The Wzi outer membrane protein mediates assembly of a tight capsular polysaccharide layer on the Acinetobacter baumannii cell surface

Identification of novel therapeutic targets is required for developing alternate strategies to treat infections caused by the extensively drug-resistant bacterial pathogen, Acinetobacter baumannii. As capsular polysaccharide (CPS) is a prime virulence determinant required for evasion of host immune defenses, understanding the pathways for synthesis and assembly of this discrete cell-surface barrier is important. In this study, we assess cell-bound and cell-free CPS material from A. baumannii AB5075 wildtype and transposon library mutants and demonstrate that the Wzi outer membrane protein is required for the proper assembly of the CPS layer on the cell surface. Loss of Wzi resulted in an estimated 4.4-fold reduction in cell-associated CPS with a reciprocal increase in CPS material shed in the extracellular surrounds. Transmission electron microscopy revealed a disrupted CPS layer with sparse patches of CPS on the external face of the outer membrane when Wzi function was lost. However, this genotype did not have a significant effect on biofilm formation. Genetic analysis demonstrated that the wzi gene is ubiquitous in the species, though the nucleotide sequences were surprisingly diverse. Though divergence was not concomitant with variation at the CPS biosynthesis K locus, an association between wzi type and the first sugar of the CPS representing the base of the structure most likely to interact with Wzi was observed.

Carbapenem-resistant Acinetobacter baumannii is listed as a critical priority bacterial pathogen by the World Health Organization (WHO) 1 , identifying it as species for which innovative therapeutic strategies are urgently needed. Extensive and pan antibiotic-resistant A. baumannii infections are not only associated with poorer patient outcomes with significant morbidity and mortality but are also difficult to eradicate from clinical environments despite infection control procedures 2 . The success of the species as a serious global pathogen can be attributed to a highly plastic genome with significant mutation rates and frequent acquisition of genes that confer extensive antimicrobial resistance, increase virulence in the host, and/or enhance survival in unfavorable conditions over long periods 3,4 . These characteristics have favored the expansion of multi-drug resistant clonal lineages, including the two globally disseminated clones, global clone 1 (GC1) and global clone 2 (GC2), emphasizing a need to better understand mechanisms for virulence and survival in host and nosocomial contexts.
Bacterial protection from external antimicrobial threats is known to be significantly enhanced by the presence of a complex polysaccharide matrix, known as the capsular polysaccharide (CPS). CPS forms the external-most layer of the bacterial cell envelope and is comprised of high molecular weight polymers of repeating oligosaccharide units (K units) 5,6 . CPS is a primary virulence determinant and its presence is critical for protection against complement-mediated killing and opsonophagocytosis 7-10 , as well as for mediating resistance to desiccation and other antimicrobials [11][12][13] . Thus, removing or damaging the CPS barrier is a promising strategy to attenuate or re-sensitize the bacterium to specific antimicrobials.
Biosynthesis of the A. baumannii CPS follows a generalized Wzy-dependent pathway for K-unit construction, polymerization and export to the extracellular side of the gram-negative outer membrane 5 . The majority of genes responsible for CPS biosynthesis and export are arranged in a cluster located at the chromosomal K locus 5 . More than 128 distinct gene clusters have been identified at this location 14 , predicting extensive structural

Results
Validation and complementation of A. baumannii wzi mutants. The AB5075-UW wildtype (Gen-Bank accession number CP008706.1), three AB5075-UW mutants carrying T26 insertions in wzi (locus tag ABUW_2898), and one AB5075-UW mutant with a T26 insertion in itrA1 (locus tag ABUW_3822) representing a CPS-negative phenotype were acquired for analysis (see Table 1). Genomic material from each mutant strain was sequenced and the precise genetic context of the T26 insertion was confirmed within the correct gene for all mutants (Fig. 1B). Whole genome sequences were further investigated for other differences from the wildtype AB5075-UW reference genome, and only one single nucleotide polymorphism (SNP) was identified in each genome assembly. The SNP was located away from any gene known to be involved in CPS biosynthesis 5 , indicating that T26 mutants shared an isogenic background with the AB5075 wildtype. Each wzi::T26 mutant was complemented by the insertion of a chimeric gene cassette into the stable attTn7 site in the