Molecular Characteristics of IS1216 Carrying Multidrug Resistance Gene Cluster in Serotype III/Sequence Type 19 Group B Streptococcus

ABSTRACT Streptococcus agalactiae is the leading cause of meningitis in newborns and a significant cause of invasive diseases in pregnant women and adults with underlying diseases. Antibiotic resistance against erythromycin and clindamycin in group B streptococcus (GBS) isolates has been increasing worldwide. GBS expresses the Srr1 and Srr2 proteins, which have important roles in bacterial infection. They have been investigated as novel vaccine candidates against GBS infection, with promising results. But a recent study detected non-srr1/2-expressing clinical isolates belonging to serotype III. Thus, we aimed to analyze the genotypes of non-srr1/2 GBS clinical isolates collected between 2013 and 2016 in South Korea. Forty-one (13.4%) of the 305 serotype III isolates were identified as non-srr1/2 strains, including sequence type 19 (ST19) (n = 16) and ST27 (n = 18) strains. The results of the comparative genomic analysis of the ST19/serotype III/non-srr1/2 strains further revealed four unique gene clusters. Site 4 in the srr1 gene locus was replaced by an lsa(E)-lnu(B)-aadK-aac-aph-aadE-carrying multidrug-resistant gene cluster flanked by two IS1216 transposases with 99% homology to the enterococcal plasmid pKUB3007-1. Despite the Srr1 and Srr2 deficiencies, which resulted in reduced fibrinogen binding, the adherence of non-srr1/2 strains to endothelial and epithelial cells was comparable to that of Srr1- or Srr2-expressing strains. Moreover, their virulence in mouse models of meningitis was not significantly affected. Furthermore, additional adhesin-encoding genes, including a gene encoding a BspA-like protein, which may contribute to colonization by non-srr1/2 strains, were identified via whole-genome analysis. Thus, our study provides important findings that can aid in the development of vaccines and antibiotics against GBS. IMPORTANCE Most previously isolated group B streptococcus (GBS) strains express either the Srr1 or Srr2 glycoprotein, which plays an important role in bacterial colonization and invasion. These glycoproteins are potential protein vaccine candidates. In this study, we first report GBS clinical isolates in which the srr1/2 gene was deleted or replaced with foreign genes. Despite Srr1/2 deficiency, in vitro adherence to mammalian cells and in vivo virulence in murine models were not affected, suggesting that the isolates might have another adherence mechanism that enhanced their virulence aside from Srr1/2-fibrinogen-mediated adherence. In addition, several non-srr1/2 isolates replaced the srr1/2 gene with the lnu(B) and lsa(E) antibiotic resistance genes flanked by IS1216, effectively causing multidrug resistance. Collectively, we believe that our study identifies the underlying genes responsible for the pathogenesis of new GBS serotype III. Furthermore, our study emphasizes the need for alternative antibiotics for patients who are allergic to β-lactams and for those who are pregnant.

correlation between non-srr1/2 strains and GBS serotypes as well as the genomic characteristics associated with srr1/2 deficiency in GBS clinical isolates. Therefore, in the current study, 41 non-srr1/2 strains from 305 serotype III clinical isolates were analyzed using comparative whole-genome sequencing to investigate the relationship between their genetic structures and antibiotic resistance and virulence.

RESULTS
Epidemiological analysis of non-srr1/2 GBS clinical isolates. A total of 1,248 GBS clinical isolates were collected between 2013 and 2016 at the Korea University Guro Hospital (Seoul, South Korea). Capsular serotypes of the isolates were identified using a latex agglutination assay and multiplex PCR analysis, revealing 86 serotype Ia, 205 serotype Ib, 30 serotype II, 305 serotype III, 11 serotype IV, 124 serotype V, 99 serotype VI, 2 serotype VII, 185 serotype VIII, and 9 serotype IX isolates (Table 1). Additionally, 192 isolates were identified as nontypeable or multitypeable strains. Consistent with our previous report (43), serotype III (24.4%) represented the predominant GBS serotype. Interestingly, serotype VIII was also found to represent a major serotype isolated from the Korea University Guro Hospital, accounting for 14.8% of the clinical isolates and increasing from 7.0% in 2013 to 19.0% in 2016 during this study period.
