The first observation that LPS deficiency causes colistin resistance in A. baumannii was made by Moffat and coauthors [58]. Laboratory-induced colistin-resistant A. baumannii derivatives contained mutations in one of the first three lipid A biosynthetic genes (lpxA, lpxC, or lpxD). Although, these mutations ranged from single nucleotide mutations to large deletions (up to 445 nucleotides), they all resulted in complete loss of LPS. Moreover, disruption of the lpxD gene by insertion of an IS element similar to the ISX03 element (IS4 family) was described in a colistin-resistant clinical isolate [58]. Shortly thereafter, the same team found that ISAba11 inactivated the lpxC and lpxA genes in colistin-resistant derivatives of A. baumannii ATCC19606 [61]. In subsequent studies, the insertion of ISAba1 or ISAba11 within the lpxC gene was described as a common event in colistin-resistant A. baumannii. As the disruption of the lpxC gene occurred in the same region (321–420 nt) in different isolates, it was suggested that this region might represent a hot spot for the integration of ISs [61,62,63,64,65,66,67,68]. Colistin resistance in A. baumannii has also been associated with various nucleotide substitutions, deletions, and insertions in all three lipid A biosynthetic genes (lpxA, lpxC, or lpxD) that cause frameshifts or result in truncated proteins that impair lipid A biosynthesis. While the described mutational events in the lpxA gene are not site-specific, non-synonymous mutations in the lpxC (P30L or S, N287D) and lpxD (E117K) genes were previously found to be present in colistin-resistant isolates from different origins [58,69,70,71,72,73,74]. Although the amino acid substitutions N287D (lpxC) and E117K (lpxD) were detected in both colistin-susceptible and colistin-resistant isolates, it is possible that these alterations in combination with a mutation in the pmrCAB operon have a synergistic effect leading to colistin resistance [69,70,71,72,73,74]. In addition, the downregulation of lpxACD expression has been observed in some colistin-resistant A. baumannii isolates, leading to decreased LPS production [68,73,74,75,76].

At the time when LPS deficiency was described as the mechanism causing colistin resistance in A. baumannii, this discovery was surprising because LPS biosynthesis was thought to be essential for the viability of Gram-negative bacteria [77]. So far, survival without LPS has been described only in a few species, such as Neisseria meningitidis, Moraxella catarrhalis, and two Acinetobacter species (A. baumannii and A. nosocomialis) [78,79,80]. Although this mechanism ensures a high level of colistin resistance [58,61,65], the frequency of mutations in the lpxACD is lower compared to changes in the pmrCAB operon in colistin-resistant A. baumannii clinical isolates [66,81,82]. The proposed explanation for the lower prevalence of LPS-deficient colistin-resistant mutants in clinical settings could be the significant negative impact of LPS loss on fitness and virulence, as well as the susceptibility of these isolates to various antibiotics and disinfectants. This was supported by the findings that the lpx mutants grew more slowly compared to the parental wild-type strains in vitro [64,66,68,81,83], while in vitro and in vivo competition tests showed significant fitness costs of colistin resistance [81]. Determination of the pathogenicity of the lpx mutants also revealed lower cytotoxicity (A549 human alveolar epithelial cells) and attenuated virulence of these strains in the animal models (Caenorhabditis elegans, Galleria mellonella, and mouse) compared to wild-type or even pmrB mutants [63,66,81]. As expected, the absence of LPS on the cell surface resulted in weak stimulation of neutrophils and, consequently, lower production of reactive oxygen species (ROS) and pro-inflammatory cytokines [66,83]. Nevertheless, the lpx mutants were more prone to killing mediated by neutrophils compared to the wild type since they were more susceptible to neutrophil-secreted lysozyme [83]. Moreover, reduced virulence of LPS-deficient A. baumannii in the host could also be explained by reduced biofilm formation, surface motility, as well as poor growth under iron limitation [66,83]. Finally, another disadvantage of LPS loss for the bacterial cell is evident in the increased susceptibility to various clinically used antibiotics, especially antibiotics used in the therapy of A. baumannii infections (ceftazidime, imipenem, meropenem, tigecycline, ciprofloxacin, amikacin, and rifampin), and various disinfectants (phenol-based disinfectants, quaternary ammonium disinfectants, sodium dodecyl sulfate, benzalkonium, chlorhexidine, deoxycholate, and ethanol) [58,63,64,65,66,68,83].

