Phenotypic and Genotypic Characteristics of Persistent Methicillin-Resistant Staphylococcus aureus Bacteremia In Vitro and in an Experimental Endocarditis Model
Yan Q. Xiong
Vance G. Fowler
Michael R. Yeaman
Barry N. Kreiswirth
Arnold S. Bayer
Division of Infectious Diseases, Department of Medicine, University of California
San Francisco, California
David Geffen School of Medicine at UCLA
Division of Infectious Diseases, Los Angeles Biomedical Research Institute at Harbor-UCLA
University of Medicine and Dentistry of New Jersey
Public Health Research Institute
Division of Infectious Diseases, Duke University Medical Center
Durham, North Carolina
Background. Persistent MRSA bacteremia (PB) represents an important subset of Staphylococcus aureus infections and correlates with poor clinical outcomes. Methods. We profiled relevant in vitro phenotypic and genotypic characteristics of MRSA isolates from 39 persons with bacteremia (21 had PB and 18 had resolving bacteremia [RB]). We also compared the intrinsic virulence and responsiveness to vancomycin of selected PB and RB strains in an experimental endocarditis model (IE). Results. PB and RB isolates differed significantly with regard to several in vitro characteristics that are believed to impact endovascular infections. PB isolates exhibited significantly more resistance to the cationic defensin hNP-1, enhanced membrane fluidity, and substantially greater adhesion to fibronectin, fibrinogen, and endothelial cells. Genotypically, PB isolates had higher frequency of SCCmec II, CC30, and spa 16; and higher rates of agr type III, cap8, tst-1, and cna carriage. Finally, a prototypic PB strain was more resistant to vancomycin treatment in the infective endocarditis model than a RB comparator strain, despite equivalent virulence profiles. Conclusions. Our findings indicate that PB isolates may have specific virulence signatures that distinguish them from RB isolates. These data suggest that methods might be developed to identify patients at higher risk for PB in real-time, thereby optimizing the effectiveness of anti-MRSA therapeutic strategies.
The persistent bacteremia (PB) syndrome is well
represented in large clinical series of persons with
methicillinresistant Staphylococcus aureus (MRSA) bloodstream
infection (PB prevalence, 20%30%), and it is especially
relevant to patients with endovascular infection [1, 2].
However, in up to one-third of cases, no readily
identifiable cause of PB is identified, despite extensive clinical
evaluation . Therefore, understanding the molecular
mechanisms and determinants of PB is essential to
optimize prevention and therapy against life-threatening S.
We hypothesize that S. aureus uses specific virulence
determinants to persist in the bloodstream and cause PB in
the context of endovascular infection. For example, the
organism must avoid immediate killing by host-defense
peptides liberated by platelets at sites of endovascular infection.
Next, the organism must evade or survive phagocytosis and
intracellular killing by neutrophil-associated oxidative and
nonoxidative killing mechanisms (including those
associated with antimicrobial peptides, such as -defensins).
During this phase, the pathogen must adhere to host cells
and ligands to colonize vascular endothelium and avoid
clearance by the reticuloendothelial system. Following
adherence, the organism proliferates and invades tissues,
creating reservoir foci. Finally, it deploys exotoxins and
exoenzymes to reemerge from these reservoir sites, reenter the
bloodstream, and hematogenously seed metastatic target organs
Because distinct infection foci are not synchronized,
organisms are continuously reemerging into the bloodstream,
accounting for persistent bacteremia. On the basis of this
hypothesized life cycle of PB isolates, we investigated a cadre of in vitro
phenotypic and genotypic characteristics considered to be
involved in these phases of endovascular pathogenesis. In
addition, we examined the in vivo virulence and antibiotic
responsiveness of MRSA isolates associated with PB (hereafter, PB
isolates) and those associated with resolving bacteremia
(hereafter, RB isolates) in a rabbit model of infective endocarditis.
PATIENTS, MATERIALS, AND METHODS
Collection of strains. All MRSA isolates were recovered from
bacteremic patients at Duke University Medical Center during
1994 1999. All isolates were susceptible to vancomycin (minimum
inhibitory concentration [MIC] range, 0.251 g/mL), with no
significant differences in MICs between PB and RB strains.
Two groups of MRSA isolates were studied. The first group was
collected from 21 patients with PB, all of whom had MRSA-positive
blood cultures for 7 days while receiving an antibiotic to which
the isolate was susceptible. The second group was collected from 18
patients with RB who had an initial blood culture that yielded
MRSA and subsequent blood cultures that yielded no MRSA 2 4
days after therapy initiation. Patients in the 2 cohorts were matched
on the basis of demographic characteristics. Both patient groups
had similar initial clinical characteristics and laboratory findings
but differed significantly with respect to characteristics associated
with clinical course and outcome [1, 4, 5].
In vitro assay for susceptibility to hNP-1. In a recent
report, we showed that PB isolates were significantly more resistant
than RB isolates to mammalian platelet microbicidal proteins in
vitro . To evaluate whether the disparity in the in vitro
susceptibility profiles of PB and RB isolates extended to other important
host-defense effectors, we assessed their in vitro responses to
hNP-1, a 3.8-kDa cationic -defensin peptide from human
neutrophils [6 8]. Nonoxidative neutrophil-mediated killing related to
this peptide family is likely important during the initial
bacteremic clearance stage of endovascular pathogenesis [6, 9].
