The genus Achromobacter currently comprises 20 species. Achromobacter species are mostly found in aquatic environments but may also be present among the intestinal microbiota of healthy persons. Furthermore, they may cause a range of human infections, in particular pulmonary infections in persons with cystic fibrosis (CF) [1, 2, 3, 4]. In a Canadian and a French study, respectively, 11% and 27% of persons with CF were tested positive for Achromobacter by bacterial culture [5, 6]. Colonization and/or infection may be persistent in persons with CF, but in some circumstances sputum can be rendered culture negative following antibiotic therapy [6]. Nevertheless, treatment is challenging due to both intrinsic and acquired resistance [2]; Achromobacter species encode OXA and AmpC type $\beta$-lactamases and efflux pumps [2, 7]. Biofilm formation, which appears to be an intrinsic ability of all strains, contributes to antimicrobial resistance and virulence [8].

Several genome assemblies have been reported, but these were limited to one to six isolates, which yields only limited insight in the occurrence of virulence genes and acquired antibiotic resistance [9, 10, 11, 12]. In this study we determined the whole genome sequences (WGS) of 101 Achromobacter isolates and report the diversity of the isolates, minimal inhibitory concentration for eight antimicrobial agents, and the presence of putative virulence genes. The isolates were obtained from CF, bronchiectasis (BE), and other diseases.

The aim of this study was to characterize the phylogenetics, antibiotic resistance, and virulence factors of clinical Achromobacter isolates based on whole genome sequencing.

WGS yielded an average of 247 contigs per isolate (range 109–714); the average coverage was 51$\times$ (range 15–109$\times$), and the total length of the assemblies varied between 5.71–7.16 MB, with a GC content between 64.29 and 68.33% (Supplementary Table 2).

MALDI-TOF, which is commonly used to identify isolates in routine diagnostic microbiology, identified 95 isolates as A. xylosoxidans, three as A. insolitus, and one as A. spanius. Two isolates were identified only to the Achromobacter genus level. Analysis of the WGS results confirmed 63/95 (66.3%) isolates as A. xylosoxidans. This was in agreement with the identification rates reported before with the used default MALDI-TOF database [2]. The remaining 32 A. xylosoxidans isolates were A. insuavis (n = 11), A. ruhlandii (n = 7), A. deleyi (n = 5), and two putative novel species, which were designated species1 (n = 8) and species2 (n = 1) in this manuscript. The isolate identified as A. spanius by MALDI-TOF was found to be A. deleyi by WGS. Two of the three isolates identified as A. insolitus by MALDI-TOF were confirmed by WGS, the third isolate was found to be A. aegrifaciens by WGS. The two isolates which were identified only to the genus level by MALDI-TOF were A. spanius and A. deleyi by WGS analysis (Supplementary Table 1). With the exception of the two novel species, all Achromobacter species have been reported previously in persons with CF [5].

Prior to this study, 485 Achromobacter STs had been reported. MLST analysis of this collection yielded 53 novel alleles and 35 novel sequence types (STs), indicating that only a portion of the genetic diversity had been assessed previously. A total of 61 STs were present among the isolates in this study. Novel alleles and STs were submitted to the PubMLST database [https://pubmlst.org/general.shtml]. A MLST-based Minimum Spanning Tree showed that the isolates clustered by species. The isolates of the different species differed by at least 6 or 7 loci. The majority of the A. xylosoxidans isolates were not closely related as only six pairs of single-locus variants were present, whereas 18 the other sequence types diverged by 2 or more loci (Fig. 1).

The five A. deleyi isolates were all recovered in the UK and belonged to a single ST; based on sequence alignments it is possible that they belonged to a single outbreak. Six of the eight novel species1 isolates belonged to ST144 and also appeared to be closely related; a definitive conclusion would require more knowledge of the population structure and mutation rates of this species. The two other isolates belonged to ST57and ST428.

