Conditionally essential genes were imported into Kyoto Encyclopedia of Genes and Genomes (KEGG) and metabolic pathways were queried for the essential genes. While 55 of the 482 genes were annotated as metabolic, these did not differentially populate any metabolic pathway represented in KEGG. These results indicate that in host environments P. gingivalis can utilize multiple redundant metabolic pathways, and loss of any one does not confer a fitness disadvantage. On the other hand, functional annotation characterization of essential genes revealed those associated with transport and binding proteins, the cell envelope, and protein fate were enriched.

To colonize the oral cavity in which there is fluid flow and shear forces, bacteria attach to biotic and abiotic oral surfaces. P. gingivalis possess a multiplicity of adhesins including the FimA- and Mfa1- component fimbriae, haemagglutinin (Hag)A, HagB, and HagC, and the leucine-rich repeat domain Internalin InlJ (Lamont and Jenkinson, 2000; Capestany et al., 2006; Kuboniwa and Lamont, 2010; Wright et al., 2013). Of the genes encoding these adhesins, mfa1, hagA, and inlJ were negatively selected at least 10-fold, as were genes for the Mfa fimbriae accessory proteins Mfa3 and Mfa4 (Figure 2). The gene encoding the major fimbrial structure, fimA, had a RPKM <4 and was not included in the analysis. It is unclear if underrepresentation of this gene in the input pool is due to a transposon insertion “cold-spot,” or whether loss of FimA renders the strain less fit in a mixed population with fimbriated cells. Work is ongoing to resolve this issue. The potential relevance of the FimA fimbrial structure is suggested by the observation that fimC, encoding an adhesive accessory protein (Pierce et al., 2009), was negatively selected. Moreover, the FimS/R two component system (TCS), which controls transcription across the fim operon was also negatively selected (discussed further below).

The Mfa fimbriae mediate attachment to other oral biofilm bacteria, in particular the accessory pathogen S. gordonii (Wright et al., 2013), and contribute to auto-aggregation and monotypic biofilm formation (Umemoto and Hamada, 2003; Kuboniwa et al., 2009a). Moreover, strains of P. gingivalis with biofilm-forming capacity have been found to be more aggressive in inducing abscesses (Clais et al., 2014). Involvement in epithelial cell colonization, may arise from the ability of Mfa fimbriae to mediate adherence to host cells (Kuboniwa and Lamont, 2010). Mfa1 can also selectively engage the dendritic cell (DC) C-type lectin DC-SIGN, leading to evasion of antibacterial autophagy and lysosome fusion, and intracellular persistence in myeloid DCs (Arjunan et al., 2016), properties that may contribute to survival in vivo. The functional roles of the Mfa3 and Mfa4 proteins have yet to be defined; however, Mfa3 is located at the fimbrial tip and is required for integration of Mfa4 and Mfa5 (Hasegawa et al., 2013), and Mfa4 may be necessary for the stability of the fimbrial structure (Ikai et al., 2015).

InlJ has been shown to be involved in adherence to abiotic surfaces (Capestany et al., 2006), and although InlJ is not required for epithelial cell internalization, InlJ protein expression is upregulated in P. gingivalis following contact with epithelial cells (Zhang et al., 2005). In Listeria, InlJ is a sortase-LPXTG anchored adhesin which is upregulated during infection in vivo (Sabet et al., 2008). Listerial InlJ can bind to a variety of human cells in vitro (Linden et al., 2008; Sabet et al., 2008), and oral colonization of mice with an inlJ mutant results in reduced Listeria levels compared to the parental strain (Sabet et al., 2005). The results of the current study indicate that InlJ is also important for epithelial colonization and vivo survival of P. gingivalis.

HagA is a large protein with a predicted molecular mass of 283.3 kDa and containing multiple contiguous direct repeats of 440–456 amino acids, each of which has hemagglutinin activity (Han et al., 1996). HagA can promote attachment to both epithelial and endothelial cells, and antibodies to the hemagglutinin domain are protective in animal models of oral infection (Frazer et al., 2006; Belanger et al., 2012). Our finding that disruption of hagA was deleterious for survival in epithelial cells and murine abscesses is consistent with the documented properties of HagA, and suggest a role for this protein in P. gingivalis pathogenicity. On the contrary, while HagB and HagC can also mediate attachment to host cells (Song et al., 2005), and HagB is considered a major virulence factor of the organism (Pingel et al., 2008), mutation in hagB or hagC did not diminish fitness in our infection models.