Phenotypic characteristics of non-srr1/2 isolates. Generally, GBS strains express one of the two SRR proteins Srr1 and Srr2, which are heavily glycosylated cell wallanchoring proteins (45)(46)(47). Thus, the loss of srr1 and srr2 expression in S9968 (ST19) and NSP15-73 (ST27) was confirmed by lectin blot analysis. The results show that a  Ia  Ib  II  III  IV  V  VI  VII  VIII  IX  NT a  Total  2013 16 (  strong glycosylated protein signal was detected in the NEM316 strain, which was lost in the isogenic srr1 deletion mutant NEM316 (Dsrr1) (Fig. 1A). A strong glycosylated protein signal was also detected in the COH1 strain, which was lost upon srr2 deletion (Dsrr2). In contrast, no glycosylated protein signals were observed in the non-srr1/2 isolates S9968 and NSP15-73, suggesting the absence of Srr1/2 proteins in the non-srr1/2 strain serotype III GBS isolates. This was confirmed following our examination of an additional five non-srr1/2 strains, all of which lacked Srr1/2 protein expression (data not shown).
Whole-genome and comparative genomic analyses of non-srr1/2 strains. To confirm the genetic organization of non-srr1/2 isolates of GBS serotype III, whole genome analysis was performed using hybrid sequencing with the Pac-Bio and Illumina MiSeq platforms. The whole genomes of S9968 (ST19), isolated from the urine of a 48year-old female patient with chronic renal failure and urinary tract infections, and NSP15-73 (ST27), isolated from the urine of a 50-year-old female patient with chronic renal failure and urinary tract infections, were analyzed (Table S1) to determine whether any genetic characteristics were unique to non-srr1/2 isolates. The circular chromosomes of S9968 and NSP15-73 were found to be 2,201,113 and 2,302,798 bp long, respectively, which were slightly longer than those of other sequenced serotype III strains, such as COH1 (CC17; 2,065,074 bp) and H002 (CC19; 2,147,420 bp). Additionally, the percent GC contents of S9968 (36.0%) and NSP15-73 (35.8%) were similar to those of COH1 (35.5%) and H002 (35.7%). Among the total 2,251 (S9968) and 2,348 (NSP15-73) genes, 2,149 and 2,246 were protein encoding, respectively. Additionally, the numbers of rRNAs and tRNAs in both strains were 21 and 80, respectively (49,50). In the NSP15-73 strain, all accessory Sec2 system genes were absent, and no other gene substitutions or mutations were detected (Fig. S2). Considering the clinical importance of clindamycin and and Srr2 on the cell surface. Cell wall proteins were isolated from Srr1-expressing GBS (NEM316) and its isogenic mutant (Dsrr1), Srr2-expressing GBS (COH1) and its isogenic mutant (Dsrr2), and non-srr1/2 isolates S9968 and NSP15-73. Srr1/2 proteins separated by SDS-PAGE were detected with biotin-conjugated WGA, followed by incubation with HRP-conjugated streptavidin. (B) GBS binding to immobilized fibrinogen. Suspensions of GBS WT (NEM316, COH1, and S9968) and isogenic mutant (NEM316 Dsrr1 and COH1 Dsrr2) strains were incubated in wells pretreated with human fibrinogen (0.1 mM), followed by detection with 0.1% crystal violet. Cell binding to immobilized fibrinogen was measured as the absorbance at 595 nm. Data are presented as the means 6 standard deviations. **, P , 0.005; ***, P , 0.001. lincomycin in treating GBS infection in patients with allergies to penicillin, we next focused on the characteristics of non-srr1/2 isolates of GBS serotype III with aminoglycoside and lincosamide resistance patterns.
Average nucleotide identity (ANI) is considered the most relevant comparative parameter used for bacterial species delineation (51). The phylogenetic analysis of S9968 compared to 21 serotype III reference strains showed that S9968 had ANI values of .99% with the other reference serotype III strains, with the exception of CNCTC8184 (98.4%) and NEM316 (98.7%), and was found to be the most closely related to H002 (99.93%) and Sag158 (99.76%), both of which belong to ST19 ( Fig. 2A).