AB5075-UW chromosome in order to restore Wzi function to natural levels with single copy number expression. The cassette (Fig. 1C) included a short sequence of the intergenic space upstream of wzi in the AB5075-UW genome, a complete copy of the wildtype wzi gene, and an aacC4 gene conferring apramycin resistance for selection. The design of the cassette to include the intergenic sequence upstream of wzi provided the putative native promoter to enable natural control of its expression under the tested growth conditions. Wzi is required for retention of CPS on the cell surface. The role of Wzi in the association of CPS with the cell surface was first investigated by comparing CPS material purified from harvested cells (cell-associated CPS) with that of the supernatant (CPS released from the cell surface) from the same culture. SDS-PAGE analysis of extracts from both fractions (Fig. 2) showed the presence of high molecular weight CPS molecules, located close to the interface between stacking and separating gels consistent with previous studies 10,17,18 . The AB5075-UW wildtype displays an abundance of cell-associated CPS with an estimated 19.5% of the total CPS material (combined relative density of cell and supernatant fractions) found in the supernatant fraction. This suggests that a low level of CPS shedding naturally occurs in this strain. CPS material was not observed for either cell or supernatant fractions of the itrA1::T26 mutant consistent with a CPS-negative phenotype, confirming the role of ItrA1 in the synthesis of the AB5075 CPS.
An estimated 4.4 fold reduction in cell-associated CPS is observed for all three wzi::T26 mutants in comparison to the wildtype, with 80-85% of total CPS material observed in the supernatant fractions. However, 15-20% of the total CPS produced remains present in the cell fraction of each strain, suggesting that loss of Wzi reduces but does not eliminate CPS on the cell surface. Each complemented wzi::T26 attTn7-wzi strain displayed CPS phenotypes equivalent to that of the wildtype indicating successful restoration of Wzi function. These results suggest that Wzi plays an important role in retaining CPS molecules on the cell surface.
Wzi mediates assembly of a dense CPS layer on the cell surface. To directly visualize the effect of Wzi loss on the integrity of the CPS layer on the intact cell surface, cell sections of the wildtype, itrA1::T26 mutant and wzi::T26-1 mutant were examined using transmission electron microscopy (TEM). Consistent with SDS-PAGE analysis, the AB5075-UW wildtype (Fig. 3) displays a dense CPS layer with a mean thickness of 85.7 nm comparable to CPS-positive phenotypes of A. baumannii isolates examined in other studies [18][19][20] . A CPSnegative phenotype is observed for the itrA1::T26 mutant as expected, whereas the cell surface of the wzi::T26-1 mutant shows a disrupted CPS layer with sparse patches of CPS on the external face of the outer membrane. This indicates that Wzi is involved in the assembly of a tight CPS layer on the cell surface.
CPS presence rather than retention on the cell surface influences biofilm formation. In previous studies, deletion of CPS biosynthesis genes in A. baumannii strains has been shown to reduce biofilm formation or significantly alter biofilm morphology 8,20 . Therefore, we conducted an examination of the ability of the wzi::T26-1 mutant to form a biofilm. Following growth for 24 h in the MBEC device (formerly the Calgary Biofilm Device), the AB5075-UW wildtype formed cell-dense biofilm communities (Fig. 4), which is consistent with previous reports for this strain 21 . In comparison, the itrA1::T26 mutant showed a significant decrease in biofilm cell density as expected for a CPS-negative strain (P < 0.0001, ANOVA). However, the wzi::T26-1 and www.nature.com/scientificreports/ wzi::T26-1 attTn7-wzi strains each formed biofilms of equivalent cell density to that of the wildtype, indicating that extracellular presence of CPS material rather than its association with the cell surface is important for the formation of biofilm.
The wzi gene exhibits unexpected sequence diversity. The requirement of Wzi to assemble a tight CPS layer on the cell surface raises the possibility of a specific interaction between Wzi and the CPS structure. Considering the heterogeneous nature of the CPS, diversity in Wzi was also examined. A total of 9342 A. baumannii genome assemblies from the NCBI non-redundant and WGS databases were acquired for analysis. Each genome assembly was found to include a wzi gene sequence sharing 100% coverage and > 82% nucleotide sequence identity with AB5075-UW wzi (Supplementary Table S1). Eight assemblies included a second wzi gene at the K locus, referred to as wzi KL , which was found together with an itrA4 gene coding for a D-galactose (D-Gal) 1-phosphate initiating transferase described in a recent study 22 .
