A sophisticated virulence repertoire and colistin resistance of Citrobacter freundii ST150 from a patient with sepsis admitted to ICU in a tertiary care hospital in Uganda, East Africa: Insight from genomic and molecular docking analyses

Abstract

Sepsis and multidrug resistance comprise a complex of factors attributable to mortality among intensive care unit (ICU) patients globally. Pathogens implicated in sepsis are diverse, and their virulence and drug resistance remain elusive. From a tertiary care hospital ICU in Uganda, we isolated a Citrobacter freundii strain RSM030 from a patient with sepsis and phenotypically tested it against a panel of 16 antibiotics including imipenem levofloxacin, cotrimoxazole and colistin, among others. We sequenced the organism’s genome and integrated multilocus sequencing (MLST), PathogenFinder with Virulence Factor analyzer (VFanalyzer) to establish its pathogenic relevance. Thereafter, we combined antiSMASH and PRISM genome mining with molecular docking to predict biosynthetic gene clusters (BGCs), pathways, toxin structures and their potential targets in-silico. Finally, we coupled ResFinder with comprehensive antibiotic resistance database (CARD) to scrutinize the genomic antimicrobial resistance profile of the isolate. From PathogenFinder and MLST, this organism was confirmed to be a human pathogen (p = 0.843), sequence type (ST)150, whose virulence is determined by chromosomal type III secretion system (T3SS) (the injectosome) and plasmid-encoded type IV secretion system (T4SS), the enterobactin biosynthetic gene cluster and biofilm formation through the pgaABCD operon. Pathway and molecular docking analyses revealed that the shikimate pathway can generate a toxin targeting multiple host proteins including spectrin, detector of cytokinesis protein 2 (Dock2) and plasmalemma vesicle-associated protein (PLVAP), potentially distorting the host cell integrity. From phenotypic antibiotic testing, we found indeterminate results for amoxicillin/clavulanate and levofloxacin, with resistance to cotrimoxazole and colistin. Detailed genome analysis revealed chromosomal beta lactam resistance genes, i.e. blaCMY-79, blaCMY-116 and blaTEM-1B, along with multiple mutations of the lipopolysaccharide modifying operon genes PmrA/PmrB, pmrD, mgrA/mgrB and PhoP/PhoQ, conferring colistin resistance. From these findings, we infer that Citrobacter freundii strain RSM030 is implicated in sepsis and resistance to standard antibiotics, including colistin, the last resort.

1. Introduction

Sepsis is a major cause of patient mortality in intensive care units globally. While Gram positive pathogens such as Staphylococcus and Streptococcus are common, most pathogens implicated in sepsis include Gram negative bacteria of the family Enterobacteriaceae along with members of the ESKAPE group. Commonly reported include Escherichia coli (E. coli), Klebsiella pneumoniae (K. pneumoniae) and emerging Citrobacter species (Chen et al., 2023; Liu et al., 2020). Being the most frequent pathogenic species Citrobacter koseri and Citrobacter freundii have been implicated in infections of the urinary tract, gastrointestinal tract, respiratory tract blood stream, causing serious septic conditions and deaths (Anderson et al., 2018; Liu et al., 2018; Song et al., 2022). Among the virulence factors determining the pathogenesis of Citrobacter spp. include fimbriae, toxins and secretions systems including type IV and type VI (Anderson et al., 2018; Liu et al., 2021). Evidence supports the emergence of multidrug resistance strains against beta lactams, aminoglycosides, sulfonamides, tetracyclines and phenicols, among others (Liu et al., 2020, 2018). Although cases of Citrobacter spp. infection are generally rare among adults, reports are emerging, implicating the role of C. freundii in sepsis even in adult humans (Chen et al., 2023). Recent clinical surveys from East Africa have reported C. freundii among sepsis-associated pathogens from Kenya (Williams et al., 2009) and Uganda (Mayanja et al., 2023), suggesting existence of possibly unnoticed strains in the clinics. Despite this emerging evidence for sepsis and AMR, the association of C. freundii with comorbidities such as sickle cell is poorly known. In addition, it is unknown whether genome of C. freundii harbors genes that could aggravate sickle cell prognosis. Furthermore, phenotypic and genotypic determinants of colistin resistance among C. freundii strains are yet to be reported from East Africa.

