Although fitness and virulence are different concepts, studies often evaluate both. Therefore, fitness and virulence will be discussed together and we will attempt to relate these traits with genetic mutations associated with colistin resistance. Some studies test clinical isolates while others expose ATCC strains to increased concentrations of colistin to generate different resistant isolates.
4.1. Colistin Resistance Due to Mutations in pmr Genes
Clinical
A. baumannii isolates often acquire colistin resistance during treatment with this antibiotic [
26,
51,
52,
53]. Two clinical
A. baumannii strains, which acquired colistin resistance after treatment with this polymyxin, were shown to have different clinical outcomes.
A. baumannii ABIsac_ColiR [
25] showed impaired virulence, as seen by loss of clinical signs of infection in the human patient [
51] and in a rat model of acute pneumonia [
52], while
A. baumannii CR17 did not lose its infecting capacity [
54]. Both strains were later further explored and colistin resistance was associated with
pmrA mutations:
A. baumannii ABIsac_ColiR with
pmrA E8D [
25] and
A. baumannii CR17 with
pmrA M12K [
55]. Strain ABIsac_ColiR was also shown to have lost a prophage, which could contribute to or explain the loss of virulence in this strain [
25]. A decreased in vitro and in vivo fitness has also been observed [
52]. Although strain CR17 remained infectious, it was also associated with a decreased virulence and fitness, as compared with the initial susceptible strain CS01 [
55]. The levels of virulence vary between
A. baumannii strains [
56], and the retention of the capacity to infect of CR17 strain might be related with the initial high virulence of the susceptible strain [
55].
Four clinical
A. baumannii strains showed an in vitro fitness cost, seen as the decreased growth rates of the resistant strains. The in vivo fitness cost was evidenced by the loss of resistance after treatment cessation. Although mutations in different genes were observed, all resistant isolates carried
pmrB mutations which also varied, including P233S, R263C, M145I, T13A, or indel AAT at position 69. These mutations could be associated with the overexpression of the PmrC phosphoethanolamine transferase, with a consequent increase of the
pmr operon transcript. Return to susceptibility to colistin occurred by different mechanisms. In two patients, there was re-emergence of the susceptible strain. In another patient, the resistant strain was lost, but unexpectedly, the apparent susceptible strain was detected to be also present during colistin treatment. This strain carried a L271R mutation in
pmrB, associated with a low fitness cost and in vivo stability in the absence of colistin. In the fourth patient, with the resistant isolate carrying the
pmrB P233S mutation, a compensatory mutation in
pmrA (L206P) was observed, making the future re-acquisition of colistin resistance highly unlikely [
26]. In a different study, the same
pmrB P233S mutation present in other clinical colistin-resistant
A. baumannii strains (Ab4451 and Ab4452) was not associated with loss of virulence or reduced fitness; compensatory mutations in the
pmrCAB locus were also not detected. Despite the fact that in vitro and in vivo fitness costs were not observed in Ab4451 and Ab4452, resistance to colistin was lost after colistin was withdrawn [
57]. In a recent study, the
pmrB P233S mutation was not associated with a fitness cost, but had an impact on the in vitro virulence, as evaluated by attenuated proteolytic activity and siderophore production, of the clinical strain C440 [
58]. While the study by Snitkin and colleagues showed mutations in other genes usually not specifically related with colistin resistance [
26], Durante-Mangoni et al. only sequenced the
pmrCAB and
lpxACD loci [
57]. Whether compensatory mutations in non-analysed genes, post-translational modifications, or physiological changes could explain the different study outcomes remains to be determined.
Two different clinical strains acquired colistin resistance after patient treatment with colistin. Ab249 and Ab347 harboured
pmrB P233S and P170L mutations, respectively. Both strains showed a reduced in vitro fitness and in vitro and in vivo decreased virulence [
53,
59]. The reduced virulence could be associated with a diminished initial cell adhesion with consequent reduced ability of the resistant strains to produce biofilm. Additionally, Ab249 carried a mutation in
lpsB (*241K), and Ab347 lost several genes while carrying a mutation in CarO (A19fs), which has been previously associated with biofilm production [
59].
