We isolated strain BP-1(T) (bedside rail) and BP-2 (nurse call button) from an intensive care room at a tertiary care hospital in Pakistan with the Eswab collection device and transport system (Becton Dickenson & Company, Franklin Lakes, NJ) in June 2016. Surfaces were swabbed in triplicate during an ongoing yearlong longitudinal sampling study. Swabs were sent to the US at room temperature and arrived in the US within ~3 days of sampling. Samples from Pakistan were received by the Burnham lab under approval of the Washington University IBC Protocol number 6,572. Swabs were cultured within 24 h of arrival in the US laboratory site. The isolates were recovered on MacConkey agar with 5 μg/ml cefotaxime (Hardy Diagnostics, Santa Maria, CA) and re-streaked on blood agar using the four-quadrant method to recover distinct single colonies. The isolates were unidentified by the MALDI-TOF VITEK MS IVD v2.3.3 (bioMerieux, Durham, NC) and Bruker BioTyper (Bruker Daltonics, Billerica, MA) mass spectrometry systems. For subsequent genomic analysis, frozen stocks of each isolate were made by creating a dense suspension of the isolate in Tryptic Soy Broth (Sigma Aldrich, St. Louis, MO) with 15% Glycerol (Hardy Diagnostics, Santa Maria, CA).

Glycerol stocks of strain BP-1(T) and BP-2 were streaked separately onto 5% sheep′s blood agar (Hardy Diagnostics, Santa Maria, CA), HardyCHROM™ ESBL Agar (Hardy Diagnostics, Santa Maria, CA), Hektoen Enteric Agar (Remel, Lenexa, KS) and MacConkey agar (Hardy Diagnostics, Santa Maria, CA) plates using four quadrant streaking. Plates were incubated for 35°C in air and then imaged. Additional streaks of blood agar plates were incubated at either 4, 25, 35, or 42°C. Blood agar plates were also stored in an anaerobic bag (BD, Franklin Lakes, NJ) at 35°C. Phenotypic assays to assess indole production, catalase, urease and motility were performed. The isolates were also evaluated on lysine iron agar (LIA) (Remel, Lenexa, KS), triple sugar iron (TSI) agar (Remel, Lenexa, KS), Bile esculin agar (Remel, Lenexa, KS) and with an oxidation/fermentation glucose test (OF glucose) (Remel, Lenexa, KS), all performed according to manufacturer′s recommendations. Each isolate was also evaluated using the Analytical Profile Index (API®) 20-E Enterobacteriaceae identification kit (bioMerieux, Durham, NC). API values of related taxa were obtained from for Kluyvera, Enterobacter cloacae and Citrobacter koseri (referred to as Citrobacter amalonaticus biogroup 1 by Farmer et al., 1981, 1985). Similar to how this information was used for Escherichia albertii delineation, characters are scored as “+” if > = 85% of the strains are positive, “−” if > = 85% of the strains are negative, “v+” if 50–85% of strains are positive and “v–” if 50-85% of strains are negative (Huys et al., 2003).

Glycerol stocks of strain BP-1(T) and BP-2 were plated onto MacConkey agar with ceftriaxone (5 μg/ml) overnight. A single colony of each isolate was separately inoculated into 5 ml of LB broth and grown until log phase at 37°C with orbital shaking at 220 rpm.

For analysis of whole bacteria by negative staining, the isolates absorbed onto formvar/carbon-coated copper grids for 2 min. Grids were washed in dH₂O and stained with 1% aqueous uranyl acetate (Ted Pella Inc., Redding CA) for 1 min. Excess liquid was gently wicked off and grids were air dried. Bacteria viewed on a JEOL 1200 EX transmission electron microscope (JEOL USA Inc., Peabody, MA) equipped with an AMT 8-megapixel digital camera (Advanced Microscopy Techniques, Woburn, MA).

For ultrastructural analysis of cross-sections of bacteria, the isolates were fixed in 2% paraformaldehyde/2.5% glutaraldehyde in 100 mM cacodylate buffer, pH 7.2 for 1 h at room temperature. Samples were washed in cacodylate buffer and postfixed in 1% osmium tetroxide (Polysciences Inc.) for 1 h. Samples were then rinsed extensively in dH₂0 prior to en bloc staining with 1% aqueous uranyl acetate (Ted Pella Inc., Redding, CA) for 1 h. Following several rinses in dH₂0, samples were dehydrated in a graded series of ethanol and embedded in Eponate 12 resin (Ted Pella Inc.). Sections of 95 nm were cut with a Leica Ultracut UCT ultramicrotome (Leica Microsystems Inc., Bannockburn, IL), stained with uranyl acetate and lead citrate and viewed on the transmission electron microscope.

