Genetic and phenotypic diversity of methicillin-resistant Staphylococcus aureus among Japanese inpatients in the early 1980s

To trace the linkage between Japanese healthcare-associated methicillin-resistant Staphylococcus aureus (HA-MRSA) strains in the early 1980s and the 2000s onward, we performed molecular characterizations using mainly whole-genome sequencing. Among the 194 S. aureus strains isolated, 20 mecA-positive MRSA (10.3%), 8 mecA-negative MRSA (4.1%) and 3 mecA-positive methicillin-susceptible S. aureus (MSSA) (1.5%) strains were identified. The most frequent sequence type (ST) was ST30 (n = 11), followed by ST5 (n = 8), ST81 (n = 4), and ST247 (n = 3). Rates of staphylococcal cassette chromosome mec (SCCmec) types I, II, and IV composed 65.2%, 13.0%, and 17.4% of isolates, respectively. Notably, 73.3% of SCCmec type I strains were susceptible to imipenem unlike SCCmec type II strains (0%). ST30-SCCmec I (n = 7) and ST5-SCCmec I (n = 5) predominated, whereas only two strains exhibited imipenem-resistance and were tst-positive ST5-SCCmec II, which is the current Japanese HA-MRSA genotype. All ST30 strains shared the common ancestor strain 55/2053, which caused the global pandemic of Panton-Valentine leukocidin-positive MSSA in Europe and the United States in the 1950s. Conspicuously more heterogeneous, the population of HA-MRSA clones observed in the 1980s, including the ST30-SCCmec I clone, has shifted to the current homogeneous population of imipenem-resistant ST5-SCCmec II clones, probably due to the introduction of new antimicrobials.

Detailed genetic characterizations of MRSA and mecA-positive MSSA strains. The results of MLST, SCCmec typing, spa-typing, toxin profiling, and acquired antimicrobial resistance gene profiling are shown in Table 2.
Overall, a diversity of MRSA isolates representing separate clones were found to be present during the early 1980s, and diverse genotypes were detected even among MRSA strains exhibiting the same ST.
Antimicrobial susceptibilities of the strains across SCCmec types. In order to consider the mechanisms for the shift in population structure of HA-MRSA strains from polyclonal to monoclonal in recent years, we compared the antimicrobial susceptibilities of MRSA strains across SCCmec types ( Table 3).
Strains that carried SCCmec types I and II were highly resistant to β-lactams including oxacillin, but those that carried SCCmec type IV were more susceptible to β-lactams despite being mecA-positive. Among the 15 strains carrying SCCmec type I, the rate of erythromycin resistance was the highest (80.0%), followed by resistance to gentamicin (66.7%), clindamycin (40.0%), and minocycline (26.7%), which were commonly used antimicrobials Phylogenetic tree based on whole-genome SNPs in CC5 strains. S. aureus strain ED98 was used as the outgroup. NJ tree was constructed based on the alignment of 3684 SNP sites. mecA-positive MRSA, mecA-negative MRSA, mecA-positive MSSA isolated in this study and reference strains were indicated in red, blue, green and black letters, respectively. The numbers of inter-strain SNP differences were visualized in a red-yellow-green gradient with red indicating the top score (> 600) and green indicating the bottom score (0 www.nature.com/scientificreports/ at that time. The imipenem susceptibility rate in SCCmec type I strains was 73.3% (11 of 15 strains) as compared with 0% in SCCmec type II strains (0 of 3 strains) (p = 0.043).
The mecA-negative MRSA strains (n = 8) were more susceptible to β-lactams, and notably, all strains were susceptible to imipenem except the isolate N89. Also, these strains showed lower MICs for aminoglycosides, minocycline, erythromycin and clindamycin, when compared with mecA-positive strains.  Fig. 1.
A neighbor-joining (NJ) tree was constructed by the alignment of 41,910 SNP sites. The percentage of the reference genome (S. argenteus MSHR1132) covered by all isolates was 62.95% (1,739,038 of 2,762,785 positions). Strains belonging to the CC30 cluster were most abundant (n = 11), followed by clonal complexes (CC) 5 (n = 9), CC1 (n = 4), and CC8 (n = 4), indicating that the population structure of MRSA strains during the early 1980s was composed of diverse clones.
