Abstract
Bacteria responsible for the greatest global mortality colonize the human microbiome far more frequently than they cause severe infections. Whether mutation and selection within the microbiome accompany infection is unknown. We investigated de novo mutation in 1163 Staphylococcus aureus genomes from 105 infected patients with nose-colonization. We report that 72% of infections emerged from the microbiome, with infecting and nose-colonizing bacteria showing parallel adaptive differences. We found 2.8-to-3.6-fold enrichments of protein-altering variants in genes responding to rsp, which regulates surface antigens and toxicity; agr, which regulates quorum-sensing, toxicity and abscess formation; and host-derived antimicrobial peptides. Adaptive mutations in pathogenesis-associated genes were 3.1-fold enriched in infecting but not nose-colonizing bacteria. None of these signatures were observed in healthy carriers nor at the species-level, suggesting disease-associated, short-term, within-host selection pressures. Our results show that infection, like a cancer of the microbiome, emerges through spontaneous adaptive evolution, raising new possibilities for diagnosis and treatment.
One Sentence Summary Life-threatening S. aureus infections emerge from nose microbiome bacteria in association with repeatable adaptive evolution.
Main Text
Communicable diseases remain a leading cause of global mortality, with bacterial pathogens among the greatest concern (1). However, many of the bacteria imposing the greatest burden of mortality, such as Staphylococcus aureus, are frequently found as commensal components of the body’s microbiome (2). For them invasive disease is a relatively uncommon event that is often unnecessary (3,4), and perhaps disadvantageous (5), for onward transmission. Genomics is shedding light on important bacterial traits such as host-specificity, toxicity and antimicrobial resistance (6–10). These approaches offer new opportunities to understand the role of genetics and within-host evolution in the outcome of human interactions with major bacterial pathogens (11).
Several lines of evidence support a plausible role for within-host evolution influencing the virulence of bacterial pathogens. Common bacterial infections, including S. aureus, are often associated with colonization of the microbiome by a genetically similar strain (12). Genome sequencing suggests that bacteria mutate much more quickly than previously accepted, and this confers a potent ability to adapt, for example evolving antimicrobial resistance de novo within individual patients (13,14). Opportunistic pathogens infecting cystic fibrosis patients have been found to rapidly adapt to the lung environment, with strong evidence of parallel evolution across patients (15–19). However, the selection pressures associated with antimicrobial resistance and opportunistic infections of cystic fibrosis patients may not typify within-host adaptation in common commensal pathogens that have co-evolved with humans for thousands or millions of years (20,21).
Candidate gene studies have demonstrated that certain regions, notably quorum-sensing systems such as the S. aureus accessory gene regulator (agr), mutate particularly quickly in vivo and in culture (22). The agr operon encodes a pheromone that coordinates a shift at higher cell densities from production of surface proteins promoting biofilm formation to production of secreted toxins and proteases promoting inflammation and dispersal (23). Mutants typically produce the pheromone but no longer respond to it (24). The evolution of agr has been variously ascribed to directional selection (25), balancing selection (26), social cheating (27) and life-history trade-off (28). However, the role of agr mutants in disease remains unclear, since they are frequently sampled from both asymptomatic carriage and severe infections (24).
Whole-genome sequencing case studies add weight to the idea that within-host evolution plays an important role in infection. In one persistent S. aureus infection, a single mutation was sufficient to permanently activate the stringent stress response, reducing growth, colony size and experimentally measured disease severity (29). In another patient, we found that bloodstream bacteria differed from those initially colonizing the nose by several mutations including loss-of-function of the rsp regulator (30). Functional follow-up revealed that the rsp mutant expressed reduced toxicity (31), but maintained the ability to cause disseminated infection (32). Unexpectedly, we found that bloodstream-infecting bacteria exhibit lower toxicity than nose-colonizing bacteria more generally (31). These results raise the question: are unique hallmarks of de novo mutation and selection associated with bacterial evolution in severely infected patients?
