Molecular Signatures Related to the Virulence of Bacillus cereus Sensu Lato, a Leading Cause of Devastating Endophthalmitis

In this study, we provided a detailed and comprehensive clinicopathological and pathogenic report of Bacillus endophthalmitis over the 8 years of the study period. We first reported the whole-genome sequence of Bacillus spp. causing devastating endophthalmitis and found that Bacillus toyonensis is able to cause endophthalmitis. Finally, we revealed significant endophthalmitis-associated virulence genes involved in hemolysis, immunity inhibition, and pathogenesis. Overall, as more sequencing data sets become available, these data will facilitate comparative research and will reveal the emergence of pathogenic “ocular bacteria.”

E ndophthalmitis is an ocular inflammatory condition within the anterior or posterior segment of the eye, which is commonly caused by bacterial or fungal infections. Despite aggressive treatment and surgical intervention, endophthalmitis frequently leads to partial or complete loss of vision. Posttraumatic endophthalmitis results from bacterial infections following ophthalmic surgery and trauma. One of the most explosive and devastating forms of posttraumatic endophthalmitis is caused by Bacillus cereus, which typically migrates throughout the eye causing a rapid and severe intraocular inflammatory response, leading to a loss of functional vision. Although posttraumatic endophthalmitis caused by B. cereus is rare, it is ranked second behind that caused by staphylococci, occurring in ϳ2 to 7% of all penetrating eye wounds (1).
A molecular understanding of the genetic basis and evolution of pathogens for host adaptation is important for the control of infection and therapeutic intervention. Advances in genome sequencing have facilitated the comparison of genomes from large numbers of bacterial isolates and revealed how the pathogenicity of bacterial clones can evolve via mutation and horizontal gene transfer (9,10). From a bacterial perspective, the acquisition of new genes provides the flexibility to adapt and take advantage of novel niches. Although an array of studies have provided insight into genetically and phenotypically diverse environmental Bacillus species strains and clinical isolates associated with gastrointestinal and ocular infections (2,6,(11)(12)(13)(14), genomic information and the key determinants of virulence leading to posttraumatic endophthalmitis in vivo remain unclear (15,16).
In this study, we present a retrospective analysis of 52 cases of posttraumatic Bacillus endophthalmitis between January 2010 and December 2018 and determine the pathogenicity of intraocular Bacillus species isolates in mouse models of bacterial endophthalmitis. To investigate the genomic characteristics and molecular signatures of intraocular strains, we sequenced the genomes of eight Bacillus intraocular isolates spanning different patients. Incorporating publicly available genome sequences for Bacillus spp., we investigated the population structure and evolution of intraocular isolates through phylogenomic analysis. We also performed pangenome-wide association studies to identify the association between gene pools and disease outcomes and to identify different sets of accessory genes associated with virulence traits in endophthalmitis. These data have important implications for our understanding of the pathogenicity in Bacillus spp. and will facilitate future epidemiology and therapeutic studies.

RESULTS
Clinicopathological characteristics of Bacillus endophthalmitis. We followed 52 patients with posttraumatic endophthalmitis attending the Affiliated Eye Hospital of Wenzhou Medical University, China, between January 2010 and December 2018. The patient inclusion criteria were as previously described (15), namely, infective endophthalmitis caused by penetrating wounds, positive Gram stains for vitreous biopsy specimens or aqueous humor positivity, Bacillus spp. that grew on the vitreous or aqueous humor, and absence of other infectious diseases. Finally, 52 cases (48 males and 4 females) with a mean age of 47.3 Ϯ 17.3 years (range: 3 to 82 years) were included in this retrospective study ( Table 1). The follow-up time was 20.2 Ϯ 9.5 months (range: 8 to 38 months). Penetrating trauma among the patients was caused by iron (n ϭ 32), pebbles (n ϭ 10), branches (n ϭ 3), electrical wires (n ϭ 2), firecrackers (n ϭ 2), glass (n ϭ 1), and other reasons (n ϭ 2). Of the patients, 92% (48/52) developed rapid and aggressive ocular pain, redness, a thick yellowish discharge, and notably reduced visual acuity. Despite aggressive therapeutic and surgical intervention, endophthalmitis generally resulted in a partial or complete