Insights from the Genome Annotation of Elizabethkingia anophelis from the Malaria Vector Anopheles gambiae

Elizabethkingia anophelis is a dominant bacterial species in the gut ecosystem of the malaria vector mosquito Anopheles gambiae. We recently sequenced the genomes of two strains of E. anophelis, R26T and Ag1, isolated from different strains of A. gambiae. The two bacterial strains are identical with a few exceptions. Phylogenetically, Elizabethkingia is closer to Chryseobacterium and Riemerella than to Flavobacterium. In line with other Bacteroidetes known to utilize various polymers in their ecological niches, the E. anophelis genome contains numerous TonB dependent transporters with various substrate specificities. In addition, several genes belonging to the polysaccharide utilization system and the glycoside hydrolase family were identified that could potentially be of benefit for the mosquito carbohydrate metabolism. In agreement with previous reports of broad antibiotic resistance in E. anophelis, a large number of genes encoding efflux pumps and β-lactamases are present in the genome. The component genes of resistance-nodulation-division type efflux pumps were found to be syntenic and conserved in different taxa of Bacteroidetes. The bacterium also displays hemolytic activity and encodes several hemolysins that may participate in the digestion of erythrocytes in the mosquito gut. At the same time, the OxyR regulon and antioxidant genes could provide defense against the oxidative stress that is associated with blood digestion. The genome annotation and comparative genomic analysis revealed functional characteristics associated with the symbiotic relationship with the mosquito host.


Introduction
The mosquito gut accommodates a diverse and dynamic microbiota [1][2][3][4][5], which has a profound impact on host metabolism, fecundity [6] and immunity [7,8]. The gut microbiome is not a random assemblage; the common core taxa belong to Proteobacteria, Bacteroidetes and Actinobacteria [2]. To better understand its structure and function in the mosquito gut ecosystem, it is necessary to characterize abundant taxa. Elizabethkingia is a genus in the Flavobacteriaceae family of Bacteroidetes and represents a separate lineage from the Chryseobacterium-Bergeyella-Riemerella branch [9]. Elizabethkingia spp. has been found to be a predominant resident in the gut of Anopheles gambiae [1,2], An. stephensi [4,10] and Aedes aegypti [11]. Lately, the strain R26 T of Elizabethkingia sp. was isolated from the midgut of An. gambiae, Ifakara strain [1]. The comparison of 16S ribosomal RNA (rRNA) gene sequences indicated closest similarity (98.6%) to that of Elizabethkingia meningoseptica. Subsequent hybridization experiments and fingerprint analyses together with biochemical tests, however, clearly separated R26 T from E. meningoseptica and E. miricola. Hence, a novel species, Elizabethkingia anophelis, was proposed as a third member in the genus Elizabethkingia [12]. Similar to E. meningoseptica, E. anophelis is a non-motile, non-sporeforming Gram negative rod with natural resistance against several antibiotics [12]. Later, the strain Ag1 of E. anophelis was isolated from the G3 strain of An. gambiae. The genomes of both strains were sequenced and annotated [13]. Here we present the in silico annotation of the genome, particularly focusing on functional categories that provide insights regarding the ecological connection between the bacteria and the mosquito host.

Genome Project History
The strains of R26 T and Ag1 of E. anophelis were isolated from the midgut of Anopheles gambiae maintained in the Faye laboratory in Stockholm University [1,12] and the Xu laboratory in New Mexico State University, respectively. The genome sequencing, assembly and gene prediction were described in [13]. The draft genomes have been deposited at DDBJ/EMBL/GenBank. The R26 T and Ag1 genomes are under the accession ANIW01000000 and AHHG00000000, respectively.

Antimicrobial and Hemolysis Assays
Antibiotics and Cecropin A from Hyalophora cecropia were purchased from Sigma-Aldrich. The radial diffusion assay was performed as earlier described by Hultmark et al [25]. Blood agar plates (horse blood) and bacterial control strains were kindly provided by Ann-Beth Jonsson's lab group. Hemolytic activity was evaluated following incubation at 37uC for up to 48 h.

