Differential development of antibiotic resistance and virulence between Acinetobacter species

ABSTRACT

The two species that account for most cases of Acinetobacter-associated bacteremia in the United Kingdom are Acinetobacter lwoffii, often a commensal but also an emerging pathogen, and Acinetobacter baumannii, a well-known antibiotic-resistant species. While these species both cause similar types of human infection and occupy the same niche, A. lwoffii (unlike A. baumannii) has thus far remained susceptible to antibiotics. Comparatively little is known about the biology of A. lwoffii, and this is the largest study on it conducted to date, providing valuable insights into its behaviour and potential threat to human health. This study aimed to explain the antibiotic susceptibility, virulence, and fundamental biological differences between these two species. The relative susceptibility of A. lwoffii was explained as it encoded fewer antibiotic resistance and efflux pump genes than A. baumannii (9 and 30, respectively). While both species had markers of horizontal gene transfer, A. lwoffii encoded more DNA defense systems and harbored a far more restricted range of plasmids. Furthermore, A. lwoffii displayed a reduced ability to select for antibiotic resistance mutations, form biofilm, and infect both in vivo and in in vitro models of infection. This study suggests that the emerging pathogen A. lwoffii has remained susceptible to antibiotics because mechanisms exist to make it highly selective about the DNA it acquires, and we hypothesize that the fact that it only harbors a single RND system restricts the ability to select for resistance mutations. This provides valuable insights into how development of resistance can be constrained in Gram-negative bacteria.

IMPORTANCE

Acinetobacter lwoffii is often a harmless commensal but is also an emerging pathogen and is the most common cause of Acinetobacter-derived bloodstream infections in England and Wales. In contrast to the well-studied and often highly drug-resistant A. baumannii, A. lwoffii has remained susceptible to antibiotics. This study explains why this organism has not evolved resistance to antibiotics. These new insights are important to understand why and how some species develop antibiotic resistance, while others do not, and could inform future novel treatment strategies.

INTRODUCTION

Acinetobacter spp. are Gram-negative, soil-dwelling Gammaproteobacteria. Despite being typically found in soil, some species within the genus also cause life-threatening human infections (1). The most clinically significant of these is Acinetobacter baumannii, which is often highly multidrug resistant (2, 3).

According to United Kingdom Health Security Agency (UKHSA), in England, the most common cause of Acinetobacter-derived bacteremia is Acinetobacter lwoffii followed by A. baumannii (30% and 21%, respectively) (4). In intensive care units, the rate of Acinetobacter infections varies from 19.2% in Asia to 3.7% in North America, although species-level epidemiology data are uncommon (5). A. lwoffii is found both in soil environments and as a common commensal of human skin (6). As well as causing bacteremia in adults, A. lwoffii can cause a variety of infections, often in immunocompromised hosts, and is a common cause of serious neonatal infections, which can lead to sepsis (7–10).

Both A. lwoffii and A. baumannii are found in hospitals and are resistant to desiccation, irradiation, and biocides (11, 12). However, A. lwoffii is generally antibiotic susceptible, in contrast to the multidrug resistance displayed by A. baumannii (4). There are few studies aimed at understanding A. lwoffii, and the reasons for its comparative sensitivity are not known.

We recently showed that the number of resistance nodulation division (RND) pumps present across the Acinetobacter genus varies and that A. lwoffii encodes fewer efflux pumps from the RND family than A. baumannii (1). These efflux pumps are important mediators of antibiotic resistance, suggesting that their absence may contribute to the difference in susceptibility to antibiotics (13). RND pumps have also been implicated in virulence and biofilm formation (14, 15).

In this study, we investigated the genomic and phenotypic differences between a range of A. baumannii and A. lwoffii strains (including clinical and type strains) to understand why two closely related species have such different responses to antibiotics. This study provides insight into the development of antibiotic resistance and differences in biology and virulence in two clinically important pathogens.

MATERIALS AND METHODS

Strains used in this study

Reference strains of A. baumannii AYE and A. lwoffii NCTC 5867 were used. In addition, representative clinical and non-clinical strains that were available to us were used in this study (listed in Table S1). All strains were cultured in lysogeny broth (LB) (Sigma) unless stated otherwise at 37°C.

Measurement of the susceptibility to antimicrobials

The minimum inhibitory concentration (MIC) of various antimicrobials to A. baumannii and A. lwoffii was determined using the agar dilution method (16) according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) (17). Antimicrobials tested included ampicillin (Sigma #A9393), cefotaxime (Fisher #10084487), chloramphenicol (Fisher #10368030), ciprofloxacin (Fisher #13531640), clindamycin (Generon #A10227), erythromycin (Fisher #10338080), fusidic acid (Sigma #F0881), gentamicin (Fisher #10224873), meropenem (TCI Chemicals #M2279), novobiocin (Fisher #15403619), rifampicin (Fisher #10533325), and tetracycline (Fisher #10460264).

Biofilm formation and susceptibility

The ability of A. baumannii and A. lwoffii to establish monospecies biofilms and the susceptibility of these biofilms to different compounds were tested. The full methods can be found in supplemental Text S1.

Whole-genome sequence analysis

All available A. lwoffii and A. baumannii whole-genome sequences were downloaded from the National Center for Biotechnology Information (NCBI) (41 and 6,127, respectively) on 20 March 2022. In addition, laboratory strains of both A. baumannii (10) and A. lwoffii (8) were whole genome sequenced using Illumina platforms and assembled using SPAdes (18) (MicrobesNG, UK). A list of strains sequenced in this study and their assembly accession numbers can be found in Table S2.