In this work, therefore, we aimed at unveiling the pathogen, its genome-based virulence and antimicrobial resistance determinants of sepsis in a patient admitted to ICU for advanced respiratory support. We hypothesized that C. freundii strain RSM030 isolated from this patient possesses virulence factors related to sepsis, which could contribute to the need for advanced respiratory support, in addition to possession of antimicrobial resistance. Here, we report a sophisticated genomic virulence repertoire, colistin resistance of Citrobacter freundii ST150, named strain RSM030 potentially causing sepsis and increasing the risk of respiratory complication in the patient from a tertiary care hospital in Uganda.

2. Materials and methods

2.1. Patients’ brief information

A 34-year-old male patient was admitted in the ICU of Mulago National Referral Hospital with sepsis and respiratory complications. The patient was supported with a ventilator, urinary catheter, nasogastric tube, central line, peripheral line and tracheal tube. The patient had no previous device, ICU history and nor was he under antibiotic regimen before admission. This patient was diagnosed with sepsis and received an intravenous (IV) injection of meropenem.

2.2. Specimen collection and isolation of bacterial pathogens

Blood sample was collected from the median cubital vein using a sterile syringe and transported under ice to Medical Microbiology Laboratory of Makerere College of Health Sciences (MakCHS) and kept under Brain heart both (BHI) containing 30% glycerol before isolation. The sample was suspended in 5 ml Buffered Peptone Water (BPW), vortexed for complete mixing, and incubated overnight for 24 h at 37C in ambient air. Thereafter, an aliquot of 10 μl from each sample was inoculated onto an in-house selective MacConkey agar containing .5μg/ml of colistin sulfate and 305μg/ml of ampicillin, followed by aerobic incubation at 35–37 °C for 24 h. The plates were then examined for the bacterial colonies based on lactose fermentation and other biochemical methods including urease, citrate fermentation test, hydrogen sulfide production, and indole motility test.

2.3. Antibiotic susceptibility testing, genome sequencing and quality control of sequence reads

A panel of 16 antibiotics was used for antimicrobial susceptibly was performed according to the Kirby-Bauer disc diffusion method, or the broth microdilution method (for colistin) (Table 1). The interpretation of susceptibility and resistance was based on the guidelines provided by the Clinical and Laboratory Standards Institute (CLSI). To determine colistin MIC, broth microdilution method (BMD) was used as the standard susceptibility test method recommended by CLSI and EUCAST for polymyxins. DNA extraction using a ZymoBIOMICS DNA Miniprep Kit (ZR D4300), according to the manufacturer’s instructions. A TruSeq DNA PCR-Free kit was used to construct the library, whole genome sequencing was performed with Illumina Novaseq 6000. Quality control of the raw reads was performed using FASTQC (v0.115) based on quality Phred score cutoff of 20 and maximum trimming error rate of 0.1.

Table 1.

The antibiotic panel, AST and colistin test results.

AST and ESBL					Colistin
Antibiotic	Amount	IZ (mm)	AST result	ESBL phenotype	Col MIC (mg/L)	Phenotype
Amoxicillin/clavulanic acid	20 μg/10 μg	17	I
Cefuroxime (CXM)	30 μg	22	S
Ceftazidime (CAZ)	30 μg	25	S
Ceftriaxone (CRO)	30 μg	25	S
Cefepime (FEP)	30 μg	29	S
Imipenem (IMI)	10 μg	24	S
Piperacillin-tazobactam (TPZ)	100 μg/10 μg	21	S
Chloramphenicol	30 μg	30	S
Gentamycin (GEN)	10 μg	16	S
Amikacin (AK)	30 μg	19	S
Trimethoprim-sulfamethoxazole	1.25 μg/23.75 μg	6	R
Ciprofloxacin (CIP)	5 μg	22	I
Levofloxacin (LVX)	5 μg	22	I
Fosfomycin (FF),	200 μg	26	S
Tigecycline (TGC)	15 μg	26	S
Colistin (CT)	10 μg (MIC = 4)	13	I	Negative	4	Resistant

AST = Antimicrobial susceptibility test, ESB = Extended spectrum beta lactamase, IZ = Inhibition zone.

