Analysis for prevalence and physical linkages amongst integrons, ISEcp1, ISCR1, Tn21 and Tn7 encountered in Escherichia coli strains from hospitalized and non-hospitalized patients in Kenya during a 19-year period (1992–2011)

Background We determined the prevalence and evidence for physical linkage amongst integrons, insertion sequences, Tn21 and Tn7 transposons in a collection of 1327 E. coli obtained over a 19-year period from patients in Kenya. Results The prevalence of class 1 integrons was 35%, class 2 integrons were detected in 3 isolates but no isolate contained a class 3 integron. Integron lacking the 3’-CS or those linked to sul3 gene or IS26 or those containing the ISCR1 were only detected in multidrug resistant (MDR) strains. The dfrAs were the most common cassettes and their prevalence was: - dfrA1(28%), dfrA12(20%), dfA17(9%), dfrA7(9%), and dfrA16(5%). The aadA were the second most abundant cassettes and their prevalence was: - aadA1(25%), aadA2(21%), and aadA5(14%). Other cassettes occurred in lower prevalence of below 5%. Prevalence of Tn21, ISEcp1, ISCR1 and IS26 was 22%, 10%, 15%, and 7% respectively. Majority of Tn21 containing integrons carried a complete set of transposition genes while class 2 integrons were borne on Tn7 transposon. The qnrA genes were detected in 34(3%) isolates while 19(1%) carried qnrB. All qnr genes were in MDR strains carrying integrons containing the ISCR1. Close to 88% of blaTEM-52 were linked to IS26 while ≥ 80% of blaCTX-Ms and blaCMYs were linked to ISEcp1. Only a few studies have identified a blaCTX-M-9 containing an ISEcp1 element as reported in this study. Multiple genetic elements, especially those borne on incIl, incFII, and incL/M plasmids, and their associated resistance genes were transferrable en bloc to E. coli strain J53 in mating experiments. Conclusions This is the first detailed study on the prevalence of selected elements implicated in evolution of resistance determinants in a large collection of clinical E. coli in Africa. Proliferation of such strains carrying multiple resistance elements is likely to compromise the use of affordable and available treatment options for majority of poor patients in Africa. There is therefore a need to monitor the spread of these highly resistant strains in developing countries through proper infection control and appropriate use of antimicrobials.


Background
Recent studies conducted in Kenya show that a significant proportion of E. coli strains from clinical specimens exhibit a strong multi-drug resistance (MDR) phenotype [1,2]. Fortunately, β-lactams, fluoroquinolones and aminoglycosides remain effective against a significant proportion of clinical E. coli strains in Kenya. However, recent studies have reported carriage of plasmid-borne aac(6')lb-cr and qnr genes among β-lactamase producers [1,2]. The qnr genes confer resistance to quinolones, while aac (6')-lb-cr confers reduced susceptibility to fluoroquinolones and aminoglycosides. Therefore, carbapenems remain some of the few alternative antimicrobials that are effective against strains harboring a combination of multiple β-lactamase (bla) genes and genes conferring broad-spectrum resistance to fluoroquinolones and amino-glycosides. Carbapenems may however not be readily available or affordable for many patients in Sub-Saharan Africa [3].
In a recent study, we reported carriage of integrons, IS elements, Tn21 and Tn7 in a collection of 27 E. coli strains obtained from hospitalised patients [1]. These strains also harbored conjugatively transferrable plasmids conferring resistance to β-lactams, fluoroquinolones, aminoglycosides and co-trimoxazole among other antimicrobials suggesting that genes encoding resistance to these antimicrobials are physically linked to each other. Carriage of physically linked elements, each containing a set of resistance genes, may increases the chances of en bloc horizontal transfer of multiple resistance determinants to susceptible strains. Carriage of multiple resistance elements may in turn confer unique advantages to the host and enable them survive a strong antimicrobial selection pressure especially in hospital settings [4]. Studies to determine the prevalence of resistance elements in a large collection of strains from Sub-Saharan Africa are still lacking. Furthermore, little is known on whether the genetic elements encountered among E. coli strains in this region are physically linked to each other. In this study, we determined the prevalence of integrons, ISEcp1, ISCR1, IS26 as well as transposons Tn21 and Tn7 in a collection of 1327 E. coli strains obtained from inpatient and outpatient populations seeking treatment in Kenyan hospitals during a 19-year period (1992-2011). We also determined genetic content of integrons and determined plasmid incompatibility groupings among strains exhibiting unique resistance phenotypes. Physical linkages among these elements and to bla genes were investigated using PCR methods. Similar analysis were done to determine if the aac(6')-lb-cr and qnr genes are physically linked to these elements.
conserved sequences (3'-CS) that contains qacEΔ1 (a truncated gene encoding resistance to quaternary ammonium compounds, and sul1 encoding resistance to sulfonamides). All the three class 2 integrons contained an identical cassette array comprising dfrA1-sat2-aadA1.

