Laboratory S. aureus strains, cell lines, and growth conditions

S. aureus strain ATCC29213 was obtained from the American Type Culture Collection. Strain JE2 and JE2 transposon mutants were from the Biodefense and Emerging Infections Research Resources Repository (BEI Resources). All bacteria in this study were maintained in LB medium supplemented with hemin, menadione, and thymidine (sup-LB) in order to support potentially relevant auxotrophic mutants [32].

THP-1 cells were obtained from ATCC and cultured according to the supplier’s recommendations at 37°C in a humidified 5% (v/v) CO2 air atmosphere in RPMI 1640 medium (Gibco) supplemented with 0.1 mg/ml l-glutamine, 0.1 mg/ml streptomycin, 100 U/ml penicillin (PAA) and 20% (v/v) Nu-Serum Serum Replacement (Corning).

S. aureus clinical isolates

Control clinical isolates of methicillin-sensitive S. aureus (MSSA) and methicillin-resistant S. aureus (MRSA), derived from individuals without CF and from sites not involved in active disease, were obtained from fully de-identified surveillance swabs [33] at the University of Washington Clinical Microbiology Laboratory following the completion of clinical testing. MSSA isolates were obtained by streaking swabs onto mannitol-salt agar, and MRSA isolates from streaking onto chromogenic MRSA selective plates (bioMerieux, Marcy-l’Étoile, France). Control clinical isolates (n = 180) were subjected to whole-genome sequencing as described below.

Clinical isolates from children with CF were previously described and characterized by whole-genome sequencing [4], constituting 1,382 longitudinally-collected S. aureus isolates collected from 246 participants. For clonally related isolates derived from the same patient (n = 237 lineages), we selected the first S. aureus isolate that was initially cultured from the patient and the last isolate temporally collected during the study period [4]. Secondary data, but no human subjects, were analyzed in this study. From this collection of paired clinical isolates, we further reduced the set analyzed in this study to a cohort of 57 lineages identified from the phylogeny of that collection [4] which spanned the genomic diversity present within the collection.

For each such lineage, we identified a matching, phylogenomically similar control strain to minimize lineage-specific effects [34] in comparisons between clinical and control strains. To accomplish this, whole-genome sequencing data from CF and control isolates were subjected to de novo assembly using ABySS 2.0 [35]. We generated a core-genome alignment containing the entire control population and the first isolate from each CF lineage using ROARY [36]. SNP-sites [37] was then used to calculate pairwise core-genome distances between CF-derived and control isolates. For each CF-derived isolate, a phylogenomically similar match from the control population was then identified using an optimal matching strategy to minimize the global distance between all matched cases and controls in the set [38] with a maximum core-genome SNP caliper of 6% and a maximum control to clinical isolate matching ratio of two. This process resulted in matching of the 57 CF-derived isolate pairs to 33 distinct control isolates. All isolates (57 CF first-collected isolates, 57 CF last-collected isolates, and 33 matched control isolates) were subjected to characterization by intracellular pathogenesis assays as described above (S1 Table).

Intracellular pathogenesis assays and selection procedures

48 hours prior to bacterial infection, THP-1 cells were differentiated into monocytes with PMA (Sigma) solubilized in DMSO at a final concentration of 10 ng/ml [39–41]. PMA-containing medium was removed from cells 24 hours after treatment. Cells were washed with incubation medium (IM, RPMI 1640 containing no amendments) and incubated for 24 additional hours.

Bacteria were expanded overnight in sup-LB broth, washed twice in Dulbecco’s phosphate-buffered saline (DPBS) (Difco) to remove soluble toxins, and cell density assessed by optical density at 600 nm (OD600). Six replicates were independently passaged in parallel for each strain.