2.4. Genome assembly, annotation and pathogen phylogrouping

Quality reads were de novo-assembled into contigs using Unicycler v0.48 (minimum contig length 300 bp) (Wick et al., 2017). These contigs were subjected to BV-BRC comprehensive annotation service (https://www.bv-brc.org/app/ComprehensiveGenomeAnalysis) to predict the closest relatives and specialty genes including AMR genes. Then the genome sequences of the closest relatives were downloaded from the National Center for Biotechnology Information (NCBI) and combined with our study genome to generate a proteome-based phylogeny using the TYGS platform (Meier-Kolthoff and Göker, 2019). Then the genomes of strains that clustered in the same clade with our strain RSM030 were chosen for average nucleotide identity (ANI) estimation using the ANI calculator accessible at https://www.ezbiocloud.net/tools/ani. To predict relevant serotype, the contigs were analyzed using SerotypeFinder-2.0 (https://cge.food.dtu.dk/services/SerotypeFinder/): the selected percentage identity threshold was set to 85%, with minimum length of 60%. Plasmids were separately assembled using plasmidSPAdes v3.15.5 (Antipov et al., 2016) and typing was performed Pathogenwatch (Argimón et al., 2021), using Inc. typing method to search for corresponding plasmids from the PlasmidFinder database https://cge.cbs.dtu.dk/services/PlasmidFinder/. The contigs were further assembled into a chromosome with default parameters of Medusa v1.3 (Bosi et al., 2015). Then the chromosome and plasmids were annotated with the Prokaryotic Genome Annotation Pipeline (PGAP) from NCBI available at https://www.ncbi.nlm.nih.gov/genome/annotation_prok/.

2.5. Genomic analysis of virulence factors

Known virulence factors were identified using the virulence factor database (VFDB) (Liu et al., 2022). Virulence factors related to siderophores and enterotoxins were analyzed in detail by subjecting the contigs to version 7 of the antibiotics and secondary metabolite analysis shell’ (antiSMASH) (Blin et al., 2023) for biosynthetic gene clusters (BGCs) mining through the polyketide synthase (PKS) and nonribosomal peptide synthetase (NRP) pathways. Molecular assembly and structural prediction of the encoded putative toxins were performed using the PRediction Informatics for Secondary Metabolomes (PRISM) 4 toolkit (Skinnider et al., 2020).

2.6. In silico target prediction of putative toxins

Both structures of the known toxin enterobactin, retrieved from PubChem (Enterobactin | C30H27N3O15 | CID 34231 - PubChem (nih.gov)), and the candidate virulence factor scaffold structurally elucidated from PRISM were converted into the structured data file (SDF) format using PyMol v2.4 available at PyMOL | pymol.org. Structures of potential molecular targets were retrieved from the Protein Data Bank (PDB, https://www.rcsb.org/). Water molecules were removed using VSTH (Mo et al., 2023) and then molecular docking was performed using Autodock Vina (Eberhardt et al., 2021) under Seamdock (Murail et al., 2021). Autodock Vina docking parameters were set to mode = 2, energy range = 5 and exhaustiveness = 10.

2.7. Genomic analysis of antimicrobial resistance profiles

Both chromosomes and plasmids were analyzed for resistance genotypes with respect to specific drugs using ResFinder (Florensa et al., 2022) with default parameters as follows; selected percentage identity threshold for ResFinder was set to 90%, minimum length of 60%, selected percentage identity threshold for PointFinder was 90% and minimum length for PointFinder was 60%. For detailed genome-based mechanisms of resistance, each genome was analyzed using the Resistance Gene Identifier (RGI) to predict resistome(s) based on homology and SNP models of the comprehensive antimicrobial resistance database (CARD) (Alcock et al., 2020).

2.8. Ethical consideration

The study obtained ethical approval from the Higher Degree and Graduate Research Ethics Committee (HDREC) of School of Biomedical Sciences, Makerere University College of Health Sciences and the Uganda National Council of Science and Technology (UNCST) (SBS-2021–47).

3. Results

3.1. Antimicrobial resistance phenotypes

From antimicrobial susceptibility testing, the strain RSM030 was susceptible to 11 antibiotics as summarized in Table 1. The results showed indeterminate phenotypes for (amoxicillin/clavulanate, ciprofloxacin, levofloxacin and colistin but resistance phenotypes for cotrimoxazole. In addition, the strain was found to be colistin resistant from broth microdilution colistin MIC results.

3.2. Genome sequencing assembly and annotation results

A total of 22,656,902 reads were generated, with maximum read sequence length of 151 bp and average length of 149 bp. Following quality control, the reads were reduced to 14,567,396. A total of 108 contigs were generated from assembly, with annotation features summarized in Table 2 From Pathogenwatch search for incompatibility group I1 (IncI1) plasmids, the first plasmid was identical to IncFII(pCRY)_1 (95.28%)), carried by Yersinia pestis biovar Microtus str. 91,001 (GenBank accession no. NC_005814). The second plasmid was identical to Col440I_1 (94.74%), characterized from Klebsiella pneumoniae strain FDAARGOS_440 (accession no. CP023920). The two RSM030 plasmids were successfully assembled, given identifiers pRSM030_p1 and pRSM030_p2 and deposited into the NCBI GenBank (accession nos CP133061.1 and CP133062.1).