To assess wzi diversity in the species, a selection of 109 wzi nucleotide sequences, including five wzi KL sequences, were extracted for further analysis. Sequences selected represented a diverse pool of A. baumannii isolates that belong to a range of clonal lineages and carry different CPS biosynthesis genes at the K locus (details in Supplementary Table S2). In addition, as the first sugar of the CPS is the base of the structure proximal to the cell surface that is most likely to interact with Wzi, the first sugar of each CPS type was also predicted. This was achieved via the identification of the Itr initiating transferase encoded by the CPS biosynthesis gene cluster in each genome (Supplementary Table S2). In A. baumannii, the linkage of the first sugar to the inner membrane lipid carrier to begin CPS synthesis is catalysed by one of six Itr initiating transferase enzymes, for which the sugar substrates have been experimentally confirmed or deduced for > 65 A. baumannii CPS structures determined to date. Itr enzymes and their associated sugar substrates are listed in Table 2.
A maximum likelihood tree ( Supplementary Fig. S1) of the 109 wzi nucleotide sequences revealed five distinct phylogenetic clades, one of which occurred as a separate lineage and included all five wzi KL sequences. Sequences from the wzi KL clade share 64-67% nucleotide sequence identity with wzi sequences from the other four clades  Table S3), suggesting an import of the wzi KL sequence from outside the species as proposed previously 22 . Therefore, a second maximum likelihood tree was constructed with only wzi sequences located at the chromosomal wzi locus. Four major phylogenetic clades were again observed defining four wzi sequence types (Fig. 5, numbered in blue). Sequences belonging to the same clade/type share > 92% nucleotide sequence identity (Supplementary Table S3) indicating a conserved relationship. The AB5075-UW wzi sequence was found to belong to a homogenous lineage within clade 1, which includes mostly isolates belonging to GC1 (ST1 and single locus variants, SLVs, of ST1). Though there is a diverse array of STs (ring B in Fig. 5) represented in each clade, GC2 (ST2 and SLVs) isolates group together in Clade 3 along with isolates belonging to ST25, another important multi-drug resistant clonal lineage 23 . A total of 64 KL types are also represented in the phylogeny (ring A, Fig. 5), and interestingly, there is no observed association between the clade and KL type, suggesting that the overall composition and/or topology of the CPS structure is irrespective of the wzi sequence.
A larger proportion of isolates within the same wzi clade encoded the same itr initiating transferase gene (shown in ring C, Fig. 5). For example, most isolates with wzi type 1 (clade 1) carry an itrA1 gene predicting N-acetylbacillosamine (D-QuiNAc4NR) as the first sugar of the CPS structure. In comparison, the itrA2 gene, which predicts a N-acetylgalactosamine (D-GalNAc) first sugar, is observed more frequently with wzi type 2 (clade 2) and wzi type 3 (clade 3), while itrA3 predicting a N-acetylglucosamine (D-GlcNAc) first sugar is more often found in isolates with wzi type 4 (clade 4). This observed trend suggests that wzi sequence types may have an association with the type of proximal sugar at the base of the CPS structure.
Co-occurrence of wzi and itr gene types. To more widely test wzi-itr gene associations, the relative frequency of wzi types co-occurring with specific itr genes was examined in 9342 A. baumannii genome assemblies (Supplementary Table S1). The number of genomes encoding each wzi type was examined relative to the number of genomes encoding each itr gene (Supplementary Table S4) and plotted in Fig. 6A. With the exception of wzi KL type 5, which only co-occurred with itrA4 (see above), almost all itr genes can be found in genomes carrying each wzi type. However, the majority of genomes (n = 7295) possess wzi type 3 (clade 3), of which 5092 (~ 70%) also carry itrA2 (Fig. 6A). The second most common was wzi type 1 (n = 1270) with 1047 genomes (82%) encoding itrA1 and 119 (9%) having itrB3. Type 2 and type 4 wzi were found more frequently in genomes with itrA2 and itrA3, respectively. www.nature.com/scientificreports/ The frequency of wzi types co-occurring with itr genes was further visualized by a mosaic plot (Fig. 6B), which displays the results of a chi-squared test for independence with colours indicating deviation from the expected frequency (residual). The plot in Fig. 6B shows significant positive residuals for seven different combinations of wzi and itr types (blue tiles), with the wzi type 3/itrA2 combination displayed as the largest observed group (shown by the largest tile size). The observed association between wzi and itr sequences suggests that Wzi types may have specificity for the proximal sugar of the CPS substrate should interaction between the two molecules exist.
