Acinetobacter baumannii Antibiotic Resistance Mechanisms

Acinetobacter baumannii is a Gram-negative ESKAPE microorganism that poses a threat to public health by causing severe and invasive (mostly nosocomial) infections linked with high mortality rates. During the last years, this pathogen displayed multidrug resistance (MDR), mainly due to extensive antibiotic abuse and poor stewardship. MDR isolates are associated with medical history of long hospitalization stays, presence of catheters, and mechanical ventilation, while immunocompromised and severely ill hosts predispose to invasive infections. Next-generation sequencing techniques have revolutionized diagnosis of severe A. baumannii infections, contributing to timely diagnosis and personalized therapeutic regimens according to the identification of the respective resistance genes. The aim of this review is to describe in detail all current knowledge on the genetic background of A. baumannii resistance mechanisms in humans as regards beta-lactams (penicillins, cephalosporins, carbapenems, monobactams, and beta-lactamase inhibitors), aminoglycosides, tetracyclines, fluoroquinolones, macrolides, lincosamides, streptogramin antibiotics, polymyxins, and others (amphenicols, oxazolidinones, rifamycins, fosfomycin, diaminopyrimidines, sulfonamides, glycopeptide, and lipopeptide antibiotics). Mechanisms of antimicrobial resistance refer mainly to regulation of antibiotic transportation through bacterial membranes, alteration of the antibiotic target site, and enzymatic modifications resulting in antibiotic neutralization. Virulence factors that may affect antibiotic susceptibility profiles and confer drug resistance are also being discussed. Reports from cases of A. baumannii coinfection with SARS-CoV-2 during the COVID-19 pandemic in terms of resistance profiles and MDR genes have been investigated.


Introduction
Antimicrobial or antibiotic resistance (AMR) has emerged as a substantial and triggering phenomenon with increasing costs for healthcare systems worldwide. In recent years it has been related to significant morbidity, mortality, and increased cost due to both prolonged length of hospitalization and treatment. Although in the last decades, there are new antimicrobial agents in our armamentarium, resistance seems to be an expanding problem with geometric evolution. Data from multicenter studies in the last decades have demonstrated that both community-acquired and nosocomial AMR are increasing alongside the increased number of older patients with primary or secondary immunodeficiencies [1,2].

Resistance to Beta-Lactams
Beta-lactams comprise of penicillins, cephalosporins, carbapenems, monobactams, and beta-lactamase inhibitors. Beta-lactam antibiotics are structurally similar to the d-Ala-d-Ala moiety of peptidoglycan, where penicillin-binding proteins bind, and thus they inhibit the transpeptidation, which is the last step in peptidoglycan synthesis [21]. A. baumannii is nowadays intrinsically resistant to penicillins and cephalosporins [22]. Resistance to beta-lactam antibiotics can be conferred through all the aforementioned mechanisms, which is inactivation by hydrolysis, increase of efflux, decrease of influx, and protection of the antibiotic target.

Beta-Lactamases
Beta-lactamases are enzymes that catalyze the hydrolysis of beta-lactam antibiotics and can be divided into four classes based on sequence motifs and differences in hydrolytic mechanism [23,24].
Other notable members of this class are the extended spectrum cefotaximases (CTX-M) and the Klebsiella pneumoniae carbapenemases (KPC) [32,33].

Class B
Class B or metallo-beta-lactamases (MBLs) are encoded by mobile DNA (plasmids, integrons) and catalyze the hydrolysis of virtually all beta-lactamases (including carbapenems), but not monobactams, conferring multidrug-resistance. These enzymes require either zinc or another heavy metal for the catalysis and are further classified in three subclasses (B1, B2, and B3) based on sequence diversity and differences in the structure of their active sites. Moreover, four types of MBLs have been described in A. baumannii, namely IMP, VIM, NDM, and SIM [34]. The detection of MBL-producing organisms with the help of phenotypic methods, such as double disc synergy test, MBL E-test, and combined disk test, has been challenging [35,36]; however, molecular methods and in particular next generation sequencing are going to shed some light on their detection [37][38][39]. Phenotypic methods are not sensitive enough and thus do not detect all MBL producing strains [36]. With the help of PCR, blaVIM-1 was detected in 14.3% of A. baumannii isolates characterized as MBL negative by E-test, highlighting the importance of introducing molecular methods into every-day practice in order to detect these hidden MBLs [40]. Recently, López et al. demonstrated that the expression of NDM lactamases does not compromise bacterial growth and is favored over other MBLs due to the lack of fitness cost leading to its worldwide dissemination among Gram-negative bacteria [41].

Class C
Class C beta-lactamases are chromosomally encoded cephalosporinases (acinetobacterderived cephalosporinase, ADC), inherent to all A. baumannii. Overexpression of these lactamases can be induced by the insertion of ISAba1 and ISAba125 sequences upstream of the encoding gene blaADC (formerly known as blaAmpC), which appear to be stronger promoters than the intrinsic promoter [42][43][44][45][46]. Insertions sequences (IS) are short transposable elements that are repeated multiple times throughout the genome, and thus are difficult to detect. An open-source bioinformatics pipeline (panISa) that utilizes whole genome sequencing (WGS) data as input has been recently developed in order to overcome this problem [47]. Several ADC variants have been described, many of which exhibit extended spectrum antibiotic resistance [48,49]. For example, ADC-30 provides resistance not only to cephalosporins, but also to carbapenems and sulbactam [50,51]. Finally, phosphoproteomic analysis revealed that dephosphorylation of ADC may lead to imipenem resistance in clinical isolates [52].