Table 2.

Assembly and annotation features of Citrobacter freundii strain RSM030. Plasmid types recovered from Pathogenwatch are shown.

Isolate	Contigs	Genome size	GC (%)	N50	Total CDS	tRNAs	rRNAs	CRISPR arrays
RSM030	108	4,928,315	51.76	122,686	4871	74	3	0
Isolate	Plasmid type		Percentage identity (%)		Hit organism			Hit accession
RSM030	IncFII(pCRY)_1		95.28		Y. pestis str. 91,001			NC_005814
	Col440I_1		94.74		K. pneumoniae str. FDAARGOS_440			CP023920

3.3. Sequence typing and phylogrouping

Multilocus sequence typing (MLST) involved seven housekeeping genes aspC, clpX, fadD, mdh, arcA, dnaG, lysP from the database https://pubmlst.org/cfreundii, and assigned our isolate the sequence type ST150, which was yet to be reported among Citrobacter spp. From ANI analysis, the closest relative strains were C. freundii ATCC 8090 MTCC 1658 NBRC 1268 (GCF_011064845.1, ANI = 98.85%), C. freundii DY2010 (GCA_020809005.1, ANI = 98.97), C. freundii DY2007 (GCF_020639355.1, ANI = 99.00%). TYGS phylogenetic analysis also clustered RSM030 in the same clade with ATCC 8090 MTCC 1658 NBRC 1268 along with DY2010 and DY2007 (Fig. 1). While strain ATCC 8090 was originally isolated from environmental samples, the two closest strains DY2010 and DY2007 were isolated from urine and blood samples respectively in China (Ye et al., 2023).

Fig. 1.

Phylogrouping of C. freundii RSM030. Phylogenetic grouping of C. freundii strain RSM030 from TYGS proteome-based phylogenetics. Comparative virulence and AMR genotypes are indicated by heatmaps. A) The presence of a virulence factor is marked by red while its absence is marked by yellow. B) The heatmap shows presence of an AMR genotype in yellow while its absence is indicated by blue.

3.4. Virulence factors carried by genome of Citrobacter freundii strain RSM030

With PathogenFinder the organism was identified as a human pathogen (p = 0.843). Virulence factors carried by its genome are generally categorized into adherence, invasion, iron uptake, secretion systems, antiphagocytosis, biofilm formation, motility, as well as colonization and immune evasion. Type 1 fimbriae fimD gene was detected, along with five ORFs for type IV pili related to Yersinia spp. The pathogen portrays invasive potential via invasion of brain endothelial cells (Ibes) (IbeB and IbeC), also common in other Citrobacter spp. and E. coli strains as shown in Fig. 1. In addition, this pathogen possesses type III secretion system (T3SS, the injectosome) and type IV secretion systems (T4SS), which constitute a strong machinery of infection and pathogenesis (Liu et al., 2021). We noted that, the T4SS apparatus is encoded by the plasmid pRSM030_p2 (https://www.ncbi.nlm.nih.gov/nuccore/CP133062.1) while T3SS is encoded by the chromosome (https://www.ncbi.nlm.nih.gov/nuccore/CP133060.1).

Confirmed from PGAP annotation, the injectosome apparatus includes virtually all essential units including the chaperones (SycD/LcrH), structural proteins or translocons (PrgH, SctC SctE, SctF) and effector components (IpaC/SipC, IpaD). We also noted the biofilm potential and reconstructed the biosynthetic pathway of poly-beta-1,6-N-acetyl-D-glucosamine (Fig. 2) from the genes encoding the catalytic reaction for each step, i.e. poly-beta-1,6 N-acetyl-D-glucosamine export porin PgaA (accession no. WME29053.1), poly-beta-1,6-N-acetyl-D-glucosamine N-deacetylase PgaB, poly-beta-1,6-N-acetyl-D-glucosamine synthase (PgaC) (accession no. WME29055.1) and poly-beta-1,6-N-acetyl-D-glucosamine biosynthesis protein PgaD (WME29056.1).

Fig. 2.

The biosynthetic pathway of poly-beta-1,6-N-acetyl-D-glucosamine, which represents a major component of biofilm. The names of enzymes catalyzing the reactions are provided with their GenBank accession numbers in parentheses.

