Genotypic diversity among multidrug resistant Pseudomonas aeruginosa and Acinetobacter species at Mulago Hospital in Kampala, Uganda

Antimicrobial susceptibility profiles

Overall, carbapenem resistance prevalence was 33 and 21% for P. aeruginosa from patients and the environment, respectively, while it was 14 and 86% for A. baumannii from patients and environment, respectively. Furthermore, all the imipenem resistant P. aeruginosa isolates were also resistant to aztreonam while 90% were resistant to piperacillin/tazobactam. On the other hand, all carbapenem resistant A. baumannii were resistant to piperacillin/tazobactam while 57% were aztreonam resistant. Of note, SXT-susceptible, carbapenem resistant A. baumannii from patients and the environment was noted, Table 2 and Additional file 1: Tables S1, Additional file 2: Table S2. Furthermore, 38% (16/42) of P. aeruginosa and 40% (8/20) of A. baumannii isolates were multidrug resistant (resistance to three or more antimicrobial classes [12]).

Beta-lactamases

Overall, 67% of the carbapenem resistant isolates were positive with the imipenem/EDTA double disk synergy test, implying the isolates were metallo-beta-lactamase producers, Additional file 1: Table S1. Hence, metallo-beta-lactamases were responsible for carbapenem resistance in majority of the carbapenem resistant isolates studied. Furthermore, 60% (37/62) of the isolates were resistant either to aztreonam or ceftazidime (extended spectrum cephalosporins), implying they were suspects for AmpC (class C β-lactamase enzyme/cephalosporinase) and extended spectrum beta-lactamases (ESBLs) production. However, none of these isolates was positive on ESBL and AmpC screening.

Genotypic diversity

Genotypic diversity within each species was investigated using Rep-PCR genotyping. Rep-PCRs produced 0–15 consistent polymorphic bands per isolate of sizes 250–9400 bp (BOXAIR-PCR), 90–8000 bp (ERIC-PCR), and 100–6557 bp (REP-PCR), Additional file 3: Figure S1.

At 100% similarity scale, BOXAIR-PCR fingerprints for A. baumannii isolates were not clustered (Fig. 1); more to this for Acinetobacter, REP-PCR fingerprints produced only one cluster (Fig. 2) while ERIC-PCR fingerprints produced two clusters (Additional file 4: Figure S2). Isolates 59 and 62 (both carbapenem susceptible) from a patient and environment, respectively, were considered genetically related in that their fingerprints were clustered at 100% similarity scale by both ERIC-PCR and REP-PCR cluster analysis, Fig. 2 and Additional file 4: Figure S2. Altogether, only five Acinetobacter isolates were clustered at 100% similarity scale. This implies low level of genetic relatedness among the isolates. More to this, the few clustered isolates also reveal a complex nature of A. baumannii multidrug resistant clones circulating at Mulago Hospital in that the isolates exhibited diverse antimicrobial susceptibility phenotypes, Additional file 2: Table S2.

For P. aeruginosa, cluster analysis at 100% similarity scale with BOXAIR-PCR, ERIC-PCR, and REP-PCR fingerprints generated one, two and three clusters, respectively, Fig. 3 and Additional file 5: Figures S3, Additional file 6: Figure S4. Isolates 20 and 21 from the environment were considered genetically related as their fingerprints were clustered by both BOXAIR-PCR and REP-PCR analysis, Fig. 3 and Additional file 6: Figure S4. As observed above for Acinetobacter, isolates from a patient and environment (1 and 25, respectively) were also clustered by ERIC-PCR analysis, Additional file 5: Figure S3. Likewise for P. aeruginosa, few isolates (10 of 42, 24%) generated fingerprints that were clustered at 100% similarity scale, implying a low level of genetic relatedness but a complex nature of multidrug resistant P. aeruginosa clones could be circulating at Mulago Hospital given the diverse antimicrobial susceptibility profiles generated by the clustering isolates, Additional file 2: Table S2.

The Simpson’s index of discrimination (D) for P. aeruginosa fingerprints was 0.9829 (BOXAIR-PCR), 0.9756 (ERIC-PCR), and 0.9791 (REP-PCR), indicating that the three genotyping methods successfully discriminated P. aeruginosa but BOXAIR-PCR and REP-PCR performed better than ERIC-PCR. For A. baumannii, Simpson’s index of discrimination was 0.9632 (BOXAIR-PCR), 0.9608 (ERIC-PCR), and 0.9474 (REP-PCR) indicating that the three methods also successfully discriminated the isolates but BOXAIR-PCR and ERIC-PCR performed better.