Genome Properties
The draft genome of type strain R26 T of Elizabethkingia anophelis is 4.03 Mbp in size with an average GC content of 35.4%. The genome was predicted to have 3687 protein coding sequences (CDS) and 44 RNA genes. A single copy of a 16 S rRNA gene is found in the genome. The genome statistics are presented in Table  S1. Among the predicted protein coding genes, 2169 were assigned a putative function and 1518 were hypothetical proteins. To characterize the functions of the genome, the draft genome was annotated at RAST using the SEED subsystems [15]. Among the protein coding genes, 1174 were assigned into 308 subsystems (Fig. 1). The gene name, locus tag, protein ID and similarity comparison between the R26 T and Ag1 genomes are presented in Table S2. The genomes of Ag1 and R26 T are almost identical with a few exceptions. A small number of predicted ORFs (185 in R26 T and 146 in Ag1) were not found in the other strain, likely because both genomes were incomplete. Strain specific genomic regions were, however, identified. Two R26 T contigs, 104 (ANIW01000041; 28,343 bp) and 107 (ANIW01000035; 5,773 bp), are not found in the Ag1 genome. The R26 T contig 104 contains six putative genes associated with conjugal elements, indicating the presence of a conjugative plasmid or transposon. In addition, a segment of 32,350 bp in the Ag1 contig 47 (AHHG01000028) is not present in the counterpart contig 16 (ANIW01000059) in the R26 T genome. The presence of putative phage genes in the Ag1-segment indicates presence, or remnants, of a lysogenic phage. The insertion/deletion between Ag1 contig 47 and R26 T contig 16 was verified by PCR and subsequent sequencing of the amplicons (Fig. S1). Further investigation is necessary to confirm the identity and nature of those strain-specific elements.

Phylogenetic Relationship to Related Genomes
Upon comparison to available genomes in the family Flavobacteriaceae, E. anophelis is closer to E. meningoseptica, Riemerella anatipestifer RH-CH-2, Chryseobacterium gleum ATCC 35910 and Chryseobacterium sp. CF314 [20] and is divergent to taxa in the genus Flavobacterium. Table 1 shows the similarity of comparable CDS between E. anophelis R26 T and six other genomes. The pattern was supported by the phylogenetic relationship inferred from the 16S rRNA (Fig. S2A) and the peptide sequences of RNA polymerase beta-subunit, rpoB (Fig. S2B). E. anophelis and E. meningoseptica are grouped as a clade sister to the genera Chryseobacterium and Riemerella, while the taxa of the genus Flavobacterium are clustered in a separate clade.

TonB Dependent Transporters
The bacterial TonB-dependent transporters (TBDTs) are specialized elaborate machinery for active uptake of rare but essential nutrients and other substrates, such as iron complexes, vitamin B12, nickel, carbohydrates and colicin [26][27][28][29][30]. The total number of TBDTs is highly variable among bacterial genomes. Recently, Mirus et al. (2009) found putative TBDTs in 347 out of 686 sequenced bacterial genomes [31]. Among the investigated taxa of Bacteroidetes, all were found to endow over 50 TBDT genes. This is in line with our finding of 59 TBDT genes in E. anophelis. In the two other bacterial genomes that were isolated from the gut of An. gambiae in the Xu lab, Pseudomonas sp. Ag1 and Enterobacter sp. Ag1 possessed 55 and 20 TBDT genes, respectively [32,33]. Different TBDTs are distinct in substrate specificity [31,34]. The protein conserved domain analysis revealed various domains associated with different substrates. The conserved domains of 14 representative TBDTs are presented in Fig. S3. The architecture of mosaic conserved domains implies sophisticated interactions of TBDTs and biopolymers. The various TBDTs endow the bacterium with an uptake system for a variety of biopolymers, which have been demonstrated in many Gram negative bacteria, including Bacteroidetes in different metagenomic settings [35][36][37]. Further investigations are necessary to characterize the substrates of the TBDTs of E. anophelis and their contributions in the gut microbial community.
To energize the transport process, TBDTs interact with the TonB complex, a cytoplasmic transmembrane assembly of the proteins ExbB and ExbD, which couples with the TonB in periplasm. In the genome of E. anophelis, two ExbB, two ExbD genes and four TonB genes were found. A gene cluster was identified in the genome, where ExbB, ExbD and TonB as well as genes encoding an ABC transporter and a tetratricopeptide repeat containing protein are co-localized. The synteny of the gene cluster is conserved in taxa from all of the four existing classes of Bacteroidetes (Fig. 2). Such conservation indicates that it originated from a common ancestor and remained as an inheritable unit due to functional relationships among these syntenic genes.