Whole-genome SNP analysis of CC30 and CC5 strains. Detailed pairwise SNPs analysis of 11 ST30 strains described in this study and 6 CC30 reference strains was performed (Fig. 2a). The percentage of the reference genome (SJTUF_J27, ST433) covered by all strains was 89.16% (2,500,756 of 2,804,761 positions) in the SNPs analysis. SNP differences ranged from 28 to 1008. All CC30 strains described in this study clustered into a single clade and were most closely related to the MSSA strain 55/2053 isolated in the United Kingdom in 1955 20 . Strains N83 and N86, which were isolated in Kumamoto in the same year and exhibited the same spa-type t1504, showed 28 SNP differences, suggesting a direct horizontal spread within the hospital. However, the SCCmec type and antimicrobial resistance gene profiles differed between these two strains, suggesting that these strains were independently acquired by each inpatient from different infectious sources. Similarly, the ST30 strains isolated in this study could be recognized as branches of a clone endemic to Japan, with only small numbers of SNPs ranging from 28 to 164 21,22 .
Next, we performed SNP analysis among 9 CC5 MRSA strains in this study and 32 reference strains (Fig. 2b). The percentage of the reference genome (ED98) covered by all strains was 89.13% (2,517,393 of 2,824,404 positions) in the SNP analysis. SNP differences ranged from 0 to 621. ST5 MRSA strains described in this study clustered into five different clades. Strains N366, N98, N345, N279, and N283 harboring SCCmec type I belonged to a single cluster, with the number of SNP differences within the cluster ranging from 30 to 84. Strains N279 and N283, which were isolated in Miyagi in the same year, showed only 30 SNP differences and similar profiles regarding spa type, toxin and antimicrobial resistance genes, suggesting transmission events within a short period. Although strains N366 and N98 strains exhibited only 41 SNP differences, these strains were isolated in locations separated by over 750 km of mostly ocean, Okinawa and Nagasaki (Table 2). According to these results, strains exhibiting ST5-SCCmec I belonging to this clade could be recognized as part of another endemic clone in Japan, as contrasted with strains 16,125, 18,412, 18,341, 10,388, 15,532, 18,583, 16,035, and 10,497, all ST228 and all isolated at a hospital during the outbreaks in Switzerland 23 , and distinguished from each other pairwise by only 0 to 7 SNPs (Fig. 2b). In contrast, only the strains N315 and N106 exhibiting ST5-SCCmec II belong to the same clade with Mu3 and Mu50, which were isolated as HA-MRSA in the 1990s 24 .
These results suggest that MRSA clones exhibiting ST30-and ST5-SCCmec I may have already spread as major endemic clones throughout Japan by the early 1980s, whereas ST5-SCCmec II had achieved only a minor presence at that time.

Discussion
Our results suggest that the population structure of Japanese HA-MRSA strains during the early 1980s was notably different from that in recent years. The early 1980s polyclonal structure included ST5-and ST30-SCCmec I clones, both of which have become uncommon recently. The recent monoclonal population structure of HA-MRSA strains in Japan, which is composed of imipenem-resistant ST5-SCCmec II clone, likely formed over the past several decades, possibly in response to the release of various new antimicrobial agents including imipenem and changes in MRSA treatment strategies from the 1980s onward.
PVL-positive ST30-SCCmec I was the most frequent genotype among Japanese HA-MRSA strains in the early 1980s. According to a previous report, nosocomial outbreaks of MRSA exhibiting PVL-positive ST30-SCCmec IV occurred frequently in the late 1980s and early 1990s in Japanese hospitals 25 , whereas this genotype comprised only a minor population in the early 1980s as represented in this study. Our results suggest that the population structure of Japanese HA-MRSA strains underwent dynamic replacement through the 1980s. This replacement, largely due to ST30 MRSA clones, has probably resulted from high genetic and phenotypic diversity among ST30 MRSA strains. Indeed, in addition to mecA-positive MRSA, ST30 strains rarely isolated today such as mecAnegative MRSA and mecA-positive MSSA, were also found among the ST30 strains analyzed here. Differences in observed SCCmec types and the presence or absence of PVL genes were also noted among these strains.