Analysis from antiSMASH and PRISM predicted two BGCs: a NRP cluster 1 and polyketide (PK) cluster 2. Structural assembly generated a scaffold (Fig. 2), which with Chemdraw v8 (Mendelsohn, 2004), we assigned the IUPAC name 2-[(2,3-dihydroxybenzoyl)amino]-3-hydroxypropanoic acid (compound 5). We further used the Biocyc database (https://biocyc.org/) to find if this metabolite is part of any known biosynthetic pathway. As expected, we found that this nonribosomal peptide is derived from the shikimic acid pathway, as an intermediate in the biosynthesis of enterobactin, as elucidated from E. coli K2 (https://biocyc.org/ECOLI/NEW-IMAGE?type=PATHWAY&object=ENTBACSYN-PWY). We then mapped the gene cluster on the chromosome and reconstructed the biosynthetic pathway for enterobactin and located the position of compound 5 in the pathway (Fig. 3), which indicates that the metabolite is at the last but one step in the pathway. While enterobactin (compound 6 is the last metabolite in the pathway, the cluster clearly shows that both compounds 5 and 6 are amenable to ATPase binding cassette (ABC)-mediated export and import via FecCD and TonB proteins.

Fig. 3.

A schematic representation of the virulence machinery comprising the most predominant facets. The biosynthetic gene cluster and pathway of compound 5 and enterobactin (6) are shown below the cytoplasmic membrane (CPM). The gene TonB is flanked between EntF and EntD, suggesting its crucial role in export of both 5 and 6.

From our molecular docking, compound 5 exhibits affinity to multiple cellular targets particularly those with functional activity in cell integrity and survival. The putative victim receptors include tubulin (PDB accession 1TUB, affinity −8.2 kcal/mol), N-terminal region of AlphaII-spectrin Tetramerization Domain of spectrin (accession 3F31, affinity −7.2 kcal/mol), erythroid beta spectrin repeats 14 and 15 (ankyrin binding domain, 5F57, affinity = −6.0 kcal/mol) plasmalemma vesicle associated protein (PLVAP, accession 8FBY, affinity −6.4 kcal/mol) and detector of cytokinesis protein 2 (Dock2, accession 3A98, affinity −6.0 kcal/mol) (Fig. 4A–E). These proteins are essential in the maintenance of red blood cells, endothelial and immune cells such as the macrophages (Chang et al., 2023; Ipsaro and Mondragón, 2010). Details of the intermolecular forces and target amino acids are presented in Fig. 4F and supplementary Table 1. We observed that compound 5 interacts with most targets mainly with hydrogen bonding, followed by hydrophobic interaction. Using SwissADME (Daina et al., 2017), we found that that compound 5 is water soluble (solubility =7.51e+01 mg/ml; 3.11e-01 mol/l) and possesses high gastrointestinal absorptivity, indicating that it can readily diffuse from the gut to the blood stream, potentially undergoing systemic distribution.

Fig. 4.

Ligand-receptor interaction of compound 5 with selected targets: A) tubulin (1TUB), B) N-terminal region of AlphaII-spectrin Tetramerization Domain of spectrin (3F31), C) erythroid beta spectrin repeats 14 and 15 (5F57), D) plasmalemma vesicle associated protein (PLVAP, 8FBY), and E) detector of cytokinesis protein 2 (Dock2, 3A98). The energy and number of amino acids involved along with the type of intermolecular forces are shown in F). Details about types of amino acids are presented in the supplementary material (Supplementary Table 1).

3.5. The genomic antimicrobial resistance repertoire of Citrobacter freundii strain RSM030

ResFinder analysis revealed three chromosome-encoded beta-lactam resistance genes i.e., blaCMY-79, blaCMY-116 and blaTEM-1B, which collectively determine resistance to amoxicillin, amoxicillin+clavulanic acid, ampicillin, ampicillin+clavulanic acid, cefotaxime, cefoxitin, ceftazidime, piperacillin, piperacillin+tazobactam, ticarcillin, ticarcillin+clavulanic acid. These were associated with mobile genetic elements of the families IS3 (ISSen4) and IS5 (ISKpn26 and cn_12161_ISKpn26), which have been demonstrated to play role in the transmission of antimicrobial resistance genes (Galiot et al., 2023). We also identified 26 single nucleotide polymorphisms (SNPs) for the RNA polymerase subunit B gene, including rpoB:p.A1193S, rpoB:p.Y1087H, rpoB:p.V282I, rpoB:p.V297A, rpoB:p.R976Y, rpoB:p.I229V, rpoB:p. P1224Q, rpoB:p.E412D, and rpoB:p.E1089A, among others. However, none of these mutations could define a significant AMR phenotype from the ResFinder database.