We addressed this question by investigating the genetic variants arising from within-patient evolution of S. aureus sampled from 105 patients with concurrent nose colonization and blood or deep tissue infection. We annotated variants to test for systematic differences between colonizing and infecting bacteria. We discovered several groups of genes showing significant enrichments of protein-altering variants indicating adaptive evolution. For genes implicated in pathogenesis, adaptive mutants were limited to infecting bacteria, while other pathways showed adaptation in the nose and infection site. Adaptive enrichments were not observed in asymptomatic carriers, nor between unrelated bacteria, indicating evolution in response to disease-associated, within-host selection pressures. Our results reveal that adaptive evolution of genes involved in regulation, toxicity, abscess formation, cell-cell communication and bacterial-host interaction drives parallel differentiation between commensal constituents of the nose microbiome and invasive infections, providing new insights into the evolution of disease in a major pathogen.
Results
Infecting bacteria are typically descended from the patient’s microbiome
We identified 105 patients suffering severe S. aureus infections admitted to hospitals in Oxford and Brighton, England, for whom we could recover contemporaneous nose swabs from admission screening. Of the 105 patients, 55 had bloodstream infections, 37 had soft tissue infections and 13 had bone and joint infections (Table 1). The infection was most often sampled on the same day as the nose, with an interquartile range of 1 day earlier to 2 days later (Table S1).
To discover de novo mutations within and between the nose microbiome and infection site, we whole-genome sequenced 1163 bacterial colonies, a median of 5 per site. We detected single nucleotide polymorphisms (SNPs) and short insertions/deletions (indels) using previously developed, combined reference-based mapping and de novo assembly approaches (30,33,34). We identified 35 distinct strains, defined by multilocus sequence type (ST), across patients (Table S1). As expected (12), colonizing and infecting bacteria were usually extremely closely related (95 patients), sharing the same ST and differing by 0-66 variants. Unrelated colonizing and infecting bacteria (10 patients) differed by 1104-50573 variants and typically possessed distinct STs (e.g. Fig. 1a, Fig. S1). After excluding variants differentiating unrelated STs, we catalogued 1322 de novo mutations within the 105 patients.
In patients with closely related strains, the within-patient population structure was always consistent with a unique migration event from the nose to the infection site, or occasionally, vice versa. Infecting and colonizing bacteria usually formed closely-related but distinct populations with no shared genotypes (74/95 patients, e.g. Fig. 1b), separated by a mean of 5.7 variants. There was never more than one identical genotype between nose-colonizing and infecting bacteria, (21/95 patients, e.g. Fig. 1c), indicating that the migration event from one population to the other involved a small number of founding bacteria (35,36). In such patients, the shared genotype likely represents the migrating genotype itself. Population structure did not differ significantly between infection types (p = 0.38, Table 1). Genetic diversity in the nose (mean pairwise distance, π = 2.8 variants) was similar to that previously observed in asymptomatic nasal carriers (33) (Reference Panel I, π = 4.1, p = 0.13), but was significantly lower in the infection site (π = 0.6, p = 10−10.0), revealing limited diversification post-infection.
In most patients the infection appeared to be descended from the nose. We used 1149 sequences from other patients and carriers (Reference Panel II) to reconstruct the most recent common ancestor (MRCA) for the 95/105 (90%) patients with related nose-colonizing and infecting bacteria. We thereby distinguished wild type from mutant alleles. In 49 such patients, we could determine the ancestral population. The nose microbiome was likely ancestral in 39/49 (80% of patients with related strains, or 72% of all patients) because all infecting bacteria shared de novo mutations in common that distinguished them from the MRCA, whereas nose-colonizing bacteria did not. In 16 of those, confidence was high because both mutant and ancestral alleles were observed in the nose, confirming it as the origin of the de novo mutation (e.g. Fig. 1d). Conversely, in 10/49 patients, bacteria colonizing the microbiome were likely descended from blood or deep tissue infections (20% of patients with related strains, or 18% of all patients) (e.g. Fig. 1f). Confidence was high for just three of those patients, and they showed unusually high diversity (Supplementary data, P063, P072, P093), suggesting that in persistent infections, infecting bacteria can recolonize the nose.