The modification of LPS is a commonly described mechanism for acquired colistin resistance in Gram-negative bacilli. In A. baumannii, PetN is added to the 4′-phosphate or 1-phosphate group of lipid A, reducing the negative charge of this LPS component and the binding affinity of colistin [57]. This type of colistin resistance is predominantly mediated by mutations in genes encoding the PmrAB two-component system, resulting in the overexpression of the phosphoethanolamine transferase PmrC [84]. The most common and diverse amino acid changes associated with colistin resistance in A. baumannii were detected in the PmrB protein. Since Adams et al. [59] observed that mutations in the pmrB gene can cause high colistin resistance (MIC greater than 128 µg/mL) in laboratory-induced A. baumannii derivatives, numerous studies have described the presence of altered PmrB proteins in colistin-resistant clinical isolates or in vitro-derived derivatives of A. baumannii. Although these nonsynonymous mutations were detected in all domains of PmrB, the greatest number were located in the histidine kinase A (HisKA) domain (predominantly at positions 226, 227, 232, 233, 235, and 263), which is responsible for autophosphorylation and the transfer of the phosphoryl group to the PmrA response regulator [85]. Accordingly, pmrB mutations could lead to the constitutive activation of PmrA, resulting in increased expression of the pmrCAB and resistance to colistin [59]. In addition, previous studies reported frequent amino acid substitutions of PmrB at position 170 (P to Y, L, Q, or S) and 315 (G to D, S, or V) in colistin-resistant isolates [68,70,76,84,86]. Although Oikonomou and coauthors [69] described the PmrB mutations (A138T, A226V, and A444V) repeated in colistin-resistant A. baumannii [70,72,73,74,76,84,85,86,87,88,89,90,91] as not responsible for colistin resistance, the involvement of A138T and A226V in this phenomenon should not be excluded. Indeed, the amino acid change at position 226 (A to V) in PmrB alone or in combination with A138T enabled high colistin resistance (64 or 128 and 256, respectively) in the tested isolates [84,88]. The amino acid substitutions within the receiver domain (REC) of the PmrA response regulator have also been described in A. baumannii as resistant to colistin (E8D, D10N, M12I, K, or V, I13M, or S, A14V, I18T, L20F, G54E, A80V, D82G, P102H, or R, F105L) [59,68,69,72,84,86,89,90,92,93,94,95,96,97,98]. Some of the PmrA mutations alone (G54E) or in combination with mutations in other genes (P102R) can confer significantly high colistin resistance to A. baumannii (>256 µg/mL or 512 µg/mL) [97,98]. To date, little data are available on the relationship between PmrC amino acid changes and colistin resistance. A comparison of PmrC amino acid sequences between colistin-susceptible and colistin-resistant isolates revealed rare changes and mostly at different positions [65,69,72,73,74,75,76,84,89,95,97]. In the study conducted by Nurtop and coauthors [72], the two most commonly described mutations in the pmrC gene (resulting in I42V and L150F) were found to be associated with an increased expression of the pmrA and pmrC genes and, consequently, colistin resistance. The PmrC substitution R109H, detected in colistin-resistant A. baumannii isolates in two previous studies [69,72], was associated with colistin resistance along with a mutation in the pmrA gene [69]. In addition, it was observed that the PmrC alteration R125P in combination with changes within the PmrB protein had a synergistic effect on colistin resistance in A. baumannii [97]. In summary, mutations in the pmrCAB locus are recognized as gain-of-function mutations because they lead to PmrC overexpression and PEtN modification of lipid A, which, in turn, results in colistin resistance [84,99]. In addition to increased expression of PmrC as a mechanism of colistin resistance in A. baumannii [65,72,73,75,76,84,85,88,92,97,98,100], the upregulation of the pmrA and pmrB genes was found in some colistin-resistant isolates [59,71,96,101,102], but to a much lesser extent [68,72,73,75,76,84,92,98]. Although this observation is to be expected as these genes are part of the same operon as the pmrC gene (pmrCAB), there are cases where no correlation was found between PmrAB and PmrC overexpression [72,73,76]. In addition, Lesho and coauthors [92] noted the overexpression of another pmrC homolog (eptA, ethanolamine phosphotransferase A) in some colistin-resistant A. baumannii isolates. Detailed analysis revealed that the gene encoding for EptA was detected only in isolates belonging to the international clone 2 (IC2), was found in ≥3 copies in a single isolate, and was distant from the pmrCAB operon in A. baumannii genomes [88,90,92]. Although the presence of the eptA gene in the bacterial genome alone does not confer resistance to colistin, the integration of ISAba1 upstream of the eptA gene could lead to increased expression of this enzyme [88]. In contrast, Gerson et al. [100] found the presence of ISAba1 upstream of the eptA gene in colistin-susceptible and colistin-resistant counterparts, but only in isolates with mutations in the eptA gene (R127L) and ISAba1 (A→T in position 1091) was overexpression of EptA detected. Interestingly, a previous study showed that disruption of the gene encoding the global regulator H-NS by ISAba125 causes high colistin resistance in A. baumannii through increased expression of the eptA gene in this mutant strain [103].