From CARD analysis, we found that the chromosome of RSM030 carries 17 antibiotic efflux pumps, namely H-NS, marA, KpnF, KpnE, mdtG, mdtB, mdtC, leuO, msbA, mdfA, kdpE, rsmA, emrB, emrA, acrA, AcrB and CRP. This is the most prevalent AMR mechanism in the strain RSM030. The next abundant AMR mechanisms include target alteration via SNPs and enzymatic modification of drug targets. CARD-detected SNPs include the E. coli UhpT and GlpT with mutations conferring resistance to fosfomycin (E350Q, E448K), Hemophilus influenzae PBP3 conferring resistance to beta-lactam antibiotics (D350N, S357N). Enzymatic target modifiers include vanG, known for glycopeptide antibiotics such as vancomycin (Stogios and Savchenko, 2020). Most importantly, we also detected the polymyxin resistance genes ArnD, ArnT EptA and EptB phosphoethanolamine transferases, which are known to confer colistin resistance via modification of the lipid A of lipopolysaccharide (LPS) component of the cell wall (Elizabeth et al., 2022). We also detected the beta lactamase families TEM (TEM-1), and CMY (CMY-116, conferring cephalosporin and carbapenem resistance (Ripabelli et al., 2020).

To further explore the polymyxin resistance genotypes, we performed manual search to identify individual genes from the PGAP annotation. We found that the chromosome (https://www.ncbi.nlm.nih.gov/nuccore/CP133060.1) carries the lipopolysaccharide modifying operon genes PmrA/PmrB, pmrD, mgrA/mgrB and PhoP/PhoQ. which are critical for the modification of polymyxin drug target (Urooj et al., 2022). From our polymorphism phenotyping with Polyphen 2, we noticed that the PrmA gene of RSM030 carries six nonsynonymous substitution mutations namely T79A, l116V, R118F, E126K, R139P and T151A, which have recently been demonstrated to confer colistin resistance in mcr-negative E. coli strains (Li et al., 2022). The mgrB gene encodes a short transmembrane protein that negatively regulates the function of pmrA/mgrB and PhoP/PhoQ (Ma et al., 2021). Mutations in mgrB have a significant role on overexpression of the PhoP/PhoQ regulon, exacerbating colistin resistance via decreased lipid A affinity to the drug, through the role of eptA and eptB (Elizabeth et al., 2022; Samantha and Vrielink, 2020). In our analysis, we found a mutation V8A, which also has been recently shown to confer colistin resistance in mcr-negative E. coli strains (Li et al., 2022).

The enzyme 4-deoxy-4-formamido-L-arabinose-phosphoundecaprenol deformylase (ArnD) catalyzes the addition of a formyl group to 4-deoxy-4-formamido-L-arabinose-phosphoundecaprenol to 4-amino-4-deoxy-L-arabinose-phosphoundecaprenol, modifying lipid A (Breazeale et al., 2005). Whereas the enzyme 4-amino-4-deoxy-L-arabinosyltransferase (ArnT) modifies lipid A by incorporating 4-amino-4-deoxy-l-arabinose (L-Ara4N), phosphoethanolamine transferase (EptA) catalyzes the addition of phosphoethanolamine to the phosphate groups of lipid A, decreasing the electrostatic interaction of lipid A with the α,γ-diaminobutyric acid (Dab) residue of the polymyxin drugs (Samantha and Vrielink, 2020).

From all our AMR analyses, we found that neither plasmid pRSM030_p1 nor pRSM030_p2 carries any AMR gene. Instead, both are fertility plasmids carrying genes for sex pilus such as type IV and type F conjugative transfer system pilins along with several mobile genetic elements including the TnpB, IS1, IS5 and other transposable elements. Although the IS1 and IS5 TEs are strongly associated with AMR genotypes (Galiot et al., 2023), they do not flank any AMR gene in both plasmids. The most remarkable cassette in the plasmid pRSM030_p2 is the T4SS, which accounts for virulence rather than AMR.

