2. Molecular mechanisms of A. baumannii involved in the physiology of persister cells
Factors influencing the level of persister cells
Several studies have investigated the formation of persister cells byA. baumannii. The study by Bart et al., (2013) was the first investigation on this type of cells in A. baumannii . Thirty-seven clinical strains from the hospital of Porto Alegre (Brazil), isolated from January to September 2012, were tested for persister cell formation with two antibiotics, polymyxin B and tobramycin (Table 1) [32]. This study showed the proportion of persister cells strongly depended on the strain and varied between 0.0007% to 10.1% and 0.0003 % to 11.84 % of the total population, for polymyxin B and tobramycin respectively [32]. However, for the majority of isolates, the most likely proportion was about 1.2% for polymyxin B and 0.5% for tobramycin.
In another study about the same hospital [33] twenty other clinical isolates were classified according to their ability to form biofilms. As known, the proportion of persister cells in biofilm mode was much higher than in planktonic mode but surprisingly, strong biofilm producers had some of the lowest persister levels (Table 1). Persister levels show a higher heterogeneity (0.2 to 7.2%) for isolates producing low biofilm compared with those producing more biofilm. Among the isolates, 5 were tested in planktonic mode at different concentrations (15 to 200 times the minimum inhibitory concentration (MIC) of meropenem). The proportion of persisters did not vary as a function of antibiotic concentration. The authors suggested that this persister level heterogeneity could have four distinct and non-exclusive origins: the individual ability to produce persistence stochastically, specific regulatory mechanisms, the metabolic structure of each isolate and the influence of other previous stresses.
Persister cells were also observed after ciprofloxacin treatment of non multidrug resistance (MDR) strong biofilm former clinical isolates [34]. This treatment also rose the level of ciprofloxacin resistance in regrown planktonic cells suggesting a link between biofilm, persistence and resistance (Table 1).
Treatments by aminoglycosides and tobramycin in particular were also shown to promote persister formation (Table 1) [35], the level of which was modulated by oxygenation level and composition of the culture medium [35], [36]. This modulation via nutriment availability was also observed on persister levels obtained after ciprofloxacin treatment (Table 1) [36]. The growth phase is another factor that was shown to strongly influence the proportion of persister cells [37], but this has not yet been demonstrated for A. baumannii . Thus, if the level of persister formation depend weakly on the culture parameters or on the antibiotic used, it was suggested to be essentially related to the strain in A. baumannii . Indeed, the diversity of the A. baumannii genomes reflects a high degree of diversity in the strains themselves (Table 1) [38]. In fact, out of the 2 500 A. baumannii genomes, only 2 221 genes are shared between each strain (core genome), out of the 19 272 that correspond to the pangenome. Strain diversity may be a main explanation for the difference in persister formation levels.
Toxin/antitoxin systems related to the formation of persister cells
The toxin/antitoxin (TA) systems are composed of a stable toxin and an unstable antitoxin, which neutralizes the activity of the toxin. TA systems are currently grouped into five types (types I to V) according to the nature of the antitoxin and the mode of interaction between toxin and antitoxin. In all cases, toxins are proteins, while an antitoxin may be either RNA or protein [39].
Initially presented as a plasmid maintenance system [40], new advances have shown that these systems act also on the physiology of the bacteria. Indeed, they allow slowing down bacterial growth in response to a stress factor, promoting thus an entry into bacterial persister state [41]. The involvement of “high persistence” genes (hipA/hipB TA operon) in the persister phenomenon was discovered in 1983 in E. coli K12 [42]. This system is the most documented for its involvement in persister cell formation. Under unfavorable conditions, a cascade of reactions activates the TA system. Indeed, the presence of stress induces activation of the stringent response, leading to 5’-diphosphate 3’-diphosphate (ppGpp) accumulation. Although not essential, ppGpp accumulation leads to the formation of polyphosphate (polyP) by the polyphosphate kinase (PPK) [43]. This polyP then interacts with the protease complex Lon, activating a complex polyP - Lon that promotes protein degradation of the HipB antitoxin from the TA. HipA toxin is thus released and inhibits protein synthesisvia an inhibition of the glutamyl-tRNA synthetase [44]. This leads to an accumulation of uncharged tRNAGlu in the cell and results in failure of protein biosynthesis. The TA systems remain partially described in A. baumannii and present, as already described, a significant variability depending on the strain [45], [46], [47].
The study of 476 genomes of A. baumannii isolates predicted 15 different TA systems. They were detected either on the strain genome or on their plasmids (ACICU, AYE, SDF, MDR-ZJ06, p3ABAYE…) [45]. The authors subcloned the toxins and expressed them in E. coli under an inducible promoter. They then investigated the toxic effect of the protein and whether it was functional. In this group, five of them appear to be functional (Table 1). These five TA systems are all type II. Three of these TA are orthologs of the bacterial and archaeal systems RelB/RelE, HicA/HicB and HigB/HigA, and two are unique TA modules, AbkA/AbkB (also called SplT/SplA) and CheT/CheA [45], [48], [49].