OXA-family $\beta$-lactamases were detected in A. deleyi, A. dolens, A. insuavis, A. ruhlandii, in 61 of the 63 A. xylosoxidans, and in putative novel species1 (Table 1); these were presumably chromosomally located [5]. The sequences of the OXA-family $\beta$-lactamases showed variability, as has been reported before [5]. The evolutionary history of the OXA-type $\beta$-lactamases was inferred using the Neighbor-Joining method and the OXA-$\beta$-lactamases clustered with the respective type strains of their species (Fig. 2) [5]. No relevant differences were observed between isolates with and without OXA-family $\beta$-lactamases regarding the MIC${}_{50}$/MIC${}_{90}$ values for the $\beta$-lactam antibiotics ceftazidime (respectively 4 mg/L and 128 mg/L vs 4 mg/L and 16 mg/L), aztreonam (both 256 mg/L and $>$256 mg/L), imipenem (2 mg/L and 16 mg/L vs 2 mg/L and 4 mg/L) or meropenem (1 mg/L and 32 mg/L vs 1 mg/L and 8 mg/L) (Table 1). This indicates that other mechanisms, such as efflux pumps, must be involved in resistance against these antimicrobials [2].

MICs for individual isolates are reported in Table 1. The most active agent was co-trimoxazole followed by imipenem. The MIC-distribution of the isolates in this study has been reported previously [23].

Only 14 isolates carried known acquired antibiotic resistance genes. One A. xylosoxidans isolate carried nine resistance genes including a GES-type Extended-Spectrum $\beta$-Lactamase (ESBL) and five aminoglycoside resistance genes; one of these aminoglycoside resistance genes was aac-(6’)-Ib-cr, which also confers resistance to several fluoroquinolones and which has been described in Achromobacter before [24]. Multiple aminoglycoside resistance genes were found in five isolates. Nine isolates had the sul1 gene, which is associated with class 1 integrons. Integrons have been described in Achromobacter previously [25, 26]; however, in three of these isolates (537297, 537315, 551610) the class 1 integron integrase was not detected. The complete integrons could not be reconstructed due to the use of short-read sequencing. Other detected resistance genes were dfrA1, dfrA12, ereA, tet(A), tet(G), bla${}_{O\mathit{}X\mathit{}A\mathrm{-}\mathrm{2}}$, and sul2, which encode resistance to trimethoprim (dfr), macrolides (ere), tetracycline (tet), $\beta$-lactam antibiotics (bla), and sulfonamide (sul), respectively (Table 1). Since most isolates did not have acquired resistance genes, high MICs must be the result of other mechanisms; a likely explanation is the presence of efflux mechanisms possibly combined with reduced porin expression.

Despite infections by Achromobacter spp. being mostly limited to persons with CF and immunocompromised hosts, many virulence factors have been proposed for these species [12, 27]. The role of many of these (putative) virulence factors in infection remains to be elucidated. Some of these factors, such as O-antigens and capsules, may play a role in disease, but are also just basic bacterial structures. For our analysis, based on the encoding sequences, the following virulence factors were studied: hlyA, yqfA1, and yfqA2, and regions 1, 6, 23, and 24, defined by Li et al. [26]. Proposed factors without a defined role in virulence were excluded from the analysis [12].

Region 1, encoding a type III secretion system that has been implied in pathogenicity in other bacteria [27], was absent in some A. xylosoxidans isolates (n = 12, 19.0%), one A. ruhlandii isolate, and all isolates of A. aegrifaciens, A. insolitus and A. spanius. The region 1 sequences in the other isolates were species-specific, but in A. xylosoxidans two variants were present. A ~5 kb region of variant 1 was replaced by a ~6.5 kb sequence. Annotation showed that the assigned gene functions of both sequences were similar.

Region 6, encoding genes for dTDP-rhamnose synthesis, an O-antigen component, was present in only 19 isolates and 6 of these belonged to the same ST. The sequences could be divided into seven variants, which clustered into four groups. One variant, present in four A. insuavis isolates, lacks the gene for CDP-glucose 4,6-dehydratase. An additional 25 isolates encoded only the glucose-1-phosphate cytidylyltransferase (Table 2, Ref. [12]).

Region 13, involved in activation of hemolysin, was present in 49 A. xylosoxidans isolates as well as all A. insuavis and species1 isolates and absent in nine A. xylosoxidans isolates including four from ST27, two from ST22 and ST496 each. The two A. insolitus isolates and the single A. aegrifaciens isolate also lacked this region. Five A. xylosoxidans isolates lacked the gene for the MSF transporter and the LysR family regulator. This was also the case for the A. ruhlandii, A. spanius, and the species2 isolate. Interestingly, the A. deleyi isolates encoded only the MSF transporter and the LysR family regulator (Table 2).

