Streptococcus endopeptidases promote HPV infection in vitro

Abstract Both cervical and throat cancers are associated with human papillomavirus (HPV). HPV infection requires cleavage of the minor capsid protein L2 by furin. While furin is present in the vaginal epithelium, it is absent in oral epithelial basal cells where HPV infection occurs. The objective of this study was to investigate whether common oral bacteria express furin‐like peptidases. By screening strains representing 12 oral Streptococcus and Enterococcus species, we identified that eight Streptococcus strains displayed high levels of furin‐like peptidase activity, with S. gordonii V2016 the highest. We constructed null mutations for 14 genes encoding putative endopeptidases in S. gordonii V2016. Results showed that three endopeptidases, PepO, PulO, and SepM, had furin‐like activities. All three mutants showed decreased natural transformation by chromosomal DNA, while the pepO mutant also showed reduced transformation by plasmid DNA, indicating involvement of these endopeptidases in competence development. The purified S. gordonii PepO protein promoted infection of epithelial 293TT cells in vitro by HPV16 pseudovirus. In conclusion, oral bacteria might promote HPV infection and contribute to HPV tissue tropism and subsequent carcinogenesis in the oral cavity and throat by providing furin‐like endopeptidases.


| INTRODUC TI ON
Oral and oropharyngeal (throat) cancers are the sixth most common cancer world-wide. Each year, about 650,000 cases are diagnosed and nearly 40% of them are fatal (Ferlay, Pisani, & Parkin, 2004).
While cervical cancer is the most common human papillomavirus (HPV)-associated cancer in women, oral and oropharyngeal cancers are the most common HPV-associated cancers in men (CDC, 2012).
In recent decades, the incidences of HPV-associated oropharyngeal cancers have been increasing consistently in industrialized countries (Chaturvedi et al., 2011;Shiboski, Schmidt, & Jordan, 2005). HPVrelated cervical cancer is a predominant cause of mortality among women in African countries (De Vuyst et al., 2013).
To infect human tissues, viruses often require activation by host peptidases such as furin. Furin is a serine endopeptidase expressed in many mammalian cells, mainly localized in the trans-Golgi network (Nakayama, 1997). It belongs to the family of subtilisin-like proprotein convertases that process latent precursor proteins into their biologically active products, and it is a calcium-dependent protease and can efficiently cleave precursor proteins at their proteolytic cleavage consensus site (-RXXR↓-) (Thomas, 2002). Proteolytic activation of viral attachment factors at the site of infection is a strategy shared among many viruses (Klenk & Garten, 1994), where many viral proproteins contain consensus sites for furin. These include the minor capsid L2 protein of HPV (Richards, Lowy, Schiller, & Day, 2006) and the envelope glycoprotein gB of cytomegalovirus (CMV) (Jean et al., 2000) and Epstein-Barr virus (EBV) (Sorem & Longnecker, 2009). Among many HPV subtypes, HPV16 is the most oncogenic.
Infection by HPV16 is initiated by binding to the receptor heparan sulfate proteoglycans (HSPGs) on exposed basement membrane. This exposes the L2 cleavage site. Following cleavage by furin, HPV16 invades the tissue (Raff et al., 2013;Richards et al., 2006;Schiller, Day, & Kines, 2010). Although HSPGs exist in nearly all tissues throughout the body (Sarrazin, Lamanna, & Esko, 2011), the expression of furin varies. In all layers of the mouse vaginal epithelium, furin expression is high (Kines, Thompson, Lowy, Schiller, & Day, 2009), while in the human oral epithelium, furin expression is low, detectable only in the upper spinous and surface layers, not in the lower basal cell layer, where HPV initiates infection (López de Cicco, Bassi, Page, & Klein-Szanto, 2002). The absence of furin in local tissues cannot explain why HPV-induced cancers commonly occur in the oral cavity and oropharynx. Apparently, a gap exists in current understanding of HPV infection. It is largely based on in vitro cell culture studies under germ-free conditions (Schiller et al., 2010), experiments with nonhuman animals (Kines et al., 2009), and level of infection reported in specific populations, such as young men in the United States and women in Africa (Curado & Boyle, 2013;De Vuyst et al., 2013).
The oral cavity is not sterile. It is colonized by billions of microorganisms, collectively referred to as the oral microbiome (McLean, 2014) mostly growing as biofilm attached to the mucosal and tooth surfaces (Kolenbrander, 2000). For HPV to infect the oral tissue, the lack of host furin in epithelial basal cells might be compensated by furin-like peptidases from the oral microbiome. In our previous publications, we reported that Streptococcus promotes HPV16 entry into oral keratinocytes and oral squamous cell carcinoma cells. These suggest a possible interplay for Streptococcus and other similar bacteria to contribute to HPV16 entry (Schwartz et al., 2012;Tao, Pavlova, Gasparovich, Jin, & Schwartz, 2015). It is, therefore, reasonable to investigate how these microorganisms could play a role in determining HPV tissue tropism, and subsequently, how they might contribute to the development of tissue-specific neoplasia.
Several anatomic sites in the oral cavity and oropharynx, including the back of the mouth, the base of the tongue, and the tonsils, are subject to HPV infection. These sites have nonkeratinized thin mucosal lining that form crypts, which house a complex microbial population. Included are various bacteria and viruses that cause periodontitis, common cold, influenza, and cancer. Tissue tropism varies among different viruses, and bacteria found at specific locations in the body might provide necessary conditions for infection by specific viruses (Doolittle & Webster-Cyriaque, 2014). Although a furin-like peptidase has been reported in the eukaryotic parasite Cryptosporidium parvum (Wanyiri et al., 2007), in hundreds of bacterial species comprising the human microbiome (Human Microbiome Project, 2012), furin-like peptidases have not been reported. Here, we continue our studies of interaction between Streptococcus and HPV16 entry capability. We report finding of three furin-like peptidases in S. gordonii, characterizing one of them, and observation of its promotion of HPV infection in vitro.

