Bacterial behavior in human blood reveals complement evaders with persister-like features

Bacterial bloodstream infections (BSI) are a major health concern and can cause up to 40% mortality. Pseudomonas aeruginosa BSI is often of nosocomial origin and is associated with a particularly poor prognosis. The mechanism of bacterial persistence in blood is still largely unknown. Here, we analyzed the behavior of a cohort of clinical and laboratory Pseudomonas aeruginosa strains in human blood. In this specific environment, complement was the main defensive mechanism, acting either by direct bacterial lysis or by opsonophagocytosis, which required recognition by immune cells. We found highly variable survival rates for different strains in blood, whatever their origin, serotype, or the nature of their secreted toxins and despite their detection by immune cells. We identified and characterized a complement-tolerant subpopulation of bacterial cells that we named “evaders”. Evaders represented 0.1-0.001% of the initial bacterial load and displayed transient tolerance. Although evaders shared some features with bacterial persisters, which tolerate antibiotic treatment, they appear to have evolved distinct strategies to escape complement. We detected the evaders for five other major human pathogens: Acinetobacter baumannii, Burkholderia multivorans, enteroaggregative Escherichia coli, Klebsiella pneumoniae, and Yersinia enterocolitica. Thus, the evaders could allow the pathogen to persist within the bloodstream, and may be the cause of fatal bacteremia or dissemination, notably in the absence of effective antibiotic treatments. Author summary for “Complement evaders” Blood infections by antibiotic resistant bacteria, notably Pseudomonas aeruginosa, are major concerns in hospital settings. The complex interplay between P. aeruginosa and the innate immune system in the context of human blood is still poorly understood. By studying the behavior of various P. aeruginosa strains in human whole blood and plasma, we showed that bacterial strains display different rate of tolerance to the complement system. Despite the complement microbicide activity, most bacteria withstand elimination through phenotypic heterogeneity creating a tiny (<0.1%) subpopulation of transiently tolerant evaders. While genetically identical to the rest of the complement-sensitive population, evaders allow the bacteria to persist in plasma. This phenotypic heterogeneity thus prevents total elimination of the pathogen from the circulation, and represent a new strategy to disseminate within the organism.


23
Bacterial bloodstream infections (BSI) are a major health concern and can cause up 24 to 40% mortality. Pseudomonas aeruginosa BSI is often of nosocomial origin and is 25 associated with a particularly poor prognosis. The mechanism of bacterial persistence 26 in blood is still largely unknown. Here, we analyzed the behavior of a cohort of clinical 27 and laboratory Pseudomonas aeruginosa strains in human blood. In this specific 28 environment, complement was the main defensive mechanism, acting either by direct 29 bacterial lysis or by opsonophagocytosis, which required recognition by immune cells. 30 We found highly variable survival rates for different strains in blood, whatever their 31 origin, serotype, or the nature of their secreted toxins and despite their detection by 32 immune cells. We identified and characterized a complement-tolerant subpopulation 33 of bacterial cells that we named "evaders". Evaders represented 0.1-0.001% of the 34 initial bacterial load and displayed transient tolerance. Although evaders shared some 35 features with bacterial persisters, which tolerate antibiotic treatment, they appear to 36 have evolved distinct strategies to escape complement. We detected the evaders for