Protein-truncating mutants are over-represented within infected patients
To help identify variants that could promote, or be promoted by, infection of the blood and deep tissue by bacteria colonizing the nose, we reconstructed within-patient phylogenies and classified variants by their position in the phylogeny. Sequencing multiple colonies per site enabled us to classify variants into those representing genuine differences between nose-colonizing and infection populations (B-class), variants specific to the nose-colonizing microbiome population (C-class) and variants specific to the disease-causing infection population (D-class). We hypothesized that B-class variants would be most enriched for variants promoting, or promoted by, infection, if such variants occur (Fig. 1g).
We cross-classified variants by their predicted functional effect: synonymous, non-synonymous or truncating within protein-coding sequences, or non-coding (Table 2, Table S2). As expected, the prevailing tendency of selection within patients was to conserve protein sequences, with dN/dS ratios indicating rates of non-synonymous change 0.55, 0.68 and 0.63 times that expected under neutral evolution for B, C and D-class variants respectively.
In a longitudinal study of one long-term carrier, we previously reported that a burst of protein-truncating variants punctuated the transition from asymptomatic carriage to invasive infection (30). Here we found a 3.9-fold over-abundance of protein-truncating variants of all phylogenetic classes in infected patients compared to asymptomatic carriers (Reference Panel I, p = 0.002, Table 2), supporting the conclusion that loss-of-function mutations are disproportionately associated with evolution within infected patients. This may reflect a reduction in the efficiency with which selection removes deleterious protein-truncating mutations, and provides evidence of a systematic difference in selection within severely infected patients.
Quorum sensing and cell-adhesion proteins exhibit adaptive evolution between colonizing and infecting bacteria
We hypothesized that variants associated with invasive infection would be enriched among the protein-altering B-class variants between the nose and infection site (Fig. 1g). Therefore we aggregated mutations by genes in a well-annotated reference genome, MRSA252, and tested each gene for an excess of non-synonymous and protein-truncating B-class variants, taking into account the length of the gene. Aggregating by gene was necessary because 1318/1322 variants were unique to single patients. The two exceptions involved non-coding variants arising in two patients each, one B-class variant 130 bases upstream of azlC, an azaleucine resistance protein (SAR0010), and one D-class variant 88 bases upstream of eapH1, a secreted serine protease inhibitor (SAR2295) (38).
We found a significant excess of five protein-altering B-class variants representing a 58.3-fold enrichment in agrA, which encodes the response regulator that mediates activation of the quorum sensing system at high cell densities (p=10−7.5, Fig. 2a, Table 3). The clfB gene encoding clumping factor B, which binds human fibrinogen and loricrin (39), showed an excess of five protein-altering B-class variants, representing a 15.9-fold enrichment that was near genome-wide significance after multiple testing correction (p=10−4.7).
Previously we identified a truncating mutation in the transcriptional regulator rsp to be the most likely candidate for involvement in the progression to invasive disease in one long-term nasal carrier (30). Although we observed just one variant in rsp among the 105 patients (3.9-fold enrichment, p=0.27), we found it was a non-synonymous B-class variant resulting in an alanine to proline substitution in the regulator’s helix-turn-helix DNA binding domain. In separately published experiments (32), we demonstrated that this and the original mutation induce similar loss-of-function phenotypes which, like agr loss-of-function mutants, express reduced toxicity, but maintained an ability to persist, disseminate and cause abscesses in vivo.
We found no significant enrichments of protein-altering variants among D-class variants, but we observed a significant excess of six protein-altering C-class variants in pbp2 which encodes a penicillin binding protein involved in cell wall synthesis (19.0-fold enrichment, p=10−6.0, Fig. S2a). Pbp2 is an important target of β-lactam antibiotics (40), revealing adaption – potentially in response to antibiotic treatment – in the nose populations of some patients.
Genes modulated by virulence regulators and antimicrobial peptides show adaptive evolution between colonizing and infecting bacteria
To improve the sensitivity to identify adaptive evolution associated with invasive infection, we developed a gene set enrichment analysis (GSEA) approach in which we tested for enrichments of protein-altering B-class variants among groups of genes. GSEA allowed us to detect signatures of adaptive evolution in groups of related genes that were not apparent when interrogating individual genes.