P. gingivalis exhibits a strong preference for iron in the form of hemin-containing compounds (Lewis, 2010; Smalley and Olczak, 2017). Consequently, multiple hemin uptake systems with differing affinities and specificities are present in the organism. In addition, P. gingivalis also possesses a functional ferrous iron transporter, FeoB (Dashper et al., 2005; Anaya-Bergman et al., 2015). Uptake mechanisms that impacted fitness included the Hmu (Lewis et al., 2006), Iht (Slakeski et al., 2000), and Hus (Gao et al., 2010) systems, and PGN_1335-PGN_1336, a proposed hemin uptake system composed of a surface lipoprotein and an outer membrane TonB-dependent receptor (Anaya-Bergman et al., 2015) (Figure 2). feoB, and PGN_0604, the gene encoding the iron storage protein ferritin, were also selected negatively. Hemin can enhance virulence of P. gingivalis in animal models (McKee et al., 1986), and loss of FeoB renders P. gingivalis avirulent in vivo (Dashper et al., 2005). Moreover, hemin levels affect the structure of LPS and its properties as a TLR4 antagonist or antagonist (Al-Qutub et al., 2006). A HmuR-deficient mutant of P. gingivalis has been shown to be deficient in multispecies community formation (Kuboniwa et al., 2009b), and a relationship between iron regulation and epithelial colonization has also been established, as mutation of the P. gingivalis Fur homolog showed significantly weaker adherence and invasion of epithelial cells (Ciuraszkiewicz et al., 2014). In addition, mutation of genes encoding PGN_1335-PGN_1336 reduces survival of P. gingivalis within epithelial cells (Anaya-Bergman et al., 2015). Collectively, the current results and the existing literature show hemin and inorganic iron uptake to be fundamental to fitness.

P. gingivalis is adapted to the environment of a polymicrobial biofilm (Kuboniwa et al., 2009b; Hendrickson et al., 2017), and intrusion of host tissues can imposes stress on the organism, in particular oxidative stress (Park et al., 2004; Xia et al., 2007). General stress response mechanisms including Heat Shock Proteins and the Clp system were not found to be important for either epithelial colonization or in vivo survival. However, transposon disruption of many oxidative stress resistance associated genes was detrimental to fitness in our screens (Figure 2); although the possibility the P. gingivalis experienced selective oxidative stress during the transition to cell culture or in the preparation of the inoculum for mouse infection can not be entirely eliminated. While many of these genes comprise the regulon controlled by OxyR a redox-sensitive transcriptional regulator (Diaz et al., 2006), oxyR itself was not negatively selected in our analysis, indicating a complex control mechanism for oxidative stress in P. gingivalis. Intracellular iron/hemin and oxidative stress are also interconnected as free iron and hydrogen peroxide produce reactive oxidative species through Fenton chemistry (Winterbourn, 1995). In addition, a hemin-limited growth environment significantly enhances OxyR activity (Xie and Zheng, 2012), and μ-oxo bisheme, a cell surface layer of the dimeric heme, protects P. gingivalis against H₂O₂ (Smalley and Olczak, 2017). Nitrosative stress is discussed further below.

Bacteria utilize TCS sense environmental conditions and respond with an appropriate transcriptional program (Goulian, 2010). P. gingivalis possess a limited number of TCS, six in 33277 along with the hybrid GppX and the orphan response regulator (RR) RprY (Naito et al., 2008). Among the TCS, disruption of genes encoding PGN_0012/PGN_0013 and FimS/FimR, along with the response regulator PGN_0774, reduced fitness in our model systems (Figure 2). As mentioned above, the FimS/R TCS controls transcription across the fim operon (Nishikawa and Duncan, 2010), consistent with a role for the FimA-fimbriae in fitness. However, a transcriptome analysis revealed that inactivation of fimS resulted in the differential expression of 10% of the P. gingivalis genome, including genes encoding seven different transcriptional regulators, and three extracytoplasmic sigma factor genes, (Lo et al., 2010), and so FimS/R could effect fitness independently of FimA expression. FimS/FimR can also control expression of the Mfa fimbriae (Wu et al., 2007) which were identified as necessary for fitness in the current study.

In addition to TCS RRs, strain 33277 contains 21 annotated transcriptional regulators, of which six were negatively selected (Figure 2). These are mostly of unknown function; however, SinR is a negative regulator of polysaccharide production in P. gingivalis monospecies biofilms (Yamamoto et al., 2013). PGN_1300 (HcpR) is discussed further below.

