High Rate of Multidrug-resistance Among Clinical Isolates of Gram-negative Bacteria in Hospitalized Patients With Febrile Diseases in Central Ethiopia

Background Infectious diseases are among the leading causes of death in many low-income countries such as Ethiopia. Without reliable local data concerning causative pathogens and antimicrobial resistance, empirical treatment is suboptimal. The objective of this study was the characterization of Gram-negative bacteria (GNB) as pathogens and their resistance pattern in hospitalized patients with infections in Central Ethiopia. Methods Patients ≥ 1 year of age with fever admitted to the Asella Referral and Teaching Hospital from April 2016 to June 2018 were included. Blood and other appropriate clinical specimens were collected and cultured on appropriate media. Antibiotic susceptibility testing (AST) was performed using the Kirby-Bauer method and VITEK ® 2. Species identication and detection of resistance-genes were conducted using MALDI-ToF-MS (VITEK ® MS) and PCR respectively. Results Of 684 study participants, 54.2% were male and median age was 22.0 (IQR: 14-35) years. Blood cultures were positive in 5.4% (n=37). In other clinical samples, cultures were positive in 60.6% (20/33), 20.8% (5/24), and 37.5% (3/8) in swabs/pus, urine and other body uids, respectively. Of 66 pathogenic isolates, 57.6% (n=38) were GNB, (39.4% (n=26) Gram-positive and 3.0% (n=2) were candida species. Among the isolated GNB, 42.1% (16/38) were E. coli, 23.7% (9/38) K. pneumoniae and 10.5% (4/38) P. aeruginosa. In total, 27/38 Gram-negative isolates were available for further analysis. Resistance rates were: ampicillin/sulbactam 92.6% (n=25), cefotaxime 88.9% (n=24), ceftazidime 74.1% (n=20), cefepime 74.1% (n=20), gentamicin 55.6% (n=15), piperacillin/tazobactam 48.1% (n=13), meropenem 7.4% (n=2), and amikacin 3.7% (n=1). The NDM-1 gene was detected in one K. pneumoniae and one A. baumannii isolate, which carried an additional OXA-51 gene. Genes for ESBLs were detected in 81.5% (n=22) isolates as follows: TEM 77.2% (n=17), CTX-M-1-group 68.2% (n=15), SHV-group 27.3% (n=6) and CTX-M-9-group 9.1% (n=2). Clinical treatment data revealed ineffective antimicrobial treatment in 13 of 18 (72.2%) patients. ESBL-positive. ndings with recently published other parts of the most common ESBL-expressing pathogen, followed by E. coli (10, 11, 22). the characterization of ESBL genes among the Gram-negative isolates in our study, TEM-type and CTX-M-1-type were most common, followed by SHV-type and least frequently CTX-M-9-type. CTX-M-2 and CTX-M-8/25 detected at all. Compared to TEM-type and SHV-type, CTX-M-type enzymes widely disseminated world-wide and many variants associated with clinically relevant functional heterogeneity described. co-expression of different ESBL types is more common among GNB harbor CTX-M-type we found a high prevalence of different ESBL enzymes with predominance of TEM- and CTX-M-1-types. More than half of Gram-negative isolates carried two or more resistance ESBL genes. In addition, carbapenemases were detected in as many as 7.4% of the Gram-negative isolates, despite the limited availability and rare application of carbapenems in the country. These ndings emphasize the need for regular microbiological investigation of appropriate specimen before initiation of empirical antibiotic treatment wherever possible. In resource-limited settings like Ethiopia with conned capacities for regular microbiological investigation, local and national


Introduction
Infectious diseases are among the leading causes of morbidity and mortality in many low-income countries such as Ethiopia (1). Bacteremia and sepsis are common and are severe complications of many different infections caused by GNB leading to high rates of mortality. Pneumonia, bacteremia, urinary tract infection, intra-abdominal infection and skin and soft tissue infection are the major GNB source of infections (2).
Despite evidence for high rates of 3GC-reistance among Gram-negative isolates in many African countries of around 50%, 3GCs are among the most commonly used antibiotics for the initial empiric treatment of severe infections and sepsis in sub-Saharan Africa. Data to estimate the mortality associated with antimicrobial resistance (AMR) in sub-Sahara African countries is limited (7). Rapid interventions such as the establishment of antimicrobial stewardship (AMS) programs, strengthening of antimicrobial susceptibility testing (AST) capacity, revision and updating of existing management protocols of infectious diseases and provision of local and national AMR data for common bacterial pathogens should have a high priority. Continued extensive empiric use of 3GC in the absence of AST might lead to a further decrease in antimicrobial activity and therefore increase the burden on resource-limited health care systems of the countries affected (8). This development can also be observed locally and previous investigations at the study site show that the effectiveness of the limited choice of available antimicrobials is diminished due to the high rate of ESBL-producing GNB (9). There is an increasing need for strategies to enhance prudent use of antibiotics in the country due to the alarming spread of ESBL-producing GNB leading to high rates of infections due to MDR, which are di cult if not impossible to treat (10,11).
The lack of systematically acquired local data concerning both the causative organisms and common resistance patterns results in ineffective empirical treatment and an unfavorable clinical outcome. Therefore, we assessed the rate and extent of drug resistance among GNB isolated from hospitalized patients with acute infectious diseases at a tertiary hospital in Central Ethiopia and the pertinent drug resistance genes.

