Multiple antibiotic resistant Staphylococcus aureus continues to be one of the most common pathogens of both hospital-associated and community-associated infections worldwide (Klevens et al., 2007; Popovich et al., 2007; Hidron et al., 2008; Kallen et al., 2010). Methicillin-resistant S. aureus (MRSA) infection, acquired immunodeficiency syndrome (AIDS) and viral hepatitis B are the three major infectious diseases worldwide and pose a serious threat to public health (Dantes et al., 2013). Vancomycin is the first-line antibiotic therapy for MRSA infections (Sieradzki et al., 1999; Deresinski, 2005; Moellering, 2005). However, increased use of vancomycin has led to the emergence of vancomycin-intermediate S. aureus (VISA) (Hiramatsu et al., 1997b). Currently, the Clinical Laboratory Standards Institute (CLSI) categorizes S. aureus as vancomycin susceptible (VSSA) (MIC ≤ 2 μg/mL), vancomycin intermediate resistant (4–8 μg/mL), and vancomycin resistant (VRSA) (MIC ≥ 16 μg/mL) (Patel, 2014). VISA has been reported more frequently worldwide and has aroused considerable concern (Hiramatsu et al., 1997a, 2002; Tenover and Moellering, 2007; Rishishwar et al., 2016).

VRSA emerged in 1997 due to acquisition of the vanA gene from vancomycin-resistant enterococci (Hiramatsu et al., 1997a; Chang et al., 2003). Previous studies indicated that spontaneous mutations play important roles in the evolution of drug resistance (Drlica, 2003; Andersson and Hughes, 2014). However, the genetic mechanism of vancomycin resistance in VISA strains has not been identified fully (Cameron et al., 2016). Several of mutations in multiple genes and gene regulation systems are needed to achieve a VISA phenotype. Several genetic alterations in two-component regulatory systems have been reported to be strongly associated with a VISA phenotype, including mutations in the vraSR operon (Mwangi et al., 2007), graRS (Howden et al., 2008b; Neoh et al., 2008; Cui et al., 2009), and walRK (Howden et al., 2011; Shoji et al., 2011). VraS may serve as a switch for the activity of the “cell wall stimulon” (Gardete et al., 2012). The vraS I5N mutation was found to confer heterogeneous vancomycin resistance when introduced into a vancomycin-susceptible MRSA strain (Katayama et al., 2009). The graR N197S mutation was suggested to convert strain Mu3 into the VISA phenotype (Neoh et al., 2008). Howden et al. confirmed that the T136I mutation in graS is also a key mediator of vancomycin resistance (Howden et al., 2008b). The G223D mutation in walK and the K208R mutation in walR were associated with increased vancomycin MICs (Howden et al., 2011; Shoji et al., 2011). The rpoB gene, which encodes an RNA polymerase subunit, also plays an important role in the evolution of VISA (Matsuo et al., 2011). The H481Y mutation in rpoB has been confirmed by allelic replacement experiments to increase vancomycin resistance during development of the VISA phenotype in the Mu3 (hVISA) strain (Matsuo et al., 2011). The majority of previous reports on the genetic mechanism of VISA development have focused on clinical MRSA strains (Doddangoudar et al., 2012). The development of vancomycin non-susceptibility might be affected by methicillin or other drug resistance, and so VISA development in methicillin-susceptible strains warrants investigation.

In this study, 39 methicillin-susceptible S. aureus and 3 MRSA strains were treated with increasing concentrations of vancomycin in vitro to investigate the genetic mechanism underlying development of vancomycin resistance. The genes (rpoB, vraS, graSR, and walRK) important for development of vancomycin non-susceptibility were analyzed after 60 days of vancomycin treatment and compared with those in the parental strains. Our results suggest that four novel mutation sites are important for VISA development: L466S in rpoB, R232K in graS, I594M in walK, and A111T in walR.

VISA strains are increasingly prevalent in the hospital setting and are a major issue in the treatment of MRSA infections. It has been suggested that the occurrence of VISA strains is relatively frequent, representing a threat to public health (Sader et al., 2009). Prolonged vancomycin exposure in patients can contribute to generation of VISA, resulting in treatment failure (Liu et al., 2011). In addition, VISA can also be generated in vitro from VSSA strains by exposure to vancomycin (Matsuo et al., 2011; Doddangoudar et al., 2012). Previous studies of the mechanisms of VISA development focused on clinical MRSA strains (Doddangoudar et al., 2012), while few studies have addressed VISA development in methicillin-susceptible strains. In this study, 29 (69%) VISA strains were generated from 42 S. aureus strains by vancomycin treatment for 60 days in vitro. Our results suggest that seven mutation sites are important for VISA development, including four novel mutations: L466S in rpoB, R232K in graS, I594M in walK, and A111T in walR.