The annotation of the A. xylosoxidans type strain (accession number NZ_LN831029) showed the presence of three different hemolysins encoded by the hlyA, aqfA1 and aqfA2 genes. The first encoded a 3296 amino acid protein. However, this region has also been annotated as an agglutinin with additional hypothetical proteins. The opposite strand encoded a putative isopropyldehydratase large subunit. Of note, A. xylosoxidans isolates do not show hemolysis on sheep blood agar. The hlyA region was absent from A. insolitus, A. insuavis, A. spanius, species1, and A. deleyi with one exception (Table 2). The region, which was also present in A. ruhlandii, appeared to be variable; a thorough assessment of this variability was not possible, because the assembly of these sequences was highly fragmented for most isolates (despite good genome coverage for sequencing). The aqfA1 gene was present in all isolates, except for one A. xylosoxidans isolate, and also the aqf2 gene was present in all isolates, except for one A. xylosoxidans. The presence or absence of the hemolysin activating region did not match with the presence or absence of any of the putative hemolysin genes; this suggests either issues with the annotation or a complex regulation of expression.

Region 23, involved in lipopolysaccharide biosynthesis, was present in 15 isolates including six A. insuavis isolates and with 3 isolates belonging to A. xylosoxidans ST27. The nearly 22 kb region was truncated after approximately 15.6 kb. The function of the lacking part is unknown. Region 24 encoding capsule production was present in all A. insuavis and species1 isolates, but not in any of the other species (Table 2).

It should be noted that expression of virulence factors was not confirmed in vitro or in vivo and that virulence features are only predicted on the bases of sequence analysis.

The 20 non-CF isolates consisted of 15 A. xylosoxidans (75%), four A. insuavis (20%) and one species1 isolate (5%) (Table 1). The CF-isolates appeared to be more diverse: 48 A. xylosoxidans (59%), seven A. insuavis, A. deleyi, A. ruhlandii and species1 (each 8.6%), two A. insolitus (2.5%), and one A. aegrifaciens, A. spanius and species2 (each 1.2%). Comparison of the average MICs for CF isolates and non-CF isolates showed that the MICs for aztreonam, imipenem and tobramycin were approximately two times higher for non-CF isolates (non-CF vs CF:153.6 and 84.1 g/L, 11.0 and 5.5 mg/L, and 58.4 and 34.5 mg/L, respectively), whereas the MICs for ceftazidime, ciprofloxacin, colistin, co-trimoxazole, and meropenem were 1.4-2.9 higher for CF isolates (CF vs non-CF: 19.3 and 12.9 mg/L, 8.9 and 3.1 mg/L, 2.0 and 1.4 mg/L, 2.4 and 1.0 mg/L, 5.0 and 3.4 mg/L) (Table 1). No relevant differences were observed in the presence of putative virulence factors.

In conclusion, only 63/95 (66.3%) of the isolates were correctly identified using routine MALDI_-TOF identification probably indicating a lack of well-typed Achromobacter isolates in the standard database. Two putative novel species were identified. Isolates from persons with CF appeared to be more diverse. Despite the high MICs the presence of acquired resistance genes is uncommon, although some isolates harbored several acquired resistance genes. The average MICs for CF isolates were lower for aztreonam imipenem, and tobramycin, but higher for ceftazidime, ciprofloxacin, colistin, co-trimoxazole, and meropenem. The putative virulence genes of Achromobacter involved in infections or colonization are variable, but no difference in putative virulence factors were observed.

CF, cystic fibrosis; ESBL, Extended-Spectrum $\beta$-Lactamase; MALDI-TOF, matrix assisted laser desorption/ionisation time-of-flight analyzer, MIC, minimal inhibitory concentration; MLST, Multi-Locus Sequence Typing; ST, sequence type.

ACF, MDA, MMT, JSE RC and MBE designed the research study. BB-T, MvW, JFM performed the research. ACF and JRB analyzed the data. ACF, JRB, MBE wrote the manuscript. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript.

Not applicable.

Not applicable.

The research leading to these results has received support from the Innovative Medicines Initiative Joint Undertaking under grant agreement number [115721-2], resources of which are composed of financial contribution from the European Union’s Seventh Framework Programme (FP7/2007-2013) and EFPIA companies in kind contribution. Spanish CF isolates were recovered from projects supported by Instituto de Salud Carlos III of Spain (grants PI12/00743 and PI15/00466) and cofinanced by the European Development Regional Fund (A Way to Achieve Europe program; Spanish Network for Research in Infectious Diseases grant REIPI RD12/0015). MD-A was partially supported by Fundación Francisco Soria Melguizo (Madrid, Spain).

The authors declare no conflict of interest.