Class D
Class D beta-lactamases, also called oxicillinases (OXA) or carbapenem-hydrolyzing class D β-lactamases (CHDLs), can inactivate all beta-lactams (mainly OXA-10 family) and comprise the main mechanism of carbapenem resistance. These enzymes are serinedependent just like Class A and C beta-lactamases [53]. In addition, class D beta-lactamases usually cannot be inhibited by clavulanic acid, sulbactam, and tazobactam [54]. There are several blaOXA genes, including blaOXA-51, blaOXA-23, blaOXA-24, blaOXA-58, blaOXA-143, and blaOXA-235. The genes encoding these enzymes can be found on both the chromosome as well as the plasmids. Moreover, Wong et al. have recently confirmed that carbapenem resistance in clinical isolates of A. baumannii is mediated by over-expression of either OXA-23 or OXA-51 through insertion of ISAba1 in their promoter region [55]. With the help of molecular techniques (PCR, sequencing), reservoirs of carbapenemaseproducing A. baumannii have been isolated from companion animals and pigeons, highlighting as a result the importance of global surveillance [56,57].

Outer Membrane Proteins
Antibiotic resistance due to beta-lactamases can be significantly enhanced when they collaborate with outer membrane proteins (OMPs). The low permeability outer membrane protein A (OmpA, 40 kDa) is the main non-specific porin in A. baumannii and has mainly a structural role [58,59]. It is speculated that OmpA takes part in the transport of antibiotics out of the periplasmatic space [60]. On the contrary, Iyer et al. demonstrated that OmpA selectively enables the uptake of small molecules like sulbactam, imipenem, and ETX2514 [61]. In mutant strains lacking OmpA (∆ompA and OmpA-like domain deletion), instability of the outer membrane and increased susceptibility to antibiotics (including penicillins and cephalosporins) has been observed [58,60,62]. Finally, Zhong et al. have recently shown that OmpA C-terminal domain is also responsible for anchoring beta-lactamases (Oxa23, GES-11) in the periplasmatic space, which may further explain the increased susceptibility in case of OmpA loss [63].
Mussi et al. have vigorously studied the carbapenem-associated outer membrane protein (CarO, 29 kDa) in carbapenem sensitive and resistant A. baumannii [64][65][66][67]. They have demonstrated that CarO serves as an uptake channel for L-ornithine and possibly carbapenems [66]. In carbapenem-resistant clinical isolates of A. baumannii insertion, elements can disrupt the expression of the chromosomal CarO gene and lead to complete loss of CarO [65,68,69]. In addition, CarO gene mutations can induce significant structural modifications in CarO and thus reduce outer membrane permeability and eventually cause drug resistance [70]. However, there are also studies that question the role of CarO in carbapenem resistance and suggest that other porin-mediated mechanisms might exist [71][72][73][74].
Srinivasan et al. studied the physiological functions of another OMP, AbuO, and demonstrated its role in multidrug resistance. AbuO is involved in the active efflux of multiple antibiotics, including ceftriaxone and meropenem [75]. Moreover, decreased expression of a different 33-36 kDa OMP is associated with carbapenem resistance [76,77]. Lastly, there have been contradicting studies regarding the involvement of the outer membrane carboxylate channel (Occ; formerly OprD) family in carbapenem resistance [78,79].

Efflux Pumps
Increased expression of efflux pumps contributes synergistically with beta-lactamases to antibiotic resistance [80]. Overexpression of AdeABC efflux pump is associated with A. baumannii carbapenem and cephalosporin resistance [81][82][83]. The AdeABC is a threecomponent efflux pump, member of the resistance-nodulation-division (RND) family. AdeB component expels antibiotics out of the cell, while AdeA is a membrane fusion protein and AdeC an outer membrane protein [84]. The substrates of AdeB can be diverse; they can range from hydrophilic to hydrophobic and can be either positively charged or neutral [83]. The expression of this efflux pump is regulated by the AdeRS two-component system. Point mutations in the adeRS operon can lead to increased expression of the pump and consequently to antibiotic resistance [85,86]. It has been shown that other efflux pumps AdeIJK and AbeM may also contribute to imipenem and cephalosporin resistance [87,88].

Penicillin-Binding Proteins
Penicillin-binding proteins (PBPs) are enzymes that catalyze the polymerization of peptidoglycan and are responsible for its insertion into the cell wall [89]. Beta-lactams bind to PBPs because they mimic their substrate. Inhibition of PBPs by beta-lactams then leads to an imbalance in cell wall metabolism and as a result to cell death [90].
The role of this resistance mechanism appears to be less significant but cannot be ignored. Although mutations in hot spot regions have been observed in PBP genes, the corresponding changes in amino acid sequences were not directly related to beta-lactam resistance [91] [96]. Finally, PBP3 mutants have been associated with resistance to meropenem, sulbactam, and cefiderocol [82,[94][95][96].
In Table 1 are summarized the genes and the respective proteins that are involved in beta-lactam resistance retrieved from the pathogen detection microbial browser for identification of genetic and genomic elements (MicroBIGG-E), a database housed by National Center for Biotechnology Information (NCBI) [97].
Although AMEs remain the principal procedure by which A. baumannii evades elimination by AGs, efflux is also an emerging problem with AG use. Seven different gene products (Table 2) result in effective efflux of AGs, including pumps, permeases, periplasmic adaptors, and two component systems (TCSs) [105]. Of note, gentamicin and netilmicin are effectively cleared by AdeABC and AbeM pumps, but efflux is significantly weaker as regards more hydrophilic AGs, like amikacin and kanamycin [106]. Changes in membrane lipids and porin expression are only minor mechanisms of AG resistance and are still under investigation [99]. Recent case reports have indicated the usefulness of next-generation sequencing (NGS) in the prompt diagnosis and appropriate treatment of A. baumannii infections by early identification of resistance genes [107].