Polysaccharide Utilization Loci
Carbohydrates serve as a major carbon and energy source for bacteria. In Bacteroidetes, utilization of complex carbohydrates (glycans) involves polysaccharide sensing, degrading and import machinery. Such activities are mainly controlled by genes in various polysaccharide utilization loci (PULs) [29,[38][39][40][41]. The starch-utilization system (Sus) represents a typical example of glycan acquisition [29,42]. The Sus like PUL consists of a cluster of genes encoding SusD (glycan-binding protein), SusC (Ton-B dependent transporter), SusE/SusF (carbohydrate-binding proteins without enzyme activity), SusA, SusB, SusG (enzymes for polysaccharide deconstruction) and SusR (an inner membraneassociated sensor-regulator system for transcriptional activation of Sus genes) [29,42]. The genome of E. anophelis contains 28 pairs of SusD and SusC homologs. Fig. 3 shows four loci where putative Sus-like genes are located. Table S3 lists predicted CAZy proteins in the E. anophelis genome, including glycoside hydrolases (GH), such as a-glucosidase, b-glucosidases, a-galactosidase, b-galactosidase, a-mannosidase, a-amylase, b-mannosidase, b-glucanase, cellulase, maltodextrin glucosidase, xylosidase, and a-dextrin endo-1,6-a-glucosidase, which indicate a broad capability of degrading polysaccharides. Recently, Kolton et al. (2013) demonstrated that taxa in the genus Flavobacterium were separated into terrestrial and aquatic clades according to geographical distribution. The terrestrial taxa have greater capacity to degrade polysaccharides [43]. Interestingly, some CAZy enzymes of R26 T have orthologs in F. johnsoniae, a terrestrial taxon, but not in the aquatic taxon F. branchiophilum. It is thus plausible that Elizabethkingia species have a similar ecological niche as the terrestrial taxa of flavobacteria.
Sucrose and fructose are the most common sugars that mosquitoes ingest from floral nectar [44,45]. Digestion of the sugar is carried out mainly by a-glucosidases that catalyze the hydrolysis of 1,4-a-glucosidic bonds to release a-glucose. The mosquito a-glucosidases have been characterized [46,47]. In addition, mosquitoes can ingest plant tissue and cellulose particles from plants, especially in arid habitats where floral nectar is scarcely available [48][49][50]. Certain polysaccharides from plants (e.g. hemicelluloses, pectins, and starch) may contain several different monosaccharides and a variety of glycosidic linkages. Utilization of the polysaccharides with such a wide structural variety and fluctuating abundance requires sophisticated mecha-  nisms to recognize and degrade them [41]. The predominance of E. anophelis in the sugar fed gut [2] and the possession of numerous Sus-like loci and GHs suggest that the bacterium may be capable of utilizing plant cellulose in the diet. Interestingly, no SusC and SusD were found in the genomes of Enterobacter and Pseudomonas strains that were isolated from the mosquito gut [32,33]. The large capacity of polymer transport and utilization implies a potential ecological association between the mosquito and the gut microbiota in which certain microbial residents may support effective utilization of various types of polysaccharides that mosquitoes take in from nectar-rich or nectar-poor plants.

