Study of susceptibility to antibiotics and molecular characterization of high virulence Staphylococcus aureus strains isolated from a rural hospital in Ethiopia

We characterised 80 Staphylococcus aureus strains isolated from human patients with SSTIs at a rural hospital in Ethiopia. Susceptibility to antibiotic of all strains was tested. The MLST method was used to type and a phylogenetic analysis was conducted employing the sequences of 7 housekeeping genes. PCR amplification was used to investigate the presence of the following virulence genes in all strains: hla (α-haemolysin), tstH (toxic shock syndrome toxin), luk PV (Panton-Valentine leukocidin), fnbA (fibronectin binding protein A) and mecA (methicillin resistance). Most of the strains were resistant to penicillin and ampicillin, but only 3 strains were resistant to oxacillin, and 1 of them was a true MRSA. The MLST results showed a high diversity of sequence types (ST), 55% of which were new, and ST152 was the most prevalent. A phylogeny study showed that many of the new STs were phylogenetically related to other previously described STs, but bore little relationship to the only ST from Ethiopia described in the database. Virulence gene detection showed a high prevalence of strains encoding the hla, fnbA and pvl genes (98.77%, 96.3% and 72.84%, respectively), a low prevalence of the tst gene (13.58%) and a markedly low prevalence of MRSA (1.25%). S. aureus strains isolated from patients in a rural area in Ethiopia showed low levels of antibiotic resistance, except to penicillin. Moreover, this study reveals new STs in Eastern Africa that are phylogenetically related to other previously described STs, and confirm the high prevalence of the pvl gene and the low prevalence of MRSA on the continent.


Introduction
Staphylococcus aureus is a gram-positive bacterium with carrier rates of 25-50% in the general population as a commensal microorganism, but it can also become an opportunistic pathogen under certain circumstances. Consequently, it is not only the cause of community-acquired infection (CAI), but is also one of the most important aetiological agents of hospital-acquired a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 The general aim of the present study was to characterise the strains isolated from a rural hospital in Ethiopia in order to determine the origin of their virulence. To this end, susceptibility to antibiotic was tested, and strains were typified by MLST and their sequences were used to study their phylogenetic evolution. Then, the presence of several virulence genes was investigated to determine the origin of the strains' virulence.

Strains
Between June 2014 and June 2018, 80 Staphylococcus aureus strains were isolated at Gambo General Rural Hospital from different human samples, mainly SSTIs but also from some patients with osteomyelitis, leprosy and pyomiositis. Most of the strains belonged to children (50), but 22 of them were isolated from adults, and age date of 8 of them could not be collected. In general, strains appeared to be highly virulent and all patients presented deep and extensive lesions, mostly ulcers, and recurrent infection in some cases.
Four different S. aureus subsp. aureus strains from the Spanish Type Culture Collection (CECT 976, CECT 957, CECT 435 and CECT 4439) were used as reference controls for different experiments.
In this study, the DNA of each strain was extracted using the UltraClean 1 Microbial DNA Isolation Kit (MoBio Laboratories Inc., 12224-250) according to the manufacturer's instructions, and this DNA was used for all analyses.
Clinical signs, medical histories and patient characteristics were collected to study any possible relationship with strain characteristics. This study was authorised by the Gambo General Rural Hospital management.

Ethics statement
After asking to the hospital, the study didn't require any ethics statement because no work was developed with human samples. Strains were isolated directly from the patients to plates.
Strains were collected not only for this study, but also for diagnosing of infection. Patient identifying information was collected by medical doctors as part of the routine hospital patient care procedure, and a number was assigned to each patient.
Information arrived at the laboratory with this number after isolating and identifying all strains. Patient consents for collecting their clinical signs, medical histories, and characteristics were obtained during de admission of the hospital as a part of the routine hospital patient care procedure.

Identification of strains
First, the preliminary identification of the strains was confirmed by MALDI-TOF (matrixassisted laser desorption/ionisation-time-of-flight) (MALDI Biotyper System, Bruker), and by partial amplification of 16S gene using the 27F and 1492R primers and sequencing, according to the procedure described by Lane (1991). All sequences were compared against each other for similarity using NCBI-BLAST and the nucleotide collection database (http://blast.ncbi. nlm.nih.gov/Blast.cgi).
Moreover, age of patients was taking into account for comparing with the prevalence of resistances found in each strain.