The predicted tertiary structure of Wzi. A single isolate from each wzi clade was chosen as a representative of each Wzi type (boxed in red in Fig. 5), and a percentage identity matrix of the translated Wzi sequences showed that the amino acid sequence identities of types 1 to 4 ranged between 88.12% and 93.75% (Table 3). A multiple sequence alignment ( Supplementary Fig. S2) indicated that sequence differences in the four Wzi types lie predominately in the central portion (amino acid range 230-320) of the 480 aa protein sequence. To explore the potential functional significance of these amino acid differences, the relationship of A. baumannii Wzi to the established tertiary structure of Wzi from E. coli O9a:K30 was examined.
The tertiary structure of the AB5075-UW Wzi (ABUW_2898; Type 1) was modeled using Phyre 2 software, which revealed a match of 100% confidence with 48% identity (94% coverage: residues 23-476) to the E. coli   15 . The predicted type 1 Wzi three-dimensional structure (Fig. 7A) consists of an 18-stranded β-barrel fold that includes a helical periplasmic bundle and an arrangement of nine loops on the extracellular side. This is consistent with the E. coli O9a:K30 Wzi structure, and an overlay of the two demonstrates a close match (Fig. 7B). Previously, an interaction between the E. coli K30 CPS structure and the extracellular loops of Wzi was demonstrated, with L3, L6 and L7 loops found to be critical for CPS assembly on the cell surface 15 . Therefore, the extracellular loops were identified in the modeled tertiary structure of the AB5075-UW Wzi (numbered L1-L9 in Fig. 7A), and the amino acid sequences of the loops were identified in the multiple sequence alignment of the four A. baumannii Wzi types (Supplementary Fig. S2). The predicted L3, L6 and L7 were found to be mostly conserved, whereas the majority of amino acid sequence differences between all types (range 230-320) were identified in the sequence of L5. Further work will be needed to directly assess the importance of the extracellular loops in Wzi function in A. baumannii.

Discussion
The A. baumannii CPS affords a remarkable amount of cell-surface diversity between different isolates with more than 128 distinct structural forms predicted 14 . The intrinsic capacity of the organism to frequently exchange and replace CPS biosynthesis genes 24 indicates that there are likely many more CPS forms yet to be discovered. This extreme heterogeneity complicates therapeutic strategies that target specific CPS structures. However, damaging or completely removing this surface barrier could present a viable alternative approach, yet research on the precise functions of A. baumannii CPS biosynthesis proteins is still developing.
In this study, we confirm that Wzi is required for the proper assembly of a tight CPS layer on the cell surface, with the deletion of wzi resulting in a reduction of cell-associated CPS with a reciprocal increase of CPS in the extracellular surrounds. Though a small amount of CPS material (~ 20%) is still observed on the cell surface in wzi mutants, it may be CPS actively passing through the outer membrane completing its export to the cell surface. Nonetheless, the integrity of the CPS barrier as a discrete layer surrounding the cell is compromised. www.nature.com/scientificreports/ Analysis of the AB5075-UW wildtype showed a small amount of cell-free CPS. In a previous study, both cellbound and cell-free CPS material has been detected for another wildtype isolate 10 , suggesting that CPS shedding may be a natural phenomenon in the species regulated by Wzi expression. Thus, it is possible that regulation of CPS retention on the cell surface may be necessary for different environmental contexts and stresses or during different stages of host infection. Further work will be needed to assess the expression of Wzi while under stress, and to confirm the promoter region and any potential regulators.