Resistance to Tetracyclines
Tetracycline antibiotics bind to the 30S ribosomal subunit and thereby inhibit protein synthesis by deterring the start of translation [108]. Resistance to tetracycline antibiotics is attributed to three main mechanisms: (i) efflux dependent on ATP, (ii) inactivation of tetracyclines by enzymes, and (iii) ribosomal protection proteins (RPPs) [109]. Table 3 describes resistance mechanisms that are unique for this antibiotic group.  Two types of efflux pumps that require energy are responsible for tetracycline resistance in A. baumannii. The resistance/nodulation/cell division (RND) family-type pumps are constitutive non-specific pumps originating from adeA, adeB, and adeC genes, which encode periplasmic adaptor subunits, permease subunits, and outer membrane pump elements, respectively [83]. RND pumps, and predominantly AdeABC, can effectively eliminate tetracyclines, while correspondingly, they mediate a substantial elevation of minimum inhibitory concentrations (MICs) for tigecyline, minocycline, and tetracycline [106]. RND pump AdeIJK seems to have a minor role in tetracycline resistance against A. baumannii, but it can act synergistically with other overexpressed efflux pumps (like AdeABC and AcrAB-TolC) and result in tigecycline resistance [115]. The second category refers to tetracycline major facilitator superfamily (MFS) efflux pumps: TetA and TetB [116]. TetA seems to lead efflux of tigecycline into the periplasm, and subsequently, RND pumps drive to elimination through the outer membrane [117].
TetM/TetW/TetO/TetS family tetracycline resistance RPPs abolish the inhibitory effect of tetracycline on protein synthesis by a non-covalent modification of the ribosomes [118]. Though rare, resistance to minocycline in A. baumannii has been attributed to ribosomal protection and tetM gene. The latter gene product mediates tetracycline release from its ribosomal binding site by a GTP-dependent mechanism, enabling the continuation of translation despite the presence of tetracycline [119].
Recent transcriptomic studies have shown that A. baumannii can swiftly become resistant to tigecycline [120]. Tigecycline has been a powerful tool against A. baumannii in the clinician's armamentarium until recently, when various plasmid-mediated tet(X) gene variants emerged. Tet(X3), Tet(X4), and Tet(X5) are monooxygenases that can inactivate all tetracyclines even tigecycline and the recently authorized eravacycline and omadacycline [121,122].