Quast (v.5.0.2) was used to quality check (QC) sequences and those with N50 values of <30,000 and >165 Ns per kbp were removed (19). FastANI (v.1.31) was used to determine average nucleotide identity of A. baumannii sequences to A. baumannii AYE (CU459141.1) and A. lwoffii sequences to A. lwoffii 5867 (GCA_900444925.1), and only sequences >95.5% were kept (20). MASH (v.2.2.2) (21) was also performed to identify any duplicate assemblies, which were then removed using a custom R script (https://github.com/C-Connor/MashDistDeReplication/blob/master/MashDistDeReplication.R). The final quality step was CheckM (v.1.1.3) (22), where sequences with >5% contamination and/or <95% completeness were removed. The final numbers of A. baumannii and A. lwoffii sequences were 4,809 and 38, respectively.

Assemblies were searched for antibiotic resistance genes (ARGs) [Comprehensive Antibiotic Resistance Database (23)], type IV pilus genes [“twitching” database using reference (24)], plasmid rep genes [database from reference (25)], and virulence and biofilm genes [“vandb” database using reference (26)] using ABRicate (v.0.8.13). The twitching and vandb databases can be found at https://github.com/emd803/Gene-Databases/tree/main. Prophages were identified in 10 random isolates of A. baumannii and A. lwoffii using PHASTER, and DNA defense systems were searched in all the genomes using DefenseFinder (v.1.0.9) (27, 28).

Selection for resistance to meropenem, ciprofloxacin, and gentamicin

To determine if A. baumannii (AB18) and A. lwoffii (AL28) could evolve resistance to three clinically relevant drugs, a selection experiment was set up using strains clinically susceptible to all three selection antibiotics. Briefly, a single colony was inoculated into 5 mL of nutrient broth (Sigma), and a 1% transfer was passaged every 24 hours in increasing concentrations of each drug or without drug as a control. Drug concentrations started at one-fourth of the MIC for the organism passaged, then increased to one-half of the MIC, MIC and 2× MIC by day 7. Populations from the terminal passage were spread onto LB agar, and individual colonies were tested for their susceptibility to antibiotics listed above, as well as moxifloxacin (Sigma #PHR1542) and ethidium bromide (Fisher #10042120). Following selection, five colonies from parental strains AL28 and AB18 were subjected to whole-genome sequencing (MicrobesNG), along with two colonies that had been passaged in nutrient broth only. Resulting sequences were compared to the appropriate parental strain.

Each whole-genome sequence was confirmed to be from the species expected using an average nucleotide identity (ANI) score above (>95%), and sequences were compared to both the ancestral strain and the cells passaged in nutrient broth only, using Snippy (v.4.6.0) to find sequence variants (29).

Measurement of twitching motility and growth

A previously described crystal violet assay was used to measure twitching motility in A. baumannii and A. lwoffii (30). Additionally, growth in LB and human serum (Merck #H4522) was measured. Full methods are listed in supplemental Text S2.

Scanning electron microscopy

Strains were grown overnight in LB, then diluted 1:50 for A. baumannii and 1:10 for A. lwoffii in LB because A. lwoffii grows to a lower final cell density than A. baumannii. Strains were grown to mid-log, washed with phosphate-buffered saline (Merck #D8537), and then resuspended in 2.5% glutaraldehyde (Sigma #354400) to fix. Cells were imaged on an Apreo 2 Scanning Electron Microscope (Thermo Fisher) at ×5,000, ×10,000, and ×25,000 magnification. Cell length analysis was performed in ImageJ (31), where the lengths of 100 randomly selected cells from each strain were measured.

Virulence in the Galleria mellonella model

Galleria mellonella larvae were injected (n = 10 larvae per condition, which was independently repeated four times) with 10⁶ bacterial cells as previously described (32), and the number of live/dead larvae was quantified across 7 days.

Comparing the virulence in a macrophage cell line in vitro

Human monocyte THP-1 cell line (American Type Culture Collection TIB-202) was cultured in Roswell Park Memorial Institute Medium with GlutaMAX (Thermo Fisher #61870–010) supplemented with 10% heat-inactivated fetal bovine serum (Life Technologies, #A5256701) at 37°C and 5% CO₂. THP-1 monocytes were differentiated to macrophages with medium containing 50 ng/mL phorbol 12-myristate 13-acetate (PMA) (Sigma #P1585) for 3 days. Cells were then left to rest for 2 days by replacing the differentiation medium with complete medium without PMA. Macrophages were infected as previously described (33), with a multiplicity of infection of 100. Extracellular bacteria were killed after 2 hours using gentamicin at either 100 µg/mL or at 1 mg/mL for AB05. Association, invasion, and proliferation (after 6 hours) were quantified. Association was determined by subtracting the number of intracellular bacteria (invasion) from the total number of bacteria associated with macrophages (and within macrophages).

RESULTS

A. lwoffii is more susceptible to a broad range of antibiotics than A. baumannii

Data from the UKHSA shows that A. lwoffii sp. isolated from patients in England were more susceptible than A. baumannii to gentamicin, ciprofloxacin, meropenem, and colistin (4). Therefore, we sought to determine if the same was true in our diverse strain collection for a range of antibiotics from different drug classes (Table 1). MICs were higher for A. baumannii than for A. lwoffii for all compounds tested. EUCAST resistance breakpoints were only available for ciprofloxacin (>1 µg/mL), meropenem (>2 µg/mL), and gentamicin (>4 µg/mL) (17). A. lwoffii was clinically susceptible in all instances, whereas for A. baumannii, all but one isolate was resistant to ciprofloxacin; three of six strains were resistant to meropenem; and all were resistant to gentamicin.