Setting

This was a follow-up surveillance study conducted at Mulago Hospital in Kampala Uganda, following the detection of carbapenem resistant Gram negative bacteria at the hospital in the period 2007–2009 [8, 10]. A total of 736 specimens from hospitalized patients were referred to the Clinical Microbiology and Molecular Diagnostics Laboratories at the College of Health Sciences, Makerere University, for culture and sensitivity testing. Additionally, 100 samples from the hospital environment [water, disinfectants in use (chlorhexidine gluconate), wet sink swabs, wet floors swabs, and swabs of cleaning materials (mops, squeezers)] were randomly collected and similarly processed for isolation of Pseudomonas and Acinetobacter species.

Identification of P. aeruginosa and Acinetobacter species

The sample collection and processing procedure was described in Kateete et al. [8]. Briefly, samples/specimens were processed within 2 h of collection for identification of Gram negative bacteria by following standard microbiological/biochemical procedures [8]. The recovered isolates were first presumptively identified based on colony morphology, Gram-staining properties and standard biochemical characteristics. Characteristic colony morphological features (i.e. colonies with characteristic spreading pattern and serrated edges, fruity sweet-grape smell, and bright green color) were used to identify Pseudomonas. Positive catalase and oxidase tests, negative triple sugar iron and glucose fermentation tests and growth at 42 °C were used to distinguish P. aeruginosa from other lactose non-fermenting Gram-negative rods. Acinetobacter species were presumptively identified based on negative motility, catalase, oxidase and glucose fermentation tests, and inability to grow under aerobic conditions.

To confirm Acinetobacter isolates to species level, the ‘Phoenix™ Automated Microbiology System’ (Becton and Dickson, Franklin Lakes, NJ, USA) was used; in addition, PCR-amplification followed by DNA sequencing of the species-specific region of the bla _OXA-51 gene intrinsic to A. baumannii [5, 36] was performed. To confirm P. aeruginosa to species level, and to determine the antimicrobial susceptibility profiles of both A. baumannii and P. aeruginosa isolates, minimum inhibitory concentrations (MICs) were performed with the ‘Phoenix™ Automated Microbiology System’ according to the manufacturer’s guidelines with minor modifications for Gram negative bacteria as described elsewhere [8, 37]. In addition, the disc diffusion method was performed to determine SXT susceptibility of A. baumannii isolates, according to the Clinical and Laboratory Standards Institute (CLSI) guidelines. Quality control and maintenance for the ‘Phoenix™ Automated Microbiology System’ were performed as recommended by the manufacturer using reference strains of E. coli ATCC 25922 and P. aeruginosa ATCC 27853. Phoenix™ default MIC-breakpoints were described in Kateete et al. [8].

Metallo-β-lactamase assays

All the carbapenem-resistant isolates were tested for metallo-β-lactamase activity as bacteria producing these enzymes are characterized by rapid spread worldwide [5]. Assays were performed with the imipenem-EDTA (Ethylenediaminetetraacetic acid) double-disk synergy test using K. pneumoniae ATCC 700603 and E. coli ATCC 25922 as indicator strains [10, 38, 39]. An overnight liquid culture of the test isolate was adjusted to a turbidity of 0.5 McFarland standard and spread on the surface of a Mueller–Hinton Agar (MHA) plate. Two imipenem discs (10 μg each) were placed on the agar 15 mm apart (center-to-center); 10 μl of 0.5 M EDTA was added to one of the imipenem disc to get a desired concentration of 750 μg. After incubating at 37 °C overnight, the difference in the inhibition zone diameter that was ≥5 mm between the imipenem disc supplemented with EDTA and imipenem disc alone was interpreted as positive for metallo-β-lactamase production [21, 39]. K. pneumoniae strains ATCC BAA-1705 and ATCC BAA-1706 served as positive and negative controls, respectively; strain BAA-1705 possesses a K. pneumoniae carbapenemase KPC-2 that is highly active against cephamycins, carbapenems, and to a substrate range of ESBLs [40]. Positive or negative results were interpreted according to the CLSI guidelines and the UK Standards for Microbiology Investigations [21, 39, 41].