loss of vision. The visual acuity of the 52 patients was as follows: 55% (29/52) with a complete loss of vision, including no light perception (NLP) (14/29) and enucleation of the eye (15/29); 17% (9/52) with visual acuity of only light perception (LP); 15% (8/52) with hand movements (HM); 4% (2/52) with visual acuity of the count finger (CF); and 8% (4/52) with limited visual acuity ranging from 20/200 to 20/40. Figure 1 shows the clinicopathological characteristics of three representative posttraumatic endophthalmitis cases (Fig. 1, rows 1 to 3) and Staphylococcus endophthalmitis controls (Fig. 1, row 4). The typical clinical manifestations were severe swelling of the eyelids, conjunctival congestion, conjunctival chemosis, corneal edema, and hypopyon in the anterior chamber ( Fig. 1, a1 to a3, anterior segment photograph). The shape, brightness, position, and after-movement of echogenic dots in the dark area of vitreous body are shown (Fig. 1, b1 to b3, B scan ultrasound images). The vitreous smear indicated a Gram-positive bacterium as the causative factor ( Fig. 1, c1 to c3). The bacteria were often arranged in pairs or chains and were typically located externally to phagocytes. Upon 24 h of vitreous humor and aqueous humor culture in sheep blood agar medium, the strains formed large gray-white opaque colonies of different sizes, irregular edges, wax-like clusters, and obvious hemolytic rings (Fig. 1, d1 to d3). Colony morphology combined with biochemical assessments (API 20E and API 50CHB strip identification) showed that the strains belonged to the Bacillus cereus group. Compared to Bacillus infection, Staphylococcus epidermidis, which is the more prevalent pathogen in posttraumatic endophthalmitis (17), caused fewer inflammatory or pathogenic ef- fects ( Fig. 1, a4 to d4). At the same time, the vitreous smear of the patients indicated that S. epidermidis can be easily phagocytized by neutrophils, which differs from Bacillus cereus ( Fig. 1, c1 to c4). Taxonomic status and pathogenicity of the intraocular isolates. To determine the taxonomic status of the eight intraocular isolates, we sequenced the whole genomes of the eight intraocular strains LY1, LY2, LY5, LY6, LY7, LY178, LY557, and LY460 (see Table S1 in the supplemental material) and performed phylogenetic analysis using RAxML (18). As shown in Fig. S1, although all intraocular isolates were members of the B. cereus group, they belonged to independent species. Bacillus strains LY1, LY5, LY6, and LY178 were members of B. thuringiensis; Bacillus strains LY2, LY7, and LY460 were B. cereus sensu stricto; and Bacillus strain LY557 is a novel strain of B. toyonensis that is the first to be reported in endophthalmitis.
To investigate whether the Bacillus isolates caused endophthalmitis, we injected wild-type C57BL/6 mice with the three representative isolates of Bacillus species and phosphate-buffered saline (PBS) controls and analyzed retinal function, histology, and inflammatory mediator expression of the vitreous chamber in vivo at 3 h, 6 h, 9 h, and 12 h postinjection as previously described (19). We measured scotopic electroretinography (ERG) responses to determine whether infection with the isolates influenced retinal function. We found that both A (57%)-and B (45%)-wave amplitudes were significantly reduced in infected eyes 6 h postinfection (hpi), (P ϭ 6.3 ϫ 10 Ϫ7 and 7.7 ϫ 10 Ϫ8 for A-wave and B-wave, respectively, Student's t test) and almost completely disappeared at 12 hpi ( Fig. 2A). To further elucidate the degree of retinal damage caused by Bacillus pathogens, we infected mice with normal ocular surface flora of Staphylococcus epidermidis and performed electroretinogram experiments. We found that Bacillus was able to cause more severe retinal damage after 6 h of infection. These results were consistent with the clinical phenotype (Fig. S2A). Compared to untreated eyes ( Fig. S2B), hematoxylin and eosin (HE) staining in injected mouse eyeball tissue revealed a large number of inflammatory cells in the vitreous cavity following infection (6 hpi in strain LY7, 9 hpi in strain LY178, and 12 hpi in strain LY557) (P Ͻ 0.01, Student's t test [ Fig. 2B and Fig. S2C]). In addition, the expression of IL-1␤ in eyeball homogenates rapidly increased at 12 hpi, while that of IL-6 and TNF-␣ increased dramatically at 9 hpi (P Ͻ 0.01, Student's t test [ Fig. 2C]). Together, these results validate that the intraocular isolates of Bacillus spp. directly caused severe intraocular inflammation and retinal damage leading to vision loss. Importantly, these results indicate that B. toyonensis LY557 is an emerging but rare ocular pathogen capable of severe intraocular inflammation and retinal damage.