Antibiotic Resistance
E. anophelis is known for its intrinsic resistance to many antibiotics [12]. The resistance mechanisms employed by bacteria are manifold, such as the enzymatic degradation of the drug, the alteration of the target drug site and the direct extrusion of the drug from the cells through efflux pumps. The E. anophelis genome appears to contain a large number of resistance genes ( Table 2,  S4). This includes a large set of multidrug efflux pumps, predominantly members of the resistance-nodulation-division (RND) and major facilitator multidrug pumps (MFS) families. The RND transporters can mediate extrusion of a broad range of substrates, including heavy metals (heavy metal efflux, HME), multidrug hydrophobe/amphiphile efflux-1 (HAE1) and toxic chemical compounds to maintain homeostasis [51,52]. The RND efflux pump is a tripartite assembly composed of an inner RND transport protein, a membrane fusion protein (MFP) and an outer membrane protein [53,54]. There are 13 sets of genes encoding the components of the RND pumps in the genome of E. anophelis. In most cases, three genes were found adjacent to each other, likely in a single operon. The synteny is conserved in the flavobacteria compared in this study as well as in other taxa of Bacterioidetes (Fig. 4, Table S4). The role of RND-efflux pumps in intrinsic antibiotic resistance has been demonstrated in Chryseobacterium and F. johnsoniae [55,56], as well as in Bacteroides fragilis [57]. The genome also encompasses genes encoding 44 MFS and two multidrug and toxic compound extrusion (MATE) transporters (Table S4). Generally, the MFS proteins mediate transport of a wide spectrum of substrates, including ions, carbohydrates, lipids, amino acids and peptides, nucleosides, and other molecules [58,59].
The synthesis of b-lactamases is the most commonly employed strategy among Gram-negative bacteria to combat b-lactam antibiotics [60]. A large set of genes conferring resistance to blactams were annotated in the R26 T genome, including 19 blactamases, four metallo-b-lactamases (MBLs) and four penicillinbinding proteins (Table S4). An overall lower degree of conservation between the compared species and R26 T was observed in this subcategory. Phylogenetic analysis of selected lactam degrading enzymes suggests that in addition to vertical inheritance, certain b-lactamase genes may be acquired by lateral   Table S5). In contrast, D505_08675 is not found in other members of Flavobacteriaceae except E. meningoseptica. Orthologs are instead present in the taxa belonging to the class Sphingobacteria of Bacteroidetes (Fig. 5B). Moreover, the MBL encoding gene D505_08350 lacks orthologs in several closely related species including E. meningoseptica ATCC 13253 (Fig. 5C), and appears to share common ancestry with the homologues in some Proteobacteria, particularly of the genus Pseudomonas, which indicates a history of lateral transmission. It is noteworthy that Pseudomonas aeruginosa produces transferable MBLs that recently have been spread to Enterobacteriaceae, likely in a clinical environment [61]. MBLs are also present in environmental microbiota, suggesting that certain taxa, including E. anophelis, potentially could act as reservoirs for lateral gene transfers in nature [62][63][64][65]. These patterns suggest that the b-lactamase superfamily is adaptive in response to various antimicrobial compounds in different eco-contexts.
E. meningoseptica endows two families of wide spectrum MBLs termed GOB [66,67] and BlaB [68]. Of note, an exceptional genetic diversity of the family members has been reported between clinical isolates from Korea [69]. In line with this, we repeatedly observed a clearly stronger homology for the lactam degrading enzymes between R26 T and E. meningoseptica strain 502 than the type strain ATCC 13253 (data not shown). This included the putative GOB/BlaB members, of which most are present E. meningoseptica 502 but only a few in ATCC 13253 [70]. This led us to compare the genome-wide CDS similarities between these two strains and R26 T . In comparison to E. anophelis, 85.3% and 69.3% of CDS in E. meningoseptica 502 and ATCC 13253, respectively, display .80% identity. A similar comparison against CDS of E. meningoseptica 502 revealed 85.1% identity to R26 T and only 58.7% to ATCC 13253. This unequivocally suggests that E. meningoseptica 502 is more related, albeit not identical to E. anophelis and raises concerns of potential misinterpretations when classifying strains of the genus. More thorough methods are hence needed in order to accurately determine whether isolates of Elizabethkingia sp. belong to either of the described species or represent a novel taxon.
The broad genetic capacity for antibiotic resistance is consistent with the observation that E. anophelis R26 T has natural antibiotic resistance to ampicillin, chloramphenicol, kanamycin, streptomycin and tetracycline [12]. In addition, ciprofloxacin that previously has proven moderately potent against clinical isolates of E. meningoseptica [71][72][73] had limited effects on R26 T growth ($50 mg 6 ml 21 ), whereas rifampicin [73,74] displayed cytotoxicity at moderate doses ($3.125 mg 6 ml 21 ). A strong resistance against the insect antimicrobial peptide Cecropin A was also seen, as no bacterial clearance was observed at the highest dose (100 mM; Fig.  S4).
The interactions among antibiotic-producing and resistant bacteria may be one of the determinants that shape and stabilize the community structure in the mosquito gut. Similar metagenomic contexts have been demonstrated in natural environments    Table S5. doi:10.1371/journal.pone.0097715.g005 [75] and host associated microbiomes [76,77]. The inhabitation of multidrug resistant E. anophelis and possibly other bacteria in the mosquito gut will affect the effectiveness of antibiotic treatment in mosquito microbiome research. Utilization of antibiotics may also perturb the community structure by only acting on sensitive bacteria. In summary, Table S4 presents the genes in the category of antibiotic resistance, efflux pumps and heavy metal detoxification, and the similarity to the orthologs in other taxa in the family Flavobacteriaceae.
The antibiotic resistance might have consequences for future work with the E. anophelis. Pathogenicity of the bacterium was recently demonstrated in a clinical case of meningitis in Africa [78] and a hospital outbreak in Singapore [79]. The isolates in both cases were resistant against a wide array of antibiotics. These case reports raised a concern regarding whether or not mosquitoes can pass E. anophelis and E. meningoseptica to humans in clinical situations and when handling mosquitoes in research. Further investigation is required to evaluate these potential risks.