TABLE 1.

MIC values for A. baumannii and A. lwoffii (μg/mL)^a

Strain	AMP	CEF	CHL	CIP	CLI	ERY	FUS	GEN	MER	NOV	RIF	TET
AB05	512	>32	256	>32	128	32	128	1,024	0.5	16	16	128
AB18	64	16	128	1	64	16	128	4	0.25	8	4	1
AB19	8	32	128	>32	64	32	128	16	0.25	8	4	8
AB20	32	16	128	2	64	64	64	4	>16	16	4	>128
AB25	1,024	>32	256	>32	128	64	128	128	8	32	4	>128
AB27	1,024	16	128	>32	32	8	64	1,024	16	8	4	>128
AL04	<1	2	2	0.06	2	<0.5	16	<2	0.03	8	0.5	0.5
AL28	<1	1	1	0.06	1	0.5	4	<2	0.12	4	0.5	0.25
AL29	<1	2	1	0.06	2	0.5	8	<2	0.03	8	0.5	0.25
AL32	<1	1	1	0.03	4	0.5	8	<2	0.12	8	0.5	0.25
AL33	<1	1	1	0.06	4	0.5	8	<2	0.12	8	0.5	0.25

^a AMP, ampicillin; CEF, cefotaxime; CHL, chloramphenicol; CIP, ciprofloxacin; CLI, clindamycin; ERY, erythromycin; FUS, fusidic acid; GEN, gentamicin; MER, meropenem; NOV, novobiocin; RIF, rifampicin; TET, tetracycline.

A. lwoffii carries fewer ARGs than A. baumannii

To explain the differences in antibiotic sensitivity between A. lwoffii and A. baumannii, whole-genome sequences were searched for the presence of ARGs using the CARD database. Following QC, there were 4,809 A. baumannii and 38 A. lwoffii genome sequences. Across the A. lwoffii genomes, 40 different ARGs were found, while 333 different ARGs were detected across A. baumannii. Due to the lack of available sequences for A. lwoffii, to quantitatively compare the presence of ARGs between the two species, a random permutation was conducted, which subsampled 38 sequences (the same number as the population of A. lwoffii sequences) from the A. baumannii population 100 times to create an average. A. baumannii encodes significantly more ARGs than A. lwoffii (P < 0.0001); the mean number of ARGs in A. lwoffii was 9 but was 30 in A. baumannii (Fig. 1a).

Fig 1.

A. baumannii encodes more ARGs than A. lwoffii. (a) Number of ARGs found per whole-genome sequence from either species. A. baumannii in pink, n = 4,809, A. lwoffii in yellow, n = 38. A random permutation and Welch’s t-test was performed to compare the average number of genes when the sample sizes were the same. ****P < 0.0001. (b) Stacked bar chart showing drug classes targeted by all antibiotic resistance genes found in A. lwoffii and A. baumannii whole-genome sequences. Only 40 different ARGs were found for A. lwoffii, whereas 333 different ARGs were found across A. baumannii, but this is likely explained by the different data set sizes of either species. EP, efflux pump-associated genes

Although there was a difference in total gene presence, the classes of antibiotics that the ARGs were active against were similar across the two species (Fig. 1b). The majority of ARGs (>50%) found in A. lwoffii and A. baumannii reduce the host’s susceptibility to beta lactams and aminoglycosides.

A. lwoffii and A. baumannii possess similar genomic signatures of horizontal gene transfer, but A. lwoffii contains more DNA defense systems

The greater antibiotic resistance levels of A. baumannii are seemingly explained by the fact that this species harbors significantly more ARGs than A. lwoffii. However, both species inhabit similar niches, cause similar types of infection, and therefore are expected to have been exposed to similar antibiotics. Variation in rates of horizontal gene transfer into and within each species might explain the difference in the numbers of ARGs they carry. To investigate this, the presence of prophage and plasmid-associated sequences, type IV pili genes for natural transformation, and the presence of DNA defense systems, which would limit the acquisition of foreign DNA, were searched for in the whole-genome sequences.

To determine whether A. baumannii and A. lwoffii harbor different numbers or types of plasmids, ABRicate was used to screen for plasmid replicons from an Acinetobacter replication initiation (rep) gene database (25). An average of four and two rep genes were found per A. lwoffii and A. baumannii genome, respectively. A random permutation and Welch’s t-test revealed that A. lwoffii contained significantly more rep genes than A. baumannii (P < 0.0001), suggesting that A. lwoffii harbors more plasmids (Fig. 2a).

Fig 2.

Signatures of foreign DNA acquisition. (a) Number of rep genes found per whole-genome sequence from either species, (b) number of DNA defense system genes per sequence. A. baumannii in pink, n = 4,809, A. lwoffii in yellow, n = 38. (c) The twitching motility of individual strains, (d) combined values of all strains tested per species, and (e) the number of type IV pili-associated genes found in the whole-genome sequences. A random permutation and Welch’s t-test were used to compare the mean number of rep, DNA defense system, and type IV-associated genes in comparable data sets of A. baumannii and A. lwoffii. A. baumannii genomes encode more genes than A. lwoffii (in all instances, ****P < 0.0001).