Prevalence of Tn21, Tn7 and IS elements
The prevalence of Tn21 was 22% while Tn7 was detected in 3 isolates that also carried class 2 integrons. Prevalence of ISEcp1, ISCR1 and IS26 was 10%, 15%, and 7% respectively. A high proportion (≥ 60%) of isolates containing the IS elements and integrons were MDR (resistant to at least 3 different classes of antimicrobials), Table 4. Isolates carrying multiple elements were more likely to exhibit an MDR phenotype than those lacking such elements (p:0.0001, CI:549.5 to 2419.6, OR:1153) and isolates from urine were more likely to harbor multiple elements compared to those from blood (p:0.0001, CI:3.1 to 5.5, OR:4.1) or those from stool (p:0.0008, CI:1.2 to 2.0, OR:1.6). Although integrons, IS elements and Tn21 were detected in isolates from all specimen-types, a high proportion (69%) of these elements were detected among strains from urine of hospitalized patients.
Physical linkage amongst genetic elements Figure 1 illustrates the strategy used for interrogation for physical linkages amongst genetic elements while Figure 2 illustrates some of the genetic associations identified in this study. Majority (69%) of integrons containing 3'-CS were physically linked to the Tn21 transposon while 75% of those containing a sul3 gene at the 3'-terminal were linked to IS26. This element was also linked to 80% of integrons lacking the 3'-CS, Table 5. Forty (40) isolates contained class 1 integrons linked to a single IS26 upstream the 5'-CS while in 12 isolates the integrons was flanked by two IS26 elements. All ISCR1 were detected only in MDR strains and were flanked by a pair of class 1 integron 3'-CS. Close to 94% of Tn21 that were linked to an integron contained a complete set of transposition genes (tnpA, tnpR and tnpM) while 89% of Tn21 with an incomplete set of these genes did not contain an integron, Table 6. All the three class 2 integrons were physically linked to Tn7.
Physical linkages between resistance genes and genetic elements Figure 2 illustrates selected examples of physical linkages between bla genes and different genetic elements. Over 40% of isolates carrying bla TEM-52 , bla SHV-5 or bla CTX-M- 14 were physically linked to the IS26, Table 7. The ISEcp1 was the most common IS element associated with bla CTX-M-14, bla CTX-M −15 and bla CMY-2. One isolate contained a bla CTX-M-9 linked to this element. In all cases, the ISEcp1 was detected upstream the bla gene, Figure 2. Thirty seven (88%) of the 42 aac(6')-lb-cr were borne on integrons containing the ISCR1 while 55% were borne on integrons linked to the IS26. Twenty four (71%) of the 34 isolates carrying a qnrA gene were resistant to nalidixic acid but not to ciprofloxacin while the other 10 isolates carrying this gene and 19 carrying the qnrB subtype were resistant to both antimicrobials, Table 8. None of the isolates tested positive for qnrS. Majority (87%) of qnr genes were physically linked to either integron-associated ISCR1 or the IS26. All Isolates carrying aac(6')-lb-cr or the qnr genes contained multiple genetic elements and were all MDR. Carriage of genetic elements or combination of elements among strains exhibiting resistance to different antimicrobials tested in this study. The antimicrobials were grouped into 8 convenient groups:-β-lactams and β-lactamase inhibitors, aminoglycosides, (fluoro)quinolones, nitrofurantoin, chloramphenicol, sulphonamides, trimethoprim, and tetracyclines.
Conjugative plasmids mediate en bloc transfer of multiple elements and resistance genes Multiple resistance genes and genetic elements associated with them were transferred en bloc to E. coli J53 in mating experiments, Table 9. Majority of such transferred were mediated by plasmids containing I1, L/M, XI, HI2 and the F-type replicons. These experiments further revealed that genes conferring resistance to tetracylines and chloramphenicol were also harbored in the same plasmids encoding resistance to β-lactams, (fluoro)quinolones and aminoglycosides. However, various gene combinations that had been determined to be physically linked using PCR could not be transferred in conjugation experiments using media containing different combinations of antimicrobials.