Genomic DNA was extracted using the bacteremia kit (Qiagen, Germantown, MD), from a suspension of ~10 colonies of strain BP-1(T) and BP-2 after overnight growth on blood agar (Hardy Diagnostics, Santa Maria, CA). 0.5 ng of DNA was used as input for constructing Nextera Illumina sequencing libraries (Illumina, San Diego, CA) (Baym et al., 2015). Sample libraries were sequenced on an Illumina NextSeq 550 to obtain ~2.5 million 2 × 150 bp reads. Raw reads had Illumina adapters removed with Trimmomatic and were decontaminated with Deconseq using the commands: “java -Xms1024m -Xmx1024m -jar $TRIMMOMATIC_HOME/trimmomatic-0.36.jar PE -phred33 -trimlog < trimlog.txt> < input forward reads> < input reverse reads> < output paired forward reads> < output unpaired forward reads> < output paired reverse reads> < output unpaired reverse reads> ILLUMINACLIP:/opt/apps/trimmomatic/0.36/adapters/NexteraPE-PE.fa:2:30:10:1:true” and “deconseq.pl -f < paired input reads> -o < paired output reads> -dbs hsref,” respectively (Schmieder and Edwards, 2011; Bolger et al., 2014). Paired forward and reverse reads were used as input for de-novo assembly with SPAdes with the command: “spades.py -k 21,33,55,77 –careful –pe2-1 < input forward reads> –pe2-2 < input reverse reads> -o < output directory>” (Bankevich et al., 2012). Assembly metrics were assessed by using QUAST on all contigs >500 bp in length using the webserver (http://quast.bioinf.spbau.ru/) (Gurevich et al., 2013). Draft whole genome sequences for strain BP-1(T) and BP-2 have been deposited with NCBI BioProject: PRJNA395420.

16S rRNA sequences from the BP-1(T) file were identified using Barrnap with the command: “barrnap –quiet < contigs.fasta>” (https://github.com/tseemann/barrnap) and manually retrieved. The sequence is on contig_69 from base pairs 22 to 1559. The complete 16S rRNA sequence was submitted to the EzBioCloud identify service on 08/31/17 (Kim et al., 2014; Yoon et al., 2017). The 16S rRNA sequences for the top 10 most similar taxa were retrieved and aligned using MUSCLE (Edgar, 2004) using the command “muscle -in < input_multifasta> -out < output_aligned_multifasta>.” The aligned 16S rRNA multifasta was made into a phylogenetic tree with 1,000 bootstraps using raxML with the command: “raxmlHPC -s < input_aligned_multifasta> -n output -m GTRGAMMA -p 100 -f a -N 1000 -x 12345.” The resulting newick tree was visualized in FigTree with bootstrap values depicted as node label. The aligned multifasta sequences were uploaded into Jalview v2.0 and colored by nucleotide (Waterhouse et al., 2009). The 16S rRNA sequence has been submitted to Genbank under accession number MG866003.

Fifty-six genomes from type strains in species of Escherichia, Shigella, Citrobacter, Salmonella, Kluyvera, Klebsiella, Kosakonia, Siccibacter and Pasteurella (Table S1) were obtained from NCBI Genomes on 09/01/17. The protein coding sequences of these genomes and strains BP-1(T) and BP-2 were annotated using Prokka with the previously described command (Seemann, 2014). The core genome, representing genes shared by 100% of all isolates in the cohort was constructed and aligned using the PRANK module within roary with the following command: “roary -e –group_limit 700000 -p 8 -i < percent identity> ^*.gff” (Page et al., 2015). Eighty percent of gene identity was used for core genome alignment of Enterobacteriaceae genomes plus the P. multocida outgroup and for core genome alignment of only the Enterobacteriaceae genomes.

Aligned multifasta files of the 48 genes shared by all Enterobacteriaceae genomes and P. multocida (Table S2) was constructed into a maximum likelihood phylogenetic tree using 1,000 bootstraps in RAxML with the following command: “raxmlHPC -s core_gene_alignment.aln -n output -m GTRGAMMA -o < P. multocida name>-p 100 -f a -N 1000 -x 12345” (Stamatakis, 2014). Trees were visualized in FigTree (http://tree.bio.ed.ac.uk/software/figtree/) with bootstrap values depicted as node labels.

Aligned multifasta files of the 750 genes shared by all Enterobacteriaceae genomes (Table S2) was constructed into a NeighborNet tree using the EqualAngle method in SplitsTree4 (v4.14.5) (Huson and Bryant, 2006).