BLAST Ring Image Generator (BRIG) software was then used to create the circular structure of the S9968 genome, which was then compared with those of H002 (STIII; ST19) and Sag158 (STIII; ST19) (52). Four large novel cluster sites unique to S9968 were found (Fig. 2B). Site 1 contained 15 genes, including 3 transposons and 1 integrase. Site 2 comprised 36 genes, including that encoding the conjugal transfer protein TraG as well as several conjugation-related genes encoding a conjugative transposon, a relaxase, a relaxosome, and a DNA primase. In addition, three LPxTG cell wall-anchoring proteins were encoded in site 2. Meanwhile, site 3 and site 4 were the contigs flanked by the IS6-like element IS1216. Site 3 harbored 19 genes, including the spw gene belonging to the ANT (9) family of aminoglycoside nucleotidyltransferases, which confers resistance to spectinomycin. The site 4 cluster, found in the srr1 gene locus, comprised 29 genes located upstream of the accessory sec2 system and spanned 11,886 bp (Table S2). In addition, PCR and sequencing analyses (Fig. S3) revealed that among the 41 non-srr1/2 isolates, only those with both aminoglycoside and lincosamide resistance (n = 11) harbored the lnu(B)-carrying fragment (Fig. 2B).
Genetic environment of IS1216 contigs in S9968. Comparative genomic analysis showed that the site 4 cluster of S9968 comprised 12 open reading frames (ORFs), which contained lsa(E) for pleuromutilin-lincosamide-streptogramin A resistance, lnu(B) for lincosamide resistance, aadK and aadE for streptomycin resistance, and aac-aph for gentamicin-kanamycin-tobramycin resistance (Fig. 3). Unlike the previously reported lnu (B)-carrying fragment in GBS (28,31), which carries aadE, spw, and lsa(E), the lnu(B)-carrying fragment of S9968 contains additional aminoglycoside resistance genes, aadK and aacaph (31). In addition, the spw gene, which was absent from the site 4 cluster, was located in the site 3 cluster.
Basic Local Alignment Search Tool Nucleotide (BLASTN) analysis further indicated that antibiotic resistance genes (12 ORFs) in the site 4 cluster showed high similarity to those of several multidrug-resistant bacteria, including Staphylococcus and Enterococcus. Among them, enterococcal plasmid pKUB3007-1 carrying antibiotic resistance genes, flanked by two ISEf1 elements, shared the highest nucleotide homology with the lnu(B)-carrying fragment of S9968 (99% maximum identity and 99.89% query coverage of the total nucleotide sequence) (53). However, it exhibited low homology with previously reported GBS lnu(B)lsa(E) transposons (31).
Virulence of non-srr1/2 isolates. Srr1 and Srr2 serve as important adhesins involved in GBS invasion via the blood-brain barrier (54). In fact, deletion of srr1 or srr2 in GBS resulted in significantly lower levels of invasiveness in human brain microvascular endothelial cells (hBMECs) than those of WT GBS strains (41). Thus, to determine whether non-srr1/2 clinical isolates exhibit impaired binding to endothelial and epithelial cells, we examined if the adherence of S9968 to hBMECs and A549 cells differed compared with that of Srr1-and Srr2-expressing serotype III reference strains (NEM316 and COH1, respectively). The mutant strains deficient in srr1 or srr2 (NEM316 Dsrr1 or COH1 Dsrr2) showed significantly diminished adherence to both hBMECs and A549 cells compared to their parental WT strains (NEM316 or COH1) (Fig. 4A and B). However, despite the defect in both the srr1 and srr2 genes, the capacity for adherence of the S9968 strain to hBMECs and A549 cells did not differ significantly from that of NEM316 and COH1.
Finally, the virulences of S9968 and Srr1-or Srr2-expressing serotype III strains (NEM316 or COH1, respectively) were compared in a mouse model of GBS meningitis. CD-1 mice (n = 10 per group) were infected intravenously (i.v.) with either NEM316, COH1, or S9968, and their survival was monitored over 72 h (Fig. 4C). The results showed that 80% of the mice infected with COH1 or S9968 and 90% of the mice infected with NEM316 died within 72 h, with no significant differences among the three groups. In contrast, srr1or srr2-deficient isogenic mutant strains exhibited  significantly lower virulence than their parental strains (NEM316 and COH1) or S9968. Mice were euthanized, and brain, blood, and lung tissues were collected to enumerate the bacterial loads. As shown in Fig. 4D to F, mice infected with srr1or srr2-deficient isogenic mutants exhibited significantly lower bacterial loads in the brain, lung, and blood than those infected with the parental strains (NEM316 and COH1) or S9968.
Collectively, these findings suggest that despite the defect in Srr1 or Srr2 expression and the failure to bind fibrinogen, non-srr1/2 clinical isolates exhibited capacities for binding to brain cells and virulence comparable to those of Srr1-or Srr2-expressing isolates.