As previously reported, the CC30 S. aureus lineage can be divided into three clusters: Clade 1 (prototype strain 55/2053; PVL-positive and penicillin-resistant MSSA), Clade 2 (prototype strain TCH60; PVL-positive CA-MRSA harboring SCCmec type IV), and Clade 3 (prototype strain MRSA252/EMRSA-16; PVL-negative HA-MRSA harboring SCCmec type II or IV) 20 . Clade 1 strains cause severe infections and were the epidemic strain type in Europe, the United States, and Australia in the 1950s [26][27][28][29][30] . The percent of S. aureus infections caused by Clade 1 strains had dramatically decreased by the mid-1960s, due to methicillin use for the treatment of penicillin-resistant strains 31  www.nature.com/scientificreports/ that all Japanese ST30 isolates clustered into a single clade including strain 55/2053, suggesting that a Clade 1 strain imported from overseas had acquired SCCmec type I, SCCmec type IV, or unknown genetic factors and had already undergone diversification in Japanese hospitals by the early 1980s. Consequently, the ST30 strains had likely spread throughout Japan as a nosocomial clone causing a regional outbreak at that time. This study shows that ST5-SCCmec I was the second-most frequent genotype among Japanese HA-MRSA strains in the early 1980s. This genotype is shared by EMRSA-3, which was the most common MRSA clone in the United Kingdom in 1987-1988 along with EMRSA-15 (ST22-SCCmec IV) and EMRSA-16 (ST36-SCCmec II) 4 . Studies conducted in South America in the late 1990s have identified the Cordobes/Chilean clone, which is genetically related to EMRSA-3 but presents differences in its pulsed-field gel electrophoresis (PFGE) pattern and spa type [32][33][34] . This MRSA clone was also detected at a high rate in hospitals in South Brazil in 2008, suggesting the potential for re-dissemination in Brazil 35,36 . Although this MRSA clone exhibiting ST5-SCCmec I has remained uncommon in regions outside of South America in recent years, continuous monitoring is needed to prevent future outbreaks.
Surprisingly, the clones of MRSA resistant to imipenem, such as strain N315, existed before imipenem entered clinical use. We previously reported strain N315, which was imipenem-resistant, tst-positive ST5-SCCmec II, as a representative strain of the New York/Japan HA-MRSA clone 37 . We reported that strains harboring SCCmec type II accounted for a large portion of MRSA in Japanese hospitals in the late 1990s 12,38 . In this study, our results show that a Japanese HA-MRSA lineage exhibiting the same genotype as strain N315 was already circulating as one of the diverse clones in the early 1980s. The phenotypic characteristics of strain N315 was multidrug-resistant, especially to imipenem. In the 1980s, multiple broad-spectrum antimicrobials entered clinical use in Japan, while imipenem/cilastatin was launched in 1987 and was being used as an anti-MRSA agent before the clinical introduction of vancomycin in 1991 in Japan. By contrast with SCCmec II strains, strains harboring SCCmec type I, which was the predominant genotype in this study and some countries including the United Kingdom in the early 1980s 12,39 , displayed a high rate of imipenem-susceptibility. It was also reported that in vitro exposure to imipenem can select for conversions of heterogeneous-to-homogeneous and Eagle type-to-homogeneous methicillin resistance in S. aureus strains via mutations to such chromosomal genes as vraSR and rpoB 6,13,24,[40][41][42][43] . Thus, the frequent use of imipenem to treat MRSA infections may have contributed to the selective pressure for imipenem-resistant ST5-SCCmec II MRSA between 1980 and 2000, and caused the dynamic population shift in Japanese hospitals from diverse imipenem-susceptible MRSA clones to the monoclonal imipenem-resistant ST5-SCCmec II MRSA. The reason why imipenem-resistant clones other than ST5-SCCmec II MRSA disappeared in the 1990s is unclear, but some not-yet-understood factors may exist that boost the survival rate of ST5-SCCmec II MRSA.