Strain typing by MLST
MLST analysis was used to typify all strains (Jolley, Bray, & Maiden, 2018). For that purpose, amplification of 7 housekeeping genes (arc, aro, glp, gmk, pta, tpi and yqi) was carried out using the PCR conditions described on the S. aureus MLST website (S1 Table), and sequences were obtained by the Sanger method using multi-capillary sequencing (ABI PRISM 3130XL, Applied Biosystems). Then, sequences for each gene from each strain were analysed with BioEdit software and compared with the S. aureus MLST database (https://pubmlst.org/saureus/). In accordance with this database, one allele was assigned to each sequence, and the combination of 7 alleles comprised a specific sequence type (ST) for each strain.

Phylogeny study
Firstly, phylogenetic analysis with all the strains and ST that belong to S. argenteus and S. schweitzeri was performed to discard that strains belong to these species rather than S. aureus.
Then, another phylogenetic analysis was realized with each ST determined previously by MLST typing to know their phylogenetic relations.
Both phylogenetic studies were conducted using the sequences obtained from the 7 housekeeping genes employed in the MLST study. First, the 7 sequences were concatenated in order to create a unique sequence for each strain, and then a multiple alignment of all these sequences was performed with the BioEdit software. G-block software (version 0.91b) was used to eliminate highly variable areas of the sequences. Next, the appropriate substitution model was selected using the jModelTest application (version 2.1.10). Lastly, a maximum likelihood tree was generated and analysed using Mega X software.

Virulence genes detection by PCR
The virulence genes analysed were hla, tst, pvl, fnbA and mecA, that codifies for α-haemolysin, toxic shock syndrome toxin, Panton Valentine leucocidin, fibronectin binding protein A and methicillin resistance, respectively. The analyses were carried out by partial PCR amplification of the genes in all strains, under the conditions described in S1 Table [27]. To detect amplified genes, PCR products were resolved using electrophoresis in 1.2% agarose gel (D1 Low EEO, Conda, 8018). The reference controls used to detect each virulence gene were: CECT 976 for hla, CECT 957 for tst, CECT 435 for lukS/F-PV, CECT 976 for fnbA and CECT 4439 for mecA, all from the Spanish Type Culture Collection (Spanish initials: CECT). In addition, a molecular weight standard (φX174-Hae III digest, Takara Bio Inc., 3405A) was included to determine the size of the amplified fragments.
On other hand, SCCmec typing was performed for positive mecA strains, using multiplex PCR described previously [28]. Furthermore, the amplified fragment was sequence and compared with Genbank database.
A relationship between MLST STs and the presence of virulence genes was investigated using Canoco software (version 5.12), conducting a constrained redundancy analysis (RDA).

Strain identification
Besides MALDI-TOF identification of Staphylococcus aureus, the 16S ribosomal sequences obtained from each sample were compared in the database and identification of the samples as S. aureus was confirmed with at least 98% identity for all of them.

Susceptibility to antibiotics
Results for the MicroScan analysis of all strains are shown in Table 1. Resistance to β-lactam antibiotics was: 88.75% to penicillin, 86.25% to ampicillin, 3.75% to oxacillin and 1.25% to cefoxitin and imipenem. No resistance was detected to ceftaroline.
With regard to β-lactam antibiotics, 3 of the total strains showed resistance to oxacillin, but only 1 of them (strain 73) showed resistance to cefoxitin and imipenem, but also to quinolones (ciprofloxacin, levofloxacin and moxifloxacin).
It is important to note that strain 73 showed the most resistance to the antibiotics used in the assay (13 of 29), and was the only strain isolated that showed resistance to cefoxitin, imipenem, trimethoprim-sulphamethoxazole and to all the MLS group (macrolides, lincosamides and streptogramines). Finally, strains that belonged to adult patients presented more number of resistances than children patients, despite the second group was higher.