| Bacterial strains, growth conditions, and plasmids
Bacterial strains and plasmids used in these studies are described in

| Furin-like peptidase assay
Since certain viruses require furin activation in order to infect, we analyzed oral and throat bacterial strains for furin-like peptidase activity. Bacteria were grown to mid-exponential phase to OD 600 of ~0.7 in THY broth for Streptococcus or Enterococcus and MRS broth for Lactobacillus, and 100 μl culture was harvested, washed, and resuspended in Na/MES buffer for analysis of furin-like peptidase activity. The fluorogenic furin substrate Boc-RVRR-AMC (Enzo Life Sciences) at 50 μmol/L in Na/MES buffer [pH 7.0, ethanesulfonic acid], 1 mmol/L CaCl 2 , and 0.01% Triton X-100] was added. The mixture was incubated for 30 min at 37°C. The cells were removed by centrifugation. The fluorescence released from the substrate in the supernatant was read with the VICTOR X5 Plate Reader of PerkinElmer (Excitation 360 nm; Emission 460 nm). For analysis of purified peptidases, the protein was added to the reaction mixture and incubated for 30 min at 37°C before fluorescence reading. The purified recombinant human furin from New England BioLabs (NEB) was used as control.

| Construction of mutant strains
The V2016 sgc, subAB, sepA, and secA2 deletion mutants were obtained by allelic exchange (Pavlova, Jin, Gasparovich, & Tao, 2013), while the remaining 10 genes were inactivated by using the streptococcal integration plasmid pSF151 (Tao, 1998). Therefore, a total of 13 genetic loci were inactivated by allelic replacement or insertion  (Sambrook, Fritsch, & Maniatis, 1989). Multiple pairs of oligonucleotides (Integrated DNA Technologies) used in this study are shown in Table 2. Chromosomal DNA was prepared by the glass bead method (Ranhand, 1974

| Purification of recombinant S. gordonii PepO
The pepO (SGO_1799) gene was amplified from S. gordonii V2016 genomic DNA using primers RW112 and RW113. The fragment was assembled into pET21a, following digest with XhoI and NdeI, upstream of a 6-HIS tag using NEB HiFi builder (NEB). The following product was transformed into E. coli DH5α cells. Isolated plasmid bearing the insert of interest was verified by transformation into E. coli BL21(DE3), induction with 0.5 mmol/L IPTG and the demonstrated ability of protein of the correct molecular mass in the lysate to bind nickel resin. Protein was purified as described in Wilkening, Chang, and Federle (2016

| Interaction of peptidases with bacteria and mammalian cells
Because the purified rPepO from bacteria displayed low peptidase activity compared with enzyme associated with bacteria, we studied interaction between the rPepO enzyme and bacteria or mammalian cells, using furin as control. Bacterial cells were those that naturally displayed a low furin-like peptidase activity, including the purified proteins in buffer without cells as control were assayed for peptidase activity as described above.