Our results show that the heterogeneous population of diverse clones observed in the 1980s shifted to the homogeneous population of ST5-SCCmec II clones from the 2000s onward among Japanese HA-MRSA isolates. However, entering the 2010s, further changes have been occurring in the population structural. It was reported that the population of Japanese HA-MRSA was shifting again in the 2010s from N315-like CC5-SCCmec II to CC8-SCCmec IV and CC1-SCCmec IV, both of which had higher susceptibility to cefotaxime, levofloxacin, clarithromycin and clindamycin 44 . The recovery of antimicrobial susceptibilities, and the history of clonal evolution of HA-MRSA strains from the 1980s to the 2010s, seems to reflect improved recent awareness of appropriate antimicrobial usage.
In this study, multiple MRSA strains exhibiting ST247-SCCmec I were isolated in the northeast area of Japan. This genotype is known as the Iberian clone, which was one of the major pandemic MRSA clones until the 2000s [45][46][47][48] . Our results show the local existence of the Iberian clone in Japan during the early 1980s. Interestingly, all ST247-SCCmec I strains in this study were resistant to imipenem. During the early clinical use of imipenem around the world, the Iberian clone may have undergone spread from the 1990s to the early 2000s. However, the Iberian clone has already been supplanted by the current major epidemic clones 49 .
Intriguingly, multiple mecA-negative MRSA and mecA-positive MSSA strains, though rarely observed today, were identified in this study, suggesting that methicillin-resistance in S. aureus strains of that time had both genetic and phenotypic diversity. These atypical S. aureus strains are known as oxacillin-susceptible MRSA (OS-MRSA) or borderline oxacillin-resistant S. aureus (BORSA) with oxacillin MICs typically equal to 1-8 μg/ mL, which have been reported from various geographic locations for over a decade 50,51 . Although the clinical instances are not frequent compared with typical MRSA, OS-MRSA could have been underestimated because of the discrepancy between the phenotypes and genotypes in clinical laboratories 50 . BORSA can appear as community-acquired infections related to previous antimicrobial drug usage 52 . Our results suggest that OS-MRSA and BORSA strains were already circulating in the early 1980s. All OS-MRSA strains in the present study were ST30 strains isolated in geographically separated regions, and accounted for 30.0% (3 of 10) of mecA-positive ST30 strains. Even though S. aureus strains with intermediate methicillin-resistance such as OS-MRSA and BORSA were frequently isolated from hospital inpatients in the early 1980s, enhanced selective pressures due to new drug developments may have eliminated them from hospital environments over the past several decades. The intrinsic mechanisms of methicillin-resistance vary from isolate to isolate. The presence or hyper-production of beta-lactamase [53][54][55] , or the quantity of native PBP proteins and their β-lactam binding affinities [54][55][56][57][58][59][60][61][62] , or mutations in several chromosomal genes (e.g., femA, femB, gdpP, yjbH, and acrB) have been presumed to mediate β-lactam resistance [58][59][60][61][62] . Further systematic characterization of these atypical phenotypes will be indispensable in identifying undiscovered genetic traits of methicillin resistance.
In conclusion, this study reveals the alteration in population structure of HA-MRSA strains from the early 1980s onward, probably due to the survival of highly drug-resistant clones that may have arisen in response to new drugs introduced over the past several decades. Our findings clarify the role of diagnostic microbiology for tracking the epidemiology of MRSA, giving important evidence for a close correlation between spread of www.nature.com/scientificreports/ drug-resistance and appropriate/inappropriate use of antimicrobials. These findings will aid efforts to prevent escalating antimicrobial resistance.