Hierarchical clustering which can directly decompose the dataset into a set of disjoint clusters was performed to trace the dynamic changes in VISA development, and investigate different vancomycin non-susceptibility evolution patterns. The 42 strains were grouped into five clusters, and the MICs of nine strains (eight from C1 and C2 and one from C3) were ≥12 μg/mL (Figure 1). Among those nine strains, only S28 and S23 were assigned to the same ST type (Table 3). 38/42 parental strains was susceptible to oxacillin, rifampicin and teicoplanin. The three MRSA strains (S3, S34, and S37) were resistant to oxacillin, and 2/42 strains (S3 and S16) were resistant to rifampicin. Strains S3 and S16 developed high MICs (10–16 μg/mL) after vancomycin treatment. Three methicillin-susceptible strains—S8, S15, and S41—showed the highest MIC of 16 μg/mL. There was no significant correlation between VISA development and initial resistance to oxacillin, rifampicin or teicoplanin. The oxacillin MICs of five strains (S3, S15, S16, S34, and S37) decreased after vancomycin treatment. It has been suggested that upon acquisition of vancomycin resistance, some strains show a concomitant decrease in oxacillin resistance (Bhateja et al., 2006), and a previous study reported that mutated graR may impair oxacillin resistance (Neoh et al., 2008). In this study, no graR mutation was detected in these five strains. The H481Y/N mutation is located in the rifampin resistance-determining region, and this locus has been reported repeatedly in clinical rifampicin-resistant S. aureus strains (Aubry-Damon et al., 1998; O'Neill et al., 2006; Mick et al., 2010). In this study, five rifampicin-resistant strains harbored the Asn (N) mutation in the 481st amino acid of rpoB (Table 3). However, the parental strains S37 and S41₆₀, which harbor Asn (N), were susceptible to rifampicin. In summary, no significant association was found between these mutations and antibiotic susceptibility changes.

The genetic basis for vancomycin resistance in VISA remains unclear. The rpoB, vraSR, graSR, and walRK genes have been reported to be highly associated with vancomycin resistance. In this study, the complete sequences of rpoB, vraS, graS, graR, walK, and walR were analyzed and compared with those of the susceptible parental strains. Seven mutations occurred more frequently in strains with high MICs (12–16 μg/mL), including L466S and H481N in rpoB, R232K in graS, R222K, A468T, and I594M in walK, and A111T in walR (Table 3). In vitro experiment and JT trend test indicated that four novel mutation sites were important for VISA development—L466S, R232K, I594M, and A111T—which were first reported in this study (Figure 2).

Alam et al. reported that rpoB H481 is the predominant locus associated with an increased vancomycin MIC (Alam et al., 2014). The mutations H481Y/N in rpoB play a dual role in rifampin and vancomycin resistance (Watanabe et al., 2011; Gao et al., 2013). Moreover, it has been suggested that the rpoB mutation itself, and not any other incidental genetic change caused by rifampin resistance, was responsible for the decreased vancomycin susceptibility (Matsuo et al., 2011). In this study, two mutations (L466S and H481N) were significantly associated with increased vancomycin MICs. Although the L466S mutation had not been reported previously, it might be important for the development of VISA.

In vraS, the 15th amino acid locus was significantly associated with MIC according to the JT trend test, but both the V15G and G15V mutations were detected in different VISA strains. Therefore, V15G/G15V seemed to be less important in VISA development.