Acinetobacter rep genes are classified broadly according to the protein family they encode [Rep_1, Rep_3 (R3), or RepPriCT_1] and specifically by homology (>95% nucleotide identity cut-off) to a collection of reference rep sequences (25). All A. lwoffii rep genes detected here belonged to the R3 group. However, since the rep database was constructed primarily for the purpose of typing plasmids in A. baumannii, there were inconsistencies when comparing the rep genes identified by ABRicate and the number of circular plasmid sequences in complete A. lwoffii genomes. ABRicate detected fewer rep genes (n = 34) than there were plasmids (n = 64) in the complete genomes (Table S3). While it is possible that some plasmids did not contain a recognizable rep gene, as has been reported for A. baumannii plasmids (25), this was unlikely to be the case for all instances here. Therefore, the NCBI annotations for all plasmids in complete A. lwoffii genomes were screened for ORFs labeled “rep,” and a further six genes not represented in the database were found, five encoding Rep_3 proteins [CP032104 1 (pALWEK1.11), CP080579 1 (pALWVS1.3), CP072552.1 (pH7-68), CP080580 1 (pALWVS1.4), and CP080643 1 (pALWEK1.16)] and one encoding Rep_1 [CP080641 (pALWEK1.14)]. In a phylogenetic tree, these genes clustered independently of previously known rep genes (Fig. S1). With these considered, all but one of the A. lwoffii rep genes clustered in R3, supporting the idea that A. lwoffii almost exclusively maintains R3-type plasmids.

The most common rep types in A. lwoffii were R3-T25/R3-T45, which were found in a total of 92% of genomes. R3-T25 and R3-T45 are 94.71% identical at the nucleotide level and therefore, although classed as different rep types using a 95% cut-off value, are very closely related. Therefore, we propose that R3-T25/R3-T45 replicons represent a native A. lwoffii plasmid family, found in almost all complete genome sequences of this species examined here. In contrast, R3-T25/R3-T45 replicons were only found in 0.4% of A. baumannii genomes. For A. baumannii, 38% of sequences contained R2-T1 and 37% encoded RP-T1 rep types. In total, A. baumannii had 82 distinct rep types, including from RP, R1, R2, and R3 groups. A full list of rep genes highlighted in both species can be found in Table S4.

In addition to ARGs, occasionally, plasmids may also carry genes for RND efflux pumps, which can export a wide range of structurally diverse compounds, including antibiotics (13), and can act as important mechanisms for antibiotic resistance. RND determinants have been seen in plasmids in A. baumannii, for example, pDETAB2 from a Chinese ICU patient isolate (34), and more recently in A. lwoffii, where AL_065, which was isolated from a hospital bed rail in Pakistan, harbored a plasmid encoding an RND transporter and periplasmic adaptor protein (35). This plasmid (CP078046.1, rep type R3-T25) is also found in A. baumannii and has the potential to disseminate RND efflux genes across A. lwoffii more broadly. The RND pump is closest in homology to AdeB (31) and may therefore represent the acquisition of an additional, adaptive RND pump, reducing the susceptibility of this strain to structurally different substrates than those exported by its native RND system: AdeIJK (1).

To determine if the relative lack of ARGs in A. lwoffii could also be related to other mechanisms of HGT, we searched for the presence of prophage DNA within genomes of both species. Both A. lwoffii and A. baumannii had prophage DNA within their genomes, as determined by PHASTER (Table S5). Therefore, both species have been previously infected by phage and have the capacity to acquire novel DNA, such as ARGs, introduced by phages.

The number of DNA defense systems across the two species was determined as this could impact their acquisition and maintenance of foreign DNA. Using DefenseFinder, A. lwoffii genomes were found to encode between 3 and 24 defense systems per genome, which was significantly more than A. baumannii which had between 1 and 14 (P ≤ 0.0001) (Fig. 2b). The types of defense systems present also differed. A. lwoffii encoded mostly type I and IV restriction modification systems, which cleave unmethylated DNA, whereas A. baumannii encodes more PsyrTA toxin antitoxin systems and antiphage systems, e.g., SspBCDE (Table S6).

Acinetobacter spp. can display twitching motility in laboratory conditions, which aids the natural transformation of DNA from the extracellular environment into the cell (24). Therefore, the ability of A. lwoffii and A. baumannii to twitch was measured. While there was strain variation in subsurface twitching motility, generally A. lwoffii twitched less (average of 0.6 cm) than A. baumannii (average of 2.5 cm) at 37°C (Fig. 2c and d), suggesting that A. lwoffii may be less naturally competent than A. baumannii; however, twitching efficiency does not always translate to natural competence.

Natural transformation uses type IV pili genes, and therefore, we also looked for the presence of genes associated with type IV pili in both species (Fig. 2e). There were significantly more type IV-associated genes found in A. baumannii genomes compared to A. lwoffii genomes (P < 0.001, Table S7).

A. baumannii readily evolved resistance to meropenem, ciprofloxacin, and gentamicin, but A. lwoffii only evolved resistance to ciprofloxacin

Since A. lwoffii has remained susceptible to antibiotics, we sought to determine whether it can evolve resistance to clinically relevant antibiotics in vitro. For context, we also included A. baumannii, which is well known to evolve drug resistance rapidly. To this end, selection experiments were set up, where susceptible strains were grown in the presence of increasing concentrations of meropenem, ciprofloxacin, or gentamicin. After 7 days, whole-genome sequencing was performed to characterize any genomic changes compared to the ancestral strain (Table S8). Sequencing accession codes can be found in Table S2, where the A. baumannii 18 ancestor BioSample was SAMN32597910 and A. lwoffii 28 ancestor was already available on NCBI as SAMEA50767168 (GCA_900444925.1).

