to 69 938) without adjuvant, 10,240 (1114 to 94 145) with MF59, and 2190 (838 to 5723) with alum. Responses of subjects within each group were similar. Neutralizing and ELISA antibody titers were highly correlated (Spearman correlation =.85), confirming that ELISA titers are valid proxies for neutralizing antibodies. CONCLUSIONS: The HPV16 L1 VLP vaccine is well tolerated and is highly immunogenic even without adjuvant, with the majority of the recipients achieving serum antibody titers that were approximately 40-fold higher than what is observed in natural infection.

Minor capsid protein of human genital papillomaviruses contains subdominant, cross-neutralizing epitopes.

Vaccination with virus-like particles (VLP), comprising both L1 and L2 of human papillomavirus (HPV) genital types 6, 16, and 18, induces predominantly type-specific neutralizing antibodies. L2 polypeptide vaccines protect animals against experimental challenge with homologous papillomavirus and cross-reactive epitopes are present in HPV L2. To assess L2-specific cross-neutralization of HPV genotypes, sheep were immunized with purified, bacterially expressed HPV6, 16, or 18 L2. In addition to neutralizing the homologous HPV type in vitro, antisera to each HPV L2 also cross-neutralized both heterologous HPV types. This suggests that unlike VLP-based prophylactic HPV vaccines, an L2 polypeptide vaccine may provide broad-spectrum protection.

Two antibodies that neutralize papillomavirus by different mechanisms show distinct binding patterns at 13 A resolution.

Complexes between bovine papillomavirus type 1 (BPV1) and examples of two sets of neutralizing, monoclonal antibodies (mAb) to the major capsid protein (L1) were analyzed by low-dose cryo-electron microscopy and three-dimensional (3D) image reconstruction to 13 A resolution. mAb #9 is representative of a set of neutralizing antibodies that can inhibit viral binding to the cell surface, while mAb 5B6 is representative of a second set that efficiently neutralizes papillomaviruses without significantly inhibiting viral binding to the cell surface. The 3D reconstructions reveal that mAb #9 binds to L1 molecules of both pentavalent and hexavalent capsomeres. In contrast, 5B6 binds only to hexavalent capsomeres, reflecting the significant structural or environmental differences for the 5B6 epitope in the 12 pentavalent capsomeres. Epitope localization shows that mAb #9 binds monovalently to the tips of capsomeres whereas 5B6 binds both monovalently and bivalently to the sides of hexavalent capsomeres approximately two-thirds of the way down from the outer tips, very close to the putative stabilizing intercapsomere connections. The absence of mAb 5B6 from the pentavalent capsomeres and its inability to prevent viral binding to the cell surface suggest that receptor binding may occur at one or more of the 12 virion vertices.

Chimeric papillomavirus virus-like particles elicit antitumor immunity against the E7 oncoprotein in an HPV16 tumor model.

Papillomavirus-like particles (VLPs) are a promising prophylactic vaccine candidate to prevent human papillomavirus (HPV) infections and associated epithelial neoplasia. However, they are unlikely to have therapeutic effects because the virion capsid proteins are not detected in the proliferating cells of the infected epithelia or in cervical carcinomas. To increase the number of viral antigen targets for cell-mediated immune responses in a VLP-based vaccine, we have generated stable chimeric VLPs consisting of the L1 major capsid protein plus the entire E7 (11 kDa) or E2 (43 kDa) nonstructural papillomavirus protein fused to the L2 minor capsid protein. The chimeric VLPs are indistinguishable from the parental VLPs in their morphology and in their ability to agglutinate erythrocytes and elicit high titers of neutralizing antibodies. Protection from tumor challenge was tested in C57BL/6 mice by using the tumor cell line TC-1, which expresses HPV16 E7, but not the virion structural proteins. Injection of HPV16 L1/L2-HPV16 E7 chimeric VLPs, but not HPV16 L1/L2 VLPs, protected the mice from tumor challenge, even in the absence of adjuvant. The chimeric VLPs also induced protection against tumor challenge in major histocompatibility class II-deficient mice, but not in beta2-microglobulin or perforin knockout mice implying that protection was mediated by class I-restricted cytotoxic lymphocytes. These findings raise the possibility that VLPs may generally be efficient vehicles for generating cell-mediated immune responses and that, specifically, chimeric VLPs containing papillomavirus nonstructural proteins may increase the therapeutic potential of VLP-based prophylactic vaccines in humans.

The papillomavirus minor capsid protein, L2, induces localization of the major capsid protein, L1, and the viral transcription/replication protein, E2, to PML oncogenic domains.