We grouped genes in two different ways: by gene ontology and by expression pathway. First, we obtained a gene ontology for the reference genome from BioCyc (41), which classifies genes into biological processes, cellular components and molecular functions. There were 552 unique gene ontology groupings of two or more genes. We tested for an enrichment among genes belonging to the ontology, compared to the rest of the genes.
Second, we obtained 248 unique expression pathways from the SAMMD database of transcriptional studies (42). For each expression pathway genes were classified as up-regulated, down-regulated or not differentially regulated in response to experimentally manipulated growth conditions or expression of a regulatory gene. For each expression pathway, we tested for an enrichment in genes that were up- or down-regulated compared to genes not differentially regulated.
The most significant enrichment for protein-altering B-class variants between nose and infection sites occurred in the group of genes down-regulated by the cationic antimicrobial peptide (CAMP) ovispirin-1 (p=10−7.8), with a similar enrichment in genes down-regulated by temporin L exposure (p=10−6.9, Fig. 2c). Like human CAMPs, the animal-derived ovispirin and temporin compounds inhibit epithelial infections by killing phagocytosed bacteria and mediating inflammatory responses (43). In response to inhibitory levels of ovispirin and temporin, agr, surface-expressed adhesins and secreted toxins are all down-regulated. Collectively, down-regulated genes showed 2.7-fold and 2.8-fold enrichments of adaptive evolution, respectively. Conversely, genes up-regulated in response to CAMPs, including the vraSR and vraDE cell-wall operons and stress response genes (43), exhibited 0.4-fold and 0.7-fold enrichments (i.e. depletions), respectively (Table 3). Thus, genes undergoing adaptive evolution are strongly inhibited by the CAMP-mediated immune response.
Genes belonging to the cell wall ontology showed the second most significant enrichment for adaptive evolution (p=10−7.0). Genes contributing to this 5.0-fold enrichment included the immunoglobulin-binding S. aureus Protein A (spa), the serine rich adhesin for platelets (sasA), clumping factors A and B (clfA, clfB), fibronectin binding protein A (fnbA) and bone sialic acid binding protein (bbp). The latter four genes contributed to another statistically significant 6.4- fold enrichment of adaptive protein evolution in the cell adhesion ontology (p=10−6.5, Fig. 3). Therefore, there is a general enrichment of surface-expressed host-binding antigens undergoing adaptive evolution.
The rsp regulon showed the most significant enrichment among gene sets defined by response to individual bacterial regulators (p=10−6.4). Genes down-regulated by rsp in exponential phase (44), including surface antigens and the urease operon, exhibited a 3.6-fold enrichment for adaptive evolution, while up-regulated genes showed 0.6-fold enrichment. So whereas rsp loss-of-function mutants were rare per se, genes up-regulated in such mutants were hotspots of within-patient adaptation in infected patients. Since expression is a prerequisite for adaptive protein evolution, this implies there are alternative routes by which genes down-regulated by intact rsp can be expressed and thereby play an important role within patients other than direct inactivation of rsp.
Loss-of-function in agr mutants represent one alternative route, since they exhibit similar phenotypes to rsp mutants, with reduced toxicity and increased surface antigen expression, albeit reduced ability to form abscesses (32). We found significant enrichments of genes regulated by agrA in two different backgrounds (p<10−4.5) and by sarA (p=10−4.6), underlining the influence of adaptive evolution on both secreted and surface-expressed proteins during infection. We found that expression of genes enriched for protein-altering substitutions was also altered in strains possessing reduced susceptibility to vancomycin, although not in a consistent direction across strains (p<10−4.7), and to pine-oil disinfectant (p=10−4.4), suggesting such genes may be generally involved in response to harsh environments.
Several genes contributed to multiple evolutionary signals, particularly cell-wall anchored proteins involved in adhesion, invasion and immune evasion (39), including fnbA, clfA, clfB, sasA and spa. These multifactorial, partially overlapping signals suggest a large target for selection in adapting to the within-patient environment (Fig. 3). The fact that we observed no comparable significant enrichments in C-class and D-class protein-altering variants (Fig. S2) indicates that these evolutionary patterns are associated specifically with the infection process.