Discussion
The current study shows that a significant proportion of clinical E. coli strains in Kenyan are resistant to important classes of antimicrobials such as β-lactams, fluoroquinolones and aminoglycosides. These results are in agreement with those published before [1,3,5]. These MDR strains were however susceptible to carbapenems. It is easy (although illegal) to purchase antimicrobials in Kenya without prescriptions or with prescriptions not backed by laboratory investigations [6]. We hypothesize that such practices may directly or indirectly lead to emergence of highly resistant strains. A high prevalence of MDR strains from urine and all specimens from hospitalized patients may reflects a corresponding heavy use of antimicrobials among this category of patients as reported in past studies [7,8]. Majority of resistances encountered in hospital isolates were also encountered in community settings probably because patients are often discharged from hospitals as soon as their conditions improve, even before they complete their treatment regiments (our unpublished observations). It is therefore possible that hospital strains find their way into community settings and vis versa. However, we do not rule out the possibility that some MDR phenotypes may arise in community settings. The high prevalence of class 1 integrons may partially be due to their association with the Tn21 that contain a complete set of transposition genes. Past studies show that dfrA7 and dfrA1 cassettes associated with Tn21borne integrons are the most prevalent dfrA-subtypes in Central, North and Western Africa [9][10][11][12]. In this study however, the prevalence of dfrA7 was much lower than that of dfrA1, dfrA12 and dfA17 in that order. The class 2 integron dfrA1/sat2/aadA1 array reported in this study is globally distributed [13]. Our results may therefore reflect regional differences or similarities in distribution of integron cassette arrays. Such differences may arise from unique antimicrobial-use patterns in different countries. This study also demonstrates an apparent correlation between carriage of dfrA17 and resistance to multiple β-lactams as has been reported in Tunisia [12,14] and from Northern Kenya among isolates from dog, cat and human specimens [5]. The reasons behind these correlations are yet to be elucidated. Carriage of different dfrA sub-types in our isolates and carriage of multiple integronassociated sul genes (sul1 and sul3) in the same isolate possibly correlates to heavy usage of sulfonamides and trimethoprim in Kenya for treatment of different infections and as prophylaxis against opportunistic infections among people with HIV/AIDS [15][16][17].
Some integrons, especially those lacking the 3'-CS and those containing a sul3 at the 3'-end, were linked to the IS26 possibly because this element mediates deletion of 3'-CS in class 1 integrons 3'-terminal [18,19]. Similar results have been published in Australia, Spain and Nigeria 1.f Figure 2 Schematic diagram illustrating examples of physical linkages amongst genetic elements and selected genes. 1a-1f: An example of physical linkages between bla genes and multiple genetic elements such as integrons, ISEcp1, and IS26. 2a-2b: An example of physical linkages between bla genes and ISEcp1. 3a-3d: An example of physical linkages between integrons and other genetic elements (such as the ISCR1 element) that are in turn linked to bla genes and (fluoro)quinolone resistant genes. 4a-4c: An example of physical linkages between Tn21 and integrons that are in turn be linked to IS elements. These illustrations are based on PCR mapping data and not sequencing. Therefore, the sizes of each gene and the distances between any two genes are not drawn to scale.  [11,12,18,19]. Our data further suggest that strains carrying IS26-associated integrons are highly MDR probably because the IS26 is also linked to other non-integron genes such as β-lactamases.
Most β-lactamases, particularly those encoding CTX-M -14 and −15 and CMY-2, were physically linked to ISEcp1. Similar reports have been published in Tunisian [20,21] but no ISEcp1 was detected upstream the bla -CTX-M-1 among our isolates as reported in a related study from the same country [22]. In one isolate, this element was found upstream the bla CTX-M-9. Reports of ISEcp1-bla CTX-M-9 linkages are rare but such linkages have been reported in Klebsiella pneumoniae isolates in Taiwan [23]. Majority of bla TEM genes, bla TEM-52 in particular, were physically linked to the IS26 as reported in Belgium and Germany [24,25]. Taken together, these results suggest that most bla genes in our isolates are in similar genetic environments as those reported globally but the genetic environment of bla CTX-M-9 and bla CTX-M-1 in our isolates appears to be different from those reported globally.
Our results further demonstrated that most bla genes are distantly linked to elements that are in turn linked to other resistance genes such as aac(6')-lb-cr and qnr. Similar reports have been published in Tunisia [20,21] and in Nigeria [11]. ISEcp1, IS26 and ISCR1 are known to mediate transposition and/or expression of multiple resistance genes in their close proximity [26][27][28][29][30][31]. Carriage of such multiple elements, each carrying a set of resistance genes may be responsible for the observed coresistance to multiple antimicrobials among our isolates.
Conjugation experiments confirmed that multiple elements were borne on narrow host-range plasmids such as IncFII, IncH12 or on broad host-range plasmids such as IncL/M. The type of conjugative plasmids in our isolates (especially those carrying plasmids containing incFtype, incHI2 and incI1 incL/M replicons) were shown to confer resistances similar to those in strains from Tunisia, [32] and from two other studies conducted in Kenya [1,5]. We hypothesis that plasmids of different incompatibility groups have acquired similar or identical sets of resistance genes and this acquisition is mediated by genetic elements such as those investigated in this study. Therefore, there is a possibility that such elements act as genetic shuttles between plasmids of different incompatibility grouping. The similarities and differences in genetic environments of bla, aac (6')-lb-cr and qnr genes reported in this study may reflect a difference in transposition activities of such elements. We further hypothesize that differences in antibiotic use patterns in different regions influence the transposition activity of such elements.