A. baumannii (AB18) mutants selected in the presence of meropenem had meropenem MICs two- to threefold above that of the parent strain MIC, from 1 to 2–4 μg/mL (Table S8). There were also MIC increases for ampicillin (four- to fivefold), ciprofloxacin (threefold), and tetracycline (threefold), with some mutants also being less susceptible to moxifloxacin (two- to threefold) and erythromycin (two- to threefold) (Fig. 3). It was noted that fewer A. lwoffii (AL28) colonies were selected for; however, when MIC testing the mutants, the increase was also threefold above the ancestral MIC from 0.015 to 0.06 µg/mL. There was no significant MIC change for the other antibiotics tested.

Fig 3.

MIC fold change results and single-nucleotide polymorphism presence for strains evolved to grow in increasing concentrations of meropenem. “Ancestor” and “WT” (broth-only) controls are compared to individual “M” isolates from the terminal passage. AB18, A. baumannii, and AL28, A. lwoffii. An MIC fold change of 1 means the strain is as susceptible or more susceptible to the drug compared with the ancestor. M, mutant; WT, wild type.

Five mutants from AL28 and AB18 were subject to whole-genome sequencing, and their sequences were compared to the ancestral strain and parental strains, which had been passaged in the same experiment in nutrient broth only. Despite A. lwoffii being able to grow at the final concentration of meropenem used in the evolution experiment, no canonical resistance mutations were seen. In fact, no single-nucleotide polymorphisms (SNPs) were found in the mutants, even though the strains passaged in nutrient broth alone encoded some SNPs, indicating an adaptive physiological (rather than genetic) change. However, for A. baumanii (AB18), all five sequenced strains had SNPs in the RND efflux transporter encoding gene adeJ and in the gene that encodes the repressor protein for this system, adeN. Three of the adeJ mutations were within the distal binding pocket of the pump, where beta-lactams bind (1). Additionally, AB18 mutant 5 had mutations in genes encoding penicillin-binding proteins 2 and 3, known to be involved in meropenem resistance (36).

For ciprofloxacin, both A. baumannii and A. lwoffii cultures evolved resistance to above the EUCAST breakpoint. In AB18, large MIC changes, between 9- and 10-fold higher than the ancestral strain, were seen for ciprofloxacin and moxifloxacin. Additionally, MIC increases were also observed for gentamicin (four- to fivefold) and erythromycin (threefold) in some mutants (AB18 M2, M3, and M5), and the tetracycline MIC was also elevated (threefold) in AB18 M2 and M3. Mutants selected in the presence of increasing concentrations of ciprofloxacin had mutations in both the target of the drug (gyrA/gyrB/parC) and RND efflux systems (ade pumps).

For A. lwoffii, in contrast to the results seen with meropenem, target site and efflux SNPs were seen in the AL28 mutants. It is also worth noting that the AL28 wild-type (WT) strains also harbored polymorphisms, despite being passaged in nutrient broth only. SNPs were found in genes such as higA1, encoding an antitoxin protein, and yfdX2, encoding a heat resistance protein. AL28 mutants had SNPs in adeJ (RND pump), adeN (RND pump regulator), atpB (ATP synthase), and gyrA and gyrB (genes coding for DNA gyrase). Presumably, the combination of SNPs in efflux-related genes and target-site genes contributed to the reduced susceptibility of the mutants to ciprofloxacin and moxifloxacin and also protected against meropenem (Fig. 4).

Fig 4.

MIC fold change results and SNP presence for strains evolved to grow in increasing concentrations of ciprofloxacin. “Ancestor” and “WT” (broth-only) controls are compared to individual “M” isolates from the terminal passage. AB18, A. baumannii, and AL28, A. lwoffii. An MIC fold change of 1 means the strain is as susceptible or more susceptible to the drug compared with the ancestor. M, mutant; WT, wild-type.

Since A. lwoffii seemed to be capable of evolving drug-resistance mutations to ciprofloxacin but not meropenem, a third experiment was conducted. Here, gentamicin was chosen which is also used clinically to treat Acinetobacter infections. All AB18 mutants had elevated MICs to gentamicin (eight or ninefold above the ancestral strain MIC), taking them from clinically susceptible to resistant (>4 µg/mL) (Fig. 5). These mutants also displayed a reduced susceptibility to ciprofloxacin and moxifloxacin, and some of the AB18 mutants (1, 2, and 4) also showed a reduced susceptibility to erythromycin and tetracycline too. The WT strains grown in broth did not encode any SNPs, whereas the mutant strains had SNPs in adeB (RND pump), adeR, adeS (genes for RND pump regulator), fusA (elongation factor B), ptsP (phosphoenolpyruvate protein phosphotransferase, which is important in sugar transport), and tetR (global regulator).

Fig 5.

MIC fold change results and SNP presence for strains evolved to grow in increasing concentrations of gentamicin. “Ancestor” and “WT” (broth-only) controls are compared to individual “M” isolates from the terminal passage. AB18, A. baumannii, and AL28, A. lwoffii. An MIC fold change of 1 means the strain is as susceptible or more susceptible to the drug compared with the ancestor. M, mutant; WT, wild-type.

As with meropenem, the A. lwoffii strain tested did not exhibit drug resistance to gentamicin or other drugs tested. However, during this experiment many SNPs were selected in both the nutrient broth-only conditions (WT1 and WT2) and gentamicin conditions (M1–M5). Mutations found only in the AL28 cells grown in gentamicin included SNPs in acr3 and arsH (arsenic resistance), chrA (chromate resistance), and merA, merD merR, and merT (mercuric transport proteins). Therefore, there was both conservative MIC differences and genomic evidence of a stress response, particularly in metal-tolerance genes.