Methods
Bacterial strains collection. This study examined a collection of 194 S. aureus strains (designated as "N" strains) that were isolated from 184 Japanese inpatients in 22 prefectures between January 1982 and December 1983. A subset of these N strains was previously reported 63 . The isolates were mainly from Fukushima (19,9.8%), Miyagi (18, 9.3%), Okinawa (18, 9.3%), and Osaka (18, 9.3%) prefectures. Specimen types were as follows: pus (88, 45%), sputum (43, 22%), blood (20, 10%), urine (15, 8%), pharynx (10, 5%), other (7, 4%), and unknown (12, 6%), suggesting that skin and soft tissue infection and respiratory tract infection were major infectious diseases in the study population. The originally stored isolates were inoculated on BBL Trypticase Soy Agar (TSA) (Beckton Dickinson Japan, Co., Ltd., Tokyo, Japan) and incubated at 37 °C for 24 h. Catalase-positive, Gram-positive cocci that were presumptively identified as staphylococci by colony morphology, were subcultured on TSA. Tube coagulase tests with rabbit plasma (Denka Seiken Co., Ltd., Tokyo, Japan) were performed, and only coagulase-positive staphylococcal strains were selected for further investigation. S. aureus was confirmed by a PCR method targeting the thermonuclease (nuc) gene after DNA extraction 64  Determination of methicillin resistance. We determined phenotypic methicillin resistance in all S. aureus strains by evaluating oxacillin and cefoxitin susceptibilities according to Clinical and Laboratory Standards Institute (CLSI) M100-S22 performance standards. In addition, all strains were genetically assessed by two different PCRs for the presence of the mecA gene 65,66 . We also confirmed the presence or absence of the mecA gene by whole-genome sequencing for phenotypically-identified MRSA strains.
The differences in the rates of imipenem susceptibility by SCCmec types were evaluated using Fisher's exact test utilizing the fisher. test function in R version 3.5.1 (R Development Core Team). Differences with p values < 0.05 were considered significant. Molecular typing of MRSA strains. spa-typing and multilocus sequence typing (MLST) were carried out as previously reported [67][68][69] . Direct sequencing of PCR products was performed for spa typing and MLST for the S. aureus strains. Sequencing reactions were performed using a Big Dye Terminator (version 3.1) Cycle Sequencing Kit with an ABI Prism 3100 genetic analyzer (Applied Biosystems, Thermo Fisher Scientific Inc., Waltham, MA, USA). After assembling both forward and reverse consensus sequences, the spa type and MLST were assigned using the RIDOM web server (http://spase rver.ridom .de/) and the PubMLST (https ://pubml st.org/organ isms/staph yloco ccus-aureu s), respectively.
After sequencing, the obtained reads were filtered and trimmed by removing bases with quality value scores of 20 or less, de novo assembly was performed using the CLC Genomics Workbench version 9 (Qiagen N.V., Venlo, The Netherlands) with the default parameters.
Assembled contigs were submitted to spaTyper 1.0 for spa-typing, ResFinder 3.0 for detection of acquired drug-resistant genes, and MLST 1.8 for MLST, which are all housed at the Center for Genomic Epidemiology (CGE) website (http://www.genom icepi demio logy.org//) 16 www.nature.com/scientificreports/ To infer the phylogenetic relationship based on whole-genome SNPs among strains in this study and 125 reference strains, assembled contigs were also submitted to CSI phylogeny 1.4 on the CGE website 76 . The complete sequences of the reference strains were accessed from the National Center for Biochemistry Information (NCBI) database. In addition, pairwise SNP analyses were performed focusing on CC5 and CC30 strains in order to elucidate relatedness with and preservation among the recent MRSA strains. Using a Newick file output from SNP analysis by CSI phylogeny, a neighbor-joining (NJ) tree was visualized using Figtree v1.4.3 (http://tree.bio. ed.ac.uk/softw are/figtr ee/). The numbers of inter-strain SNP differences were constructed in a red-yellow-green gradient with red indicating the top score (> 600) and green indicating the bottom score (0).

Data availability
The read data for whole-genome sequencing analysis of strains in this study have been deposited in GenBank under accession number DRA010146.