A negative correlation was found between PmrAB-related colistin resistance and the fitness and virulence of A. baumannii in the host. The colistin-resistant A. baumannii isolates showed lower fitness in vitro and in vivo and reduced virulence potential in animal models of infection compared to their colistin-susceptible parental strains [62,92,93,104,105,106,107,108]. This could be explained by the downregulation of metabolic and antioxidant proteins, virulent porins OmpA and CarO, and the system responsible for biofilm formation in colistin-resistant A. baumannii [107,109,110]. In addition, some studies reported a correlation between colistin resistance and decreased biofilm formation ability [108,110]. In contrast to the initially reported negative correlation, additional studies showed unchanged fitness [63,64,68,100,111,112] and pathogenicity of colistin-resistant A. baumannii [63,81,100,111]. Interestingly, two studies described the emergence of colistin resistance in A. baumannii isolated from patients exposed to colistin therapy and the subsequent disappearance of this resistance after the discontinuation of colistin [111,113]. Durante-Mangoni and coauthors [111] observed that colistin-resistant pmrB-mutated isolates were comparable to wild type in in vitro and in vivo assays, whereas Snitkin et al. [113] hypothesized that resistant isolates were outcompeted by colistin-susceptible isolates due to lower in vivo fitness costs. In addition, a comparison of five longitudinal colistin-resistant A. baumannii isolates from the same patient indicated an increase in growth rate as well as virulence in the mouse lung infection model during colistin therapy [114]. Jones and coauthors [114] explained this phenomenon by more pronounced resistance to ROS in late colistin-resistant isolates. Overall, these data suggest that no clear conclusion can be made about the correlation of colistin resistance due to pmrAB mutations and biological costs in A. baumannii. Although some pmrAB mutations responsible for colistin resistance initially appeared to be maladaptive to bacterial cells, prolonged exposure to the selective agent (colistin) may have allowed the emergence of compensatory changes at different regulatory levels and remedied a deficit in fitness and virulence [63,104,113,114]. In addition, in this type of research, the genetic background should be taken into account as the results obtained from different isolates containing the same PmrB alteration P233S were different [107,108,111,112]. The studies comparing the behavior of the pmrAB mutants with lpxACD mutants have undoubtedly confirmed that the LPS modification causes lower fitness and virulence costs than LPS deficiency [63,64,81]. Most studies that examined colistin-resistant A. baumannii showed that PmrAB alterations had no significant impact on the antibiotic resistance profile of these isolates [64,68,69,84,92,112]. Consistent with the above observations, a systematic review concluded that LPS modification mediated by the pmrAB mutations is the major in vivo mechanism of colistin resistance [82].

