Analysis of the small RNA transcriptional response in multidrug-resistant Staphylococcus aureus after antimicrobial exposure

Abstract

The critical role of noncoding small RNAs (sRNAs) in the bacterial response to changing conditions is increasingly recognized. However, a specific role for sRNAs during antibiotic exposure has not been investigated in Staphylococcus aureus. Here, we used Illumina RNA-Seq to examine the sRNA response of multiresistant sequence type 239 (ST239) S. aureus after exposure to four antibiotics (vancomycin, linezolid, ceftobiprole, and tigecycline) representing the major classes of antimicrobials used to treat methicillin-resistant S. aureus (MRSA) infections. We identified 409 potential sRNAs and then compared global sRNA and mRNA expression profiles at 2 and 6 h, without antibiotic exposure and after exposure to each antibiotic, for a vancomycin-susceptible strain (JKD6009) and a vancomycin-intermediate strain (JKD6008). Exploration of this data set by multivariate analysis using a novel implementation of nonnegative matrix factorization (NMF) revealed very different responses for mRNA and sRNA. Where mRNA responses clustered with strain or growth phase conditions, the sRNA responses were predominantly linked to antibiotic exposure, including sRNA responses that were specific for particular antibiotics. A remarkable feature of the antimicrobial response was the prominence of antisense sRNAs to genes encoding proteins involved in protein synthesis and ribosomal function. This study has defined a large sRNA repertoire in epidemic ST239 MRSA and shown for the first time that a subset of sRNAs are part of a coordinated transcriptional response to specific antimicrobial exposures in S. aureus. These data provide a framework for interrogating the role of staphylococcal sRNAs in antimicrobial resistance and exploring new avenues for sRNA-based antimicrobial therapies.

Fig 1

Heat map of sRNA responses. This heat map includes 80 sRNAs where a standard deviation (SD) of log₂ expression levels across the samples of greater than 2 was detected across the individual sRNA feature. The heat map demonstrates clear antimicrobial-responsive sRNA transcriptional patterns. The details are presented in Data Set S2 in the supplemental material. Abbreviations: none, no antibiotic; cefto, ceftobiprole; linez, linezolid; tige, tigecycline; vanco, vancomycin.

Fig 2

Heat map of mRNA responses. This heat map includes 124 mRNAs where a standard deviation of log₂ expression levels across the samples of greater than 2 was detected across the individual mRNA feature. The heat map demonstrates clear time- and strain-specific mRNA transcriptional patterns. The details are presented in Data Set S3 in the supplemental material. Abbreviations: none, no antibiotic; cefto, ceftobiprole; linez, linezolid; tige, tigecycline; vanco, vancomycin.

Fig 3

Nonnegative matrix factorization of the sRNA data set. The results of the 8-class NMF are presented. The 20 different experimental conditions are represented, and their relative membership to each of the 8 classes is demonstrated. Black indicates no membership in the relevant class, while increasing color intensity indicates greater membership. Each experimental condition contributed predominately to one class; however, membership of other classes also occurred. Abbreviations: none, no antibiotic; cefto, ceftobiprole; linez, linezolid; tige, tigecycline; vanco, vancomycin.

Fig 4

sRNA membership in class 4, 5, 6, and 7 responses. The membership of each of the 20 sample conditions is demonstrated in the miniature heat maps at the top of each class. In the larger heat maps for each class, the degree of membership for the top 25 statistically significant sRNAs is demonstrated for classes 4, 5, 6, and 7. Black indicates no membership of the relevant class, while increasing color intensity indicates greater membership. Note, only 23 sRNAs were statistically significant in class 5. (A) The class 4 response, primarily related to ceftobiprole exposure, demonstrates a small number of class-specific sRNAs, including sRNA116 and sRNA259. (B) Class 5 was primarily related to vancomycin exposure and demonstrated 2 highly class-specific sRNAs, sRNA97 and sRNA356. (C) Class 6 was clearly linked to linezolid exposure, and a large number of class-specific sRNAs were identified. (D) Class 7 appeared to be tigecycline responsive and demonstrated a small number of class-specific sRNAs, including sRNA144. Abbreviations: C, ceftobiprole; L, linezolid; T, tigecycline; V, vancomycin. The first column of the miniature heat maps is the no-antibiotic sample.

Fig 5

Heat maps of antibiotic-responsive sRNAs. Shown are miniature heat maps of 15 sRNAs that were upregulated after antimicrobial exposure. These are colored based on the log₂ expression level compared to the gene average, but scaled so that the sample furthest from the mean is fully red or fully blue. Very clear and consistent antimicrobial responses are demonstrated, especially for linezolid-responsive sRNAs. Abbreviations: C, ceftobiprole; L, linezolid; T, tigecycline; V, vancomycin.

Fig 6

Heat maps of strain-specific sRNAs. Shown are miniature heat maps of differentially regulated sRNAs in JKD6009 (VSSA) compared to those of JKD6008 (VISA). These are colored based on log₂ expression level compared to the gene average, but scaled so that the sample furthest from the mean is fully red or fully blue. Abbreviations: C, ceftobiprole; L, linezolid; T, tigecycline; V, vancomycin.