Intracellular pathogenesis assays were based on previously described lysostaphin protection assays [42]. Briefly, bacteria were combined with differentiated THP-1 cells at a multiplicity of infection (MOI) of 10 in IM for 1 h, with aliquots of inoculum plated to empirically determine the number of bacteria applied. The medium was replaced with RPMI 1640 containing 50 μg/ml lysostaphin (Sigma), and selections were performed at 37°C with 5% atmospheric CO2. 3 h or 48 h post-inoculation, host cells were washed with DPBS and lysed with water. Serial dilutions of lysate were plated onto sup-LB agar to evaluate the number of viable bacteria. At least 4 independent replicates were used to assess the number of viable cells at both 3 and 48 hours. Viable bacteria counted at 3 h were normalized to empiric counts of the bacteria initially applied to determine the proportion of inoculated cells remaining viable after early macrophage entry (measurement of “macrophage invasion”). Viable organisms counted at 48 h were normalized to the number of viable organisms counted at 3 hours for the same isolate (“macrophage survival index”). Measurements of invasion phenotypes were normalized to in-run controls of the parental JE2 or ATCC29213 strains, as appropriate, to control for possible batch effects, whereas the macrophage survival index was inherently normalized for invasion efficiency. Invasion testing of clinical isolates and lineage-matched controls were normalized to an in-run JE2 control. Parental JE2 and ATCC29213 strain invasion was measured as absolute (non-normalized) values collected in a single experiment, due to lack of an external comparator.

For passaging experiments, procedures were completed essentially as described for intracellular pathogenesis assays, except that an initial MOI of 100 was used, lysate was collected only at 48 h, and bacteria were transferred to sup-LB broth for overnight expansion prior to the next passage. As the bacteria became more efficient at infection with increased passaging, it was necessary to empirically reduce the MOI over time to prevent host cells from undergoing lysis during infection while maintaining selective pressure for mutants with increased cytotoxic capability. Consequently, for the first 8 passages, an MOI of 100 was used, after which the MOI was reduced to 10.

Whole-genome sequencing and variant analysis

Whole-genome sequencing was performed as previously described [43] using MiSeq and NextSeq500 platforms with 300 cycle chemistries (Illumina, San Diego, CA). Briefly, variant calling was conducted as before [43] utilizing reference genomes identified as those most closely matching the de novo assemblies of sequenced isolates. Mutations were manually verified using the Integrative Genomics Viewer [44] prior to annotation of sequence features. Common gene names were established using BLAST [45] searches of target gene sequences against S. aureus reference genomes and by cross-correlating gene identifiers against pan-genome identifiers in the AureoWiki database [46]. MLST was determined using mlst-2.0.3 (https://github.com/tseemann/mlst) in conjunction with the PubMLST database [47].

Bacterial genome-wide association studies

Three complementary classes of bacterial GWAS were conducted using clinical isolates and matched control strains to identify associations between bacterial genotypes and macrophage pathogenesis phenotype measurements.

To assess the association of specific genotypes with phenotypes in this phylogenetically diverse collection, we performed reference-free unitig analysis with pyseer (v1.3.10) [48]. We adjusted for population structure using a mixed effects model with an approximate maximum likelihood phylogeny constructed from the ROARY core-genome alignment using Fastree2.1.11 [49]. Phenotype measurements for macrophage invasion and macrophage survival index were provided as continuous variables. Both CF isolates and matched controls were included in this analysis.

Different mutations occurring in a gene may confer equivalent phenotypic effects encoded by distinct genotypes. Unitig-based analysis has less power to detect this class of genotype-phenotype association, which is better assessed using kernel-based models of association between the collective “burden” of mutations within a gene and a phenotype. We therefore performed burden testing using the pyseer rare variants method, considering all non-synonymous mutations as equally weighted and using a mixed effects model incorporating the same phylogenetic tree as above. Both CF isolates and matched controls were included in this analysis.

To identify de novo mutations occurring within clonally related S. aureus lineages from an individual host, whole-genome sequences of the first- and last-collected clinical isolate from each patient were compared. For each gene in the pangenome, a multivariable linear regression analysis was performed in R version 4.2.2 to assess the independent association between the presence of a non-synonymous de novo mutation in that gene (primary predictor variable) with the absolute value of change in measured phenotype between first and last isolates (outcome variable), with adjustment for changes in two other phenotypes (covariates): biofilm formation (measured as previously [50], S1 Table) and the alternate macrophage invasion phenotype measured in this study (i.e., intracellular survival for analysis for invasion activity, and vice versa). Only CF isolates were included in this analysis.