Conclusions
This study reports carriage of multiple genetic elements in MDR E. coli strains and their association with selected resistance genes. Strains carrying such elements are likely to be well adapted to survive deleterious effects of combined antimicrobial therapy. Furthermore, such MDR strains have a potential to increase morbidity and mortality among patients. It is therefore important to launch  (47) 14 (11) PCR methods were used for screening for three genes that are crucial for transposition of Tn21. The tnpA encodes a Tn21-like transposase, the tnpM encodes a putative transposition regulator. Integrons are incorporated into the Tn21 framework adjacent to the tnpM gene. The tnpR encodes a resolvase. Table 7 Analysis for physical association between bla genes and various genetic elements Analysis for physical linkages between bla genes and various genetic elements. The bla content of the isolates analyzed had been determined in a past study [3]. Table shows the number of isolates carrying the three (fluoro)quinolone resistance genes and the proportion of such strains in which these genes were physically linked to various genetic elements and to bla genes. a: Distribution of the aac(6')-lb-cr and qnr genes among strains fully susceptible to β-lactams, among those resistant to TEM-1 or SHV-1 with a narrow substrate-range and among those carrying genes encoding broadspectrum β-lactamases such as bla SHV-5 , bla SHV-12 , bla CMY and bla CTX-Ms . Table 9 Horizontal transfer of genetic elements and associated resistance genes from clinical strains (donors) to E. coli J53 (recipient)

Resistance profiles among donor and transconjugants
Resistance to selected antimicrobials among donors Physically linked genetic elements or resistance genes detected in donors and recipients Other genes whose linkages were not determined   surveillance programs and to put up measures to curtail the spread of these highly resistant strains. There is also a need to compare the genomes of strains encountered in Africa with those from other parts of the world.

Isolates
The 1327 non-duplicate isolates were obtained sequentially from 13 healthcare facilities in Kenya between 1992 and 2011 (19-year period) from 654 hospitalized and 673 non-hospitalized patients. These isolates comprised of 451 strains from patients with urethral tract infections (UTI) and those with urinary catheters while 371 were from blood of patients with septicemia. Another 505 strains were from fecal specimens of patients with loose stool, watery and bloody diarrhea. Only one isolate per specimen per patient was included for further analysis. Among the isolates investigated in this study, 912 had been analyzed for bla genes in a a past study [3] while 27 had been analyzed for selected genetic elements [1]. Ethical clearance to carry out this study was obtained from the KEMRI/National Ethics Committee (approval number SSC No. 1177).