Since the first report of the phosphoethanolamine transferase-encoding mcr gene (mcr-1) in E. coli in China [119], the presence of this gene and its variants has been demonstrated in many Gram-negative bacteria distributed worldwide [60]. To date, ten different mcr gene families (mcr-1 to mcr-10) with more than 100 variants have been described in bacteria isolated from animals, food, humans, and the environment [60,120]. In A. baumannii, the mcr-1 and mcr-4.3 variants are most commonly detected. The mcr-1 has been reported in clinical isolates from Asia (Pakistan, Iraq, and China) and Africa (Egypt) at sporadic frequency (n = 1–3) with the exception of samples collected from hospitals in Iraq (up to 89) [121,122,123,124,125,126]. The earliest mcr-4.3-positive isolate of A. baumannii was recovered from the cerebrospinal fluid of a patient with meningitis in Brazil in 2008 [127], which preceded the mcr discovery by Lui and coauthors [119]. Subsequently, mcr-4.3 was detected in pig feces from a slaughterhouse in China [128] and in isolates from the Czech Republic [129,130]. The studies from the Czech Republic suggest that food imports from Latin America (frozen turkey livers from Brazil) and Asia (frog legs from Vietnam) may represent the primary route of transmission of mcr-carrying A. baumannii to Europe and thus to European hospitals [129,130]. As some studies showed that the recombinant expression of mcr-4.3 in E. coli did not alter colistin MIC [131,132], while another study indicated that the heterologous expression of mcr-4.3 could ensure colistin resistance through LPS modification in A. baumannii [127], it is not possible to draw a firm conclusion about its role in colistin resistance. Moreover, a comparative analysis revealed that the mcr-4.3-harbouring plasmids in A. baumannii share a common origin for this structure. It was found that these plasmids are untypable and cannot be transferred to other bacteria by conjugation or transformation [128,129,130]. Although mcr-1 and mcr-4.3 are predominant, other mcr variants have also been described in clinical and environmental samples of A. baumannii, as in a study from Iraq where the mcr-2 and mcr-3 genes were found. A large number of these isolates carry a single mcr gene or a combination of different mcr families (mcr-1, mcr-2, and mcr-3) [122]. Finally, it is important to highlight that most of the mcr-carrying A. baumannii isolates are MDR [121,122,124,125,126,127], and there are few antibiotic-susceptible isolates [128,129].

Antibiotic heteroresistance is defined as the presence of a resistant subpopulation within a population susceptible to a given antibiotic [140]. Since the first report of colistin heteroresistance in clinical isolates of A. baumannii from Australia [141], this phenomenon has been described in many studies with prevalence ranging from 1.84 to 100% [142,143,144]. Hawley and coauthors [142] found a higher rate of heteroresistance in isolates from patients treated with colistin, suggesting that previous colistin therapy may be a risk factor for the induction of heteroresistance. Data indicating resistance stability within the surviving subpopulation after cultivation under non-selective conditions were conflicting in different studies, suggesting a possible species-specific dependence [140,141,142,145]. Interestingly, Hong et al. [140] observed isolates that exhibited a heteroresistance phenotype only at low antibiotic concentrations alongside the typical colistin-heteroresistant isolates that emerged at exposure to high colistin concentrations. The previously described mechanisms of colistin heteroresistance in A. baumannii are the same as those of colistin resistance (LpxACD, PmrCAB, and efflux pumps) [73,140,143,145,146]. Amino acid changes in LpxC (S186R) and LpxD (N148K and T289I) were associated with partial loss of LPS in heteroresistant strains [143], while another study showed upregulation of the pmrCAB operon in combination with mutations in the pmrA and pmrB genes in resistant subpopulations of A. baumannii [146]. The overexpression of efflux pumps and the synergistic effect of EPIs and colistin against the resistant subpopulation of heteroresistant A. baumannii clearly demonstrated the involvement of efflux pumps in this phenotype [143,145]. Of particular concern is the fact that conventional susceptibility testing identifies heteroresistant isolates as susceptible to colistin, resulting in colistin treatment failure [143]. As population analysis profiling (PAP) is recognized as the gold standard for detecting heteroresistance, the introduction of the mini-PAP method with colistin at >2 mg/L into clinical practice has been recommended [147]. Moreover, the prevalence of heteroresistant isolates clearly exceeds the occurrence of colistin-resistant A. baumannii [148]. Moreover, under selection pressure, a resistant subpopulation of heteroresistant populations could become predominant and lead to a resistant cell population [145]. Accordingly, isolates identified as colistin-heteroresistant have been proposed for colistin-based combination therapy instead of colistin monotherapy [144]. Although the phenomenon of colistin heteroresistance has been studied mainly in A. baumannii of nosocomial origin, it has also been detected in samples from hospital wastewaters [73,149].

Another phenomenon observed in some colistin-susceptible A. baumannii isolates exposed to colistin is colistin dependence. After exposure to colistin, a colistin-dependent subpopulation of cells becomes dependent on this antibiotic for optimal growth [150]. Previous findings have suggested the colistin-dependent phenotype as a survival response to colistin pressure and an intermediate stage between colistin susceptibility or heteroresistance and even extreme resistance to colistin [65,150].