Candidate genes implicated in invasion and survival phenotypes in vivo by GWAS were selected for functional testing by prioritizing those with high levels of association in a single GWAS model and those that were independently identified in more than one model, including associations with lower individual levels of significance. For unitig analyses, we selected genes having three or more component unitigs achieving Bonferroni-adjusted levels of significance with alpha values of 0.05. For burden testing analyses of both phenotypes, all genes achieving Bonferroni-adjusted significance using an alpha value of 0.05 were selected. For analysis of de novo mutation within clinical isolate lineages, we selected all genes with p-values<0.05 and association scores in the top 5%. Lastly, candidate genes showing overlap between GWAS analyses for a phenotype (those with a one or two unitigs meeting Bonferroni-adjusted significance and further identified as undergoing de novo mutation between a first-last pair) were selected for validation.

S. aureus CRISPRi knockdown

We generated a CRISPRi [31] vector for use in S. aureus by modifying our earlier system for CAS9-mediated counterselection [51]. Briefly, dCAS9 from pdCAS9 [52] was cloned under expression of the constitutive p23 promoter into plasmid pCN50wt [53] as described previously [51], followed by addition of a synthetic sgRNA under control of the PRAB17 promoter as previously [51]. To facilitate screening for successful guide RNA cloning, a GFP expression cassette was ultimately cloned into the BSAI site of the sgRNA to generate plasmid pCRISPRi (S2 Table). Guide RNAs were designed against coding sequences using CRISPOR [54] with screening for off-target effects performed against the S. aureus USA300_FPR3757 genome, and selecting guide RNAs designed in the “reverse” orientation of the chosen coding sequence. Guide RNA templates were synthesized as ssDNA oligonucleotides (IDT, S2 Table) with appropriate tail sequences, then phosphorylated, annealed, and cloned into the BSAI site of pCRISRPi as described elsewhere [51]. After transformation into E. coli, plasmid was purified from a transformant lacking fluorescence and silencing vectors were subsequently transformed into S. aureus JE2 as described elsewhere [51]. Gene knockdown was assessed using gene-targeted real-time PCR (S2 Table) of cDNA prepared from mid-log phase growth cells, normalizing gene-specific expression to that of gyrA. Gene expression levels in individual CRISPRi mutants were compared to relative gene expression obtained for a CRISRPi knockdown targeted to a biologically irrelevant sequence (GFP) [55] using the ΔΔCt method (S3 Table).

Quantitative macrophage pathogenesis phenotypes of CRISPRi mutants were assessed as above, but including 10 μg/mL chloramphenicol in all culture media to select for retention of plasmids, and comparing metrics to those of a GFP-silenced CRISPRi strain [56]. The GFP-silenced CRISPRi strain was found to be equivalent to those of the wild-type strain carrying a GFP-expressing plasmid built on the same vector backbone.

Complementation studies

Genes for complementation studies were PCR amplified from JE2 genomic DNA (S2 Table) and cloned into vector pCN50wt. Where possible, the native promoters of cloned genes were used to drive gene expression by including the intergenic DNA falling between the gene of interest and the upstream gene. Three genes (saeS, B7H15_07805, and B7H15_05310) were assumed to be expressed within an operon on the basis of a gene in the same orientation terminating <30 bp upstream, and were therefore expressed under control of the constitutive Pcap promoter using PCR to add that element [57]. Transposon mutants were transformed with matched complementation vectors and with pCN50wt expressing superfolder GFP (ordered as a synthetic gene cassette, S2 Table) [58], which served as a comparator for phenotypic studies. Measurements of macrophage invasion and macrophage survival index were directly compared between complemented and GFP vector-transformed transposon mutant strains assayed in parallel, at least in quadruplicate.

Macrophage viability assessments

The number of viable and dead macrophages was assessed 48 hours after S. aureus infection using tryptan blue exclusion assays as described elsewhere [59], with measurements taken at least in triplicate.

Growth rate analyses

Growth rates were monitored by absorbance at OD₆₀₀ during incubation at 37°C in sup-LB broth, using a BioTek Synergy H1 Plate reader (Agilent), with measurements taken in 10 minute intervals and experiments performed at least in triplicate.

Intracellular pathogenesis of macrophages is elevated in S. aureus isolates from people with CF

We ascertained whether S. aureus from chronic airway infections in people with CF were subject to selection for enhanced intracellular macrophage pathogenesis in vivo. We utilized a previously described collection of clonally related S. aureus isolate pairs collected longitudinally over time from pediatric participants with CF [4], selecting a subset of 57 distinct clinical first- and last-collected isolate pairs that represented the genomic diversity present in the collection (S1 Table). To provide phylogenetically matched control strains for comparison, we compared each CF clinical isolate against a population of S. aureus control strains that were not derived from active disease and which originated from individuals without CF in order to identify those with the most closely matched genomes (n = 33).