D148Q mutation in graR was reported previously to be important for development of high-level resistance (Neoh et al., 2008; Doddangoudar et al., 2012). In this study, only the S41 strain contained the D148Q mutation in graR (MIC = 16 μg/mL), while 28/42 (66.7%) VSSA parental strains initially harbored Gln (Q) at locus 148. Therefore, we speculated that the D148Q mutation in graR did not significantly affect the development of vancomycin resistance. The S79F mutation is important in the development of VISA strains (Neoh et al., 2008; Shoji et al., 2011; Doddangoudar et al., 2012; Hafer et al., 2012). Among our strains, only S27 carried the S79F mutation, and exhibited a MIC of 8 μg/mL. According to the JT trend test, no mutation in graR and four mutations (L26F, I59L, Y223D, and R232K) in graS were significantly related to elevated vancomycin MICs. However, the L26F/F26L, I59L/L59I, and Y223D mutations were detected in both VISA and VSSA strains. Thus, these mutation sites were unlikely to be responsible for VISA development. The R232K mutation occurred frequently (3/9) in strains with high MICs (12–16 μg/mL) and was absent in VSSA strains. This suggests that the R232K mutation is involved in the development of vancomycin-intermediate resistance.

G223D mutation in walK and K208R mutation in walR were associated with increased vancomycin MICs (Howden et al., 2011; Shoji et al., 2011). However, these genetic changes are not observed frequently in clinical and laboratory S. aureus strains (Howden et al., 2008a,b; Kato et al., 2010). Shoji et al. collected 39 clinical VISA strains from various countries worldwide and then analyzed the complete sequences of vraSR, graSR, clpP, and walRK. Nine of the 39 (23%) VISA strains from four countries harbored R222K and A468T mutations in walK. The L10F and A243T mutations in walK and P216S in walR have been reported previously in VISA strains (Shoji et al., 2011; Hafer et al., 2012). In this study, 3/9 VISA strains with MICs ≥ 12 μg/mL harbored the R222K mutation, S3 and S16 initially carried Lys (K) at locus 222. The A468T mutation occurred in the same three VISA strains as R222K, while other ten strains carried Thr (T) including seven VISA strains initially (Table 3). Therefore, A468T seemed to be important for VISA development. According to the JT trend test, R222K, I594M in walK and A111T in walR were significantly associated with vancomycin resistance. Although the mutations I594M in walK and A111T in walR have not been reported to date, they were frequently detected (4/9) in strains with MICs ≥ 12 μg/mL. The presence of I594M and A111T in the same strain may result in development of the VISA phenotype.

Overall, seven amino acid changes—L466S, H481N, R232K, R222K, A468T, I594M, and A111T—were detected frequently in our VISA strains. Although the four novel non-synonymous mutations L466S, R232K, I594M, and A111T have not been reported to date, we speculate that these mutations may play an important role in development of VISA strains, together with the H481N, R222K, and A468T mutations, synergistically promote and maintain high-level vancomycin resistance. Several other mutations were detected in both VISA and VSSA strains or in only VSSA strains. Whether these mutations affect development of a VISA phenotype remains unknown. Notably, certain VISA strains did not harbor any important mutations within the sequenced genes. The increased MICs of these strains may be caused by mutations in other undetected/unidentified genes, such as the proteolytic regulatory gene clp (Shoji et al., 2011) and the accessory gene regulator agr (Sakoulas et al., 2002, 2005).

After removal of vancomycin for 60 days, the MICs of all strains decreased. The L10F mutation in walK has been reported previously in VISA strains, thus, the reverse mutation F10L in walK might be associated with loss of vancomycin resistance. These findings suggest that VISA development is affected by a complex gene regulatory network, although other genes or pathways might be involved in decreased MICs observed. The presence of stop codons in vraS and graR was reported to be related to loss of vancomycin non-susceptibility (Doddangoudar et al., 2012). However, no stop codon was detected in this study.

In conclusion, this study demonstrated that prolonged vancomycin exposure leads to development of vancomycin-intermediate resistance in methicillin-susceptible S. aureus strains. Compared with the susceptible parental strains, four novel missense mutations—L466S in rpoB, R232K in graS, I594M in walK and A111T in walR—were detected in high-level vancomycin-resistant strains. Our results also suggest that rpoB, graS, walK and walR are more important than vraS and graR in the evolution of vancomycin non-susceptibility.

YJ, XH, and RW conceived and designed the experiments. YW, XL, LJ, and WH. performed the experiments. XX, YJ, XH, and RW analyzed the data, YW and YJ. wrote the manuscript. All authors reviewed the manuscript.

This work was supported by the Fundamental Research Funds for the Central University (TD2012-03), the Special Fund for Forest Scientific Research in the Public Welfare (201404102), Natural Science Foundation of China (51108029) and a “One-Thousand Person Plan” award.