Adaptive evolution in pathogenesis genes is found only in infecting bacteria
Having identified adaptive evolution differentiating nose-colonizing and disease-causing bacteria, we next asked whether the mutant alleles were preferentially found in the nose or infection site. We used 1149 sequences from other patients or carriers (Reference Panel II) to reconstruct the genotype of the MRCA of colonizing and infecting bacteria respectively in each patient. This allowed us to sub-classify B-class variants by whether the mutant allele was found in the nose-colonizing bacteria (BC-class) or the disease-causing bacteria (BD-class).
A priori, we had expected the enrichments of adaptive evolution to be driven primarily by mutants occurring in the disease-causing bacteria (BD-class). One group of genes showed a signal of such an enrichment among BD-class variants specifically. Genes belonging to the BioCyc pathogenesis ontology were marginally genome-wide significant in BD-class variants, showing a 3.1-fold enrichment (p=10−4.6) and a statistically insignificant 1.7-fold enrichment in BC-class variants (p=0.13). BD-class mutants driving this differential signal arose in toxins including gamma hemolysin and several regulatory loci implicated in toxicity and virulence regulation: rot, sarS and saeR.
Surprisingly however, we found that all other significantly enriched gene sets were driven by mutant alleles occurring both in colonizing and infecting bacteria (Fig. S3). This indicates there are common selection pressures in the nose and infection site during the process of infection within patients, leading to convergent evolution across body sites. So while adaptation in pathogenesis genes appears specifically invasion-associated, other signals of adaptation in severely infected patients are driven by selection pressures, which might compensate for an altered within-host environment during infection, that are as likely to favor mutants in nose-colonizing bacteria as infecting bacteria.
Signals of adaptation are specific to infected patients and differ from prevailing signatures of selection
Two lines of evidence show that the newly discovered signatures of within-host adaptive evolution, both in infecting and nose-colonizing bacteria, are unique to evolution in infected patients. To test this theory against the alternative explanation that our approach merely detects the most rapidly evolving proteins, we searched for similar signals in alternative settings: evolution within asymptomatic carriers, and species-level evolution between unrelated bacteria.
There was no significant enrichment of protein-altering variants in any gene, ontology or pathway among 235 variants identified from 10 longitudinally sampled asymptomatic nasal carriers (Reference Panel III, Fig. S4, Table S3). To address the modest sample size, we performed goodness-of-fit tests, focusing on the signals most significantly enriched in patients. We found significant depletions of protein-altering variants in carriers relative to patients in the rsp, agr and sarA regulons (p=10−4.0) and the pathogenesis ontology (p=10−3.2, Table S4).
Nor were the relative rates of non-synonymous to synonymous substitution (dN/dS) higher between unrelated S. aureus (Reference Panel IV) in the genes that contributed most to the signals associated with adaptation within patients: agrA, agrC clfA, clfB, fnbA and sasA. Although synonymous diversity was somewhat higher than typical in these genes, the dN/dS ratios showed no evidence for excess protein-altering change in these compared to other genes (Fig. S5). Accordingly, incorporating this locus-specific variability of dN/dS into the GSEA did not affect the results (Fig. S6). Taken together these lines of evidence show that the ontologies, pathways and genes significantly differentiated between colonizing and infecting bacteria arise in response to selection pressures specifically associated with infected patients, and are not repeated in asymptomatic carriers or species-level evolution.
Discussion
We have discovered that common, life-threatening infections of S. aureus are frequently descended from bacteria colonizing the human microbiome. These infections are associated with repeatable patterns of bacterial evolution driven by within-patient mutation and selection. Genes involved in pathogenesis, notably toxins and regulators, showed evidence for adaptation in infecting but not nose-colonizing bacteria. Surprisingly, other signatures of adaptation occurred in parallel in nose-colonizing and infecting bacteria, affecting genes responding to cationic antimicrobial peptides and the virulence regulators rsp and agr. Such genes mediate toxicity, abscess formation, immune evasion and bacterial-host binding. Adaptation within both regulator and effector genes reveals that multiple, alternative evolutionary paths are targeted by selection in infected patients.