CF isolates and lineage matched controls were assayed for intracellular pathogenesis phenotypes in differentiated THP-1 cells, which have previously been used as a model for S. aureus macrophage interactions [28,56,60]. Macrophage intracellular pathogenesis can be regarded either with respect to the proportion of bacteria that are able to enter and evade initial killing within host cells after phagocytosis (which in this study we term “invasion”), or to the rate of ongoing bacterial survival within host cells over longer periods of time (which here we term “survival”). We therefore assessed these properties separately: invasion was measured as the fraction of viable intracellular bacteria recovered from the initial inoculum 3 hours after sterilization of the extracellular compartment, whereas survival was assessed 48 hours after initial THP-1 host cell infection as the fraction of viable intracellular bacteria recovered at that time relative to the number ascertained at 3 hours.

We found that early S. aureus isolates obtained from individuals with CF by serial surveillance (i.e., those from the time of initial detection) exhibited capacities for macrophage invasion similar to those measured in phylogeny-matched control isolates from the general population. However, the final isolates collected from those same patient lineages after persistent respiratory infection (median 22.9 months, interquartile range 5.1 months), had significantly higher metrics of invasion than the early isolates (Fig 1A). 32 of the isolate pairs showed significantly (p<0.05, two-tailed T test) increased invasion phenotypes over time, while 14 demonstrated decreases, and 11 showed no significant change (S1 Table and S2A Fig). The aggregate increase in intracellular invasion phenotypes result was significant both when comparing groups of later isolates against the first clonally-related isolate collected from the same patient (p = 0.041, paired two-tailed T test) and against phylogeny matched controls from unaffected individuals (p = 0.005, paired two-tailed T test). In contrast, 20 of the isolate pairs showed significantly (p<0.05, two-tailed T test) increased levels of macrophage intracellular survival over time, while 20 evidenced decreases and 11 had no significant change in that phenotype. (S1 Table and S2B Fig). Differences between groups of first, last, and control isolates did not differ significantly among comparator populations in macrophage survival when considered in aggregate (Fig 1B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Macrophage pathogenesis phenotypes of S. aureus lineages from individuals with CF over time and phylogenetically matched control strains.

Phenotypes of S. aureus strains isolated from respiratory cultures of individuals with CF at the time of initial isolation during serial surveillance are compared to the last-collected strains from same patient, and to phylogeny-matched strains from unaffected controls. Measures of macrophage invasion, relative to in-batch testing of the JE2 control strain (A), and macrophage intracellular survival (B) are separately reported. Statistical significance was assessed using paired two-tailed T tests.

https://doi.org/10.1371/journal.ppat.1012394.g001

These findings indicate that, at the time of initial detection, S. aureus strains infecting individuals with CF begin with capacities for macrophage invasion similar to strains from general population that are not involved in chronic disease, but that selective pressure for that phenotype occurs in the CF airway in vivo. Persistent survival within macrophages does not appear to be as strongly selected in CF respiratory infections, although it may be important in a subset of individual strains or clinical cases.

Genomic analysis identifies candidate genes associated with intracellular macrophage pathogenesis in vivo

In order to identify genetic variation associated with enhanced macrophage pathogenesis phenotypes in vivo, we performed bacterial GWAS [48]. For both invasion and survival phenotypes, three complementary models of genetic association were applied (S1 Fig).

First we performed unitig-based testing, a reference-free method which identifies unique, variable-length sequences that are associated with a phenotype and incorporates statistical adjustment for bacterial population structure [61]. This analysis (S1A Fig) identified 179 sequences in 110 genes significantly associated with macrophage invasion (S4 Table). However, the unitigs most highly associated with macrophage invasion comprised an apparent haplotype, indicating insufficient sample diversity to resolve a specific candidate risk allele by this method and requiring prioritization of genes comprising multiple significantly associated unitigs for functional validation. No unitig sequences were significantly associated with survival.

Second, we conducted burden testing [62], which considers the collective impact of independent, non-synonymous variants occurring within a single annotated gene. This analysis (S1B Fig, S5 Table and S6 Table) identified five genes significantly associated with macrophage invasion (purR, folB, sirA, B7H15_03165, B7H15_06690) and three with macrophage survival (recG, B7H15_09405, B7H15_06245).