Mevalonate Pathway Utilization for Isoprenoid Synthesis
Isoprenoids comprise the largest group of organic molecules in nature and are involved in a wide variety of biological functions. The universal isoprenoid precursor and building block isopentenyl pyrophosphate (IPP) and its isomer dimethylallyl pyrophosphate (DMAPP) can be synthesized by two independent pathways, the classical mevalonate (MVA) pathway and the alternative 2Cmethyl-D-erythtritol 4-phosphate (MEP) pathway. While the MVA pathway is present in higher eukaryotes, plant cytoplasm and archaea, the MEP pathway is generally found in eubacteria and the plastids of plants and apicomplexan parasites. Remnants of enzyme sequences for either pathway are commonly found among bacterial genomes, and in a few bacteria both pathways are active (reviewed in [80]).
Interestingly, opposite to most eubacteria, E. anophelis and flavobacteria appear to utilize the MVA pathway for isoprenoid biosynthesis. Exemplified in Fig. 6, E. anophelis and F. branchiophilum possess all but one of the enzymes known in the MVA pathway and none in the MEP pathway, while Enterobacter sp. Ag1 and Pseudomonas sp. Ag1 have all enzymes in the MEP pathway. A comparison with six other related bacteria indicated this to be a general feature within the family Flavobacteriaceae (Table S6). Whereas the phosphomevalonate kinase (EC 2.7.4.2) seems to be missing, which is not uncommon in bacteria, the phosphorylation of phosphomevalonate is likely mediated via an alternate route in this family [81,82]. A nonorthologous kinase in the galactokinase/ homoserine kinase/mevalonate kinase/phosphomevalonate (GHMP) family could possibly play the role [82,83]. In the E. anophelis genome, a homoserine kinase coding gene (D505_ 04189) and a galactokinase coding gene (D505_05014) were identified. Vinella et al. showed that the function of the MEP pathway is dependent on the ability to form Fe-S complexes in the IspG and IspH enzymes [84]. Fe-S complexes can be destroyed by nitric oxide (NO) [85], suggesting that possession of the MVA pathway would be conducive to living in a NO-rich environment. Because the production of NO is an essential part of both the mosquito anti-parasitic and anti-bacterial response [86][87][88], the MVA pathway utilizing E. anophelis and E. meningoseptica could have an advantage over other bacteria which have the MEP pathway in the mosquito midgut populations.