Phylogenetic structure and evolution of intraocular strains. A maximum likelihood (ML) tree of 129 intraocular isolate-related genome sequences was created based on 192,201 core genome single nucleotide polymorphisms (SNPs) in the 1,184 core genes (Fig. 3A). Phylogenetic analysis showed a wide diversity of clearly defined lineages within the genus. We divided 129 genomes into 4 lineages, including 8 subclusters (SC1 to -8) based on the analysis of the core ML tree using RAMI (20) (Fig. 3A). Principal-component analysis (PCA) using 5,368 accessory genes of intermediate frequency (5% to 95%) clearly distinguished the four groups (Fig. 3B), which were comparable to phylogenetic lineages. These data sets provide whole-genome support for the theory that the phylogroups constitute discrete bacterial populations that are evolving independently. As shown in Fig. 3A, although our eight intraocular isolates were distributed in lineages 1, 2, and 3, they were mainly comprised (6/8) of lineage 1. Using a pangenome approach, the Bacillus population had an open pangenome, indicating a far larger gene pool within the Bacillus population and frequent gene gain and loss events (Fig. 3C). When comparing the total number of unique genes to a given number of genomes, we found that lineage 1, which included intraocular isolates, had more unique gene numbers than lineage 4, another dominant lineage of the Bacillus genus (P ϭ 7.3 ϫ 10 Ϫ8 , Wilcoxon rank sum test).
The analysis of virulence genes has formed a core narrative in our understanding of pathogen evolution. Several genetic loci have been identified as virulent based on the virulence factor database (VFDB) (21), which includes gene clusters involved in enzyme activity, immune evasion, immune inhibition, and toxins. Although these virulence determinants have been characterized in B. cereus ATCC 10987 and ATCC 14579 genomes, their distribution across members of the genus remains unclear. From the clustering distribution of pathogenic genes (Fig. 3D), it is evident that the virulence genes form variable patterns between lineage 1 and lineage 4. Analysis of the number of virulence factors between the two lineages indicated a significantly higher prevalence of virulence genes in lineage 1 (P ϭ 4.38 ϫ 10 Ϫ16 , Wilcoxon rank sum test [ Fig. 3E]), which implied that this lineage has higher pathogenicity than lineage 4. In addition, we sequenced the whole genomes of two gastrointestinal isolates and collected 13 genomes of publicly available gastrointestinal strains (Fig. S1). We found that the intraocular strains were enriched in lineage 1, in which isolates showed more plasticity and pathogenicity, while gastrointestinal isolates were enriched in lineage 4 (P Ͻ 0.01, hypergeometric test [ Fig. 3F]).

DISCUSSION
Through sequencing and analysis of the genomes of eight intraocular isolates of Bacillus spp., we have gained insight into the phylogenetic evolution and molecular signatures of intraocular pathogens. To our knowledge, this is the first and largest collection of genomic sequences of bacterial isolates from patients with Bacillus endophthalmitis. Although some case studies covering whole-genome isolates of bacterial pathogens in endophthalmitis have been performed (26,27), this study uniquely focused on opportunistic Bacillus pathogens that cause devastating eye infections. Initial bacterial isolates were sampled in 2010, and sampling was continued, allowing us to obtain a comprehensive picture of the genetic alterations and adaptability of microorganisms to infect the eye and to facilitate future studies of endophthalmitis epidemiology and therapeutics.
In this study, we defined the genomic features and phylogenetic diversity of Bacillus spp. causing devastating endophthalmitis. All genomes revealed high quality and similar benchmark data, including genome size, GC content, and the number of predicted genes (see Table S1 in the supplemental material). The completeness of each whole-genome sequencing was more than 99%, and we have low contamination levels in our sequencing libraries. The high coverage of our genome assembly helped obtain complete and accurate genome and gene analysis. Phylogenetic analysis and intravitreal injections showed that the intraocular isolates belong to B. cereus, B. thuringiensis, and a novel strain of B. toyonensis which was closely related to Bacillus toyonensis sp. nov., a new species of the Bacillus cereus group (28). Interestingly, the phylogenetic position of Bacillus cereus LY7 is closely related to the gastrointestinal isolate RIMV/BC/ 126 (29), which was originally submitted as Bacillus cereus but which was modified in August 2018 due to the average nucleotide identity (ANI) data of Bacillus wiedmannii strain FCC41 (30). This further indicates the increasing application of whole-genome sequences to discriminate information and provide taxonomic identification that surpasses previous typing methods. These observations highlight the need for further studies regarding the genomics and epidemiology of Bacillus spp. in endophthalmitis.