Whole genome sequences of strains BP-1(T) and BP-2 were compared against each other using the mummer method of average nucleotide identity via the jspecies webserver (http://jspecies.ribohost.com/jspeciesws/) on 09/18/17 (Richter et al., 2016). The contig file for strains BP-1(T), BP-2 and the set of Enterobacteriaceae genomes (Table S1) were compared using the mummer method of average nucleotide identity via the standalone pyANI tool (https://github.com/widdowquinn/pyani) with command “python3 average_nucleotide_identity.py -i < input_fasta_directory> -o < outout_directory> -m ANIm –nucmer_exe < mummer_path> -v –f.” The Hadamard matrix representing the multiplication of the percentage identity and alignment length was converted into a dendrogram using scipy.cluster.hierarchy.dendrogram. within python3 using “euclidian” metric and “weighted” method. The dendrogram was viewed in matplotlib.

Antimicrobial susceptibility testing was performed on the isolates using Kirby Bauer disk diffusion on Mueller Hinton Agar (Hardy Diagnostics, Santa Maria, CA) according to CLSI standards (CLSI, 2013). Results were interpreted using CLSI M100 criteria for Enterobacteriacace (CLSI, 2013).

Antibiotic resistance genes (ARGs) from the draft whole genome assemblies of strains BP-1(T) and BP-2 were annotated by submission to the ResFinder webserver (https://cge.cbs.dtu.dk/services/ResFinder/) on 05/18/17 (Zankari et al., 2012). We annotated sequence divergent ARGs using Resfams with the command “annotate_functional_selections.py –Resfams_only -contigs < contig file> -o < output directory>” (Gibson et al., 2015). Putative resistant determinants were then mapped to observed phenotypic resistance in accordance with Resfams and ResFinder classification. Contig_4 in the BP-1(T) assembly was submitted to BLASTN against the nr/nt database on 01/22/18 (Camacho et al., 2009).

Strain BP-1(T) and BP-2 were annotated for protein coding sequences by Prokka using the command: “Prokka < input fasta> < output directory>” (Seemann, 2014). Contigs 54 and 72 from BP-1(T) had BLAST similarity compared against Contig 42 from BP-2 using default settings on WebAct (webact.org/WebACT/generate) (Abbott et al., 2007). BLAST similarity was viewed using EasyFig for sequences with e value 0.001 and length >100 bp (Sullivan et al., 2011).

Both strains BP-1(T) and BP-2 appeared identical on agar plates. BP-1(T) formed circular, shiny, mucoid, gray colonies and were non-hemolytic on blood agar plates (Figure 1A). The colonies were small, blue and rigid on HardyCHROM™ ESBL Agar (Figure 1B). BP-1(T) colonies were bright pink on MacConkey agar (i.e., lactose fermenting) with an opaque zone surrounding the edge of the colony (Figure 1C). Coral colored colonies were produced on HE agar (Figure 1D).

Analysis by negative staining and transmission electron microscopy revealed that BP-1(T) and BP-2 are rod shaped and ~2.7 μm in length (Figures 2A,B). Cross sections from resin embedding also show rod shaped bacteria (Figures 2C,D) with two electron dense lipid bilayers of the outer membrane and cytoplasmic membrane (Figure 2E). No evidence of flagella was observed, but pili-like structures were present between a group of isolates (Figures 2F,G).

Strains BP-1(T) and BP-2 had identical results from phenotypic assays (Table 1). Aerobically, BP-1(T) showed no growth at 4°C, scant growth at 25 and 42°C, but robust growth at 35°C (Table 1). BP-1(T) and BP-2 both showed robust growth anaerobically at 35°C using an anaerobic bag (Table 1). The isolates were negative for motility, oxidase, indole, catalase and urease. Both strains BP-1(T) and BP-2 were positive for gas production in TSI and LIA reactions (Table 1). The strains grew on bile esculin agar and can ferment glucose (Table 1).