109
Laboratory and clinical strains display diverse survival rates in human blood 110 Using a highly standardized method in human whole blood (HWB), we examined the 111 survival of six P. aeruginosa strains with distinct toxin repertoires and serotypes (Fig   112   1). These assays included the commonly used laboratory strains PAO1 [24], PA14 [25] 113 and PA7 [26], which belong to distinct phylogenetic lineages/groups . PAO1 and PA14 114 both possess the type III secretion system (T3SS) which they use to translocate ExoS 115 or ExoU, respectively, into target host cells. PA7 lacks T3SS genes, but encodes the 116 pore-forming toxin Exolysin A [26,27]. In addition to these laboratory strains, we 117 included in the survey three P. aeruginosa strains recently isolated from infected 118 patients [27][28][29]. E. coli CF7968, a derivative of the K12 laboratory strain [30] was 119 added as a control (Table S1). Bacteria were incubated for 3 h in HWB from healthy 120 donors, and bacterial survival was assessed in ten independent experiments by 121 counting colony-forming units (CFU) (Fig 1A). Strains showed clearly distinguishable 122 and reproducible survival rates. Approximately 10% of the laboratory and reference 123 strains, PAO1 (ExoS + ) and PA7 (ExlA + ), survived, whereas only around 1% of the most 124 sensitive strain, PA14 (ExoU + ), was still present following the 3-h incubation in HWB. 125 Very different survival rates were measured for the three recent clinical strains, YIK 126 (ExoU + ), CLJ1 and IHMA87 (ExlA + ), ranging from 0.05% for CLJ1 to complete 127 tolerance for YIK. Only the non-pathogenic laboratory strain E. coli CF7968 was 128 completely eliminated, with no detectable CFU after 3 h exposure to HWB. The nature 129 of the strain's virulence factors (T3SS versus ExlA) did not appear to confer any 130 significant benefit for survival, as similar sensitivities were measured for ExlA + strains 131 7 and T3SS + strains (e.g. IHMA87 versus PA14, or PA7 versus PAO1). In addition, 132 survival in HWB did not correlate with a given serotype, as highly variable survival rates 133 were measured for the three O12 strains (Fig 1B). Lack of the O antigen was 134 detrimental for the bacteria, as illustrated by the hypersensitivity of E. coli CF7968. most other strains showed similar limited levels of cytotoxic potential with no significant 147 differences ( Fig S2). Thus, except for YIK, the extent to which the bacterial strains 148 tested were recognized by and destroyed circulating leukocytes could not explain the 149 different survival rates measured in HWB. 150 We next assessed the capacity of the strains to cope with the complement system in 151 plasma (Fig 2A) the same level of bacterial elimination was observed as with untreated HWB (Fig 2C), 178 indicating that NETs play a negligible role in the process observed here. In contrast, 179 PA7 and PAO1 elimination appear to involve some internalization by phagocytes, as 180 their survival was consistently increased in Cytochalasin D-treated blood ( Fig 2C). 181 Because the clearance of these two strains was also complement-dependent, we 182 conclude that they are eliminated through opsonophagocytosis. Blocking phagocytosis 183 had no impact on the other complement-sensitive strains CLJ1, IHMA87, PA14, and 184 E. coli CF7968, suggesting that they are killed through direct MAC-induced lysis.

185
Importantly, even though most of the strains were highly sensitive to complement, we 186 recurrently detected a subpopulation of survivors corresponding to <1% or even 187 0.002% of the initial inoculum, in HWB and plasma, respectively (Fig 1A and 2A). 188 These results suggest that a small bacterial subpopulation, that we termed "evaders", 189 differ in phenotype from the majority of the population, and display increased tolerance  Complement evaders display persister-like features 193 As indicated above, only the laboratory strain E. coli CF7968 was entirely eliminated 194 upon exposure to complement. The subpopulation of P. aeruginosa evaders in plasma 195 ranged from 0.1% down to 0.002% of the initial bacterial load, depending on the strain 196 (Fig 2A). We further investigated this intriguing difference in sensitivities using the three failed to grow, even after 6 h. We verified that the drop off in killing rate was not due to 203 depletion of complement activity after 2 h by retesting the used plasma. The used 204 plasma was still sufficiently active to kill > 10 7 P. aeruginosa PA14 cells during a 1-h 205 incubation ( Fig S3A). Thus, the plasma had a residual bactericidal capacity, sufficient 206 to eliminate a population at least 4-log more numerous than the number of evaders.

207
As further evidence that evaders are not simply a result of bacterial overload of the

212
To further phenotypically characterize evaders, we re-cultured the survivors recovered 213 from a first incubation in plasma and challenged their progeny. As shown in Fig 3B, 214 following re-culture, the bacterial population had a similar sensitivity profile to 215 previously unchallenged cells. In some cases, the number of evaders in these repeat 216 challenges was below the limit of detection in our experimental settings (e.g. CLJ1 217 after 2 h). This apparent paradox is a hallmark of antibiotic persisters [42]. Thus, 218 neither evaders nor persisters reflect the emergence of resistant mutants, rather the 219 evader phenotype is transient and reversible.