It is well established that Bacillus spp. produce a range of tissue-destructive exotoxins that contribute to devastating outcomes in endophthalmitis (4)(5)(6)(7)(8). However, recent studies into the pathogenesis of Bacillus-induced endophthalmitis have identified several other factors that contribute to the poor outcome of Bacillus endophthalmitis. Through whole-genome analysis, we have provided a robust framework to redefine the species clusters of this genus, mapping genetic traits and scoring the significance of pathogenic genes based on their presence or absence. Through analysis of the genes gained or lost from intraocular lineages, it is clear that four virulence factors (plcA-2, InhA-3, InhA-4, and hblA-5) and a flagellar cap protein, FliD (fliD), are overrepresented in intraocular isolates compared to gastrointestinal isolates. This comparison should be interpreted with caution since it was underpowered (8 intraocular isolates versus 15 gastrointestinal isolates). However, the function of these genes may be related to the ability of intraocular strains to cause devastating endophthalmitis. In addition, to investigate the expression and function of plcA-2, we used pET-28a expression vectors to express recombinant His-tagged plcA-2 in Escherichia coli and performed purifications using the Äkta pure chromatography system (see Fig. S5 in the supplemental material). Phosphatidylinositol-specific phospholipase Cs (PI-PLCs) are secreted by various Gram-positive bacterial pathogens, including Bacillus spp. and Listeria monocytogenes (31). Due to the different protein structures and alternative cleavage activity of glycosyl-PI (GPI)-anchored proteins (24,32), PI-PLCs from different species dysregulate the immune response through different mechanisms: Bacillus enzymes downmodulate dendritic cell function and T-cell responses (33), while enzymes from L. monocytogenes reduce phosphatidylinositol (3,4,5)-trisphosphate (PIP3) levels on preautophagosomal structures, leading to the prevention of autophagic flux, favoring the escape of cytosolic bacteria from host autophagic defenses (34). Metallopeptidase InhA, the major component of the exosporium, is essential for efficient spore escape from macrophages through permeability alterations in the macrophage membranes that liberate the mature form of the crucial escape effector NprA (35,36). Although Callegan et al. showed that the role of the BL toxin in intraocular B. cereus infection was limited (8), Beecher et al. demonstrated that purified hemolysin BL (HBL) is highly damaging to retinal tissue in vitro and that the intravitreal injection of HBL into rabbits produces symptoms of B. cereus endophthalmitis (4,5). B. cereus possesses flagella which allow the organism to migrate within the eye during intraocular infection. When B. cereus infects the eye, it rapidly migrates from the initial site of injection into the vitreous and to the anterior segment within 6 to 12 hpi. The absence of motility affects toxin production; hence, nonmotile Bacillus is less pathogenic (16,37). Taken together, the acquisition of plcA-2, InhA-3, InhA-4, hblA-5, and fliD suggests that escaping host immune responses, particularly from macrophages, increases virulence and may be central to the ability of intraocular strains to cause devastating endophthalmitis. It is often observed that as pathogens acquire virulence determinants, they increasingly adapt to specific hosts (38)(39)(40). The application of whole-genome sequencing to the study of host adaptation has revealed remarkable patterns of evolution in Yersinia (41), Klebsiella (42), Salmonella (38,43), and Pseudomonas (44). Bacillus is of particular importance as it gains functional genes during host infection within intraocular-pathogenic lineages. Thus, the dramatic genetic changes demonstrated across the different lineages form a paradigm rather than accidental events and appear to emphasize the emergence of pathogenic "ocular bacteria." Conclusions. In this study, we reveal how a clinical collection of bacteria sampled from acutely infected patients constitutes a valuable basis for understanding the pathogenicity and evolution of pathogens in vivo. To our knowledge, this is the first report of Bacillus isolates from human intraocular infections to be sequenced, and B. toyonensis strain LY557 has the ability to cause complete retinal function loss, in a manner comparable to B. thuringiensis and B. cereus. Notably, we identified several molecular signatures of virulence and motility genes that were overrepresented in intraocular isolates, which may facilitate host adaption. Sampling should now be continued, as the results obtained facilitate comparative studies as more sequencing data sets become available. This may aid the design of future intervention strategies in the clinical setting.

MATERIALS AND METHODS
Clinical characteristics. Three patients' eyeballs were observed, and all underwent complete ophthalmic examinations and B-scan ultrasonography. Vitreous smears were stained with Gram's solution. Colony morphological and bacterial microstructures were observed after culturing on blood plates at 37°C for 24 h.