Last, we examined non-synonymous de novo mutations occurring between temporally first- and last-collected CF clinical isolates. This method of comparison provides the most robust control for bias related to residual population structure and reference annotation. This strategy (S1C Fig) identified 16 (S7 Table) and four genes (clfA, fnbA, sdrE, and a hypothetical protein, S8 Table) whose spontaneous mutation in vivo was associated with changes in macrophage invasion and survival, respectively. To assess the functional impact of such mutations, we examined the proportion of stop-gain mutations and indels (which are more likely to confer loss-of-function effects) to missense changes (which could confer carry either loss-of-function or gain-of-function effects) to assess for statistical imbalance between these classes for each gene. We found that three of the 16 invasion-associated genes (19%) and three of four of the survival-associated genes (75%) exhibited a statistically significant (p<0.05, Fisher Exact Test) excess of loss-of-function associated mutation types, although the missense mutations occurring in other genes may also exert similar effects. No genes demonstrated a proportional excess of missense mutation or patterns of mutation at common residues that would be expected from gain-of-function effects.

Together, these analyses implicate variation in 127 unique candidate genes as being correlated with macrophage pathogenesis in vivo, and suggest that mutations occurring in vivo frequently confer loss-of-function effects.

Serial passaging selects mutants with elevated intracellular pathogenesis in macrophages in vitro

To select for mutants having greater capacity to invade and survive within host macrophages under well-controlled laboratory conditions, and in well-defined bacterial genetic backgrounds, we separately conducted serial passaging experiments of S. aureus strains in the presence of THP-1 cell line-derived macrophages. We characterized the initial capacity for macrophage pathogenesis of two well-defined and phylogenomically distinct S. aureus reference strains, JE2 (Sequence Type 8, Clonal Complex 8) and ATCC29213 (Sequence Type 5, Clonal Complex 5). Viable bacteria from both strains were recovered three hours after macrophage invasion; however, a significantly greater proportion of JE2 were internalized at three hours relative to ATCC29213 (Fig 2A, p<0.0001, unpaired two-tailed T test). After normalizing for these differences in initial invasion, the natural capacity of these strains for ongoing intracellular survival at 48 hours also differed, with a significantly greater fraction of initially internalized JE2 surviving relative to ATCC29213 (Fig 2B, p<0.0001, unpaired two-tailed T test). To assess the fate of infected macrophages, we quantified their absolute number and the proportion of dead cells 48 hours after S. aureus was applied. Compared to an uninfected macrophage control, neither significant reduction in the number of macrophages (p>0.12, two-tailed T test) nor significant increases in the proportion of dead macrophages (p>0.45, two-tailed T test) were observed after infection with either S. aureus strain, indicating that S. aureus did not kill infected macrophages by lytic or non-lytic mechanisms.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Macrophage pathogenesis phenotypes of S. aureus laboratory strains ATCC29213 and JE2.

Phenotypes of strains were assessed for macrophage invasion (A) and macrophage intracellular survival (B). Statistical significance of measurements was assessed using unpaired two-tailed T tests. Macrophage invasion in this analysis was performed in a single experiment and determined as an absolute value, due to a lack of an external comparator strain for normalization.

https://doi.org/10.1371/journal.ppat.1012394.g002

From these differing natural baselines, six independent replicates from both strains were then passaged to select for intracellular macrophage pathogenesis phenotypes using lysostaphin protection assays [42,56], which selectively kill bacteria that have not been internalized by eukaryotic cells. After 12 passages, isolates were obtained from each expanded population and subjected to further characterization (S9 Table).

While strain JE2 showed greater baseline capacity for macrophage invasion than ATCC29213, that phenotype increased with passaging of both strains, derivatives of which ultimately achieved similar levels of activity (Fig 3A). In contrast, we observed significant strain-level differences in intracellular survival after selection (Fig 3B): ATCC29213, but not JE2, showed markedly increased survival in macrophages after passaging. For passaged ATCC29213 isolates, the fraction of bacteria recovered from macrophages after 48 hours exceeded that at the time of invasion, indicating some degree of bacterial replication following macrophage infection. The survival capacity of two JE2 derivatives decreased slightly with passaging, possibly indicating greater selective pressure for invasion phenotypes and mutual exclusivity of invasion and survival traits in that strain. Growth rate analysis of passaged mutants was conducted (S10 Table). Half of the replicates (two ATCC29213 derivatives and four JE2 mutants) had growth rates not significantly different from their parental strains. Four replicates (three ATCC29213 and one JE2) had doubling times significantly greater than their parents, whereas only one replicate from each lineage replicated significantly faster.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Macrophage pathogenesis phenotypes of passaged S. aureus laboratory strains.