| Interaction of PepO with bacterial cell wall peptidoglycans
Two bacterial cell wall peptidoglycans were used to interact with furin and the S. gordonii PepO protein. One is from Bacillus subtilis and the other is from S. pyogenes. These two cell walls were tested because B. subtilis cell wall was available commercially (Sigma), and S. pyogenes cell was isolated in our laboratory for another project.
S. pyogenes NZ131 was grown overnight in THY broth. Cells were isolated by centrifugation and suspended in 0.1 mol/L Tris-HCl pH 6.8 with 0.25% SDS. The cells in solution were placed inside a conical tube and incubated in a boiling water bath for 2 hr to kill the bacteria and inactivate wall enzymes. Bacterial debris was collected were mixed with these bacterial cell wall peptidoglycans for 60 min at room temperature and assayed for furin activity.

| Activation of PepO by proteases
Since interaction with bacteria or cells boosted the furin-like peptidase activity of the human furin and purified PepO of S. gordonii, we tested if other proteases could activate PepO. Briefly, proteases including trypsin, chymotrypsin, papain, pepsin, pronase, protease, and proteinase K (from Sigma) solutions were made at three concentrations, 1 mg/ml, 100 μg/ml, and 10 μg/ml in Na/MES buffer

| Statistical analysis
The paired 2-tailed Student t test with a confidence limit of p < .05 determined the level of significance between two comparative

| Bacterial genes encoding furin-like peptidase activity
We analyzed 12 oral and throat bacterial strains representing 11 Streptococcus and 1 Enterococcus species (Table 1) for furin-like peptidase activities using a fluorogenic furin substrate Boc-RVRR-AMC (Enzo Life Sciences) (Cruz, Biryukov, Conway, & Meyers, 2015). Upon incubation of bacterial cells with the substrate, 8 of the 11 streptococcal strains tested were positive for furin-like activity (73%; Figure 1a), whereas the Enterococcus strain was negative. S. gordonii V2016 displayed the highest furin-like peptidase activity. We also tested 19 oral Lactobacillus strains for furin-like activity ( Table 1) and found that 18 were negative (no higher than the buffer control), and 1 (L. rhamnosus OLB25b) was positive. Thus, in comparison to a panel of Lactobacillus strains, the Streptococcus species were more likely to express extracellular furin-like enzymes.
To identify genes encoding furin-like peptidases in Streptococcus, we used the human furin protein sequence to BLAST against the genomic sequences of the Streptococcus genus. The subtilisin-like serine protease was found as a homolog. However, blasting the S. gordonii genome (Vickerman, Iobst, Jesionowski, & Gill, 2007) with the furin sequence did not yield significant similarity. We then performed in silico analysis of S. gordonii genome for all genes encoding putative peptidases and identified 47 genes. As furin is an endopeptidase, we dismissed genes encoding exopeptidases (aminopeptidases and carboxypeptidases), leaving 14 genes to consider as candidates for furin activity. Among them, two genes, SGO_0316 and SGO_0317, F I G U R E 1 (a) Furin-like activity of strains representing 12 oral and throat Streptococcus and Enterococcus species. (b) Furin-like activities of S. gordonii V2016 and its 13 peptidase and the secA2 mutants (one measure per sample). (c) Furin-like activities of S. gordonii and its furinlike peptidase-negative mutants. **Statistically very significant difference (p < .01) between different groups: the wild-type, single, double and triple mutants, and the buffer control. Each bar represents the mean of triplicate values ± standard deviation are linked, and thus, a total of 13 genetic loci were subjected to mutagenesis with the kanamycin resistance marker to achieve allelic exchange (Pavlova et al., 2013) or inactivation by insertion duplication with the plasmid pSF151 (Tao, 1998) (Table 1). In addition, to identify if the Sec2 system of S. gordonii (Bensing & Sullam, 2009) is required for secretion of these furin-like peptidases without a typical signal peptide, we constructed a secA2 deletion mutant.
Each mutant was subjected to the furin-like activity assay with the fluorogenic furin substrate. Results showed that inactivation of 3 genetic loci, SGO_1799, SGO_0664 and SGO_0652, each had reduced furin-like activity (Figure 1b). We used chromosomal DNA isolated from one mutant as a substrate for natural transformation by another mutant containing a different antibiotic resistance marker to generate double-and triple-protease mutants (Tao, 1998).
These mutants showed further reduced furin-like activity compared with the wild-type and single-gene knockout mutants (Figure 1c The SGO_0652 locus encodes a 347-amino acid with a calculated molecular weight of 38.1 kDa. Without a signal peptide, this locus is homologous to the conserved sepM gene in S. mutans, which encodes a membrane-associated peptidase required for processing competence stimulating peptide (CSP) for competence development and quorum sensing (Biswas, Cao, Kim, & Biswas, 2015;Hossain & Biswas, 2012). With one transmembrane domain, S. gordonii sepM is also predicted as a membrane-associated peptidase. The genomic location shows that S. gordonii sepM appears to be the last gene in an operon with two other genes, which encode phosphopantetheine adenylyltransferase (coaD) involved in coenzyme A biosynthesis and rRNA methyltransferase. It is unknown why the three seemingly unrelated genes co-transcribe in an operon.
To test if these genetic loci in other Streptococcus species also encode peptidases with furin-like activity, we transformed chromosomal DNA isolated from S. gordonii pepO, pulO, and sepM mutants into S. sanguinis SK36. We were able to obtain transformants with S. gordonii pepO and pulO DNAs, but not with sepM DNA (Table 1).
Furin assay with these two S. sanguinis peptidase mutants showed reduction of furin-like peptidase activity in comparison with the wild-type strain SK36. Therefore, at least two furin-like endopeptidases, PepO and PulO, may be present in multiple oral streptococcal species.