220
As complement evaders were detected for all sensitive P. aeruginosa strains, we 221 tested whether this behavior could be extended to other Gram-negative bacteria. To 222 that aim, we selected strains of seven Gram-negative species (Table S1)  to the one recorded for P. aeruginosa ( Fig 3C). For these strains, the proportion of 230 evaders withstanding complement-mediated lysis after 3 h of incubation ranged from 231 1% to 0.0002% of the initial population. When individual evader colonies were re-232 cultured, a population as sensitive as the parental one was recovered, as seen for P. 233 aeruginosa. Unexpectedly, for K. pneumoniae, one evader colony out of the three that 234 were randomly selected gave rise to a resistant population ( Fig 3D). Thus,  between the two states ( Fig 4A). Therefore, exponentially growing bacteria had a 252 similar or higher capacity to produce evaders, suggesting that the emergence of 253 evaders is unrelated to dormancy before complement challenge. To assess whether 254 metabolic shut-down could increase the proportion of evaders, prior to exposure to 255 plasma, P. aeruginosa IHMA87 was treated with the protonophore cyanide m-256 chlorophenylhydrazone (CCCP), which uncouples oxidative phosphorylation.  (Fig 4E and 4F). Altogether, these results suggest that survival to plasma treatment is 272 not dependent on growth rate but requires energy and active metabolism. As we had demonstrated that the central driver of bacterial clearance from blood was 277 the complement system, mainly through its direct lytic activity, we next investigated 278 survival in plasma of a cohort of twelve clinical strains isolated from patients with BSI 279 (Table S1 and S2) to determine their capacity to form evaders. Like the data obtained 280 with the initial six selected strains, BSI isolates displayed differences in survival rates 281 in plasma, of up to five orders of magnitude. Four strains (PaG2, PaG5, PaG6, and 282 PaG10) were tolerant to complement killing, displaying > 50% survival, with some even 283 able to multiply in these conditions (PaG2, PaG6 and PaG10) ( Fig 5A). In contrast, for 284 other strains (PaG8, PaG9, PaG14) just < 0.02% of the initial population survived (Fig   285   5A). The limited number of strains and the high diversity of serotypes identified (Table   286 S1) would make any attempt to correlate bacterial survival in plasma with strain 287 serotype too speculative. Even though some strains were highly sensitive to 288 complement, none were fully eliminated. To verify that these surviving cells 289 corresponded to complement evaders, we also assessed their survival kinetics in 290 plasma (Fig. 5B). The three isolates PaG3, 9, and 17 presented a tolerant phenotype, 291 with a constant rate of elimination, never reaching a plateau even after 4 h incubation.

292
In contrast, a biphasic killing curve was recorded for the five other isolates, the kinetics 293 of the curve varied from strain to strain, sometimes reaching a plateau after just 1 h, 294 whereas for others the death rate started to slow from the 3-h time point. As indicated 295 above, in some cases, surviving cells were scarce and bellow the limit of detection   Complement resistance is the main driver of survival in the bactericidal environment 320 that human blood represents. Nevertheless, we observed that to fully resist the 321 immune system, strains had to display both complement resistance and cytotoxicity 322 toward immune cells, as documented for the highly virulent YIK strain recently isolated 323 from a 49-year-old individual with no known immunodeficiency [29].

334
The most prominent result from our study is the evidence that strains highly sensitive 335 to plasma can nevertheless escape complement's bactericidal activity by forming 336 phenotypic variants, or evaders. In some instances, complement evaders represent 337 < 0.01% of the initial population, with survival kinetics reaching a plateau reminiscent 338 of bacterial persisters following antibiotic challenges [42]. As with antibiotic persisters 339 [42], we found that complement evaders do not harbor genetic mutations and that they 340 lose their complement-tolerant phenotype upon elimination of the stress. However, 341 complement evaders differ from antibiotic persisters through several features. First, 342 persisters are more numerous in stationary-phase cultures [45], whereas the fraction 343 of evaders is similar or even higher in exponentially growing cells. Second, we 344 observed no growth defect/delay of these rare cells following removal of the stress.  The bacterial strains used in this study are listed in Table S1. Bacteria were grown in To inactivate complement, HWB was centrifuged for 5 min at 400 g to isolate plasma, 424 which was subsequently heat-inactivated for 30 min at 56 °C. Meanwhile, cells were 425 washed twice with RPMI and pelleted by centrifugation for 5 min at 400 g.   to convert initially non-normally-distributed data into a normally-distributed dataset.

492
GraphPad Prism was used to create graphs.  University is also acknowledged. The funders had no role in study design, data 515 collection or analysis, decision to publish, or preparation of the manuscript.