Derivatives of JE2 and ATCC29213 after passaging with selective pressure for macrophage pathogenesis assessed for macrophage invasion, relative to in-batch testing of the matched parental strain (A), and macrophage intracellular survival (B). Statistical significance was assessed using unpaired two-tailed T tests.

https://doi.org/10.1371/journal.ppat.1012394.g003

Whole-genome sequencing detected a total of 15 genes that were mutated in passaged isolates (S11 Table). Most genes (n = 10) were mutated in only a single isolate, but five (femA, putative Abi-domain protease B7H15_05310, putative autolysin B7H15_09680, vraS, and ccpA) were each independently mutated in two to five evolved ATCC29213 isolates. Six of the mutated genes were affected by at least one unequivocally inactivating mutation (i.e., stop-gain or frameshift), consistent with loss-of-function effects, while the remainder were affected by missense mutations whose effects were less interpretable. No genes were mutated in common between ATCC29213 and JE2 over the course of passaging. Two JE2 replicates did not have detectable coding sequence or promoter mutations, suggesting epigenetic changes or mutations in remote regulatory elements.

Comparison of candidate genes identified by in vivo and in vitro approaches

Among 127 candidate genes identified by GWAS and 15 identified by passaging, only a single gene (clpX) was implicated by both in vivo and in vitro approaches. Given that 5,854 unique genes were considered in the GWAS analysis, this degree of overlap did not exceed that expected by random chance (p = 0.228; significance threshold for number of overlapping genes for by SuperExactTest [63], method 2). We conclude that distinct gene subsets were identified by in vivo and in vitro analyses.

Validation of candidate genes affecting macrophage intracellular pathogenesis

We next functionally evaluated candidate genes identified by in vitro and in vivo strategies for their ability to influence intracellular pathogenesis phenotypes in the THP-1 macrophage model. All genes identified by in vitro studies were targeted; however, given the large number of candidates implicated by analysis of in vivo isolates, we prioritized genes identified by GWAS for validation as those with the highest levels of association in a single analysis and those that were identified by more than one GWAS model. In total, the most significant 35 of 121 (28.9%) candidate invasion genes and all seven (100%) candidate survival genes met prioritization criteria for functional testing.

When available, we first tested the phenotypic impact of candidate genes using defined, isogenic transposon knockout mutants of strain JE2 [30]. Transposon mutants existed for 30 of 41 candidate genes prioritized from in vivo selection and 10 of the 15 of candidate in vitro genes. To further control for assay variability, we included transposon mutants for three genes that GWAS identified as having no association with intracellular pathogenesis, and which displayed similar intracellular pathogenesis phenotypes to the wild-type parental strain. We subsequently considered candidate genes to have a measurable impact on macrophage pathogenesis phenotypes if their corresponding transposon mutant effected a change that was statistically different from replicate measurements of the parental JE2 strain (defined as having a median measurement outside the 95% CI of controls).

Transposon disruption of 28 (93%) of in vivo candidates, six (60%) of in vitro candidates, and none of the uninvolved control genes significantly increased S. aureus invasion of THP-1 macrophages, while the remainder did not measurably alter that phenotype (Fig 4A). In contrast, S. aureus survival within macrophages was significantly decreased by disruption of only four (13%) of the in vivo candidates, none of the in vitro candidates, and none of the uninvolved control genes, with no enhancement of that phenotype observed (Fig 4B). Cumulatively, transposon disruptions of all but five candidate genes impacted macrophage pathogenesis. All such mutations improved macrophage invasion; however, four also negatively affected S. aureus survival within macrophages.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Transposon disruption of candidate macrophage pathogenesis genes.