| Interacting with cells and chymotrypsin activates S. gordonii rPepO
We

| PepO inactivation reduces competence in S. gordonii
While generating the double and triple mutants, we noticed reduction of transformation efficiency with chromosomal DNA in these S. gordonii peptidase-defective mutants. Therefore, we studied their genetic transformation. As shown in Figure

| S. gordonii PepO promotes HPV infection
We studied HPV16 pseudovirus (PsV) entry into the 293TT cells to evaluate the role of furin-like activity of S. gordonii PepO in viral activation. Since the 293TT cell has endogenous furin activity, to study the role of exogenous furin-like peptidase, we added the furin inhibitor decanoyl-RVKR-chloromethylketone (CMK) to suppress its endogenous furin activity (Richards et al., 2006). As shown in Figure 4Aa and Ba, without CMK treatment, the HPV16 PsV dis-  (Figures 4Ab and 4Bb). The S. gordonii PepO protein-treated group (Figure 4Ad) showed a higher HPV16 PsV entry than the human furin-treated one (Figure 4Ac).

| D ISCUSS I ON
It has been reported that bacteria colonizing the human airway activate influenza viruses by cleaving the viral surface glycoprotein hemagglutinin with peptidase (Böttcher-Friebertshäuser, Klenk, & Garten, 2013). However, there have been no reports on HPV activation by bacteria. Unlike influenza viruses, activation of HPV requires the cleavage of its L2 protein at the furin consensus site (-RXXR↓-) (Richards et al., 2006;Schiller et al., 2010). To our knowledge, this is the first report of a furin-like peptidase in bacteria capable of promoting HPV16 infection in vitro.