In summary, A. baumannii AB18 was able to evolve resistance in 7 days to three clinically relevant antibiotics, which provided elevated MICs not only to that antibiotic but also to drugs from other classes, such as fluroquinolones, tetracyclines, aminoglycosides, and macrolides. Furthermore, AB18 went from clinically susceptible to resistant, as defined by EUCAST breakpoints, in each instance. However, for A. lwoffii, clinical resistance was seen only for ciprofloxacin. These results show that A. lwoffii has a more limited capacity to evolve resistance to antibiotics, and due to the diversity of efflux-related mutations in A. baumannii, this may be due to the lack of RND systems in A. lwoffii.

A. lwoffii forms less biofilm, and the biofilm is more susceptible to antibiotics and biocides than those formed by A. baumannii

Antibiotic susceptibility is known to be decreased when bacteria exist within a biofilm and Acinetobacter often forms biofilm to aid survival in the clinical environment (37). Therefore, the biofilm-forming capacity and susceptibility of biofilm to antibiotics were determined. In static conditions, A. baumannii strains formed significantly more biofilm on average than the A. lwoffii strains (P < 0.001), median OD₆₀₀ values of 3.39 and 0.53, respectively (Fig. S2a and b). However, when biofilm was formed under laminar flow conditions, there was no significant difference in the amount of biofilm formed between the two species (Fig. S2c).

When the genomes were searched, for genes previously associated with biofilm formation (26), A. lwoffii sequences were found to have a mean of one gene per sequence, whereas A. baumannii had a mean of eight genes per genome sequence (Fig. S2d; Table S9). In the case of csuABC pili genes, which mediate adhesion to human cells, >89% of A. baumannii sequences searched had all three of these genes, whereas they were all absent in the A. lwoffii genome sequences. However, as this database was created using genes from A. baumannii, biofilm-associated genes exclusive to or uncharacterized in A. lwoffii would not have been found using this approach.

Given that a biofilm lifestyle is associated with decreased susceptibility to antibiotics, the MIC and minimum biofilm eradication concentration (MBEC) was determined for representatives of both species (Table 2). For both species, the MBEC values were generally higher than the MIC values; for example, for AB20, the cefotaxime MBEC was 10-fold higher than the MIC. However, the effect was less evident in A. lwoffii (AL04), where there were instances where the MBEC and MIC values did not significantly change (chlorhexidine, meropenem, and triclosan). Furthermore, in general, the A. lwoffii (AL04) MBEC values were lower than those of A. baumannii (AB20). Therefore, while the biofilms formed by both strains were less susceptible to antibiotics and biocides, the biofilm formed by A. baumannii (AB20) afforded greater protection than that formed by A. lwoffii.

TABLE 2.

MBEC^a and minimum broth inhibitory concentrations of antibiotics and biocides in A. lwoffii (AL04) and A. baumannii (AB20)

	AL04		AB20
	MBEC	MIC	MBEC	MIC
Cefotaxime (μg/mL)	256	2	8192	16
Chlorhexidine (%)	0.0017	0.0008	<1	0.0035
Ciprofloxacin (μg/mL)	16	0.06	128	2
Meropenem (μg/mL)	0.06	0.03	<128	0.25
Oxacillin (μg/mL)	8	16	<4,096	512
Tetracycline (μg/mL)	256	0.5	1,024	2
Triclosan (μg/mL)	0.5	0.5	<128	1
Rifampicin (μg/mL)	8	0.5	64	4

^a MBEC, minimum biofilm eradication concentration.

A. lwoffii has a longer cell morphology than A. baumannii

Thus far, it is clear that A. lwoffii is more susceptible to antibiotics than A. baumannii in both static and biofilm conditions, and this is likely due to a reduced ability to evolve and acquire resistance, which may be underpinned by the presence of more DNA defense systems and fewer RND efflux pumps. Given the lack of research into A. lwoffii, the basic biology of the two species under laboratory conditions was assessed.

To determine whether there were any morphological differences between these two species, two strains of A. baumannii (AB05 and AB18) and two strains of A. lwoffii (AL04 and AL28) were imaged using scanning electron microscopy. A. lwoffii had significantly longer cells than A. baumannii (n = 100 cell measurements per strain) (Fig. 6; Fig. S3).

Fig 6.

Average cell length (μM) of A. baumannii (AB) and A. lwoffii (AL) strains imaged by the Apreo 2 Scanning Electron Microscope. A. lwoffii had statistically longer cells than A. baumannii strains. ****P ≤ 0.0001, one-way analysis of variance with Tukey’s multiple comparisons. The whiskers on the box plot show minimum and maximum values obtained.

A. baumannii grows more readily in both LB and human serum than A. lwoffii

Additionally, the growth of both species was compared at 37°C, 30°C, and 25°C. In LB, A. lwoffii grew to a lower final density than A. baumannii at all temperatures. Growing at cooler temperatures generally increased the length of the lag phase. The mean generation times (Table S10) were generally faster at 30°C for A. lwoffii, while A. baumannii grew fastest at 37°C. While A. lwoffii grew to a lower final optical density (OD) than A. baumannii (Fig. S4), the generation times of AL28, AL32, and AL33 grew at rates comparable to the A. baumannii strains.