The importance of proper assembly of CPS on the cell surface for biofilm formation was also assessed as previous studies have shown that the itr initiating transferase is critical for the development of uniform biofilm structures 8 . Interestingly, the retention of CPS on the cell surface was shown to have little importance for biofilm, though a marked reduction was observed for the itrA1 mutant as expected 8,20 , indicating that CPS presence rather than cell-surface retention is important. However, in A. baumannii, K-unit oligosaccharides that make up the CPS are also used for O-glycosylation of proteins 8 , therefore loss of O-glycosylation via an itr deletion may account for biofilm defects. Though a role for Wzi in biofilm formation was not established, a previous study has shown that an AB5075 wzi (ABUW_2898) mutant displays decreased resistance to normal human serum 25 . This suggests that the proper assembly and retention of the CPS on the cell surface is likely to play a more significant role in the evasion of human immune defenses.  Table 3. Percentage amino acid sequence identity of wzi types. a Type 1 = AB5075-UW (ABUW_2898 in CP008706.1), type 2 = ATCC17978 (AUO97_12210 in CP018664.1), type 3 = ACICU (DMO12_03033 in CP031380.1), type 4 = LAC4 (BBX32_12830 in CP018677.1). www.nature.com/scientificreports/ The chromosomal K locus is a recombination hotspot 24 , and while nearly all wzi genes are located away from this genomic region in A. baumannii (Fig. 1A), sequence diversity in wzi has so far gone undetected. The amount of sequence variation at the wzi locus was a surprising finding, exceeding that observed for the intrinsic oxaAB gene with > 95% nucleotide sequence identities identified across all genomes (Supplementary Table S1). However, as five different wzi types were identified amongst > 9300 A. baumannii genomes, it is possible that multiple imports of the wzi gene into the species have occurred. The results in this study indicate that variation at the wzi and K locus is not concomitant, suggesting that the two regions are evolving independently. However, a significant association between the wzi type and the first sugar of the CPS structure was observed. This raises the possibility that acquisition, and successful maintenance of K locus types is restricted by the itr initiating transferase gene due to the possible dependence on the wzi gene type to produce a tight CPS layer on the cell surface. Further studies will be needed to confirm if Wzi types have stringent substrate specificities.

Materials and methods
Bacterial strains and cultivation. Acinetobacter baumannii AB5075-UW (K25 capsule type 26 ) and derivative T26 insertion mutants (Table 1) were obtained from the A. baumannii AB5075 transposon mutant library 27 . Bacteria were routinely grown in Luria-Bertani (LB) media at 37 °C with or without shaking at 200 rpm, and complemented strains were selected used 50 ug/ml apramycin. Optical density (OD) was measured by absorbance at 600 nm using a densitometer.
Whole genome sequencing and analysis. Whole genome sequences were obtained by Illumina Next-Seq 500 using the Nextera-XT Library at the Forensic and Scientific Services Laboratory (QLD Health, Brisbane). Paired-end short read data were de novo assembled into contigs using the SPAdes algorithm optimized via the Unicycler v 0.4.8 assembly pipeline 28 . The locations of transposon insertions were identified using Clustal Omega 29 . Snippy v 3.2 (https:// github. com/ tseem ann/ snippy) was used to determine the number of single nucleotide polymorphisms (SNPs) between mutant sequences and the complete reference genome sequence of wild type AB5075-UW (GenBank accession number CP008706).
Complementation of A. baumannii wzi::T26 mutants. Complementation of wzi::T26 mutants was performed using an overlap extension PCR protocol to generate a chimeric cassette (6.5 kb) targeted for insertion into the attTn7 site in the AB5075 chromosome 30 . Briefly, primary PCR reactions using oligonucleotides (listed in Supplementary Table S5) consisting of additional sequence homologous to the intended adjacent amplicon were used to generate individual cassette fragments (Fig. 1C). These fragments included 2 kb of sequence identical to either side of the attTn7 insertion site, the predicted wzi promoter region and wzi gene from AB5075 (GenBank accession number CP008706.1, base range 2,905,472 to 2,906,917) and an aacC4 gene (GenBank accession number KM670336.1, base range 139,244-140,020) conferring resistance to apramycin. Amplicons were assembled into a single cassette using a secondary PCR reaction with the outermost oligonucleotides. PCR was performed using a high-fidelity DNA polymerase (Phusion, Thermo Fisher), and the assembled cassette was Sanger sequenced for confirmation.