The signatures of within-patient adaptation that we found differed from prevailing signals of selection at the species level. This discordance means that infection-associated adaptive mutations within patients are rarely transmitted, and argues against a straightforward host-pathogen arms race as the predominant evolutionary force acting within and between patients. Instead, it supports the notion of a life-history trade-off between adaptations favoring colonization and infection distinct from those favoring dissemination and onward transmission. As such, invasive disease may be analogous to cancer in multicellular organisms, representing an ever-present risk of mutations in the microbiome favored by short-term selection but ultimately incidental or damaging to the bacterial reproductive life cycle.
Nor did we see these signatures of bacterial adaptation and excess loss-of-function mutations in healthy nose carriers, indicating that risk factors for invasive infections, such as a weakened or over-activated immunological response, comorbidities or medical interventions, may create distinctive selection pressures in infected patients. As in cancer, the effects of such risk factors may be mediated, at least in part, through the selection pressure they exert on the microbiome.
The existence of signatures of adaptive substitutions associated with invasive disease raises the possibility of developing new diagnostic techniques and personalizing treatment to the individual patient’s microbiome. The ability of genomics to characterize the selective forces driving adaption within the human body in unprecedented detail provides new opportunities to improve experimental models of disease. Ultimately, it may be possible to develop therapies that utilize our new understanding of within-patient evolution to target the root causes of invasive disease from the bacterial perspective.
Author contributions
BCY, study design, sample collection, DNA extraction, bioinformatics, analysis, writing. C-HW, bioinformatics, analysis, writing. NCG, JRP, sample collection, DNA extraction. KC, EL, SP, DNA extraction. AS, JC, TG, ZI, bioinformatics. RB, RCM, study design, interpretation. JP, DWC, TEAP, ASW, MJL, study design, sample collection, interpretation. DHW, study design, analysis. DJW, study design, analysis, writing.
Table S1. List of all cultures included in the site, the site of infection (and any known source if bloodstream), number of isolates sequenced from each site, ST or CC by in silico MLST, number of variants found at each site and the mean pair-wise difference comparing isolates.
Table S2. List of all variants found within patients with S. aureus disease, location on shared reference (MRSA252), or position and reference genome name and accession number if variant could not be localized on MRSA252. Each variant is described by the alleles found, its location in gene, the predicted effect on gene product and the location of the variant on the phylogenetic tree.
Table S3. List of all variants found within long term asymptomatic carriers, location on shared reference (MRSA252), or position and reference genome name and accession number if variant was not localized on MRSA252. Each variant is described by the alleles found, its location in gene and the predicted effect on gene product.
Acknowledgments
We would like to thank Ed Feil, Stephen Leslie, Gil McVean and Richard Moxon for helpful insights and useful discussions. Sequencing reads uploaded to short read archive (SRA) under BioProject PRJNA369475. RNA-Seq data relating to isolate from P005 (aka ‘patient S’) previously submitted under BioProject PRJNA279958.
The views expressed in this publication are those of the authors and not necessarily those of the funders. This study was supported by the Oxford NIHR Biomedical Research Centre, a Mérieux Research Grant, the National Institute for Health Research Health Protection Research Unit (NIHR HPRU) in Healthcare Associated Infections and Antimicrobial Resistance at Oxford University in partnership with Public Health England (PHE) (grant HPRU-2012-10041), and the Health Innovation Challenge Fund (a parallel funding partnership between the Wellcome Trust (grant WT098615/Z/12/Z) and the Department of Health (grant HICF-T5-358)). T.E.P. and D.W.C. are NIHR Senior Investigators. D.J.W. and Z.I. are Sir Henry Dale Fellows, jointly funded by the Wellcome Trust and the Royal Society (Grants 101237/Z/13/Z and 102541/Z/13/Z). B.C.Y is a Research Training Fellow funded by the Wellcome Trust (Grant 101611/Z/13/Z). We acknowledge the support of Wellcome Trust Centre for Human Genetics core funding (Grant 090532/Z/09/Z).
References and Notes
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.
- 8.
- 9.
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.
- 17.
- 18.
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.
- 46.
- 47.
- 48.
- 49.
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.
- 55.
- 56.
- 57.
- 58.
- 59.
- 60.
- 61.
- 62.
- 63.
- 64.
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