Resistance to Fluoroquinolones
Quinolones are bactericidals with a broad spectrum that are characterized by a bicyclic core formation bearing resemblance to 4-quinolone. Quinolone antibiotics are mostly fluoroquinolones displaying efficacy against both Gram-negative and Gram-positive pathogens [123]. Regarding their mechanism of action, quinolone antibiotics interrupt DNA replication by averting bacterial DNA from loosening and being cloned. Quinolones exert their action by inhibiting the ligase activity of the type II topoisomerases, DNA gyrase, and topoisomerase IV, which normally induce supercoiling in collaboration with DNA nucleases. Disrupting ligase activity, bacteria remain with double-stranded DNA breaks and thus are led to cell death. Notably, quinolones primarily affect gyrase activity, while toxicity against topoisomerase IV is secondary (i.e., there is no proof of sole parC mutations without concomitant alterations in gyrA) [123]. Quinolone resistance occurs via three different mechanisms: (i) target mutations in gyrase and topoisomerase IV, which weaken the respective quinolone-enzyme interactions; (ii) plasmid-borne resistance mediated by Qnr proteins, the AMEs AAC(6 )-Ib-cr and AAC(6 )-Ib-cr5, and by plasmid-encoded efflux pumps; and (iii) chromosome-derived resistance resulted by either low expression of porins or overexpression of chromosome-encoded efflux pumps [123][124][125]. A recent review reported resistance of A. baumannii to fluoroquinolones between 50% and 73% of cases, while the respective resistance in developing countries during the last years displayed a marked increase reaching 75% to 97.7% [98,126]. Table 4 lists all recorded quinolone-specific resistance mechanisms for A. baumannii. Quinolone resistance occurring secondary to overactive RND pumps has been long documented and is quite common [128]. Mutations in the TCS elements regarding both regulator (AdeR, primarily with polymorphisms D20N, A91V, A136V, and P116L) and sensor (AdeS, mainly due to polymorphisms G30D, A94V, G103D, G186V, and T153M) of AdeABC pump result in higher fluoroquinolone efflux. Mutations in adeR and adeS seem to trigger overexpression of the AdeABC efflux system (especially of the permease subunit adeB gene) and are associated with resistance to ciprofloxacin, norfloxacin, and ofloxacin [128]. Beyond AdeABC, efflux systems AdeIJK and AdeFGH are also utilized by A. baumannii to remove fluoroquinolones from the cell and result in high MICs. In addition, contribution of the multidrug and toxic compound extrusion (MATE) transporters AbeM and AbeS to A. baumannii resistance against quinolones is disputable. AbeM pump primarily affects the hydrophilic fluoroquinolones, norfloxacin and ciprofloxacin, rather than the hydrophobic ones (such as ofloxacin). In general, these non-RND efflux pumps confer low-level resistance to fluoroquinolones, even though some studies suggest otherwise [124,129].
Quinolone resistance-determining regions (QRDRs) refer mainly to alteration of target sites in gyrase (Ser83Leu, Gly81Asp, and Ser81Leu mutations preventing quinolones from binding its alpha-subunit) and topoisomerase IV (mutations Ser80Leu, Glu84Lys, and Gly78Cys, and Ser84Leu in its subunit C). Although a single point mutation in DNA gyrase is usually not enough for resistance to fluoroquinolones in A. baumannii (maybe only against levofloxacin; single parC mutations link with ciprofloxacin resistance), concurrent mutations within QRDR regions of the gyrA and parC genes are linked with significantly higher level of quinolone resistance [130,131]. Alterations in gyrB and parE genes are of minor significance [126].
Plasmid-mediated quinolone resistance (PMQR) has been recently identified as a clinical problem against A. baumannii infections, but generally confers still low-level (≤10fold) resistance [123,132]. Qnr genes qnrAI (the first PMQR gene to be identified), qnrB, qnrB19, and qnrS encode members of the pentapeptide-repeat protein family (homologous to McbG and MfpA proteins), which originally inhibit gyrase action through competition with DNA for binding, leading to a decrease of DNA binding with topoisomerase and thus protecting enzyme-DNA complexes from the action of quinolones [97,132]. Due to the aforementioned DNA homology, these pentapeptide repeat-containing Qnr proteins can also interact with gyrase and topoisomerase IV and prevent quinolones from cleavage, leading eventually to aggregation and subsequent accumulation of double-stranded DNA breaks that would be, in this case, lethal to A. baumannii [133]. PMQR by AMEs AAC(6 )-Ib-cr and AAC(6 )-Ib-cr5 is attributed practically to mutant variants of 6' aminoglycoside acetyltransferase-Ib gene ("-cr" implying ciprofloxacin resistance; W102R and D179Y mutations) that encode enzymes acetylating the C7 unsubstituted nitrogen in the piperazine ring of norfloxacin and ciprofloxacin [134]. The third category of plasmid-borne quinolone resistance refers to efflux pumps, but no such case has been documented for A. baumannii so far. QepA (quinolone efflux pumps A1 and A2; contributing to resistance to norfloxacin, enrofloxacin, levofloxacin, and ciprofloxacin among rmtB-positive E. Coli) and OqxAB (quinoxaline-di-N-oxide olaquindox) are the first described pumps of this category [127,133].
Chromosome-derived resistance to fluoroquinolones associates either with low influx rhythm by downregulated or dysfunctional porins, or with efflux pumps that are overactive. In fact, porins are virulence factors of Gram-negative bacteria that regulate cellular permeability through the outer membrane and are merely linked with carbapenem resistance (e.g., in Pseudomonas aeruginosa isolates with low OprD expression). Porin genes like ompA, omp25, omp33, oprC, oprD, oprW, dcap-like, and carO have been implicated in this resistance mechanism. Low relative expression of Omp25 and CarO porins seems to correlate with resistant A. baumannii strains, but resistance to quinolones has not been justified [111,135]. In the case of fluoroquinolones, chromosome-encoded efflux pumps (described in Table 2) and porin alterations alone do not seem to render significant clinical resistance to A. baumannii [123,129].