Strain typing by MLST
The results of MLST typing indicated high diversity, and according to the database, over half of the strains (N = 44, 55%) presented new STs (N) (S2 Table). Most of the new STs [38] presented new allele combinations that did not correspond to any ST in the database, while for the remaining 6 STs, there was no match in the MLST database for at least one of the 7 sequences studied and they could not be assigned to any described allele. All these new STs were submitted to database, where a new number of ST was assigned for each one. We also defined a number of new alleles described for each gene: 2 for the arc, aro and yqi genes, 1 for glp and pta, and 0 for gmk and tpi.
Among the STs previously defined in the database (45%), 10 only included 1 strain, another two comprised 2 strains (ST5 and ST3224) and ST121 and ST15 comprised 3 and 4 strains, respectively. The sequence type ST152 was the most prevalent (15 strains) in this study, although with a low percentage (18.75%) with respect to all the strains isolated ( Table 2).

Phylogeny study
A phylogenetic study was conducted using concatenated sequences of MLST housekeeping genes for each strain.
First phylogenetic study confirmed that none of the tested strains belonged to S. argenteus or S. schweitzeri.
About the second phylogenetic study, after aligning all sequences and discarding all hypervariable regions, an analysis was performed with jModelTest software, selecting the general time-reversible model (GTR) and including invariable sites (+I) and rate of variation across sites (+G) as the best substitution model. This model was used to generate a maximum likelihood (ML) tree with Mega software.
In the ML tree (Fig 1), 15 clusters could be differentiated. The largest group comprised 11 ST, six of them were new, but very similar to other already described. One of these groups (ST N30, N31 and N32) appeared external to the others with a different but closely related origin of the tree. ST N8 also appeared in an individual branch, but with a common branch point to the rest of the strains. The reference strain from Ethiopia (ST727) appeared in a group with ST N10, and this group was clearly distant from the rest, with a common branch except for the most external group mentioned previously.

Virulence gene detection
The results of PCR virulence detection are shown in Table 2. The most prevalent gene was hla, which was detected in all strains except number 24 (thus, 98.77% were α-haemolysin positive strains). In contrast, strain number 73 was the only one that presented the mecA gene (1.25% of strains) and one of the three strains did not present the fnbA gene (together with strains 15 and 72), indicating that 96.3% of strains were positive for fibronectin binding protein A. In addition, a high percentage of strains (72.84%, 59 strains) showed the pvl gene. In contrast, the tst gene showed a low prevalence with only 11 strains (13.58%).
On other hand, results of SCCmec typing determined that strain 73 belong to type IVa, according with the expected size and its sequence (S1).
A correlational analysis using Canoco software indicated that in general there was no relationship between the presence of virulence genes and MLST STs, with an explained cumulative fitted variation of 61.75% with 2 axes, especially in new STs, where virulent genes were highly dispersed (S2).