Extracellular polysaccharides of P. gingivalis are controlled by a complex multilevel regulatory system (Bainbridge et al., 2015). Such regulation may be necessary for the context-dependent coordination of polysaccharide levels, as capsule production can impede attachment and initial colonization of P. gingivalis (Davey and Duncan, 2006; Irshad et al., 2012), but is important for survival and resistance to killing by host immune cells (Singh et al., 2011). In addition to the polysaccharide production regulator SinR, genes PGN_0223 to PGN_0229, which encode proteins involved in both LPS and surface polysaccharide synthesis (Aduse-Opoku et al., 2006; Bainbridge et al., 2015), were identified as conditionally essential for fitness. Mutation of PGN_0223 results in a shortened O antigen and a significant increase in monospecies biofilm formation (Nakao et al., 2006). While P. gingivalis strain 33277 does not produce a capsule (Laine and Van Winkelhoff, 1998) it does produce disorganized extracellular polysaccharide (Maeda et al., 2008). Hence, these data would indicate that the presence of polysaccharide, and not capsule per se, is essential for fitness, at least subsequent to attachment and biofilm formation.

As an asaccharolytic organism, P. gingivalis relies on proteolytic activity to produce peptides from proteins as both a carbon and nitrogen source (Lamont and Jenkinson, 1998; Guo et al., 2010). A number of proteinases are thus produced by the organism, and a class of cysteine proteases, the gingipains, are considered of primary importance in virulence. Gingipains can have arginine (RgpA, RgpB) or lysine (Kgp) specificity, and gingipains account for a large proportion of the extracellular proteolytic activity of P. gingivalis (Potempa et al., 1997; Guo et al., 2010). RgpA and Kgp also possess haemagglutinin domains homologous to those of HagA (Potempa et al., 1997; Lamont and Jenkinson, 1998; Fitzpatrick et al., 2009; Guo et al., 2010). In addition to provision of nutritional substrates, gingipains are involved in processing of cell surface proteins and degradation of host molecules including immune effectors and matrix components. Gingipains released within host cells can also degrade host cell signaling molecules (Zhou et al., 2015; Barth and Genco, 2016). Mutants of P. gingivalis that are deficient in gingipain production display attenuated virulence in animal models, and antibodies to gingipains are protective in these in vivo models (O'Brien-Simpson et al., 2000, 2001; Kuboniwa et al., 2001; Pathirana et al., 2007; Wilensky et al., 2013). Mutations in the arginine-specific protease genes rgpA and rgpB, were identified in our assays (Figure 2). It is important to note here that RgpA and RgpB can be present in the periplasm and on the cell surface, as well as secreted extracellularly (Potempa et al., 1997; Veillard et al., 2013). Further, a feature of competitive fitness assays is the loss of an extracellular function by one mutant can be compensated in trans by other mutants that retain the property. Collectively, this would indicate that it is the periplasmic and cell-associated activities of RgpA and RgpB that make a significant contribution to fitness. Interestingly, although Kgp is thought to make a more significant contribution to pathogenicity than RgpA/B (De Diego et al., 2014), disruption of kgp did not reduce fitness in our models. Kgp plays a less important role in surface protein processing compared to RgpA/B (Kadowaki et al., 1998), and hence this role in maintaining surface integrity may be the most important contribution of the gingipains to fitness. Other proteinases identified were PrtQ and the trypsin-like protease PrtT which is involved in pathogenicity in the murine lesion model (Kesavalu et al., 1996).

The tetratricopeptide repeat (TPR) motif is a protein-protein interaction module found in multiple copies in a variety of functionally different proteins (Cerveny et al., 2013). In P. gingivalis, loss of the TPR protein TprA renders the organism less virulent in the murine subcutaneous model of infection (Kondo et al., 2010). TprA interacts with TapA, TapB, and TapC, and this complex has been shown to be cooperatively involved in abscess formation (Kondo et al., 2010). Disruption of tprA along with tapA (PGN_0152) negatively impacted fitness in both of our model systems, consistent with a role for the associated proteins in both abscess formation and epithelial cell colonization. PGN_1227, PGN_1323, and PGN_2067, which encode additional TPR motif proteins, were also selected negatively (Figure 2). The functions of these proteins have yet to be determined; however PGN_1227 expression is increased in communities of P. gingivalis with S. gordonii (Hendrickson et al., 2017).