Complementation was performed using natural transformation of the chimeric cassette into wzi-deficient transposon mutants (wzi::T26-1, wzi::T26-2, and wzi::T26-3). LB media was inoculated with a single colony of a mutant strain then incubated at 37 °C until OD 600nm 1.0 was achieved, then cultures were diluted 1:100 in Tryptone medium (5 g/L). Equal volumes of diluted culture and cassette DNA (200 ng/ml) were mixed and then spotted on Tryptone medium containing 2% low electroendosmosis agar in a microcentrifuge tube. Following www.nature.com/scientificreports/ overnight incubation at 37 °C, cells were plated onto LB agar containing 50 mg/ml apramycin and resulting colonies were screened by PCR to identify transformants. Successful transformants were confirmed by PCR across insertion sites and Sanger sequencing.

CPS extraction and SDS-PAGE visualization.
Prior to extraction, overnight cultures at equivalent OD 600nm were centrifuged to separate cells from the supernatant. CPS attached to the cell surface was purified from whole-cell lysates (cell fractions) using the phenol:chloroform extraction method described previously 17,31 . CPS shed from the cell surface was extracted from the supernatant fractions of the same cultures by adding 4 × volume of ice-cold ethanol and precipitating at − 20 °C overnight. Samples were then centrifuged 12,000 rpm at 4 °C for 15 min, and pellets were resuspended in sterile milli-Q water. Equivalent volumes of purified CPS samples were subjected to SDS-PAGE (4% stacking and 16% separating tricine gels), and gels were stained overnight in Alcian blue buffer (25% isopropanol, 7% acetic acid solution and 0.05% w/v Alcian blue) as described previously 32 . Gels were imaged using a ChemiDoc XRS gel imaging system. Densitometric analysis was performed using ImageJ software 33 .

Bioinformatics analysis.
A total of 9459 genome assemblies listed under taxonomy Acinetobacter baumannii were downloaded from NCBI (https:// www. ncbi. nlm. nih. gov/ assem bly/) on the 26th of August, 2021. The presence of the intrinsic oxaAB gene was used to confirm the species as baumannii as performed previously 14 , and oxaAB-negative genomes were removed from further analyses. Genomes were initially screened for wzi type 1 from AB5075-UW (locus tag ABUW_2898 in GenBank accession number CP008706.1). Matches with < 100% coverage and/or < 80% identity were then assessed for sequence quality using QUAST (http:// bioinf. spbau. ru/ quast). Poor quality genomes, defined using parameter described previously in Wyres et al. 2020 (> 300 contigs and/or total length of genome < 3.6 Mb), and genomes where wzi was either found broken across more than one contig or < 100% coverage to wzi type 1, were also removed. Therefore, a total of 9342 genome assemblies (listed in Supplementary Table S1) were included in further analyses. Representative wzi nucleotide sequences were extracted from 109 genome assemblies (see Supplementary  Table S2 for accession numbers and traits). K locus types were assigned using command-line Kaptive v 0.7 (https:// github. com/ katho lt/ Kapti ve) with the A. baumannii KL reference database and default parameters 14 . Sequence types were assigned using the A. baumannii Pasteur Multi-locus sequence typing (MLST) scheme available at https:// pubml st. org/ bigsdb? db= pubml st_ abaum annii_ paste ur_ seqdef.
A multiple pairwise wzi nucleotide sequence alignment was constructed using MUSCLE (https:// www. ebi. ac. uk/ Tools/ msa/ muscle/), and used to infer a maximum likelihood tree based on 1000 replicates using RAxML v 8.0.0 with the GTRGAMMA model 34 . The resulting tree was visualized, mid-rooted, and annotated using the iTOL web-interface 35 and edited using Adobe Illustrator. Stacked bar and mosaic plots were created in RStudio v 1.2.5033 36 using ggplot 37 and vcd 38 packages.
Percentage identity matrices and multiple pairwise alignments to visualise nucleotide and protein sequences were constructed using CLUSTAL Omega (https:// www. ebi. ac. uk/ Tools/ msa/ clust alo/). Phyre 2,39 was used to predict the tertiary structure of the Wzi protein from AB5075-UW via sequence threading to the E. coli K30 Wzi reference (PDB: 2YNK). The predicted structure was visualized using Chimera software 40 .