As evaluated using the API 20E, BP-1(T) and BP-2 were positive for arabinose, amygdalin, melibiose, sacharose, rhamnose, sorbitol and mannitol fermentation, but negative for inositol fermentation (Table 2). BP-1(T) and BP-2 can utilize citrate and decarboxylate ornithine for metabolism in addition to utilizing disaccharides (melibiose) (Table 2). These biochemical test results were unable to assign a high confidence genus identification to strains BP-1(T) and BP-2. The analytic profile index (API) algorithm indicates that strains BP-1(T) and BP-2 most resembles Kluyvera spp. and to a lesser extent E. cloacae and Citrobacter koseri/farmeri (Table 3). Analysis of the biochemical utilization profile from these related taxa indicate that strains BP-1(T)/BP-2 and Kluyvera cryocrescens ATCC 33435 are both negative for arginine dihydrolase activity but E. cloacae and C. koseri are positive for it. Additionally, strains BP-1(T)/BP-2, E. cloacae and C. koseri can ferment D-sorbitol but K. cryocrescens ATCC 33435 cannot. Both BP-1(T) and BP-2 were tested as non-motile but K. cryocrescens ATCC 33435, E. cloacae and C. koseri are positive. Accordingly, BP-1(T)/BP-2 can be uniquely distinguished from its closest neighbors based on being negative for arginine hydrolase activity, positive for D-sorbitol fermentation and non-motile.

We used a phylogenomic approach to better delineate the taxonomic context of strains BP-1(T) and BP-2 within the Enterobacteriaceae. Submission of the 1,512-base pair 16S rRNA locus from strain BP-1(T) to the EzBioCloud taxonomic database indicated that strain BP-1(T) had the highest identity to Citrobacter farmeri CDC 2991-81(T), C. amalonaticus CECT 863(T), C. sedlakii NBRC 105722(T) and C. rodentium NBRC 105723(T) (Table 4). The maximum likelihood tree from alignment of these 16S rRNA sequences placed BP-1(T) in-between a group of Citrobacter species [C. sedlakii NBRC 105722(T), C. rodentium NBRC 105723(T), C. farmeri CDC 2991-81(T) and C. amalonaticus CECT 863(T)] and the other Enterobacteriaceae taxa including Salmonella enterica subsp. Arizonae ATCC 13314(T), Kosakonia sacchari SP1(T), C. europaeus 97/79(T), Enterobacter hormaechei susbp. steigerwaltii DSM 16691(T), an uncharacterized Enterobacter strain termed LGIT s RIT-PI-d and Pluralibacter gergoviae JCM 1234(T) (Figure S1). The greatest amount of variation in 16S rRNA identity occurred at the likely V3 region, around nucleotides 426–450 of the BP-1(T) sequence, using standard E. coli nomenclature (Chakravorty et al., 2007). The 6mer of GAGAAT in the BP-1(T) sequence from sites 426 to 435 is unique in comparison to the other sequences at this position. Although the GAG nucleotides from 426 to 428 are shared by K. sacchari SP1(T)/P. gergoviae JCM 1234(T) and the AT site at 434–435 is shared by C. farmeri CDC 2991-81^T, the A residue at position 429 is discriminatory, as all other taxa contain a thymine at this site (Figure S2). As 98.7–98.65% is a proposed standard for species delineation via rRNA gene sequencing, we endeavored to gain further insight into the taxonomic identity of BP-1(T) and BP-2 based on their whole genome sequence characteristics.

To determine the relative taxonomic identity of strains BP-1(T) and BP-2 beyond 16S rRNA similarity, we constructed an equal angle neighbor net tree produced from the aligned 750 core genes shared between type strains in Enterobacter, Siccibacter, Escherichia, Enterobacter, Kosakonia, Kluyvera, Klebsiella, Shigella, Salmonella and Citrobacter at >80% identity (Table S1). Analysis of the network indicates that strains BP-1(T) and BP-2 cluster together and distinctly from these major Enterobacteriaceae taxa (Figure 3A).

A maximum likelihood tree from the 48 genes shared between the same Enterobacteriaceae genomes and a Pasteurella multocida outgroup with >80% identity again placed strains BP-1(T) and BP-2 in a cluster separate from existing genera. The BP-1(T)/BP-2 branch separates the genomes analyzed into two major clades composed of Escherichia [excluding Escherichia vulneris NBRC 192420(T), which other groups have suggested is misclassified (Walk et al., 2009)], Shigella, Citrobacter, Salmonella in one branch and Klebsiella, Kluyvera, Enterobacter, Siccibacter and Kosakonia on the other branch (Figure 3B).

Pairwise Average Nucleotide Identity (ANI) analysis using MUMer between strains BP-1(T) and BP-2 indicate that they are the same species, as the ANI value of 99.97% is greater than the proposed species cutoff of 95% (Thompson et al., 2013). The dendrogram of a distance matrix representing the product of ANI and percentage of the genome alignment created a clustermap that is similar to the rooted Phylogenetic tree with BP-1(T) and BP-2 separating the Escherichia, Shigella, Citrobacter and Salmonella clade from the Klebsiella, Kluyvera, Enterobacter, Siccibacter and Kosakonia branch (Figure 4). While the ordering of the two Citrobacter clades, Salmonella and Escherichia/Shigella is similar between the ANI dendrogram and the core genome phylogenetic tree, in the ANI dendrogram Kosakonia and not Kluyvera/Klebsiella is the closest genera to BP-1(T)/BP-2.