ESBL and AmpC screening

Isolates that exhibited resistance to extended-spectrum cephalosporins (ceftazidime and aztreonam) were subjected to the double disk synergy test to detect production of ESBLs as described elsewhere [26], by placing ceftazidime, cefotaxime, aztreonam and ceftriaxone discs (30 µg each) at 20 mm distance (center-to-center) from amoxicillin (20 µg) and clavulanic acid (10 µg) discs. ESBL activity was inferred if the cephalosporin zone was expanded by the clavulanic acid containing disk. Ceftazidime and/or aztreonam resistant isolates were also tested for production of plasmid-mediated AmpC as previously described [26] using MHA plates and ceftazidime (30 µg) and cefoxitin (30 µg) discs placed 20 mm (center-to-center). AmpC-production was considered when isolates showed blunting of ceftazidime zone of inhibition adjacent to cefoxitin disc.

Genotyping, electrophoresis and analysis of DNA fingerprints

Genotyping was performed with Rep-PCR assays (REP-, BOXAIR- and ERIC-PCRs). Initially, we followed the procedure described by Syrmis et al. [35] to optimize the concentration of Mg++, template DNA, Taq polymerase and PCR primers, using four unrelated isolates of P. aeruginosa and A. baumannii. Chromosomal DNA used as templates in PCRs was extracted by the Cetyltrimethyl Ammonium Bromide (CTAB) method [42, 43] and dissolved in 100 μl of sterile Tris–EDTA (TE) buffer. 100 ng of the extracted DNA was used as template in the PCRs. REP-, BOXAIR- and ERIC-PCRs were performed using previously described primers and conditions [32, 44], with minor modifications to suit our setting. Ten microliters each of the PCR products was size-fractionated by agarose gel electrophoresis (1.5% w/v agarose) at constant voltage of 40 V in 1× TAE buffer (Tris–acetate pH 8.3, 40 mM Tris, 20 mM acetic acid, 1 mM EDTA) for 8 h in a horizontal electrophoretic apparatus (Bio-Rad Inc.). For quality control and uniformity in DNA migration during electrophoresis, all the gels were electrophoresed for equal amount of time. DNA size markers contained a mixture of the 50 base pairs (bp) DNA ladder (50–1350 bp size range) plus λ DNA-HindIII digest DNA ladder (2027–23,130 bp visible size range) (New England Biolabs Inc.). REP-, ERIC-, and BOXAIR-PCR profiles were visualized under ultraviolet (UV) light after staining the gels with ethidium bromide for 30 min. Images were captured and stored as TIFF files using BioDoc-it Imaging system (UVP, Cambridge, UK).

DNA fingerprints were analyzed with the BioNumerics software version 4 (Applied Maths NV, Sint-Martens-Latem, Belgium) as described elsewhere [34, 35, 45, 46]. The similarity-analysis of the results of each genotyping assay was calculated using Dice coefficient; cluster analysis of the similarity matrices was generated using Unweighted Pair Group Method with Arithmetic Mean (UPGMA) algorithm. The relationship between profiles was expressed as dendrograms; only visible bands were used to construct the similarity matrices and dendrograms. Variation in band intensity was not considered as genetic difference. Isolates were considered identical when all the scored bands in each pattern had similar migration patterns/distance. As few isolates shared similar migration patterns, the criterion for relatedness was taken as profiles with ≥85% similar migration patterns/distance [35]. The Simpson’s index of diversity [47, 48] was used to estimate discrimination indices (D values) for the three typing methods.

Quality control

Controls for Rep-PCR genotyping included known well characterized strains K. pneumoniae DSMZ 9377, E. coli ATCC 25922 and P. aeruginosa ATCC 27853. We optimized Rep-PCR assays using the procedure described for Yersinia [46] and adapted it to P. aeruginosa and A. baumannii. Genotyping was performed at least twice in separate assays to ensure reproducibility of the patterns; BOXAIR-PCR, ERIC-PCR, and REP-PCRs were repeated for five unrelated isolates of P. aeruginosa and A. baumannii. PCRs and electrophoresis were performed in two separate trials starting from the same DNA preparation, using the same PCR reagents. None of the isolates showed qualitative differences in the fingerprints. We also investigated the effect of the DNA preparation method using chromosomal DNA extracted with two in-house methods of CTAB and boiling; although CTAB-prepared DNA consistently yielded fingerprints with stronger band intensities, no significant difference in band sizes was detected suggesting stability of our Rep-PCR fingerprints.