Fold-difference in measurements of macrophage invasion, relative to in-batch testing of the parental JE2 strain (A), and macrophage intracellular survival (B) for specified transposon mutants. Candidate genes are grouped according to the strategy by which they were identified, and sorted within groups by median value. Values from in-run JE2 strain controls are shown, and JE2 transposon mutants in genes not implicated in macrophage pathogenesis were used as an additional control group. Boxes indicate median, 25^th, and 75^th percentiles with whiskers extending to the most distant value < 1.5 times the interquartile range. Red horizontal lines indicate the 95% confidence interval of all JE2 control measurements, with significant mutants defined as having a median measurement outside this range. Values higher than the 95% confidence intervals indicate increased macrophage pathogenesis phenotypes, and those below the interval correspond to decreased macrophage pathogenesis phenotypes. Genome-wide association study (GWAS).

https://doi.org/10.1371/journal.ppat.1012394.g004

Given the large number of targets identified, we performed genetic complementation studies on a subset of genes showing the largest effect sizes for the invasion phenotype, and the complete, but more limited number of genes exhibiting significant survival effects (S12 Table). We transformed transposon mutant strains with appropriate complementation vectors and compared relevant phenotypes against matched controls transformed with a GFP-expressing plasmid. Complementation reverted the expected macrophage invasion phenotype in 12 of the 16 transposon mutants tested and the macrophage survival phenotype of three of the four mutants tested, providing further evidence for their biological impact. Failure of some mutants to yield the expected phenotype after complementation may reflect mismatches in the dosage or regulation of plasmid-borne genes, and thus does not necessarily indicate that the tested gene lacks contributions to macrophage pathogenesis.

To expand our scope of testing to factors not available as transposon knockouts, including potentially essential genes, we developed a constitutively expressed CRISPRi [31] gene silencing vector for use in S. aureus. Using high specificity guide RNA designs [54], nine additional isogenic knockdowns were successfully generated in JE2 (S3 Table) and their effects on intracellular pathogenesis assessed relative to CRISPRi knockdown of an irrelevant control target.

As anticipated, knockdowns did not quantitatively influence phenotypes as prominently as gene knockouts. Nevertheless, all tested candidate genes significantly impacted one or both macrophage pathogenesis phenotypes. One in vivo candidate gene significantly increased invasion and five others significantly decreased it (Fig 5A). Similarly, two of the in vivo candidates positively impacted macrophage survival, while five reduced it (Fig 5B). Knockdown of the single gene identified in common between in vivo analyses and passaging, clpX, also notably impaired survival.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. CRISPRi knockdown of candidate macrophage pathogenesis genes.

Fold-difference in measurements of macrophage invasion, normalized to in-batch testing of the parental JE2 strain carrying a GFP-targeted CRISPRi construct(A), and macrophage intracellular survival (B) for specified CRISPRi gene knockdowns. Candidate genes are grouped according to the strategy by which they were identified, and sorted within groups by median value. JE2 carrying a GFP-expressing plasmid built on the same vector backbone as silencing constructs was used as an additional control group. Boxes indicate median, 25^th, and 75^th percentiles with whiskers extending to the most distant value < 1.5 times the interquartile range. Red horizontal lines indicate the 95% confidence interval of all control measurements from JE2 with GFP-targeted CRISPRi, with significant mutants defined as having a median measurement outside this range. Values higher than the 95% confidence intervals indicate increased macrophage pathogenesis phenotypes, and those below the interval correspond to decreased macrophage pathogenesis phenotypes. Genome-wide association study (GWAS).

https://doi.org/10.1371/journal.ppat.1012394.g005

Between the two validation approaches, 37 of the 39 tested candidate genes from in vivo analysis and seven of 11 genes from in vitro passaging impacted one or both macrophage pathogenesis phenotypes, with a single gene shared in common between those two sets.

Features of S. aureus isogenic mutants with altered survival phenotypes

Counts of intracellular bacteria could be impacted by changes in bacterial survival, host-cell killing, or bacterial replicative capacity. We therefore assessed these features in isogenic mutants exhibiting differences in the macrophage survival phenotype. Relative to wild-type JE2, no mutant strain evidenced significant differences in the total number of macrophages (p>0.29, two-tailed T test) or the proportion of dead host cells (p>0.20, two-tailed T test) observed 48 hours after infection, indicating no measurable alterations to host cell death by lytic or non-lytic mechanisms. With only a single exception (saeS), isogenic mutants did not have significantly slower growth rates than matched comparator strains (S13 Table). We conclude that reduced intracellular survival phenotypes observed in isogenic mutants largely reflect differences in bacterial viability within macrophages, rather than reduced bacterial replication or impaired killing of the host cell population.