Both strains BP-1(T) and BP-2 had phenotypic resistance to all cephalosporins tested except cefepime and cefotetan (Table 5). This susceptibility profile suggests carriage of an extended-spectrum β-lactamase (ESBL). Both isolates were resistant to ampicillin and aztreonam but susceptible to carbapenems (meropenem and imipenem) and β-lactam/β-lactamase inhibitor combinations (pipercillin-tazobactam, ceftazidime-avibactam and ampicillin-sulbactam; Table 5). The isolates were resistant to gentamicin but susceptible to amikacin (Table 5). Each isolate was additionally resistant to trimethoprim-sulfamethoxazole and susceptible to doxycycline, minocycline and tigecycline (Table 5). BP-1(T) was susceptible to fosfomycin but BP-2 had intermediate resistance (Table 5). Both isolates shared the same antibiotic resistance genes. ResFinder identified bla_SHV−12, bla_TEM−1b, qnrB2, aac(6′)IB-cr, strB, aac(6′)-IIC, aacA4, aph(3′)-Ia, strA, sul1, sul2, dfrA18, catA2 and ere(A) (Table 5). Resfams identified the above genes except for sul1, sul2 and dfrA18, but found an additional Class A β-lactamase (Table 5). In BP-1(T) the class A β-lactamase was on contig 4, believed to be chromosomal as the highest blastn match was the complete genome of C. amalonaticus FDAARGOS and had 96% identity with 68% query coverage to a HER family β-lactamase (WP_023281079.1). Preceding it was a tetR response regulator with similarity to a protein from Klebsiella sp. RIT-PI-d (WP_049839935.1).

To gain further insight into the potential for antibiotic resistance gene dissemination from strains BP-1(T) and BP-2, we analyzed the draft genomes of BP-1(T) and BP-2 for contigs that had ARGs co-localized with mobile gene elements. An IS1380 family transposase was found co-localized in both genomes with an erythromycin resistance gene ere(A), a putative rifampin resistance gene that was unidentified by ResFinder or Resfams, a putative aminoglycoside N-acetyltransferase unidentified by Resfinder or Resfams, a hypothetical protein and aac(6′)-IIC (Figure 5).

Superficieibacter (Su.per.fi.ci.e.i.bac′ter. L. fem. n. superficies surface; N.L. masc. n. bacter a rod: N.L. masc. n. Superficieibacter a rod from a surface).

Cells are facultative aerobic Gram-negative bacilli visualized without chains but occasionally in clusters under electron microscopy. Non-motile with no evidence of a flagella. Cells are mesophilic with robust growth at 35°C, scant growth at 25 and 32°C and no growth at 4°C. Cells can grow anaerobically at 35°C. Cells are catalase positive and can produce gas in the TSI and LIA slants. Cells can ferment arabinose, amygdalin, melibiose, sacharose, rhamnose, sorbitol, lactose, glucose and mannitol. Strains can be differed from related genera by being negative for arginine dihydrolase, indole production, acetoin production and motility but positive for sorbitol fermentation. Members belong to class Gammaproteobacteria, order Enterobacteriales and family Enterobacteriaceae. The type species of the genus is Superficieibacter electus.

Superficieibacter electus (e.lec'tus. L. part. adj. electus chosen).

Main attributes are applicable from genus. Colonies form: circular, shiny, mucoid, non-haemolytic gray colonies on blood agar plates; small, blue and rigid colonies on HardyCHROM™ ESBL Agar; bright pink colonies on MacConkey agar; Coral colonies on HE agar. Cells are rod shaped and ~2.7 μm in length. 16S rRNA sequence of the strain BP-1(T) had highest percentage identity to C. farmeri CDC 2991-81(T) (98.63%) C. amalonaticus CECT 863(T) (98.56%), Citrobacter sedlakii NBRC 105722(T) (97.74%) and Citrobacter rodentium NBRC 105723(T) (97.74%). Strain BP-1(T) and BP-2 form a cluster with each other based on core genome phylogeny with related genera in the family Enterobacteriaceae. GC percentage of strain BP-1(T) and BP-2 is 52.4%. The type strain is BP-1(T) (= ATCC BAA-2937, = NBRC 113412).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.