DISCUSSION
The geographic distribution of GBS serotypes varies across countries and over time. This study comprised capsular serotyping of 1,248 GBS clinical isolates collected over 4 years at the Korea University Guro Hospital, located in the southwestern region of Seoul, South Korea, which is inhabited by a diverse Asian population. Consistent with our previous report, serotype III was the most predominant serotype detected among the 10 capsular serotypes. Previously, serotype III was reported to account for 57.7% of invasive infections in neonates, 84.3% of infections in geriatric adults (55), and 21 to 28% of infections in pregnant women (56).
Although the cause of this increased invasiveness associated with serotype III GBS compared to other serotypes has not been fully elucidated, many clinical studies suggest that the lower level of production of CPS-specific antibodies during GBS infections with serotype III may be partially responsible. For example, neonatal infections with serotype III are likely associated with poor maternal transfer of serotype III CPS-specific IgG through the placenta, compared to other serotypes. Resistance to various antibiotics caused by the overprescription of antibiotics is considered another factor contributing to the climbing number of GBS infections in both neonates and adults (57). Among the 1,248 clinical isolates examined in the current study, we observed a pattern of high lincosamide resistance in serotype III. Further genomic analysis revealed that some serotype III isolates acquired more than two IS1216 elements encoded by several antibiotic resistance genes. The first IS1216 element, located in the site 3 cluster, carried only aminoglycoside resistance genes, while the second IS1216 element, located in the site 4 locus, carried both aminoglycoside and lincosamide resistance genes.
IS1216 found in the site 4 cluster carried multiresistance genes, including lsa(E) and lnu(B) for lincosamide resistance and aadK, aacA-aphD, and aadE for aminoglycoside resistance. Although lincosamides such as clindamycin are recommended alternatives for pregnant women and patients who are allergic to b-lactams (16), resistance to this class of drugs has been reported in GBS clinical isolates since the 1990s, which has been shown to be caused primarily by the lnu(B) gene alone or together with lsa (31,61). Similarly, combination treatment with a b-lactam and an aminoglycoside is commonly administered for at least the first 2 weeks of infective endocarditis (62) and prosthetic joint infection (63) treatment; however, the emergence of resistant strains suggests that these agents should be used with caution.
Transposable elements may alter gene expression via insertion within the coding region or the promoter region to enhance or reduce host virulence. For instance, ISEfa4 of E. faecalis disrupts the ddl gene encoding the housekeeping D-alanine-D-alanine ligase, the absence of which results in reduced glycopeptide resistance in E. faecalis (64). In the current study, non-srr1/2 clinical isolates showed the transposition of IS1216 in the accessory Sec system exporting LPxTG-anchoring glycoproteins (Srr1 or Srr2). Due to the insertion of IS1216, srr1 was completely deleted in 11 non-srr1/2 isolates.
The expression of Srr1 or Srr2 promotes GBS attachment to human brain endothelial cells via interaction with the fibrinogen Aa chain (48). Previous in vivo studies with targeted knockdown of the srr1 or srr2 gene have shown that srr1 or srr2 deficiency in GBS results in reduced virulence and colonization compared to those of parental WT strains (41). Similarly, the S9968 strain exhibited significantly lower binding of immobilized human fibrinogen than srr1or srr2-expressing strains (NEM316 or COH1, respectively); however, its binding of endothelial and epithelial cells was comparable to that of the Srr1-and Srr2-expressing strains. Additionally, the virulence of S9968 did not differ significantly from that of Srr1-or Srr2-expressing strains in a murine model of invasive GBS infection. Comparative analysis of its genome with that of NEM316 or its closely related strain H002 revealed additional LPxTG motif-containing proteins, antigen I//II and BspA, in the site 2 gene cluster of S9968. Although the involvement of these proteins in GBS pathogenesis was not investigated in the current study, previous studies have reported that BspA and BspC are critical adhesins in GBS, which interact with the gp340 protein on human epithelial cells and the host cytoskeleton component vimentin, thereby contributing to GBS meningitis pathogenesis (65,66). Thus, it is hypothesized that the loss of Srr1 or Srr2 in S9968, and the resulting decrease in fibrinogen binding, might be compensated by the acquisition of other cell wall-anchoring proteins, such as the BspA isotype. In fact, four isolates (NSP14-66, NSP14-76, NSP14-161, and NSP14-28) showed fibrinogen binding activity similar to or higher than that of Srr1-or Srr2-expressing serotype III strains (NEM316 or COH1). Thus, further investigation is required to define the pathogenesis of non-srr1/2 isolates as well as effective therapeutic strategies.