AbkA/AbkB : Among these five potential TA systems, the AbkA/AbkB system was the only system studied for its involvement in persister formation in A. baumannii . It is mainly carried by plasmid but does not appear to be directly involved in plasmid maintenance [50], [51], [52]. The toxin AbkB is an endoribonuclease that inhibits translation, and AbkA as its cognate antitoxin [50]. The genes abkA/abkB were over-expressed (2.7 fold change) in persister cells under antibiotic treatment (10 × MIC of imipenem) [53]. The strains overexpressed AbkA and presented a high proportion of persister while a drastic decrease was observed in the other strains without this system. It is also interesting to note that a significant correlation was observed between the presence of abkA and blaOXA-24 genes in carbapenem-resistant isolates obtained from patients in Iran [48]. In this case, it was suggested a potential role for these systems as plasmid stabilization systems, contributing to the evolution of antibiotic resistance in A. baumannii [45], [49].
Concerning the other TA systems, their function was hypothesized regarding their identity with already described TA systems in different bacteria.
HigA/B : Indeed, for HigB/HigA, no link has been yet established for its involvement in A. baumannii persisters. However, in Mycobacterium tuberculosis it is involved in stress response [54]. In Pseudomonas aeruginosa, the toxin HigB regulates virulence factor production (e.g pyocyanin production, pyochelin, swarming mobility) and promote biofilm formation where a high proportion of cells are known to persist [55], [56]. HigA antitoxin represses the expression of a key transcriptional regulator of the virulence systems by directly binding to their promoter regions, and it controls pyocyanin synthesis, type III secretion system (T3SS), and type VI secretion system (T6SS) expression [56]. In A. baumannii, the HigB2/HigA2 system was found on the 11kb plasmid pAB120 also carrying genes for resistance to carbapenems [57]. The hetero-oligomer has a feedback control on its expression. It was also modulate by environmental stresses and was shown to be overexpressed in stationary phase or under iron deficiency [57]. Both HigB/HigA and AbkA/AbkB are the most abundant systems, and only detected to be carried by plasmids up to now. They are predominantly present in isolates from the international clones one and two (IC-I and IC-II) [45], [48], [57], [58], [59].
CheT/CheA : Regarding A. baumannii CheT/CheA system, CheA presented 25 % identity with the E. coli DinJ antitoxin, and has 37 % identity with an N-acetyltransferase superfamily [48]. . The predicted structure of the protein seems to show that CheA possesses helix-turn-helix (HTH) domains characteristic of DinJ. In E. coli , DinJ/YafQ system is involved in the formation of persister cells [60]. Kill-rescue assays on A. baumannii isolates showed that CheT/CheA served as a TA module: the HTH domain of the protein CheT acted as a toxin and the GNAT domain of CheA protein neutralized the HTH effect acting thus as an antitoxin [48].
RelB/RelE : The RelB/RelE system is detected at genomic level in some A. baumannii clinical strains (AC1633, ACICU, MDR-ZJ06, TCDC-AB0715, TYTH1 or AB 5075) [38], [45], [46], [61]. In E. coli , the target of the RelB toxin is the A ribosomal site. The blockage of this site promotes the cleavage of the messenger ribonucleic acid (mRNA) during translation [62]. This allows a translation stop and thus a stop of protein production[62].
HicA/B: The last TA system hicA/hicB was shown to be overexpressed in A. baumannii ATCC 17978 upon ciprofloxacin treatment (50 × MIC), suggesting an involvement of this system in the response of bacteria to antibiotic shocks [63]. In E. coli K12, the toxin HicA cleaves mRNA and also the tRNA in a ribosome-independent manner and HicA is reported to be inactivated by HicB [64]. In Burkholderia pseudomallei , this system was suggested to be involved in the formation of persister cells [65].
Finally, a Zeta/Epsilon system has been predicted and detected in the plasmids p2ABTCDC, pAbSK-OXA-82, pABTJ1 and pACICU2 [45]. Zeta toxins are notably larger (360 amino acids) than their currently known functional orthologs (270 amino acids). The Zeta toxin in A. baumannii could block cell wall biosynthesis by phosphorylation of the peptidoglycan precursor uridine diphosphate-N-acetylglucosamine (UNAG) [66]. Indeed, Zeta would form an ATP-UNAG-Zeta complex. This complex will block the action of MurA, responsible for peptidoglycan biosynthesis [67], which is the initial step in membrane biosynthesis [66]. Zeta then appears to have a strong implication in cell multiplication and would be an interesting target, as it stops bacteria to grow rapidly. Although not an antitoxin-toxin system, PPK kinase is also involved in protein degradation, particularly in antitoxin degradation. A study in A. baumannii investigated the effect of polyP via a mutant of PPK [68]. In this study, ATCC 17978 strains deficient in PPK1 (Dppk1::Apr) and supplemented with PPK1 (Dppk1::Apr/PJL02-ppk1) were studied. As expected, the authors showed that PPK1 is essential for polyP formation in vivo . The PPK1 mutant had a reduced motility by a destabilisation of the pili structure, inhibited biofilm formation and decreased persister under antibiotic (40 × MIC of Ampicillin) but also under other stresses such as H2O2, heat shock and starvation stress. These phenotypes were restored by complementation. Interestingly, metabolomics analysis revealed that PPK1 was associated with glycerophospholipid metabolism and fatty acid biosynthesis. The authors suggested thus alteration of glycerophospholipid metabolism could alter the membrane charge, resulting in a resistance of bacteria to the tested stresses [68].