Bacterial strains and experimental bacterial endophthalmitis. The representative strains used in experimental mouse endophthalmitis models included Bacillus sp. strains LY557, LY178, and LY7, which are clinical strains isolated from patients diagnosed with posttraumatic endophthalmitis. Three inoculated strains were cultured for 6 h at 37°C and 250 rpm and harvested during the log phase of growth. All female wild-type C57BL/6 mice, which have no retinal degeneration such as rd1 and rd8, were purchased from the Animal Center of Wenzhou Medical University (Wenzhou, China). Mice were kept in a constant-temperature environment with a 12-h on-or-off cycle and provided water and a normal diet. For the experiments, 6-to 8-week-old female C57BL/6 mice were anesthetized with a combined preparation of ketamine (70 mg/kg body weight) and xylazine (5 mg/kg body weight). Experimental bacterial endophthalmitis models were established by injecting 1 l of phosphate-buffered saline (PBS; pH 7.4) containing 100 CFU bacteria into the vitreous chamber of the eyes. Contralateral eyeballs were injected with 1 l of sterile PBS as a control group.
Histopathological analysis. Infected eyeballs and contralateral eyeballs were harvested from each mouse at 3, 6, 9, and 12 hpi. Each eyeball was incubated in Davison's fixative for 36 h at room temperature. Eyeballs were dehydrated using graded ethanol, paraffin embedded, sectioned, and stained with H&E. Sections are representative of at least 3 eyes observed at each time point from a minimum of 3 independent experiments.
Electroretinography. ERG was used to detect retinal function in the mice infected with bacteria at different time points. After dark adaptation, mice were anesthetized with ketamine (70 mg/kg body weight) and xylazine (5 mg/kg body weight) under a dim red light. A gold-wire loop electrode was placed over the cornea once the pupils were dilated. Full-field ERGs were performed using a standard Ganzfeld dome stimulating Inflammatory mediator expression. The expression levels of inflammatory mediators in infected and contralateral eyeballs were detected by enzyme-linked immunosorbent assays (ELISAs). Each eyeball was harvested at the indicated time points and homogenized in sterile PBS containing protease inhibitor cocktail (Bimake, Houston, TX, USA). Concentrations of IL-6, TNF-␣, and IL-1␤ were estimated using mouse ELISA kits (Dakewe, Shenzhen, Guangdong, China). The lower limits of detection for the ELISAs were as follows: TNF-␣, 15.6 pg/ml; IL-1␤, 31.3 pg/ml; and IL-6, 15.6 pg/ml. Values represent means Ϯ SEM for n Ն 6 eyes at each time point from Ն3 independent experiments.
Genome sequencing and assembly. Genomic DNA of the eight intraocular isolates was prepared from purified bacterial cultures using the TruSeq Nano DNA HT Sample prep kit and sequenced on an Illumina HiSeq X platform, generating 150-bp paired-end reads. Sequence reads from each isolate were de novo assembled using SPAdes v.3.13 (45). To evaluate the quality of each assembly, we used QUAST v5.0.2, BBMap, and CheckM v1.0.18. Publicly available sequences for 125 Bacillus reported isolates were downloaded from GenBank. To obtain uniform and consistent annotations for core and pangenome analyses, all sequences were annotated using Prokka v.1.12 (46). Core-genome identification and phylogenetic analysis. Two core genomes were constructed. The first contained proteins common to all B. cereus genomes, while the second contained proteins common to intraocular isolate-related genomes. Pangenome estimates for the study genomes were performed using Roary v.3.12 (-e -n -i 90 -cd 99) (47), with a minimum percent identity for blastp of 90% and an alignment of all core genes (present in 99% of isolates). Single nucleotide polymorphisms (SNPs) in the core genes were extracted and used to construct a maximum likelihood tree using RAxML with 100 bootstraps and a midpoint root (-s -f a -x 12345 -p 12345 -N 100 -n core -T 4 -m GTRCAT). We used RAMI (20) to analyze the ML tree of 129 genomes and identified distinct clusters based on patristic distance (P ϭ 0.01, i.e., branch length between isolates).
Comparative analysis of virulence genes. Putative virulence genes were collated from the Virulence Factors of the Pathogenic Bacteria Database (21). Intraocular pathogen-associated studies of all annotated virulence genes were performed using Fisher's exact tests and Bonferroni's correction of P values. An alignment for each of the extracted candidate virulence determinant genes was constructed using Clustal Omega (48) and visualized using ESPript (49).
Statistical analysis. Statistical analyses were performed in R. PCA was performed using the prcomp function. Differences in the total number of unique genes between distinct lineages and acquired virulence genes between intraocular and gastrointestinal isolates were assessed using Wilcoxon rank sum tests. We performed pangenome-wide association studies (PGWAS) using Scoary v1.6.16 (-g -t -c I -p 0.05).
Ethics approval and consent to participate.