Resistance to Macrolides-Lincosamides-Streptogramin Antibiotics
Macrolide antibiotics are of little use in A. baumannii infections. Azithromycin, but no other macrolide, seems to inhibit mucin production, suggesting efficacy against ventilatorassociated pneumonia. Azithromycin for nosocomial pneumonia due to A. baumannii for pediatric and adult patients in intensive care units (ICUs) is used in combination with other antibiotics. In the latter case, azithromycin's mechanism of action differentiates from its established way of action (inhibiting translation and protein biosynthesis by attaching to the 50S subunit): suppression of ERK/JNK pathway phosphorylation (between extracellular signal-regulated kinases and c-Jun N-terminal kinases) and nuclear translocation of NF-κB (nuclear factor kappa-light-chain-enhancer of activated B cells) [136]. According to MicroBIGG-E database, resistance to macrolides in A. baumannii is attributed to (i) three 23S rRNA (adenine(2058)-N(6))-methyltransferases, encoded by erm(B), erm(C), and erm(F) genes), (ii) ABC-F type ribosomal protection protein Msr(E) or msr(E), and (iii) two macrolide 2'-phosphotransferases encoded by mph(A) and mph(E) [97]. The first two classes lead to resistance by modification of the target site, while the third class results in macrolide inactivation. In particular, Mph(A) and Mph(E) seem to confer resistance to erythromycin, clarithromycin, azithromycin, and oleandomycin, but only in the presence of specific regulatory proteins [137].
Additional literature search revealed several references to efflux pumps involved in macrolide resistance in A. baumannii infections. WGS confirmed recently the potential role of multidrug efflux MFS transporter AmvA (encoded by amvA) in erythromycin resistance [138,139]. Another WGS study identified transferable genetic elements in macrolideresistant A. baumannii strains encoding the proton motive macrolide efflux MFS transporter and ribosomal protection protein mef(E)/mel (originally found in resistant streptococci) in an operon of MEGA-element (a DNA fragment named after the macrolide efflux genetic assembly). Unlike other mechanisms discussed above, MefE pumps do not result in higher MICs for lincosamides or streptogramins [140][141][142]. The tripartite MacA-MacB-TolC transporter in A. baumannii intersects inner and outer membranes so as to actively extrude macrolides. Main components of this transmembrane machine are the ATPase MacB, belonging to the ABC superfamily, and the membrane fusion protein MacA, conferring synergistically resistance to azithromycin and roxithromycin [143]. In addition, detection of the SMR family transporter AbeS (homologue of E. Coli EmrE) links with erythromycin resistance [144]. Another pump involved in macrolide-resistant A. baumannii is AdeABC efflux pump, regulated by base substitutions in the adeRS operon. Conversely, inactivation of the latter operon negatively affects biofilm formation and prompts decreased expression of AdeABC [145].
Lincosamides compounds lincomycin, clindamycin, and pirlimycin are of limited use as single antimicrobial agents in A. baumannii infections. Their mechanism of action resembles that of macrolides, by binding the ribosomal 50S subunit and specifically its 23S portion to a site corresponding to the peptidyl-transferase center. Consequently, plasmid-and transposon-derived 23S rRNA methyltransferases belonging to the erm family (erythromycin ribosome methylase) elicit combined resistance to macrolides, lincosamides, and streptogramins B (i.e., MLSB phenotype) [146]. Although resistance to lincosamides can occur through ribosomal modification, efflux, and drug inactivation, reports from A. baumannii strains are limited to efflux pumps. RND efflux pumps AdeABC, AdeFGH, and AdeIJK have been documented to significantly increase MICs of all lincosamides in several A. baumannii isolates [129,[147][148][149]. Far from pumps, WGS revealed the link between ABC transporter Msr(E) and macrolide phosphotransferase mph(E) with MLSB phenotype [150][151][152].