Due to the capacity of both species to cause bacteremia in humans, we also sought to understand how well both species survive and grow in human serum. Growth was compared in human serum with and without complement proteins [normal human serum (NHS) and heat inactivated serum (HIS), respectively]; both species grew more slowly in serum than LB (Table S11; Fig. S5). Of the two A. lwoffii strains tested, AL04 had a prolonged lag but did grow in both HIS and NHS, although growth rate was better in HIS. AL28 did not grow in serum and formed clumps, making OD measurements problematic. A. baumannii AB05 and AB18 grew as well in normal serum and as they did in HIS. AB18 grew significantly (P = 0.0098) better than AB05 in HIS. All other conditions were not significantly different.

Survival in human serum was also measured to determine whether, although not actively growing, strains could remain viable in the presence of serum. All strains, except AL28, could survive in NHS, and by 24 hours, CFU per milliliter was similar in both the serum and the LB control (Fig. 7a).

Fig 7.

(a) The survival of AB and AL strains in both LB (dashed line) and normal human serum (continuous line) over 24 hours. (b) Survival of Galleria mellonella after inoculation with either AB (pink) or AL (yellow). A phosphate-buffered saline injury control is included in black. A log-rank test was conducted, where only AB05 was significantly more virulent than the other conditions (P < 0.0001). (c) The number of virulence-associated genes found in A. baumannii (pink) and A. lwoffii (yellow) whole-genome sequences (****P < 0.0001). A random permutation and Welch’s t-test was used and shows that A. baumannii encodes significantly more virulence genes than A. lwoffii. Attachment (d),invasion (e), and proliferation (f)of both species in human THP-1 macrophages were measured. SE01 is a positive control of Salmonella enterica Typhimurium. Comparative statistics (one-way analysis of variance) were performed, but no conditions were significantly different. AB, A. baumannii; AL, A. lwoffii.

A synthetic wound model (38) also showed that A. lwoffii strains did not grow as well as A. baumannii strains. This supports the fact that A. lwoffii survived poorly in the presence of human serum as AL04 and AL28 did not grow (Table S12).

A. baumannii is more virulent than A. lwoffii in vivo and in vitro

We also sought to determine whether there was a difference in the in vivo virulence capacity of the two species and chose to use the well-characterized infection model organism, Galleria mellonella, which has an innate immune system (32). When G. mellonella larvae were infected with 1 × 10⁶ A. baumannii or A. lwoffii cells, more larvae were killed when infected with A. baumannii (AB05 and AB18) than A. lwoffii (AL04 and AL28), which correlates with what has been seen previously (32). However, this was only statistically significant for AB05, A. baumannii AYE; AB05 was significantly more virulent than AB18 in this model (Fig. 7b; P < 0.0001, log-rank test). By 48 hours, the probability of larvae survival was <25% for AB05 infection, whereas it was >95% for AL28.

Since A. baumannii AYE was more virulent in vivo than A. lwoffii, we also probed the whole-genome sequences for the presence of virulence genes. A. baumannii genomes encoded significantly more virulence genes than A. lwoffii genomes (P < 0.0001; Fig. 7c; Table S13) when using a random permutation and t-test to compare two equally sized sample sets. As before, virulence genes have been mainly characterized in A. baumannii, and therefore, the database is biased to highlight virulence genes in A. baumannii.

Finally, to determine virulence in vitro, strains were incubated with a human macrophage cell line, THP-1. A. baumannii strains (AB05 and AB18) were able to attach to and subsequently invade THP-1 cells (Fig. 7d and e). However, after 6 hours, proliferation was also measured, and there was no difference in the number of CFUs between invasion and proliferation, suggesting A. baumannii was not actively growing within the cells but could survive at least for the period of the assay (Fig. 7f). In contrast, neither of the A. lwoffii strains tested could attach to or invade human macrophage cells in vitro.

DISCUSSION

The emerging pathogen, A. lwoffii, is the leading cause of Acinetobacter-derived bloodstream infections in England and Wales, followed by the extensively studied A. baumannii (4). However, A. baumannii has developed widespread multidrug resistance, while A. lwoffii has remained sensitive to almost all antibiotics. While research into A. baumannii is increasing and more is known about its antibiotic resistance, there remains a knowledge gap in understanding the emerging opportunistic pathogen, A. lwoffii. This work aimed to explore differences in the two species in terms of their antibiotic susceptibility, infectivity, and basic biology. We have shown that A. lwoffii is more susceptible to drugs used to treat Acinetobacter infections than A. baumannii, is less virulent, and does not evolve drug resistance to the same degree as A. baumannii.

This work confirmed previous data suggesting A. lwoffii is susceptible to antibiotics, while A. baumannii is commonly multidrug resistant (4) and showed the difference in phenotype is caused by A. lwoffii encoding fewer resistance genes than A. baumannii. Both species are found in similar environments such as on the human body, although A. baumannii is not considered to be part of a healthy skin microbiome (6, 39). As they are both found within the hospital environment, it is peculiar that resistance (either by mutation or the horizontal acquisition of ARGs) has not been commonly selected for in A. lwoffii.

The lack of ARGs in A. lwoffii may be due, at least in part, to the presence of DNA defense systems that are absent in A. baumannii, such as a greater number of restriction modification systems. The presence of more DNA defense systems in A. lwoffii suggests that this species is more stringent about the DNA it maintains (40).