Discussion
In contrast to the rest of the world, the picture of the spread of Staphylococcus aureus infections in Africa is unclear, especially in Ethiopia, where the few studies available were performed in densely populated urban areas [11,15]. The study reported here is the first staphylococcal epidemiology study to be conducted in a rural region in Ethiopia, with low population density. Between 2014 and 2018, in Gambo General Rural Hospital, we isolated 80 S. aureus strains from clinical samples. All of them caused highly virulent infections with deep and extended lesions, and recurrent infections in some cases.
One of the most important factors in S. aureus is the presence of antibiotic resistance. Penicillin was the first and most effective antibiotic against S. aureus infections, but also the first resistance emerged [29]. In our study, most of the strains (88%) showed resistance to penicillin and ampicillin, a similar percentage to that found in other studies conducted in Ethiopia [30]. Moreover, 3 of the total isolated strains [38, 43 and 73] exhibited resistance to oxacillin and 1 of them (strain 73) was also resistant to imipenem and cefoxitin. This result suggested that strain 73 was a true MRSA, and this fact was confirmed by amplification and detection of mecA gene by PCR, which confers resistance to methicillin (MRSA) and to oxacillin, imipenem and cefoxitin, the latter being a good surrogate marker for MRSA detection. Moreover, type IVa SCCmec was determined for strain 73. In addition, MRSA strain 73 was the only strain resistant to quinolones. It has been suggested that fluoroquinolones themselves may predispose patients to infection with MRSA [31]. The percentage of MRSA and resistances to quinolones (ciprofloxacin, levofloxacin or moxifloxacin) obtained in our study was notably lower (1.25%) than that reported in other studies conducted in Ethiopia [32]. This difference may be due to the fact that our study was carried out in a rural area with predominantly community-acquired infections, whereas the other two studies were conducted in a crowded hospital using samples mainly collected from surgical wounds, suggesting hospital-acquired MRSA strains.
Moreover, the collected strains were tested against ceftaroline. All of them were susceptible, probably because it is a fifth-generation cephalosporin that is resistant to S. aureus β-lactamase and has a high affinity for PBPs, including the PBP2a present in MRSA [33].
The total percentage of aminoglycoside-resistant strains detected (15-18%) was similar to that found in other studies conducted in Africa [34].
The results for the presence of tetracycline resistance showed that 52.5% of the tested strains were resistant solely to tetracycline and 6.25% to tetracycline and minocycline. These results are similar to those described as average in the African continent [34].
The analysis of resistance to the MLS b group (macrolides, lincosamides and streptogramine b ) determined that 28.75% of strains were only resistant to the lincosamide clindamycin which is very similar that results found in other studies conducted in South Africa (30%) [35]. Resistance to both clindamycin and erythromycin (18%) could be explained by the presence of genes erm, which includes resistance to macrolides, lincosamides and streptogramin b [36]. All strains were susceptible to Sinercyd, a combination of streptogramins A and B, the first of which is not affected by erm ribosomal modifications.
In contrast to the high prevalence of trimethoprim-sulphamethoxazole resistance reported in the literature (43-84%), we obtained a low percentage of resistant strains (1.25%) [34]. However, research on S. aureus imported from Africa strongly suggests that trimethoprim-sulphamethoxazole resistance is emerging around the globe [37].
The use of antibiotics in a rural area in Ethiopia is more restricted than in the country's large cities and economic centres, and under these conditions, we would expect to find lower Maximum likelihood tree. This phylogenetic diagram was generated using sequences from the 7 housekeeping genes from the MLST analysis, and includes all ST found in this study, but also other STs as reference: the only sequence from the MLST database from Ethiopia (727), the first ST described (ST1), and one of the most common STs found in Africa (ST30), apart from ST152 and ST121 also found in this study.
https://doi.org/10.1371/journal.pone.0230031.g001 levels antibiotic resistance. Our results confirmed this hypothesis and an analysis of the data showed that resistances were more common in adult patients, who would have been treated more frequently than children with antibiotics. This finding has also been described in urban areas and in developed countries, and in our case was especially notable in strain 73, the most multiple-resistant strain, which was obtained from an adult leprosy patient who had previously been treated repeatedly with different antibiotics.
Sequencing genomes from different S. aureus strains revealed that the diversity of genes was high, with 22% belonging to variable regions, rendering the pathogen highly versatile [38]. Hence, surveillance of all S. aureus infections around the world is very important to improve our knowledge of their virulence, predict the evolution of strains and infection characteristics and prevent the global spread of multi-resistant or harmful S. aureus.
One important tool for surveillance studies is MLST, which uses an online database not only for identification but also to provide knowledge about molecular epidemiology and the global spread of virulent or antibiotic resistant isolates of bacterial pathogens [39]. This analysis has become the method of choice for molecular typing of many bacteria, including S. aureus, and for comparing strains around the world [40,41].