Resistance to Polymyxins
As mentioned before, Gram-negative bacteria, such as A. baumannii, bear a semipermeable outer membrane for insertion of essential elements and clearance of toxic compounds. Lipopolysaccharides (LPSs) reside on the outer surface and contain a negatively charged hydrophobic lipid A, which in turn interacts with the cationic non-ribosomal lipopeptides of polymyxins B and E (widely known as colistin). This interaction results in destabilization of the outer membrane, uptake of the polymyxins into the periplasm, and increased permeability by disrupting the integrity of both outer and inner membranes. Although the detailed mechanism of action is unknown, the hydrophobic tail of polymyxins seems to be crucial for the induction of membrane damage, suggesting a detergent-like mode of action [153].
Mechanisms of resistance to polymyxins in A. baumannii include (i) drug target alteration by LPS lipid A modification subsequent to mutations in the pmrCAB operon and mcr genes; (ii) mutations of lpxA, lpxC, and lpxD genes -encoding acyltransferases essential to lipid A biosynthesis and associated with lipid A deficiency; (iii) lpsB, lptD, and vacJ expression associated with permeability defects and osmotic resistance of the outer membrane, subsequently leading to markedly elevated MICs for polymyxins; (iv) insufficient concentration of cofactors constitutional for LPS formation, like biotin, which are essential for susceptibility to polymyxins; and (v) efflux pumps [153][154][155][156].
Operon pmrCAB includes pmrC gene encoding a phosphoethanolamine (PEA) transferase, along with pmrA and pmrB that encode the PmrA/PmrB TCS. Mutations in the latter TCS (and especially of pmrA) induce the overexpression of pmrC. PEA residue addition to the 4 -phosphate group site of lipid A hepta-acylated form is the most commonly reported modification in polymyxin-resistant A. baumannii, resulting in removal of negative charges and thus lowering the affinity of LPS to this drug class. The addition of 4-amino-L-arabinose and/or galactosamine to the phosphate groups of lipid A and to residues within the core oligosaccharide acts similarly. Likewise, PmrA/PmrB TCS can lead to polymyxin resistance by upregulating the transcription of NaxD deacetylase and therefore by modifying LPS lipid A by means of the deacetylated β-galactosamine. In addition, mutations in miaA gene (encoding the posttranscriptional regulator tRNA dimethylallyl diphosphate transferase) seem to act synergistically with mutations in pmrA towards producing polymyxin-resistant phenotypes [157,158]. In the same context, mutations in the pmrC paralogues eptA-1 and eptA-2 have been associated with increased expression and colistin resistance [153,159]. Other PEA transferases that affiliate with polymyxin-resistance are encoded by mcr-1, mcr-4, and mcr-4.3 genes. Noteworthily, resistance emerging from MCR PEA transferases that was thought to be originally chromosomal (and therefore limiting its rapid distribution and dissemination) has been recently reported to be also carried by plasmids [160][161][162].
Biotin, which is essentially involved in fatty acid synthesis, contributes in a dosedependent manner to LPS synthesis. Deletions in lpsB, a gene encoding a glycosyltransferase involved in biotin synthesis, have been associated with colistin resistance [163].
A recent WGS study on colistin-resistant isolates of A. baumannii identified linkage with vacJ, zndP, pldA, ttg2C, and pheS genes. VacJ is linked with the Vps-VacJ ABC transporter system, which is in charge of preserving LPS and phospholipids at the outer and inner leaflet of the outer membrane respectively. Correspondingly, PldA is designated to remove phospholipids in the outer leaflet of the outer membrane, while ZndP is a zinc-dependent peptidase A, upstream of PldA, that also has a pivotal role in the outer membrane processing. Ttg2C is an efflux ABC transporter upregulated in response to phenol exposure (an organic solvent that solubilizes the cell wall, acting equivalently with polymyxins). Moreover, mutations in pheS that encodes phenylalanine-tRNA ligase subunit alpha seem to play a role in colistin resistance [164]. Beyond the two aforementioned ABC transporters, and as stated before in text, overexpressed RND efflux pump MexAB-OprM correlates with colistin resistance [156,165].
Although decreased membrane permeability has been mainly attributed to carbapenemresistance in A. baumannii, low expression of porin OmpA has been associated with chloramphenicol, aztreonam, and nalidixic acid resistance [60]. Another gene, abrp, encoding a C13 family peptidase has been linked with decreased cell membrane permeability, faster cell growth rate, and decreased susceptibility to chloramphenicol, tetracycline, minocycline, doxycycline, tigecycline, and fosfomycin [170]. Conversely, overexpression of global regulator SoxR led to increased susceptibility to chloramphenicol, by downregulating abeS, abeM, adeJ, and adeG, although adeB and craA expression remained intact [171]. Additionally, deletion of valine-glycine repeat G (vgrG gene), a component of the Type VI Secretion System (T6SS), which is considered as a crucial virulence factor of A. baumannii, led to reduced resistance to chloramphenicol [172].
Rather bizarre combinations of colistin with anti-Gram-positive antibiotics, like the oxazolidinone linezolid, have demonstrated efficacy and have been utilized in cases of MDR A. baumannii [173]. Protein biosynthesis inhibition is mediated by different mechanisms as regards oxazolidinones, chloramphenicol, and lincosamides, but their target sites on the 23S ribosomal RNA subunit may be overlapping. The most prevalent mechanisms for linezolid resistance are multidrug efflux pumps, base substitutions in domain V of the ribosomal RNA 23S (conferring resistance to oxazolidinones, macrolides, lincosamides, streptogramins, phenicols, pleuromutilins, and glycopeptide antibiotics), and/or the presence of a transmissible Cfr(B) 23S ribosomal RNA methyltransferase (conferring resistance to oxazolidinones, streptogramins, phenicols, and lincosamides). Interestingly, plenty of microorganisms carry multiple gene copies that encode the 23S rRNA subunit, while the observed linezolid resistance corresponds to the ratio of wild type to mutant 23S rRNA [156,174]. According to the Comprehensive Antibiotic Resistance Database (CARD), resistance to linezolid can also occur by mutations in the 50S ribosomal subunit P-site (which attaches the peptidyl-tRNA to the developing polypeptide chain), by the MFS antibiotic efflux pump LmrS (capable of extruding oxazolidinones, macrolides, phenicols, aminoglycosides, and diaminopyrimidine antibiotics), ClcD (a cfr-like 23S ribosomal RNA methyltransferase linked with resistance to oxazolidinones, lincosamides, streptogramins, phenicols, and pleuromutilins), and ABC-F ATP-binding cassette ribosomal protection protein encoded by poxtA (conferring resistance to oxazolidinones, tetracyclines, macrolide antibiotic, lincosamides, streptogramins, phenicols, and pleuromutilins) [156].

Resistance to Glycopeptide and Lipopetide Antibiotics
Zeocin is a formulation of phleomycin D1, a glycopeptide antibiotic of the bleomycin family. Resistance to zeocin is linked with phleomycin/bleomycin binding protein Ble-Sh (ble-Sh gene), while resistance to bleomycin is usually conferred by the binding protein Ble-MBL (metallo-beta-lactamase-associated ble gene) [175][176][177].
The outer membrane of Gram-negative bacteria is impenetrable to large glycopeptide molecules, and therefore A. baumannii displays intrinsic resistance to vancomycin. However, VanD-type vancomycin resistance histidine kinase VanS (vanS-D gene) has been reported in A. baumannii infection [97]. Contrariwise, colistin combination with glycopeptides (vancomycin, teicoplanin, and telavancin) and lipopeptides (like daptomycin) has been successfully employed in MDR A. baumannii infections [178]. Studies on glycopeptide-and lipopeptide-resistant A. baumannii isolates are lacking.