Study design and inclusion
This prospective operational research was conducted from April 2016 to June 2018 at the Asella Referral and Teaching Hospital (ARTH), a tertiary hospital in the town of Asella, located in the central part of Ethiopia. The bacterial cultures were performed in Ethiopia, the secondary and molecular biological investigations were performed in Germany.
A trained study team identi ed eligible patients. During the study period, all patients ≥1 year of age presenting for treatment at the ARTH with fever de ned as body temperature >37.5°C according to tympanic measurement were offered inclusion in the study. After written informed consent by patients or legal guardians, blood cultures (BC) were drawn from all participants. Previously initiated antibiotic treatment was not an exclusion criterion. In addition to BC, appropriate clinical samples according to the patient's symptoms and the treating physician's decision were collected for further microbiological investigation, according to local standard operating procedures (SOPs) and national treatment guidelines. Socio-demographic and clinical data were collected by using a standardized questionnaire.
Ethical approval and consent to participate The appropriate ethical review boards of Arsi University (reference number A/U/H/S/C/120/6443/2017), the Oromia Regional Health Bureau (reference number BEFO/AHBTFH/1-8/2017), and Düsseldorf University Hospital (reference number 5729) approved the study. The ethical clearance for samples transportation between Ethiopia and Germany was obtained from the National Ethical Review Board of the Ministry of Science and Technology (reference number 310/204/2017). Before inclusion in the study, written informed consent to participate in the study was obtained from each participant or, in the case of minors, from their legal guardians.

Blood culture
From every child and adult participant, between approximately 5 mL and 10-20 mL of blood were drawn, and inoculated into an aerobic blood culture bottle (in-house production for the rst 200 participants and thereafter commercially available blood culture bottles were used). The culture was incubated at 37°C for a maximum period of ve days. After 24 hours and at the end of the incubation period, Gram staining and sub-culturing on blood, MacConkey, and chocolate agar in candle jar or a 5% CO 2 enriched atmosphere were performed. Biochemical identi cation tests were subsequently done based on the Gramstaining result. Due to limited resources, only one aerobic culture was performed per patient. Regular quality controls for sensitivity were performed by incubation of standard laboratory strains of different bacteria.

Urine Culture
In participants with possible urinary tract infection, mid-stream urine specimens were collected using sterile urine cups and 1 μL of the urine was inoculated onto blood and MacConkey agar. After 24 hours incubation at 37°C, bacterial colonies were further processed if the colony count was ≥10 5 CFU/mL.

Swabs and other body uid culture
If appropriate to the participants' complaints, swabs or aspiration of other body uid were collected using appropriate sterile sampling devices (swabs, sterile syringes etc.) and cultivated on blood and chocolate agar in 5% CO 2 enriched atmosphere. The isolates were further processed following the local laboratory's SOPs.

Antimicrobial susceptibility testing
Following the identi cation of Gram-negative bacteria, Kirby-Bauer AST was performed after cultivation on Mueller Hinton Agar. The procedure and interpretation followed the European Committee on Antimicrobial Susceptibility Testing (EUCAST) recommendations. The microbiological result was reported to the treating physicians at the ARTH as soon as possible to ensure appropriate patient care.

Con rmation of species identi cation and antimicrobial susceptibility test
Bacterial isolates were preserved at -80 °C in Microbank ® vials (Pro-Lab Diagnostics Inc., Toronto, Canada). Subsequently, samples of all isolates were exported to Germany for con rmation of species identi cation and AST as well as molecular biological analysis of resistance mechanisms. Species identi cation was con rmed by using matrix-assisted laser desorption/ionization-time of ight mass spectrometry (MALDI-TOF, Vitek ® MS, bioMérieux, Marcyl'Étoile, France) and the local AST result was con rmed by VITEK ® 2 (bioMérieux).