Intrigued by the MIC-observed colistin resistance phenotype, we further wanted to understand the potential risk of colistin resistance of the chromosome-encoded mcr-like protein EptA. We then used antibiotic candidate protein homology search with the Comer web server (Dapkūnas and Margelevičius, 2023) to predict the most relevant AMR gene from the databases. Here we note that although chromosome-borne, the RSM030 EptA protein (GenBank accession no. WME28802.1) portrays significant homology with polymyxin resistance proteins such as the EptA (https://www.rcsb.org/structure/5FGN, Neisseria meningitidis), the intrinsic colistin resistance enzyme ICR(Mc) (https://www.rcsb.org/structure/6BNC, Moraxella catarrhalis), and MCR-1-S (https://www.rcsb.org/structure/7WAA, E. coli). We went on to elucidate the functional features of this EptA (WME28802) by scanning the peptide sequence through InterProScan (https://www.ebi.ac.uk/interpro/search/sequence) and found that the protein carries the eptA-like-N EptA-like_N (IPR012549), EptA_B_N (PF08019) domains, the sulfatase_N (IPR000917 and sulfatase (PF00884) domains and belongs to the PEA transferase family, which are molecular signatures shared among the mcr genes. Most importantly it shares the same catalytic alkaline-phosphatase-like, core domain superfamily and the alkaline phosphatase subunit A (G3DSA:3.40.720.10) with in the mcr-1 gene, which confers colistin resistance (Pathak et al., 2020) (Fig. 5).

Fig. 5.

Domains and other functional signatures shared by phosphoethanolamine transferases EptA, LptA of RSM030 and mcr-1/2 genes from E. coli.

4. Discussion

Citrobacter spp. are emerging among common pathogens associated with septic shock and mortality in Africa (Lewis et al., 2019). Here we report Citrobacter freundii ST150, with virulence factors suggesting its causal association with patient sepsis. Our findings are coherent with reports from recent studies demonstrating the role of Citrobacter spp. in sepsis among ICU patients including those with respiratory difficulties from various parts (Bae et al., 2018; Cortes et al., 2023; Ranjan and Ranjan, 2013). The contributing virulence factors include the invasion of brain endothelial cells (Ibes), secretion systems (T3SS, T4SS) and biofilm through the operon pgaA-D, which catalyzes the biosynthesis of biofilm component polysaccharide poly-N-acetylglucosamine (PNAG), also described in Acinetobacter baumanii (Upmanyu et al., 2022). This suggests that the pathogen can transcytose via the motility apparatus flaA and toxin such as the compound 5, mediate sepsis while protecting itself from phagocytosis by means of biofilm and antiphagocytic capsule (glf), among other mechanisms. The versatility of the targets for compound 5 suggests that the toxin can bind to and deform multiple host proteins, thus increasing its potential to spread through the blood, respiratory system, and the brain, with effective immune evasion. The protein PLVAP is essential in stabilizing not only the endothelial wall but also the endosomes of immune cells (Chang et al., 2023). Thus, the ability of compound 5 to potentially deform PLVAP strongly suggests that the pathogen can dismantle and evade endosomes, translocate into the cytosol and multiply with less or no antigen presentation by macrophages. Similarly, the interaction of compound 5 with spectrin is an indicator of an advanced survival mechanism of Citrobacter freundii strain RSM030 in the blood stream, which entails effective RBC ambush, coupled with iron hijack via compound 5, enterobactin and other iron acquisition facets. From our BGC, pathway and in-silico docking analyses, we envisage the role of the shikimate/enterobactin biosynthetic pathway as critical not only for the generation of siderophores but also production of a conglomerate of intermediate metabolites dedicated to overwhelming the immune system by causing direct cell mortality, in addition to devastating RBC destruction for enhanced iron release into the planktonic population of an overgrowing C. freundii, anchored within the biofilm matrix. Moreover, the high solubility of compound 5 is an indicator of its ability to systemically spread through the blood stream and cause erythrocyte and immune cell depopulation, among other toxic effects. Although insufficiently substantiated in humans, C. freundii has been reported to cause serious anemia in animals including catfish (Bandeira Junior et al., 2018), which could be attributed to its possession of the virulence factors such as compound 5 and enterobactin as highlighted in our findings. In addition, a clinical case report on C. braakii from Japan revealed an association between the pathogen and sepsis with critical anemia (Yumoto et al., 2017). Decline of RBC count could be explained by the tubulin and spectrin-docking hemolytic role of compound 5 and subsequent iron acquisition. These mechanisms, combined with the role of antiphagocytosis gene (uge), capsule biosynthesis and transport gene (glf) and biofilm formation marked with PNAG (Polysaccharide poly-N-acetylglucosamine rich exopolysaccharides by the gene pgaC and YjbE, account for a sophisticated virulence repertoire contributing to the septic condition and the need for advanced respiratory support observed in this patient. We also stress that the multitude of virulence factors of this Gram-negative bacterium, integrating type IV and T3SS apparatus (the injectosome), increase its potential to migrate from the gut, through the blood, the urinary system, the nervous system and the respiratory system, among other parts. These findings suggest that the virulence of C. freundii is underrepresented, and that there is a high need for experimental establishment of the molecular mechanisms of C. freundii pathogenesis.