Resistance to Rifamycins
Resistance to rifampin (also known as rifabicin) in A. baumannii infections has been linked with mutations in the rpoB gene, which encodes rifamycin sensitive beta-subunit of RNA polymerase and averts RNA elongation just after adding the first nucleotides. Beyond rifampin, RpoB associates with resistance to all rifamycins (rifabutin, rifaximin, and rifapentine) [179]. Enzymatic modification by rifampin ADP-ribosyltransferases Arr, Arr-2, Arr-3, and Arr-4 impels to inactivation of rifamycins' 23-OH position using NAD+ [97,180]. As mentioned above in text, RND multidrug efflux pumps AdeIJK and AcrAB-TolC are also potential mechanisms of rifamycin-resistant isolates [156]. Conversely, fhuE gene encodes an outer membrane protein transporter that is involved with iron acquisition and is upregulated in low iron conditions and seems to associate with increased susceptibility to rifabutin (200-fold more potent than rifampin). The previous finding was confirmed by successful in vivo treatment of extreme drug resistant A. baumannii with rifabutin/colistin combination [181].

Resistance to Fosfomycin
Intrinsic fosfomycin resistance in A. baumannii occurs predominantly as a result of two mechanisms: (i) fosfomycin efflux MFS transporter AbaF encoded (abaF gene); (ii) fosfomycin resistance glutathione transferases FosLL, FosA3, and FosA3/FosA4 family (encoded by fos, fosA3, and fosA, respectively), and FosB1/FosB3 family fosfomycin resistance bacillithiol transferase (encoded by fosB) [97,182]. Disruption of abaF has displayed an increase in fosfomycin susceptibility and a decrease in biofilm formation and virulence, suggesting a major role for this pump [183]. Of note, FosA impels the binding of glutathione to Fosfomycin C1 and thereby renders it inactive, while FosB opens enzymatically the epoxide rind of Fosfomycin after employing either bacillithiol or L-cysteine [156].
Fosfomycin sensitive pyruvyl transferase MurA has been recently proposed as a potential drug target for MDR A. baumannii. This transferase, whih has no counterparts in eukaryotes, catalyzes the pathway of peptidoglycan biosynthesis and is indispensable for cell integrity, while it is highly sensitive to inhibition from fosfomycin [184].

Resistance to Diaminopyrimidines-Sulfonamides
Antifolate antibiotics exert their action by inhibiting purine metabolism, and thereby DNA and RNA synthesis. Trimethoprim is a dihydrofolate reductase (DHFR) inhibitor (blocking tetrahydrofolic acid formation by dihydrofolic acid, an essential step of folate biosynthesis, which in turn is involved in the generation of many nucleotides and amino acids), while sulfonamides are known dihydropteroate synthase (DHPS) inhibitors (halting the conversion of para-aminobenzoate or PABA to dihydropteroate; including sulfamethoxazole, commonly used in combination with trimethoprim, and sulfonamides) [185]. Resistance against diaminopyrimidines in A. baumannii infections is mainly conferred by trimethoprim-resistant dihydrofolate reductases DfrA1, DfrA5, DfrA7, DfrA10, DfrA12, DfrA14, DfrA16, DfrA17, DfrA19, DfrA20, DfrA27, and DfrB1 (encoded by dfrA1, dfrA5, dfrA7, dfrA10, dfrA12, dfrA14, dfrA16, dfrA17, dfrA19, dfrA20, dfrA27, and dfrB1, respectively) [97]. A recent study of A. baumannii in turkey and chicken raw meat revealed that prevalence of dfrA was as high as 71% [186]. The latter antibiotic resistance gene seems to be a typical part of cassettes belonging to class I integron [187]. RND efflux pump MexAB-OprM consists of MexA (the membrane fusion protein), MexB (the carrier located in the inner membrane), and OprM (the transporter positioned in the outer membrane). MexAB-OprM associates with lack of susceptibility to most antibiotics, including trimethoprim and even colistin, when expressed in high levels [156,165]. As mentioned above in text, efflux pumps AdeIJK and LmrS confer resistance to trimethoprim [148]. Another pump, the RND plasmid-borne OqxAB, links with resistance to trimethoprim, quinolones, tetracyclines, glycylcycline, and nitrofurans [188,189].
In general, A. baumannii is intrinsically resistant to sulfonamides in approximately 71.3% of isolates [98]. Resistance, in these cases, is being materialized through two main mechanisms: (i) sulfonamide-resistant dihydropteroate synthases encoded by sul1 and sul2, and (ii) RND efflux pump MexAB-OprM with or without MexR mutations [156,190].