ESBL and carbapenemase genes detection
After identi cation and AST, DNA of the GNB was extracted by suspending a pure colony grown on MacConkey agar in 200 μL of Tris-EDTA pH 8.0 buffer. The suspension was then heated at 95°C for 20 min, followed by centrifugation at 10,000 rpm for 10 min. Then, 150 µl of the supernatant was transferred into a new tube and stored at -20°C until PCR testing could be performed.
Bacterial strains with suspected production of ESBL or carbapenemase according to the AST result were investigated by PCR. We used the PCR protocols described by Strauß et al. for identi cation of the β-lactamase bla CTX-M , bla SHV and bla TEM genes (12) and an in-house PCR protocol established for the detection of the carbapenemase genes bla NDM-1 and bla OXA-51 , as described by Wendel  Data were entered and analyzed using IBM SPSS Statistics for Windows version 25 (IBM Corp., Armonk, NY, USA). Fisher's exact and Pearson's chi-squared tests were used for the analysis of quantitative variables and mean, median and standard division were used for continuous variables. Differences were considered statistically signi cant at p < 0.05.

Quality control
For quality control, the control strains listed in Table 2 were used during the resistance gene analysis.

Results
In total, 684 study participants were included; 54.2% were male and the median age was 22 (interquartile range: 14-35) years. More than half of the participants were from rural areas and 167 (24.4%) could not read or write (Table 3). The overall bacterial detection rate across all sample materials was 12.1% (83/684). In 9.6% (n=66) the isolated bacterium was considered a pathogen and in 2.5% (n=17) coagulase-negative staphylococci (CoNS) were isolated and considered as clinically irrelevant because of the high likelihood of contamination.
Regarding carbapenemases, a single bla NDM-1 from one isolated K. pneumoniae and a combination of bla NDM-1 plus bla OXA-51 from an isolate of A. baumannii were detected.

Effectiveness of empirical antibiotic treatment
The results of the locally performed Kirby-Bauer disc diffusion test were available from 25 study participants with E. coli and K. pneumoniae isolates. Eighteen of these 25 study participants received an empirically initiated antibiotic treatment at the time of sampling. The Kirby-Bauer AST revealed high levels of resistance against commonly used antibiotics, rendering 72.2% (13/18) of the initiated antibiotic treatments ineffective. In particular, 72.0% (18/25) of the isolated GNB were resistant to 3GC, 60.0% (15/25) to uoroquinolones and 48.0% (12/25) to gentamicin. The results of the initial clinical evaluation, empirical antibiotic treatment and AST result according to Kirby-Bauer disc diffusion test are summarized below (Table 9).