After screening 12 representative strains of Streptococcus and
Enterococcus species for furin-like peptidases, we found that S. gordonii V2016 displayed the highest furin-like peptidase activity among tested strains. S. gordonii not only colonizes the human oral cavity but also the throat (Frandsen, Pedrazzoli, & Kilian, 1991), which is a hot spot for HPV infection-associated cancers. By analyzing the S. gordonii genome, we identified 14 genes encoding endoproteases and constructed null mutations in these genes. Defects in any of the three genes, pepO, pulO, and sepM, showed reduction in furin-like peptidase activity ( Figure 1B). Although the initial screening showed that the pepO mutant had the most reduction, repeated assays with triplicate samples showed comparable reductions in furin-like peptidase among these three mutants ( Figure 1C).
PepO, a member of the M13 family of zinc metalloendopeptidases, has been well studied in several different bacteria. Streptococcus parasanguinis PepO is similar to the human endothelin-converting enzyme 1 (ECE-1) (Oetjen et al., 2001).
PulO is homologous to the prepilin peptidase (PilD) involved in Type IV pili production in other bacteria. These pili were found recently in the Gram-positive bacterium Streptococcus sanguinis as a part of filamentous nano-machine for bacterial motility (Gurung et al., 2016). In S. pneumoniae, the type IV pilus mediates DNA binding during natural transformation (Laurenceau et al., 2013). The S. gordonii CSP induces PulO expression (Vickerman et al., 2007). We found that inactivation of pulO retarded S. gordonii growth rate by about 50%.
Among these three furin-like peptidases, PulO and SepM are membrane proteins that are bound to the bacterial cell surface.
This may limit their ability to interact with viruses. PepO is not membrane-bound, so it may be released into the environment to interact with viruses in distance. Therefore, we focused our efforts on characterizing S. gordonii PepO.
After cloning and purification of S. gordonii PepO, we found it had a low activity against the fluorogenic furin-specific substrate.
Apparently, protein maturation may be required. By interacting with different bacteria, cells, bacterial peptidoglycans, and multiple different proteases, we found that chymotrypsin enhanced S. gordonii PepO in a dose-dependent manner (Figure 2e). This suggests that maturation of S. gordonii PepO may require a chymotrypsin-like protease. Interestingly, interaction with two bacterial peptidoglycans increased the activity of rFurin (Figure 2c). The mechanism is unknown, but it may be due to bacterial peptidoglycan-associated protease. For example, the B. subtilis peptidoglycan with buffer alone displayed high furin-like activity.
Without a signal peptide, PepO normally would be considered as a cytoplasmic protein. However, because inactivation of pepO substantially reduced the furin-like peptidase activity, PepO appears to be present outside the cell. Since furin-like peptidase activity is not reduced in the secA2 mutant, a Sec or SecA2 independent pathway (Bensing & Sullam, 2009) might secrete PepO. In Streptococcus pneumoniae (Agarwal et al., 2013), S. pyogenes (Honda-Ogawa et al., 2017) and Porphyromonas gingivalis (Ansai et al., 2003), PepO is secreted by an unknown pathway to facilitate bacteria invasion into host cells.
Because inactivation of pepO substantially reduced competence in S. gordonii, the intended function of PepO might be associated with competence development. Although S. gordonii pepO expression is not induced by CSP (Vickerman et al., 2007), at least two peptidases in Streptococcus, ComA (Ishii et al., 2010) and SepM (Biswas et al., 2015), have been reported to promote competence by processing CSP to its mature form. The mechanism by which S. gordonii PepO facilitates genetic transformation is unknown, but it appears to play this role intracellularly, because adding purified PepO to competent pepO mutant culture did not restore its transformation to the level of the wild type. This is different from the SepM peptidase in S. mutans, which works extracellularly to process CSP (Hossain & Biswas, 2012). Nonetheless, linking PepO to bacterial competence is novel, as currently identified bacterial competence factors have not included PepO (Straume, Stamsås, & Håvarstein, 2015). Although PepO was identified to inactivate CSP in S. pneumoniae, it was not proven to regular competence in this bacterium (Bergé, Langen, Claverys, & Martin, 2002).
The proprotein convertase furin not only is essential for activation of certain viruses such as HPV (Richards et al., 2006;Schiller et al., 2010) but also promotes cancer development and metastasis (Jaaks & Bernasconi, 2017).

ACK N OWLED G M ENTS
This study was supported, in part, by the National Institutes of Health grant R01-AI091779 and the Burroughs Wellcome Fund Investigators in the Pathogenesis of Infectious Diseases Award.

RW was partially supported by the American Heart Association
Predoctoral Fellowship 15PRE22710027.

CO N FLI C T O F I NTE R E S T
None declared.