SARS-CoV-2 and Resistant A. baumannii Coinfections
During the COVID-19 pandemic, coinfection with A. baumannii secondary to SARS-CoV-2 infections has been reported multiple times in literature. An isolate with OXA-23 has been responsible for an outbreak in COVID-19 ICUs of a tertiary Japanese hospital [191]. Incidence of secondary infections (mainly lower respiratory tract infections) attributed to A. baumannii has been reported to be as high as 1% of hospitalized COVID-19 patients in an Italian hospital [192]. The same incidence (1%) has been reported by an early study in hospitalized patients from Wuhan, China [193]. A concurrent study from Wuhan reported A. baumannii coinfection in one out of 69 hospitalized patients (1.4%) with COVID-19 [194]. In addition, a recent study from a French ICU calculated coinfection with A. baumannii at 1.1% (1 out of 92; susceptible to third generation cephalosporins) in severe SARS-CoV-2 pneumonia patients [195]. Higher incidence of MDR A. baumannii coinfection has been documented in a study from Egypt (2.7%; seven out of 260 COVID-19 hospitalized patients; susceptible only to tigecycline and fluoroquinolones; bearing resistance genes NDM-1, TEM, and CTX-M) [196]. A recent study in hospitalized COVID-19 patients from Spain showed that coinfection with A. baumannii was apparent in 2.4% of hospitalized patients (17 out of 712; 16 out of these 17 patients in ICU) and it has been the strongest determinant of mortality with OR 9.329 (95% CI: 2.289 to 38.02; p = 0.002; higher than sex, bacteriemia, and number of comorbidities) [197]. A study from an Iranian ICU reported coinfection with MDR A. baumannii in 17 out of 19 COVID-19 patients (89.5%; no MBL-producing strain; 52% resistance rate to colistin), while all these patients died [198]. A. baumannii was detected in 20% of samples acquired from COVID-19 patients in an ICU from Beijing, China. All of these cases were identified during late ICU admission [199]. Of interest, a recent study from a Mexican ICU assigned to COVID-19 patients revealed clonal dispersion of MDR AdeABCRS+ A. baumannii that was only susceptible to gentamicin, nitrofurans, and phenicols [200]. Steroids and long ICU stays, but not MDR infections, have been associated with higher mortality in an Italian retrospective analysis of 32 COVID-19 ICU patients. In the latter cohort, four patients have been coinfected with MDR A. baumannii (12.5%) [201]. CRAB outbreaks in COVID-19 patients seem to be effectively treated with a pre-optimized two-phage cocktail [202]. This finding is important, because OXA-23 CRAB infections have become a serious threat against SARS-CoV-2 infected patients in ICUs [203].

Conclusions
As justified above in text, A. baumannii may acquire antibiotic resistance through several distinct mechanisms: by altering the antibiotic target site, by controlling the passage of antibiotics through its membranes, and by enzymatic modification of antibiotics, rendering them neutralized. Secondary to innate mechanisms of antibiotic resistance that are de facto conferred by genes, A. baumannii may facilitate antibiotic resistance through various mechanisms linked with its virulence: outer membrane proteins (like porins), cell envelope factors (like LPS and the capsule around its bacterial surface), specific enzymes (like phospholipases C and D, and glycan-specific adamalysin-like protease CpaA), quorum sensing, and biofilm formation (BfmRS TCS regulating Csu pili, Csu expression regulated by the GacSA TCS, biofilm-associated proteins BapAb, synthesis of the exopolysaccharide polyβ-1,6-N-acetylglucosamine PNAG, acyl-homoserine lactones through AbaR receptor, and AbaI autoinducer synthase), by attaining twitching motility via type IV pili, micronutrient acquisition systems (like siderophores and iron transporters FecA and FecI, ZnuABC transporter, and ZigA GTPase incorporating a zinc-scavenging system, resistance-associated macrophage protein NRAMP for manganese transportation), type II (with CpaA and lipases LipA and LipH as effectors), and type VI protein secretion systems [1].
Utilization of next-generation sequencing techniques has helped clinicians to decipher the molecular mechanisms of antibiotic resistance in PDR isolates of A. baumannii, and WGS has emerged as a powerful tool in the clinician's armamentarium. In the future, application of WGS and other NGS techniques at diagnosis could provide useful insights on the microbiologic behavior and virulence of each case of critical A. baumannii infection. Apart from timely diagnosis, detailed mapping of the genetic background of each isolate could provide useful information as regards antibiotic treatment (in terms of precision medicine) and identification of novel therapeutic targets [102,107,204].
Of special note is the contribution of molecular methods, such as PCR and NGS, to the prompt identification of MDR A. baumannii epidemics. Specifically, the detection of integrase genes (intI1, intI2) is significantly associated with the epidemic character of A. baumannii as well as antibiotic resistance to multiple antimicrobials [205][206][207]. In the Netherlands, integrons are found in 43% of MDR and 85.7% of XDR A. baumannii strains [208], while in China and Iran, the prevalence of integrase genes in total A. baumannii isolates is higher and reaches 69.6% and 94.3%, respectively [209,210]. This correlation of integrase gene presence and antibiotic resistance highlights the importance of regular surveillance in order to prevent outbreaks of hospital acquired MDR A. baumannii infections.
Except from specific clinical scenarios, MDR isolates are usually treated successfully with synergistic therapeutic combinations with either beta-lactamase inhibitor sulbactam or colistin, while last-line schemes involve combination of colistin with rifampin or polymyxin B with tigecycline. In the first case, we aim for inhibiting bacterial DNA-dependent RNA polymerase along with disruption of the bacterial membranes, while in the latter case, we aim for inhibiting bacterial protein biosynthesis (by attaching to the 30S rRNA subunit) in combination with loss of the bacterial membranes' integrity. Of course, tigecycline remains the first choice for MDR strains isolated in the setting of the ICU, while phages (like Bϕ-C62) are an alternative against increased antibiotic resistance in MDR isolates [98]. Clinical experience has confirmed efficacy of trimethoprim/sulfamethoxazole against carbapenemresistant isolates, while isolates resistant to both carbapenems and sulbactam are preferred to be treated with either minocycline or doxycycline [98,211]. XDR and PDR strains remain a serious clinical problem and identification of resistance mechanisms of this staggering microbe will elucidate pathogenetic mechanisms and propose new therapeutic targets and agents.

Conflicts of Interest:
The authors declare no conflict of interest.