Discussion
To date, infectious disease are one of the most common causes of morbidity and mortality in resource-limited settings such as Ethiopia (1), but the availability of epidemiological data about causative pathogens and distribution of AMR remains limited. In this study, we found a high rate of MDR in GNB isolated from febrile patients. As previously described from Uganda, MDR GNB are the main cause of sepsis in febrile cancer patients with more than 50% of the sepsis episodes being caused by E. coli infections (15). The culture positivity rate of 5.4% from blood cultures was low, which might partially be explained by the fact that the causative pathogen was non-cultivable in a proportion of patients (haemoparasites, viruses, or non-cultivable bacteria). Similar ndings were reported from South Africa (18). In patients with febrile illness with unknown source of infection in comparison to other clinical diagnoses, the likelihood of positive blood cultures was lowest. The highest yield of bacterial cultures was reported from swabs in patients with SSTI. There was a trend of a decreased or elevated leucocyte count or an increased CRP level to being associated with blood culture positivity. However, these associations were not signi cant and the predictive value of these parameters to guide blood culture diagnostics is insu cient. Thus, our nding strengthens a previous report from Italy that CRP level alone is not su cient to predict blood culture positivity (19). In this time of wide-spread AMR and the associated risk of failing antibiotic therapies, blood culture diagnostics are essential to guide management of bacterial infections. If the broad application of blood culture diagnostics is not possible due to resource limitations, the application of other parameters such as procalcitonin or monocyte distribution width (MDW) could help to guide the rational use of blood cultures (20,21).
In our study, the resistance rates of isolated GNB against commonly used antibiotics was high, severely con ning the effectiveness of aminopenicillins in combination with betalactamase-inhibitors or 3GC for empirical treatment of infections possibly caused by GNB. The lowest resistance rates were found for the carbapenem meropenem and amikacin. These results are consistent with the results of other recently published data from different parts of Ethiopia, in particular from Addis Ababa (11,22), Jimma (23) and Bahir Dar (10). The different resistance rates to certain antibiotics re ect the frequency of antibiotic prescriptions. While aminopenicillins and 3GCs are applied very frequently, carbapenems are hardly used due to their high cost and limited availability. Of note, while the resistance rate for amikacin was very low, many of the isolated GNB were resistant to gentamicin. This difference might be explained by the frequent application of gentamicin at the study center. In contrast, amikacin is virtually not applied at all (24,25).
Our data revealed overall frequencies of ESBL and carbapenemase production of 81.5% and 7.4% among the isolated GNB, respectively. Of note, all isolates of K. pneumoniae were ESBL-positive. These ndings are consistent with recently published data from other parts of Ethiopia, where K. pneumoniae has also been shown to be the most common ESBL-expressing pathogen, followed by E. coli (10,11,22).
Regarding the characterization of ESBL genes among the Gram-negative isolates in our study, TEM-type and CTX-M-1-type were most common, followed by SHV-type and least frequently CTX-M-9-type. CTX-M-2 and CTX-M-8/25 were not detected at all. Compared to TEM-type and SHV-type, CTX-M-type enzymes are more widely disseminated world-wide and many variants associated with clinically relevant functional heterogeneity have been described. Thus, coexpression of different ESBL types is more common among GNB which harbor CTX-M-type enzymes (26).
Both, CTX-M-1 group and SHV-group were abundantly detected in K. pneumoniae, whereas in E. coli the SHV-group was detected only in 8% of the isolates (Table 8). This nding matches the report by Ogutu et al. that SHV-type is the predominant ESBL enzyme in K. pneumoniae and TEM-type is the most prevalent in E. coli (27).
A single bla NDM-1 carbapenemase gene in a K. pneumoniae isolate and bla NDM-1 plus bla OXA-51 carbapenemase genes in an A. baumannii isolate were detected. The expression of a bla NDM-1 in an A. baumannii isolate has previously been reported from the southwestern part of Ethiopia (28), but to our knowledge no case of of bla NDM-1 presence in K. pneumoniae has been reported to date. In general, this nding is not surprising since the bla NDM-1 carbapenemase gene in K. pneumoniae and A. baumannii have been commonly reported from other eastern African countries like Kenya, Uganda (29), Egypt (30) and Sudan (31).
In this study, most of the antibiotics initiated by the treating physicians for empirical treatment in the participating patients were ineffective according to the AST. However, these data have to be interpreted with caution since mostly microbiological culturing was performed after initiation of empirical antibiotic therapy and evaluation of the clinical success rate of the empirically initiated antibiotic therapies was not part of this investigation. Nevertheless, antibiotic resistance impairing the success of empirical antibiotic therapies seems alarmingly common and local epidemiological resistance data should be taken into account before initiation of antibiotic therapy (32). An adaptation of the local empiric antibiotic therapy strategy could be all the more necessary as the excessive use of 3GC could be one of the main reasons for the spread of ESBL-producing bacteria (33).
Selectively utilizing antibiotics based on the AST result is not only favorable for optimal treatment success but it also plays a major role in combating the spread of MDR bacteria. Adequate microbiological culturing before initiation of empirical antibiotic treatment is necessary to enable AST-based antibiotic therapy. As resources are limited and supply of laboratory materials is unreliable at the study site as at many other sites in low-income countries, it may not always be possible to perform comprehensive microbiological testing. In such cases, at least surveillance studies with the subsequent establishment of resistance statistics should be carried out to enable calculated antibiotic therapies adapted to the local resistance status. This might help to reduce the imprudent use of antibiotics (8).
A limitation of this study might be an impaired sensitivity of blood cultures, since only one set of locally prepared blood culture bottles was used for blood culturing. The limited availability and high cost of commercially available blood culture bottles prevent the regular use of such products on site. The molecular resistance testing was not possible for all Gram-negative isolates due to loss upon storage and transport. Our study did not investigate whether infections caused by ESBL-or carbapenemase-producing GNB were associated with reduced success of antibiotic therapies or increased mortality.

Conclusion
In this study performed at a tertiary hospital in Ethiopia, we isolated GNB from patients with infectious diseases with high rates of MDR including 3GCresistance, which are the most commonly used substances for empirical antibiotic treatment. Thus, 72% of the initiated antibiotic therapies in this cohort should be considered ineffective according to AST. As an underlying cause for extensive drug resistance, we found a high prevalence of different ESBL enzymes with predominance of TEM-and CTX-M-1-types. More than half of Gram-negative isolates carried two or more resistance ESBL genes. In addition, carbapenemases were detected in as many as 7.4% of the Gram-negative isolates, despite the limited availability and rare application of carbapenems in the country. These ndings emphasize the need for regular microbiological investigation of appropriate specimen before initiation of empirical antibiotic treatment wherever possible. In resource-limited settings like Ethiopia with con ned capacities for regular microbiological investigation, local and national programs for MDR surveillance might be helpful to reduce the futile use of antibiotics and improve the treatment outcome of calculated antibiotic therapies.
The implementation of new strategies like MDW-or PCT-guided culturing could help to enable the resource-saving use of microbiological culture methods. Strengthening of antimicrobial stewardship programs and appropriate policies to regulate the use of antibiotics and to monitor the emergence of resistant bacteria is thus needed without delay.

Declarations Funding
The rst author of this study, Tafese Beyene Tufa, was supported with stipends by the Bayer Foundation Talents for Africa Scholarship Program and by the Heinz Ansmann Foundation for AIDS research. There was no in uence of the funding organization on analysis or interpretation of the described data.