P. gingivalis strains including 33277 can conjugally transfer both chromosomal DNA and conjugative transposons through the action of tra gene homologs of the type IV secretion system (Tribble et al., 2007). Strain 33277 contains three clusters of tra genes with more than one ortholog of many of the components. Mutants in traA, traG, traI, traJ, traK, traN, traO, traQ all exhibited reduced fitness for epithelial colonization and in vivo survival (Figure 2). These data would indicate that adaptation through horizontal gene transfer is an important process for survival of P. gingivalis in host environments.

To provide additional insights into P. gingivalis fitness, and to partially corroborate the Tn-Seq dataset, specific allelic replacements were constructed in four genes that were operationally essential for fitness. These genes encoded proteins representing major functional classes including metabolic enzymes (PGN_1444), transcriptional regulators (PGN_1300), RNA processing (PGN_0770), and genome stability (PGN_1200). PGN_1444 was the most strongly negatively selected gene in the in vivo mouse abscess model, while PGN_0770 and PGN_1200 were the third and fourth most negatively selected in the epithelial cell colonization model. PGN_1300 was negatively selected over 200-fold in both assays. The specific mutants phenocopied the Tn-Seq library results and showed a defect for epithelial colonization as well as a reduced competitive index in the mouse abscess model (Figures 3, 4). Dissection of adherence and invasion properties, further showed that PGN_0770 and PGN_1444 were dispensable for gingival epithelial cell attachment, but were necessary for internalization/intracellular survival (Figure 3).

PGN_1444 is annotated as a carbamoyl phosphate synthetase, an enzyme that catalyzes the ATP-dependent synthesis of carbamoyl phosphate from glutamine or ammonia, an important early step in the synthesis of pyrimidine and citrulline as a precursor of arginine (Cunin et al., 1986). The catabolism of arginine is an important metabolic pathway for P. gingivalis (Masuda et al., 2002), and arginine in the culture medium increases fimbrial expression and monotypic biofilm formation (Cugini et al., 2013). In Streptococcus pneumoniae, lack of carbamoyl phosphate synthase activity resulted in reduced ability to release NO and H₂O₂ (Hoffmann et al., 2006). Moreover, in Francisella tularensis carbamoyl phosphate synthetase is required for inhibition of the neutrophil respiratory burst and for intramacrophage growth (Schulert et al., 2009). Hence, while the subject requires further study, PGN_1444 may contribute to P. gingivalis fitness through generating arginine that can be used for growth and as a metabolic cue for biofilm formation, and by increasing resistance to the host professional phagocytes.

PGN_1200 is annotated as a Replication-associated recombination protein MgsA/RarA (DNA-dependent ATPase). This protein, along with RecA, is involved in the rescue of stalled replication forks, and therefore prevents genomic instability (Shibata et al., 2005). A potential role in fitness represents a novel functionality for this protein.

PGN_1300 is a transcriptional regulator of the HcpR (Crp/Fnr) family involved in resistance to nitrosative stress. In the oral cavity nitrosative stress is particularly relevant due to the high intake of dietary nitrate. The protective mechanisms against nitrosative stress are poorly understood in P. gingivalis; however, HcpR is required for growth with nitrite and nitric oxide (Boutrin et al., 2012; Lewis et al., 2012). HcpR has also been shown to play a significant role in sustaining P. gingivalis viability within epithelial and endothelial cells (Lewis et al., 2012). The current study confirms and extends these findings to show the importance of HcpR in the survival of P. gingivalis in vivo. In addition to HcpR, PGN_0004, and PGN_0959 are induced under nitric oxide stress (Boutrin et al., 2012) and these genes were also negatively selected in both the mouse and TIGK infection models. PGN_0004 is annotated as a NAD⁺-dependent sirtuin deacetylase CobB, the activity of which can also impact gene transcription (Zhou et al., 2017). Similarly, PGN_0959 is also annotated as a transcriptional regulator. In addition, rubreythrin (PGN_0302), present in both screens, can confer resistance to both oxidative and nitrosative stress (Mydel et al., 2006).

PGN_0770, rnZ, is annotated as RNase Z, a zinc phosphodiesterase, which displays tRNA 3′-processing endonuclease activity, and is involved in tRNA maturation in organisms that do not contain a chromosomally encoded CCA determinant (Pellegrini et al., 2003). RNase Z is widely distributed among bacteria (Condon and Putzer, 2002). In E. coli RNase Z plays a significant role in mRNA decay (Perwez and Kushner, 2006), and controls the levels of 6S RNA, a stable sRNA, and an important transcription regulator that acts by binding to the sigma 70-containing holoenzyme of RNA polymerase (Chen et al., 2016).