We found a high diversity of STs in general and 44 strains with new STs, suggesting that they did not belong to the same outbreak. This wide diversity has been reported previously in a review of MLST studies conducted in Africa which mainly focused on Central and West Africa. The review identified ST30, ST121 and ST152 as the most prevalent S. aureus isolates, all of which were MSSA [15]. We found two of these (ST121 and ST152) in our study, and ST152 was the most prevalent of the two, with 15 strains (18.75%). The MLST database describes 31 isolates of ST152, most of them from Central-West and North Africa, but none of them from Ethiopia. It is noteworthy that the strains identified as ST152 were collected in different years, which could confirm the stability of this ST in this region.
In our study, ST15 and ST121 were also found with 4 and 3 strains, respectively. The MLST database describes 486 isolates of ST15 and 104 isolates of ST121, and both have been found in different parts of the world, including Africa, mainly in the Western region.
Some of the new STs were similar to other, previously described allelic profiles but contained 1 or 2 different alleles: 9 of them were similar to ST121, another 7 to ST5, 6 to ST15 and only 1 to ST152.
The only ST in Ethiopia described in the MLST database is ST727. Hence, ours is the first study in the Eastern region of Africa to identify not only new STs but also STs previously described in other regions of Africa. Moreover, the count of new alleles found in each of the 7 genes showed that the most variable gene in our strains was probably the last one (yqi), while gmk and tpi seemed to be more conserved genes.
Despite the high diversity observed with MLST analysis, phylogenetic studies showing that most of the strains were grouped into clusters and that many of the new STs were phylogenetically related to other, previously described STs. However, a comparison with the only ST described in Ethiopia according to the MLST database demonstrated that our strains were clearly dissimilar (except ST N10) and presented a very different evolution.
The virulence analysis was focused on detecting the main virulence genes of S. aureus: hla, tstH, lukS/F-PV, fnbA and mecA. This analysis indicated a high percentage (72.84%) of PVLpositive strains. One explanation for this high percentage might be that the clinical symptoms were mainly cutaneous (SSTI), because PVL is usually associated with furuncles, cutaneous abscesses and severe necrotic skin infections [12,42,43]. However, this high percentage of PVL-positive strains is in agreement with previous studies conducted in other African countries (usually on MSSA), suggesting that there is a high prevalence of S. aureus strains presenting genes encoding PVL on this continent [44,45,46]. It is considered an endemicity that could become a reservoir for the emergence of PVL-positive strains. Moreover, it is important to note that in our study, we found a relationship between ST152 and the presence of PVL. This relationship was described in the first studies with a CA-MRSA as a sporadic PVL producer, mostly located in Central Europe. However, later studies in Nigeria and Mali with PVLpositive MSSA ST152 isolates have suggested that ST152 divergence might be the result of a MSSA clone originating in Africa that migrated to Central Europe and acquired antibiotic resistance there, similar to MRSA [46]. This could explain the prevalence of these infections in travellers from Africa, and the strong relationship between ST152 and PVL producers [47,48].
The hla gene was detected in all strains except one [24]. The high prevalence (98.75%) of this gene was similar to that found in other studies on S. aureus infections in Africa [49,50] and throughout the world [51]. It is also important to note that strain 24 was also the most phylogenetically differentiated from the rest.
Only three of the studied strains did not show the fnba gene [15, 24 and 72]. This is a high percentage (96.3%) of fnba-encoding genes but is in agreement with previous studies conducted in other African countries [52,53]. The high prevalence of α-haemolysin and fibronectin binding protein A is not unusual and could explain the high virulence of these infections, as they play a crucial role in colonisation and infection [16,24].
We found a low prevalence (13.58%) of strains encoding the tst gene, which is similar to that found in another study in Congo (17.5%) [54] and to the prevalence of carriers and clinical isolates in Europe (15-25%) [55,56]. Although TSST-positive infections were initially associated with menstruation, the origin of staphylococcal TSS is diverse, and the most common foci of infection in non-menstrual cases in developed countries are SSTIs [57,58]. Moreover, it is important to emphasise that most of the samples (9 out of 11; 82%) were collected in the same year (2014), suggesting that there was a high rate of acquisition of this gene during that period.
Statistical analysis showed that there was no relationship between the presence of virulence genes and MLST STs. This could be explained by differences in the acquisition of virulence genes described previously and variability in MLST sequences, but might also be due to the high diversity of STs found.
To sum up, the present study shows a high diversity of S. aureus strains in the same rural region, which enhances our knowledge of this kind of infection in Eastern Africa.

Conclusions
➢ This is one of the first studies on Staphylococcus aureus epidemiology in a rural region in Ethiopia.
➢ The prevalence of resistance, especially of MRSA, was low compared with that reported in studies conducted in other parts of Ethiopia and Africa.
➢ MLST analysis showed high diversity of STs and 44 strains with new STs, most of which (new and previously defined STs) are described for the first time in Ethiopia.
➢ Phylogeny studies determined phylogenetic relationships between the new STs and the previously described STs, but far from the only Ethiopian ST in the database.
➢ Virulence gene detection showed a high prevalence of strains encoding pvl, hla and fnba, a low prevalence of the tss gene, and a markedly low prevalence of MRSA.
➢ We found no relationship between STs and the presence of virulence genes, but did observe a relationship between ST152 and the PVL gene, highly defined in other parts of Africa.