From AST, C. freundii RSM030 showed susceptibility to cefuroxime, ceftazidime, ceftriaxone, cefepime, imipenem and other antibiotics shown in Table 1. This indicates the appropriateness of these drugs for this patient, especially the carbapenems, and it could support choice of meropenem in the inherent treatment. However, the strain exhibited phenotypic resistance to trimethoprim/sulfamethoxazole, with indeterminate phenotypes for ciprofloxacin, levofloxacin, and amoxicillin/clavulanic acid. Recently from Western Uganda, unsequenced C. freundii isolates from neonatal sepsis with resistance to cotrimoxazole and carbapenems (Zamarano et al., 2021). However, despite the indeterminate results observed in this work, we observed intrinsic resistance genotypes encoded by the chromosome of RSM030 covering all these drugs, in addition to beta lactams (Fig. 1), which suggests that the strain could be resistant to these drugs. In addition, the closest relatives shown in Fig. 1 were found resistant to the same antibiotics (Ye et al., 2023). The indeterminate results in our study could be attributed to the fact that we did not check for MIC of these drugs, which was performed in the study by Ye et al. (2023). Moreover, while the closest strains showed indeterminate phenotypes on polymyxin B MIC, our strain was resistant to colistin (polymyxin E). We speculate that this discrepancy could be attributed to structural differences between polymyxin B and colistin.

The possession of multiple chromosomal polymyxin resistance genotypes in the RSM030 genome could account for the observed colistin resistance.

The mutation in the ArnT gene has been implicated in colistin resistance, recently identified to potentiate lipid A A-Ara4N pathway modification (Masood et al., 2021). Better substantiated mechanisms of colistin resistance involve the modification of lipid A, a component of the lipopolysaccharide, which is a major constituent of the cell wall. These modifications involve the biochemical addition phosphoethanolamine (PEtN) by phosphoethanolamine transferase and 4-amino-4-deoxy-L-arabinose (L-Ara4N) by the gene ArnT, which is also described from other studies (Masood et al., 2021; Samantha and Vrielink, 2020). These genotypes corroborate our phenotypic findings of colistin resistance from MIC. Moreover, the PmrA/PmrB and PhoP/PhoQ two component systems and the regulator mgrA/mgrB have been strongly associated with polymyxin resistance in enterobacteria such E. coli and Klebsiella pneumoniae, among other Enterobacteriaceae species (Li et al., 2022, 2023; Urooj et al., 2022). However, the relevance of these genes to polymyxin resistance phenotype in Citrobacter spp. is still elusive. Here we tap notable RSM030 chromosomal point mutations related to well-known colistin resistant pathogens from species other than Citrobacter freundii (Li et al., 2022; Shprung et al., 2021). Thus, the detection of PmrA and mgrB mutations in C. freundii strain RSM030 is an alert for an existence of colistin resistant pathogens causing unnoticed nosocomial infection from Uganda, which may portray a potential risk of disseminated multidrug resistance in the hospitals in the country and the region. This calls for urgent AMR surveillance initiatives encompassing Citrobacter spp. among other pathogens of the family Enterobacteriaceae. In addition, our findings are strongly suggestive of the involvement of EptA gene in colistin resistance, which stresses that an mcr- gene is unnecessary in the resistance of colistin, which supports the assertion that mcr-negative pathogens rely on intrinsic chromosomal colistin resistance mechanisms.

5. Conclusions

C. freundii is poorly studied and perhaps less associated with clinical cases as compared to Citrobacter koseri and most members of the family Enterobacteriaceae. However, from our findings, we realize that Citrobacter freundii is as dangerous as any member of the family, possessing a sophisticated virulence repertoire, resistance to common antibiotics such as (beta lactams, sulfonamides and fluoroquinolones) and, most importantly, resistance to the last resort antibiotic colistin. For the first time from East Africa, we report C. freundii strain RSM030 as a pathogen, novel sequence type ST150, with the potential to cause systemic infection and resist standard antibiotic drugs. We hereby recommend serious Citrobacter surveillance initiatives with the same emphasis given to other Enterobacteriaceae pathogens such as E. coli and Klebsiella pneumoniae.

Data availability

The complete genome of strain RSM030 was deposited in the NCBI database under the accession numbers CP133060.1 (chromosome), CP133061.1 (plasmid pRSM030_p1) and CP133062.1 (plasmid pRSM030_p2). Raw sequence reads can be made available upon request.