In contrast, decreased in vitro fitness and virulence was not observed in a clinical
A. baumannii strain with a deletion of one amino acid in PmrB (ΔI19) [
58]. In another study, where clinical strains were exposed to sub-MIC concentrations of colistin, while
pmrB S17R mutant showed a slight decrease in fitness and virulence, these changes were not observed in strains with
pmrB 17_26dup or T235I [
60].
An in vitro-induced colistin-resistant derivative of
A. baumannii ATCC 19606, called RC64, showed an increased in vitro [
61] and in vivo fitness cost as well as impaired virulence [
62]. The decreased fitness in the resistant strain has been associated with the down-regulation of several proteins, including outer membrane proteins, chaperones, protein biosynthesis factors, and metabolic enzymes [
61]. The mutations R134C and A227V have been identified in the
pmrB of this strain [
62]. Low-iron conditions, such as those found in the human serum, were related with the decrease in vitro fitness [
63] of strain RC64 [
62] and also clinical colistin-resistant strains [
64] with different
pmrB mutations, which were not directly correlated with the reduced growth phenotype [
63]. Another colistin-resistant derivative of
A. baumannii ATCC 19606, with A227V
pmrB, showed decreased in vitro and in vivo fitness, as well as attenuated virulence, although this was not observed in all tested models [
65].
A slight decrease of the in vitro fitness of a colistin-resistant derivative of
A. baumannii ATCC 17978 was associated with
pmrB G272D [
46].
Colistin-resistant clinical strains recovered during colistin therapy revealed a reduced in vitro fitness as determined by the growth rates and by pairwise competitions assays with their susceptible counterparts. These strains presented mutations in
pmrB S17R, T232I, R263L, Y116H and/or
pmrA1 E8D, with the highest fitness decrease associated with co-presence of
pmrB Y116H and
pmrA1 E8D [
24,
66]. Some of the strains, collected from the same patient over the course of colistin treatment (which has its bactericidal effect due to production of hydroxyl radicals) were further studied; the study included the susceptible parental strain as well as five resistant strains. Both in vitro and in vivo assays revealed that, after an initial loss of fitness and virulence, colistin-resistant isolates progressively increase their fitness as well as virulence under oxidative stress. This study also shows that in vitro results do not necessarily correlate with the in vivo outcome [
66].
4.2. Colistin Resistance Due to Mutations in lpx Genes
Fewer studies report on the fitness and virulence of colistin-resistant
A. baumannii due to mutations in the
lpxACD locus. In one of the studies, in vitro and in vivo fitness costs were detected in the
lpx mutants, with the Δ
lpxD mutant showing the highest in vitro fitness cost, as compared with the wild-type
A. baumannii ATCC 19606. Virulence was also evaluated in A549 human alveolar cells and in a mouse model, and in
Caenorhabditis elegans, with the
lpx mutants showing decreased virulence, seen as decreased mortality rate or reduced inhibition of fertility, respectively [
65]. In another study, Wand and colleagues showed that single mutations in the
lpxA (E216*),
lpxC (I253N, F191L or A82E), or
lpxD (K318fs) genes, or inactivation of
lpxC (
lpxC::IS
Aba1), obtained after colistin exposure to clinical
A. baumannii strains, were associated with a reduced fitness and avirulence in
Galleria mellonella [
60]. As described above, these studies also evaluated the biological costs and effect on virulence associated with mutations in the
pmrB gene. Comparing the results, the influence of mutations in the fitness and virulence was more pronounced in the
lpx mutants [
60,
65].
Mu and colleagues showed that mutation
lpxA I76K and disruption of
lpxC or
lpxD by IS
Aba1 confer fitness costs to colistin-resistant derivatives of
A. baumannii ATCC 17978, but additional mutations in other genes, such as
hepA or
rsfS, contributed to an increased resistance to colistin, as well as a (partially or completely) compensated fitness cost of these mutants. Additional costs were observed when mutants were grown in serum [
46].
Table 1 summarizes the outcome of fitness and virulence assays in
A. baumannii strains with mutations in known colistin-resistance genes.