We have used immunofluorescent staining and confocal microscopy to examine the subcellular localization of structural and nonstructural bovine papillomavirus (BPV) proteins in cultured cells that produce infectious virions. When expressed separately, L1, the major capsid protein, showed a diffuse nuclear distribution while L2, the minor capsid protein, was found to localize to punctate nuclear regions identified as promonocytic leukemia protein (PML) oncogenic domains (PODs). Coexpression of L1 and L2 induced a relocation of L1 into the PODs, leading to the colocalization of L1 and L2. The effect of L2 expression on the distribution of the nonstructural viral proteins E1 and E2, which are required for maintenance of the genome and viral DNA synthesis, was also examined. The localization of the E1 protein was unaffected by L2 expression. However, the pattern of anti-E2 staining was dramatically altered in L2-expressing cells. Similar to L1, E2 was shifted from a dispersed nuclear locality into the PODs and colocalized with L2. The recruitment of full-length E2 by L2 occurred in the absence of other viral components. L2 was shown previously to be essential for the generation of infectious BPV. Our present results provide evidence for a role for L2 in the organization of virion components by recruiting them to a distinct nuclear domain. This L2-dependent colocalization probably serves as a mechanism to promote the assembly of papillomaviruses either by increasing the local concentration of virion constituents or by providing the physical architecture necessary for efficient packaging and assembly. The data also suggest a role for a nonstructural viral protein, E2, in virion assembly, specifically the recruitment of the viral genome to the sites of assembly, through its high-affinity interaction with specific sequences in the viral DNA.

Papillomavirus is resistant to desiccation.

There is strong epidemiologic evidence for sexual transmission of high-risk genital human papillomavirus (HPV) types. However, it is unclear if infection may also be transmitted indirectly via fomites. To assess this possibility, the in vitro infectivity after desiccation was compared for pseudotype HPV-16 virions, a model for high-risk type genital HPV, and bovine papillomavirus type 1 (BPV-1), a papillomavirus known to be transmitted via fomites. The 2 viruses had similar resistance to desiccation in cell extracts, retaining approximately 100%, 50%, and 30% of infectivity when dehydrated for 1, 3, and 7 days, respectively, at room temperature. Pseudotype HPV-16 and BPV in cell extracts were completely inactivated by autoclave treatment and susceptible to 70% ethanol but were resistant to EDTA or incubation at 56 degrees C for 1 h. The data suggest that further study of nonsexual spread of high-risk genital HPV via fomites is warranted.

Human papillomavirus type 16 virus-like particles expressed in attenuated Salmonella typhimurium elicit mucosal and systemic neutralizing antibodies in mice.

Attenuated strains of Salmonella are attractive live vaccine candidates for eliciting mucosal as well as systemic immune responses. The ability to induce immune responses in the reproductive tract may be critical for the effectiveness of a prophylactic vaccine against genital human papillomaviruses (HPV), which are important etiologic agents in the development of cervical cancer. To examine the potential of a live Salmonella-based vaccine to prevent genital HPV infection, the L1 major capsid protein from HPV type 16 (HPV16) was constitutively expressed in the PhoPc strain of Salmonella typhimurium. As demonstrated by electron microscopy, the L1 protein expressed in these bacteria assembled into virus-like particles (VLPs) that resemble authentic papillomavirus virions. This is the first demonstration that papillomavirus VLPs can self-assemble in prokaryotes. BALB/c mice were immunized with the HPV16 L1 recombinant PhoPc strain by the oral and nasal routes. Despite a low stability of the L1-expressing plasmid in vivo, a double nasal immunization was effective in inducing L1-specific serum antibodies that recognized mainly native, but not disassembled, VLPs. These antibodies effectively neutralized HPV16 pseudotyped virions in an in vitro infectivity assay. Conformationally dependent anti-VLP immunoglobulin A (IgA) and IgG were also detected in oral and vaginal secretions, indicating that potentially protective antibody responses were elicited at mucosal sites. Recombinant attenuated Salmonella expressing HPV capsids may represent a promising vaccine candidate against genital HPV infection.

Characterization of a human papillomavirus type 16 variant-dependent neutralizing epitope.

We have determined that three type-specific and conformationally dependent monoclonal antibodies, H16.E70, H16.U4, and H16.V5, neutralize pseudotype human papillomavirus type 16 (HPV16) virions in vitro. H16.U4 and H16.V5 neutralized pseudotype virions derived from the German HPV16 variant 114K and the Zairian variant Z-1194 with equal efficiency. In contrast, neutralization of Z-1194 pseudotype virions by H16.E70 was two orders of magnitude weaker than neutralization of 114K pseudotype virions. This difference correlated with enzyme-linked immunosorbent assay reactivity of H16.E70 to L1 virus-like particles of the two variants. A substitution at residue 282 of L1 was responsible for this differential reactivity, suggesting that this residue constitutes part of the H16.E70 epitope.

Novel structural features of bovine papillomavirus capsid revealed by a three-dimensional reconstruction to 9 A resolution.

The three-dimensional structure of bovine papillomavirus has been determined to 9 A resolution by reconstruction of high resolution, low dose cryo-electron micrographs of quench-frozen virions. Although hexavalent and pentavalent capsomeres form star-shaped pentamers of the major capsid protein L1, they have distinct high-resolution structures. Most prominently, a 25 A hole in the centre of hexavalent capsomeres is occluded in the pentavalent capsomeres. This raises the possibility that the L2 minor capsid protein is located in the centre of the pentavalent capsomeres. Inter-capsomere connections approximately 10 A in diameter were clearly resolved. These link adjacent capsomeres and are reminiscent of the helical connections that stabilize polyomavirus.