In addition to fewer ARGs, A. lwoffii also has less readily evolved resistance to three clinically relevant drugs compared to A. baumannii, following passage in increasing concentrations of antibiotics over the course of 7 days. We note that, while the growth rate of these organisms is similar, the total generation number may have differed slightly. However, the difference in resulting drug susceptibility is nevertheless significant. Drug-resistance mutations often occur in the drug’s target: penicillin-binding proteins for meropenem (36), DNA gyrase for ciprofloxacin (41–43), and the ribosome for gentamicin (44). This was the case for A. baumannii here. In the one instance where A. lwoffii evolved resistance, to ciprofloxacin, drug target mutations were also observed. A limitation of using Snippy and short-read sequencing technology is that some genetic mutations may have been missed, for example, recombination and gene inversion events. Additionally, by working with the assemblies, we also were unable to search for gene amplification events. Ciprofloxacin mutations often occur in the quinolone resistance-determining regions (QRDR) of GyrA, GyrB, and ParC (43). The A. baumannii mutations in gyrA were in the QRDR (amino acids 65–104), but A. baumannii mutations in parC and A. lwoffii in gyrB, however, were not within the QRDRs.

Additionally, mutations were captured in RND efflux pumps that export the compounds used for selection. For example, A. baumannii meropenem mutants had adeJ mutations, and beta-lactams bind to the distal pocket of AdeJ (44, 45). Fluoroquinolones can be exported by all three Ade pumps in A. baumannii (39), which explains why mutations in all three Ade systems were seen, including mutations that affected the regulators of these systems. Gentamicin is exported by AdeB and can bind to both the proximal and distal binding pockets, but Y77, T91, and S134 are thought to be essential for gentamicin binding to the proximal pocket of AdeB (46). Given the proximity of the A. baumannii AdeB mutations (amino acids 97 and 136) in this study to those reported in the literature (46), it is likely that these mutations led to increased gentamicin export via AdeB. Mutations in AdeRS have been reported to increase AdeABC expression, for example, A91V in AdeR and A94V in AdeS (47). This study also captured the A91V SNP in AdeR, which sits in the signal receiver domain, as well as other mutations in AdeRS, indicating that AdeRS may be being modulated to increase AdeABC expression and the extrusion of gentamicin.

The mutant evolution experiments clearly show that A. lwoffii has a reduced capacity to evolve resistance to antibiotics compared to A. baumannii, where it only evolved resistance to ciprofloxacin. This could be because A. lwoffii only encodes one tripartite RND system (AdeIJK) (1). Another example of a species with one tripartite RND system is Neisseria gonorrhoeae; however, in this organism, drug-resistance mutations and the acquisition of resistance genes are common (48). RND efflux pumps have an underpinning role in the development of resistance via other molecular mechanisms (44). For example, in other species of Gram-negative bacteria, deletion of efflux pumps reduces the mutation selection frequency (44, 49). In addition, mutations within efflux pumps often occur first evolutionarily and allow for the development of more canonical drug target mutations, which may have been the case in this study (50, 51). The reduced efflux capacity of A. lwoffii could therefore limit the selection of drug-resistance mutations. This is further supported by the fact that in A. baumannii, drug-resistance mutations were found across all three tripartite systems, indicating their important role in resistance evolution. Another potential mechanism for the lack of resistance development could be more stringent DNA repair mechanisms in A. lwoffii, for example mismatch repair to inhibit the recombination of non-homologous DNA. However, further work would be needed to confirm this (52).

When looking at infection-related phenotypes, A. baumannii was more virulent than A. lwoffii in vitro, and one of the A. baumannii strains was more virulent in vivo. It was already known that certain A. baumannii strains could infect macrophages and persist within their vacuoles, but this was the first time this experiment had been done using A. lwoffii, where none of the strains tested could persist within macrophages (33). This could indicate that it is easier to clear A. lwoffii infections. Both virulence potential and resistance to antibiotics have been linked with phase variation in A. baumannii AB5075, where opaque colonies were more drug resistant and virulent (53). While not observed in this study, A. lwoffii may also be able to undergo phase variation, which could affect their virulence and resistance phenotypes.

In summary, A. lwoffii is more susceptible to antibiotics than A. baumannii due to a lack of acquired and evolved resistance. Additionally, despite causing more bloodstream infections, the strains tested in this work were also less virulent than the A. baumannii strains. Therefore, the incidence of A. lwoffii bloodstream infections in the United Kingdom may be related to the increased opportunity of A. lwoffii (a skin commensal) to get inside the body and cause infections as opposed to A. baumannii and A. calcoaceticus, which are not considered commensals. Ultimately, an open question remains surrounding why A. lwoffii does not seem to be developing drug resistance in the clinic, and more work is needed to elucidate if this results from a lack of efflux systems and/or more stringent DNA repair and defense, or other factors. While the widespread antibiotic susceptibility of A. lwoffii allows for successful clinical outcomes, there are sporadic cases of drug-resistant A. lwoffii, highlighting the possibility that drug resistance could emerge (10, 54). It is, therefore, important to fully chart the development of this emerging pathogen to limit the development of drug resistance.

ETHICS APPROVAL

This study did not require ethical approval.

DATA AVAILABILITY

Whole-genome sequences were either publicly available and downloaded from National Center for Biotechnology Information (NCBI, https://www.ncbi.nlm.nih.gov) or obtained from strains sequenced in this study. Raw reads generated from this work can be found on NCBI under project accession number PRJNA918592.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/msphere.00109-24.

Supplemental Text and Figures. msphere.00109-24-s0001.docx.

Texts S1 and S2 and Figures S1-S5.

Supplemental Tables. msphere.00109-24-s0002.xlsx.

Tables S1-S13.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.