Antimicrobial susceptibility profiles
Susceptibility profiles for all isolates were determined using antibiotic discs (Cypress diagnostics, Langdorp, Belgium) on Mueller Hinton agar (Oxoid, Louis, Mo. USA) using the Laboratory Standards Institute guidelines (CLSI) [33].
Detection of genetic elements Figure 1 illustrates the strategy used for detection and characterization of integrons and transposons. Detection of class 1, 2 and 3 and determination of carriage of 3'conserved sequences (3'-CS) in class 1 integrons was done as described before [34,35]. Class 1 integron variable cassette region (VCR), the region in which the resistance gene cassettes are integrated, was amplified as previously described by Dalsgaard et al. [35] while that of class 2 integrons was amplified as described by White et al. [36]. The VCRs of integrons lacking the typical 3'-CS was determined using a PCR walking strategy published before [37]. Identification of integron cassette identity was done using a combination of restriction fragment length polymorphism (RFLP), sequencing and published bioinformatics tools [38,39]. Detection of the ISEcp1, ISCR1, Tn21 and Tn7 elements was done as described in published studies [34,35]. Analysis for Tn21 transposition genes:-tnpA, tnpR and tnpM genes was done as previously described by Pearson et al. [40]. The primers used in this study are presented in Table 10.
Detection of aac(6')-lb-cr and qnr genes Screening for aac(6′)-Ib-cr gene that confers crossresistance to fluoroquinolones and aminoglycoside was done using a combination of PCR, RFLP and sequencing as described by Park et al. [41]. The isolates were also screened for genes conferring resistance to quinolones: -qnrA, qnrB and qnrS using PCR and sequencing strategies previously described by Wu et al. [42].

Interrogation for physical linkages between genetic elements and resistance genes
Physical linkages between integron and the transposons were determined using a combination of published primers targeting 5'-conserved sequences (5'-CS) of class 1 integrons and those targeting the tnpM of Tn2 or those specific for tnpA7 of Tn7, Figure 1. A combination of primers targeting IS elements and those targeting the 5'-CS or the 3'-termini of integrons were used for interrogation for physical linkages between integrons and IS elements. A combination of primers specific for various genetic elements and consensus primers for bla SHV or bla TEM, [43,44], bla CTX-M [45], bla CMY [46] and bla OXA

OXA-2R ATYCTGTTTGGCGTATCRATATTC
Primers used for screening various genetic elements and for interrogating physical linkages between different genetic elements and between such elements and bla genes or (fluoro)quinolone resistance genes. Y = T or C, R = G or A, S = G or C, K = G or T. [47,48] were used for determination of physical linkages between bla genes and different genetic elements. Primers for aac(6')-lb-cr and qnr genes were used in combination with those for different genetic elements to analyze for their physical association. A long-range polymerase [LongAmp W Taq DNA Polymerase, (New England Biolabs, USA)] was used in all reactions for physical linkages. A slow ramping rate of between 0.2°C/sec and 0.3°C/sec was set for the annealing step. The extension time was set at 72°C for 2 min and a final extension of 72°C for 15 min was carried out after 35-40 cycles of denaturation, annealing and extension.

Conjugation experiments
Conjugation experiments using sodium azide resistant E. coli strain J53 as the recipient were done as previous described [49]. Susceptibility to antimicrobials and determination of genetic element content of the transconjugants was determined using similar methods as those used for the corresponding donor strains. Plasmid incompatibility groupings were determined using the scheme of Carattoli et al. [50].

Statistical analysis
For the purpose of analysis, both intermediate and resistant results for antibiotic susceptibility testing were grouped together as "resistant". Differences in proportion of isolates bearing different elements was analyzed using the Chi test (χ2) while the Fisher's exact test was used for smaller sample sizes. The Odds Rations (OR) and the 95% confidence intervals (CIs) accompanying the χ2 tests were determined using the approximation of Woolf. The null hypothesis was rejected for values of p ≥ 0.05. Statistical analysis was performed using Statgraphics plus Version 5 (StatPoint Technologies, INC, Warrenton, VA, USA).