In vitro generation and type-specific neutralization of a human papillomavirus type 16 virion pseudotype.

We report a system for generating infectious papillomaviruses in vitro that facilitates the analysis of papillomavirus assembly, infectivity, and serologic relatedness. Cultured hamster BPHE-1 cells harboring autonomously replicating bovine papillomavirus type 1 (BPV1) genomes were infected with recombinant Semliki Forest viruses that express the structural proteins of BPV1. When plated on C127 cells, extracts from cells expressing L1 and L2 together induced numerous transformed foci that could be specifically prevented by BPV neutralizing antibodies, demonstrating that BPV infection was responsible for the focal transformation. Extracts from BPHE-1 cells expressing L1 or L2 separately were not infectious. Although Semliki Forest virus-expressed L1 self-assembled into virus-like particles (VLPs), viral DNA was detected in particles only when L2 was coexpressed with L1, indicating that genome encapsidation requires L2. Expression of human papillomavirus type 16 (HPV16) L1 and L2 together in BPHE-1 cells also yielded infectious virus. These pseudotyped virions were neutralized by antiserum to HPV16 VLPs derived from European (114/K) or African (Z-1194) HPV16 variants but not by antisera to BPV VLPs, to a poorly assembling mutant HPV16 L1 protein, or to VLPs of closely related genital HPV types. Extracts from BPHE-1 cells coexpressing BPV L1 and HPV16 L2 or HPV16 L1 and BPV L2 were not infectious. We conclude that (i) mouse C127 cells express the cell surface receptor for HPV16 and are able to uncoat HPV16 capsids; (ii) if a papillomavirus DNA packaging signal exists, then it is conserved between the BPV and HPV16 genomes; (iii) functional L1-L2 interaction exhibits type specificity; and (iv) protection by HPV virus-like particle vaccines is likely to be type specific.

Assessment of the serological relatedness of genital human papillomaviruses by hemagglutination inhibition.

To assess the potential for cross-protection among genital human papillomavirus (HPV) types in virus-like particle (VLP)-based vaccinations, inhibition of HPV VLP-mediated hemagglutination by rabbit antisera raised against HPV type 6b (HPV-6b), HPV-11, HPV-16, HPV-18, HPV-31, HPV-33, and HPV-45 was analyzed. Only highly homologous types (HPV-6b and HPV-11, and HPV-18 and HPV-45) exhibited detectable serological cross-reaction for the class of antibodies that inhibit virion-to-cell surface binding. However, analysis of neutralizing monoclonal antibodies to several animal and human papillomaviruses indicated that over half of these antibodies do not prevent cell surface binding, but these latter antibodies do not appear to be more cross-reactive in enzyme-linked immunosorbent assays than those that mediate inhibition of hemagglutination. The data strongly suggest that while there may be limited cross-protection between highly (>85% L1 amino acid identity) homologous types, protection by HPV VLP-based vaccines will be predominantly type specific.

Papillomavirus L1 capsids agglutinate mouse erythrocytes through a proteinaceous receptor.

Virus-like particles (VLPs) composed of L1 derived from bovine papillomavirus type 1 (BPV-1), several human papillomavirus types, or cottontail rabbit papillomavirus (CRPV) agglutinated mouse but not human or rat erythrocytes. Treatment of mouse erythrocytes with trypsin prevented hemagglutination (HA) by BPV-1. Sera from rabbits immunized with native CRPV VLPs, which protect against experimental CRPV infection, exhibited high titers of antibodies that inhibited CRPV VLP HA activity, while sera from rabbits immunized with denatured CRPV VLPs or native BPV VLPs, which do not protect against CRPV infection, were not inhibitory. Testing for HA inhibition is a rapid and simple method for examining the serological relatedness of papillomaviruses and measuring protective antibody titers after VLP vaccination.

Neutralization of bovine papillomavirus by antibodies to L1 and L2 capsid proteins.

We have generated four mouse monoclonal antibodies (MAbs) to bovine papillomavirus virions that bound type-specific, adjacent, and conformationally dependent epitopes on the L1 major capsid protein. All four MAbs were neutralizing at ratios of 1 MAb molecule per 5 to 25 L1 molecules, but only three effectively blocked binding of the virus to the cell surface. Therefore, antibodies can prevent papillomavirus infection by at least two mechanisms: inhibition of cell surface receptor binding and a subsequent step in the infectious pathway. The neutralizing epitopes of the bovine papillomavirus L2 minor capsid protein were mapped to the N-terminal half of L2 by blocking the neutralizing activity of full-length L2 antiserum with bacterially expressed peptides of L2. In addition, rabbit antiserum raised against amino acids 45 to 173 of L2 had a neutralizing titer of 1,000, confirming that at least part of the N terminus of L2 is exposed on the virion surface.