Antioxidant Capacity
Hematophagous mosquitoes take a blood meal for egg production, the digestion of which changes the gut conditions drastically. Catabolism of hemoglobins results in the release of a large quantity of free heme in the gut lumen. The pro-oxidation of heme increases oxidative stress in the gut environment (reviewed in [89]). A gene encoding the heme-degrading protein HemS (D505_03742) was present in the genome with orthologs found in E. meningoseptica and C. gleum, but not in F. branchiophilum. The HemS protein has been shown to degrade heme in vitro and is required for the defense against oxidative stress upon hydrogen peroxide exposure in Bartonella henselae [90]. The genome also encompasses four hemolysins, one hemolysin D secretion protein and six hemolysin translocator HlyD proteins. These gene products may assemble hemolysin transporters, as demonstrated in E. coli [91,92]. Hemolytic activity has been demonstrated in the fish pathogen F. psychrophilus [93] and E. meningoseptica [94]. In agreement with these findings, we found that E. anophelis also displays hemolytic activity and the ability to grow in the blood agar (Fig. S5). Following 24 h of incubation, E. anophelis caused a distinct brown discoloration of the blood agar (Fig. S5A). Despite a somewhat different appearance compared to Streptococcus pneumonia (Fig. S5B), the discoloration caused by either of these strains is a typical sign of a-hemolysis. Additional incubation for 24 h resulted in a clear, brown-colored zone adjacent to the growing E. anophelis bacteria. This effect was dependent on the bacterial density as single colonies caused discoloration but no clearance of the agar at this time point. Upon overall comparison, the peptide sequences of the orthologous genes were found with high similarity in E. meningoseptica, C. gleum, and to a less extent in the more divergent F. branchiophilum, indicating a broad conservation within the family Flavobacteriaceae. It is possible that the presence of hemolysins and heme degrading genes aids in the mosquito blood meal digestion. Similar synergism has been observed in Ae. aegypti [6].
Prokaryotic cells employ two redox-sensing regulons, OxyR and SoxRS, to sense oxidative stress signals and subsequently activate defense mechanisms [95]. The SoxRS regulon has not been found in Bacteroidetes [96,97]. The oxyR gene (D505_12281) is, however, present in the E. anophelis genome and is located in proximity to the antioxidant genes catalase (D505_12286) and manganese superoxide dismutase (MnSOD) (D505_12271). This genomic arrangement is unique to E. anophelis; no synteny was found in other flavobacteria in this study, although oxyR was found in other flavobacteria. In addition, other genes in the oxidative stress category are present, including those encoding alkyl hydroperoxide reductases (AhpC/ F), DNA-binding protein from starved cells (Dps), thioredoxins (Trxs), glutaredoxins (Grxs) and glutathione (GSH) peroxidase. OxyR controls the expression of catalase, ahpC, ahpF and dps in response to H 2 O 2 [97,98]. The genetic antioxidant capacity may contribute to the persistence of E. anophelis in the gut [2].

Summary
The genome annotation provides insights into the capabilities of E. anophelis, and sheds light on its symbiotic relationships with the mosquito host and other members of the microbiome. The predominance of E. anophelis in the gut ecosystem of mosquitoes may represent an evolutionary fit. Firstly, the presence of a large number of TonB dependent receptors with numerous substrate specificities endows the bacteria with a sufficient capacity to acquire and utilize various biopolymers. Particularly, the polysaccharide utilization system supplies consumable carbohydrates for both microbial residents and host following the intake of nectar as well as cellulose. Secondly, the multidrug efflux pump genes of the RND and MATE superfamilies allow the bacteria to extrude heavy metals, microbicides and other toxic chemical compounds and maintain a homeostatic internal environment. Various antibiotic resistance mechanisms may serve as guardians for maintaining community structure in the gut microbial community. The interactions among antibiotic-producing and resistant bacteria may be one of the determinants that shape and stabilize the community structure. Similar metagenomic contexts have been demonstrated in natural environments [75] and host associated microbiomes [76,77]. On the other hand, as an opportunistic pathogen, the multidrug resistance of E. anophelis renders infections very troublesome. Finally, the antioxidant capacity endows E. anophelis with defense against oxidative stress associated with blood digestion. The genome database provides a reference for further characterization of the mosquito gut microbiome and its impact on mosquito life traits.  Figure S4 Drug resistance and growth inhibitory capacity of E. anophelis. Representative figure of the drug resistance displayed by E. anophelis R26 T . Plates were cast using a mixture of 5610 4 bacteria in 6 ml Lysogeny broth and 1% SeaPlaque agarose (FMC BioProducts). Ciprofloxacin (100 mg/ml) and Cecropin A (100 mM) were added in a two-fold dilution series counter-clockwise in 2 mm wide holes with the lowest dose in the center and allowed to diffuse at ambient temperature for 30 min before incubation at 37uC for 24 h or until growth was apparent. E. coli strain D31 was used for comparison. (TIF) Figure S5 E. anophelis displays a-hemolytic activity. (A) Representative images taken from above (left plate) or underneath (right plate) at 48 h post inoculation. Black and white arrows depict the brown discoloration and clearance of the blood agar, respectively that were observed adjacent to the bacteria. The density dependence of the clearance zone was clear when observing the plate from underneath (right plate, white arrow). The right panel depicts the brown discoloration caused by individual colonies. (B) Control strains displaying the different types of hemolysis. (TIF)  Table S4 Drug resistance genes in E. anophelis and similarity to related genomes. Numbers reflect percent identity of the best hit in related genomes to each gene in R26 T . E. m., Elizabethkingia meningoseptica; C. g., Chryseobacterium gleum; F. b., Flavobacterium branchiophilum.