The results presented in this study describe IS1216 carrying lincosamide and aminoglycoside resistance genes in GBS, which might be acquired from other common multidrug-resistant bacteria such as Enterococcus spp. or Staphylococcus spp. Since all GBS clinical isolates have been reported to express either Srr1 or Srr2 protein on their cell wall, they have been considered potential vaccine candidates (43). However, this study reports the presence of GBS clinical isolates that do no express Srr1/2 following the insertion of IS1216. Thus, new strategies to develop a versatile GBS protein-based vaccine are warranted. In addition, considering the high level of resistance to lincosamide in GBS, alternative antibiotic agents should be considered for patients who are allergic to b-lactam antibiotics. Collectively, the results of this study provide important genomic evidence related to antibiotic resistance in clinical GBS isolates, which could facilitate the development of more effective treatment options.

MATERIALS AND METHODS
Chemical reagents. All chemicals used in this study were purchased from Sigma-Aldrich (St. Louis, MO, USA) unless otherwise indicated.
Bacterial strains. The study was approved by the Korea University Guro Hospital (KUGH IRB number 2016GR0265). All clinical isolates (n = 1,248) were collected between 2013 and 2016 at the Korea University Guro Hospital (Seoul, South Korea), and all non-srr1/2 clinical isolates used in this study are listed in Table S1 in the supplemental material. GBS reference strains and their Srr-deficient isogenic mutants, NEM316, NEM316 Dsrr1, COH1, and COH1 Dsrr2, were kindly provided by Paul Sullam (University of California, San Francisco, San Francisco, CA, USA).
Capsular and genetic analyses of GBS isolates. Capsular serotyping was conducted using the Strep-B-Latex kit (Statens Serum Institute, Copenhagen, Denmark) and a multiplex PCR assay (Table S3) as described previously (67,68). The gene cluster flanked by two IS1216 elements was detected by PCR specific for the targeted region from pol to aadE (Fig. 3 and Fig. S3). The srr1 and srr2 genes were also detected by PCR using specific primers. All primers are listed in Table S4.
Multilocus sequencing typing. GBS multilocus sequence typing (MLST) was performed as described previously (69). Seven housekeeping genes (adhP, pheS, atr, glnA, sdhA, glcK, and tkt) were amplified and sequenced using MLST primer sets (Table S5), and PCR cycles were as follows: 1 cycle at 95°C for 5 min; 30 cycles of 95°C for 20 s, 56.2°C for 20 s, and 72°C for 1.5 min; and 1 cycle of 72°C for 5 min. The identified alleles were submitted to the S. agalactiae MLST database (http://pubmlst.org/sagalactiae/) for the assignment of sequence types (STs). In addition, each ST sharing six of the seven MLST loci with another ST in the group was clustered into a CC as described previously (70).
Antimicrobial susceptibility testing. The broth microdilution test was used to determine the antimicrobial susceptibilities of GBS non-srr1/2 clinical isolates to amikacin (AMK), gentamicin (GEN), kanamycin (KAN), spectinomycin (SPT), streptomycin (STR), tobramycin (TOB), clindamycin (CLI), and lincomycin (LIN) as described previously (71,72). Briefly, GBS was inoculated in Todd-Hewitt broth supplemented with yeast extract (THY) and incubated overnight at 37°C. The culture of GBS grown overnight was diluted to achieve turbidity of an optical density at 600 nm (OD 600 ) of 0.05 to 0.06, and 100 ml of the GBS suspension was then added to each well of a round-bottom 96-well plate (SPL, Pocheon, South Korea) with 2-fold serially diluted antibiotics. The breakpoints according to the criteria of the European Committee on Antimicrobial Susceptibility Testing were used to interpret the final MICs (73).
Detection of Srr1 and Srr2 proteins using the lectin blot assay. GBS cell wall extracts, prepared as described previously (41), were separated using 8 to 12% Bis-Tris gels (Invitrogen) and transferred to nitrocellulose membranes. The membranes were then incubated with biotin-conjugated wheat germ agglutinin (WGA; Vector Labs, Burlingame, CA, USA), followed by incubation with horseradish peroxidase (HRP)-conjugated streptavidin (0.2 mg/ml; Sigma-Aldrich). Signals were visualized using a chemiluminescent Western blotting substrate (Thermo Scientific, Waltham, MA, USA), and data were acquired using the Bio-Rad ChemiDoc Touch imaging system (Bio-Rad Laboratories, Hercules, CA, USA).
Binding of GBS to immobilized fibrinogen. Purified human fibrinogen (3 mg; Sigma-Aldrich) in phosphate-buffered saline (PBS) was immobilized in 96-well microtiter plates (SPL), and the wells were then blocked with 350 ml of a casein-based blocking solution (Sigma-Aldrich) for 1 h at room temperature. The plates were then incubated with 100 ml of the GBS suspension (10 10 CFU/ml) for 30 min at 37°C and washed to remove unbound bacteria. Wells were stained with crystal violet (0.5% [vol/vol]; Sigma-Aldrich) as described previously (48), and the absorbance was measured at 595 nm using a spectrometer (Epoch 2; BioTek, Winooski, VT, USA).
Whole-genome sequencing, genome assembly, and annotation. The GBS clinical isolate S9968 (STIII; non-srr1/2; MLST ST19) was sequenced using PacBio single-molecule sequencers (Pacific Biosciences, Menlo Park, CA, USA), performed by Macrogen Inc. (Seoul, South Korea). De novo assembly was implemented using Hierarchical Genome Assembly Process version 3 (HGAP3). The genome annotations were performed using the PROKKA pipeline (v1.13), and gene functions were identified using eggNOG (74,75). The GenBank accession number for the genomic sequence of the GBS strain S9968 is SAMN15246708. Comparative genomic analysis was performed by analyzing ANI for nucleotide-level comparisons with 21 serotype III strains retrieved from the NCBI GenBank database (Table S6) (51). The phylogenetic tree of 22 GBS strains, 21 serotype III reference strains, and the S9968 clinical isolate was constructed using MEGA 6.0 based on the ANI values (52). Genome-wide visualization of coding sequence identities between S9968 and other genomes of serotype III strains was performed using BRIG.
GBS adherence assay. Human brain microvascular endothelial cells (hBMECs; Lonza, Basel, Switzerland) and human alveolar epithelial cells (A549; ATCC, Manassas, VA, USA) were cultured in 24-well cell culture plates. When the cells reached approximately 90% confluence, they were washed before bacterial infection. GBS was then added to the cells at a multiplicity of infection (MOI) of 10 for 1 h. The monolayer was washed six times with PBS, followed by lysis with 0.05% trypsin-EDTA and 0.25% Triton X-100 after a 30-min incubation at 37°C (76). The lysates were serially diluted and plated onto blood agar plates to enumerate the bacteria. The adherent GBS bacteria were then calculated as follows: (recovered CFU/original CFU) Â 100%.
Mouse model of meningitis. Animal experiments conducted in this study were approved by the Committee on the Use and Care of Animals at the Korea Energy Research Institute (KAERI-IACUC-2019-008 and KAERI-IACUC-2021-002) and were performed according to accepted veterinary standards. A murine model of hematogenous GBS meningitis was described previously (41). Groups (n = 10 per group) of outbred 6-week-old male CD-1 mice (OrientBio, Suwon, South Korea) were injected via the tail vein with 100 ml of PBS containing 5 Â 10 7 CFU of GBS (NEM316, NEM316 Dsrr1, COH1, COH1 Dsrr2, or S9968). Mouse survival was monitored for 72 h. At the experimental endpoint, the remaining mice were euthanized, and blood and brain tissues were collected. The tissues were homogenized, and the blood as well as the brain and lung homogenates were plated on blood agar for the enumeration of bacterial CFU.
Statistics analysis. Differences in the bacterial strains in the in vitro assays were evaluated using unpaired two-tailed Student's t test. The survival of mice was determined using Kaplan-Meier survival analysis; data were representative of results from three independent experiments and expressed as the means 6 standard deviations using GraphPad Prism version 6.0 (GraphPad Software Inc., La Jolla, CA, USA). A P value of ,0.05 was considered statistically significant.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. FIG S1, TIF   . The funders had no role in the study design, data collection and interpretation, or the decision to submit the work for publication.
We have no conflicts of interest to report. All authors have submitted the ICMJE form for disclosure of potential conflicts of interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.