Purified hNP-1 was purchased from Peptide International.
Bacterial susceptibility to hNP-1 was assayed in vitro, as
described in detail elsewhere . Results are expressed as the
percentage of colony-forming units in the initial inoculum
(105 cfu) that survived exposure to hNP-1 after a 2-h
In vitro measurement of adherence to host endovascular
ligands (fibrinogen and fibronectin) and host cells (endothelial cell
and platelets). Tissue culture plates were coated with purified
human fibrinogen (50 g/mL; Sigma Chemical) or fibronectin (50
g/mL; Sigma Chemical) and washed with phosphate-buffered
saline (PBS; pH 7.2) [10, 11]. Plates were then treated with 3% bovine
serum albumin (Sigma Chemicals) to prevent nonspecific adhesion
and were washed again with PBS before organism seeding.
Human umbilical cord veins were obtained and maintained as
previously described . Endothelial cell monolayer plates
were washed twice with prewarmed Hanks balanced salt
solution before organism seeding.
In assays to detect fibrinogen, fibronectin, and endothelial cell
binding, PB or RB isolates that had been cultured overnight were
added at a final inoculum of 5 103 cfu/mL to control plates
without either matrix ligands or endothelial cells, to
fibrinogencoated or fibronectin-coated plates, or to endothelial cell
monolayer-coated plates. The plates were incubated for 1 h
under static conditions, as previously described [10 12]. Unbound
bacteria were removed by washing with PBS, and tryptic soy agar
was added. Adherence was expressed as the percentage of the
initial inoculum bound.
Bacterial adherence to platelets was tested as previously
described . Fresh rabbit platelets (109 platelets/mL) were
specifically labeled with CellTracker Red (2.5 mmol/L).
Overnightcultured PB or RB isolates (109 cfu/mL) were labeled with a
fluorescent dye, Syto13 (2 mmol/L). Binding studies were done
by mixing Syto13-labeled S. aureus cells (108 cfu/mL) and
CellTracker Redlabeled platelets (107 platelets/mL). The percentage
of S. aureus cells bound to platelets was calculated as previously
Measurement of S. aureus cell membrane fluidity. We
previously showed that cationic peptide-resistant S. aureus strains
(e.g., strains resistant to tPMP-1) exhibited significantly higher
degrees of membrane fluidity . Such resistance is believed to
contribute to both initial bloodstream and progressive stages of
endovascular infection . Therefore, we tested the membrane
fluidity of the strain sets, using the fluorescent probe
1,6diphenyl-1,3,5-hexatriene (DPH) [7, 14]. Protocols for DPH
incorporation into target cell membranes, measurement of
fluorescence polarization, and calculation of the degree of
fluorescence polarization index are described in detail elsewhere .
As the polarization index decreases, the degree of membrane
fluidity increases .
E-test. Macro E-tests (AB Biodisk) were performed to
screen for heterovancomycin-intermediately resistant S. aureus
(hetero-VISA). The assays were done as recommended by the
manufacturer. Strains with vancomycin MICs of 8 g/mL
were defined as hetero-VISA .
Multilocus sequence typing (MLST) and clonal complexes
(CCs). MLST is a powerful typing tool for defining the
molecular epidemiology of S. aureus because it can identify specific
MLST clones and group strains into unique CCs . It has
recently been shown that such bacterial genotypes are associated
with distinct clinical outcomes . In addition, CC5 and CC30
have been shown to be significantly associated with more-severe
S. aureus infection . The protocol for MLST has been
published previously .
Staphylococcal protein A (spa) typing. Sequence analysis
of the tandem repeat region in the spa subtypes of MRSA helps
further discriminate the MLST profiles into distinct lineages
. We have previously shown that spa type 16 is associated
with more-complicated outcomes in patients with S. aureus
bacteremia . The spa typing of the current strain set was
performed by means of previously published methods .
Staphylococcal cassette chromosome mec (SCCmec)
classification. SCCmec typing distinguishes MRSA strains into 5 major
classes (types IV) and provides another genotypic assessment to
further refine strain differentiation on the basis of MLST and spa
typing. SCCmec typing was performed for several important
reasons: (1) our prior pulsed-field gel electrophoresis profiling
suggested that PB isolates were of 2 predominant and related lineages
(A1 and A2) ; (2) because isolates were obtained from a single
referral center (Duke University Medical Center), it was likely that
clonal bias would be identified; and (3) specific SCCmec types may
be associated with specific clinical syndromes (e.g., skin and
softtissue infections related to SCCmec type IV) . SCCmec typing
was performed on the 39 MRSA strains, using methods described
Multiplex PCR for S. aureus virulence genes. Coordinated
sets of virulence genes or gene networks are believed to be
involved in 1 pathogenic step during MRSA infection (table 1)
[23, 24]. To compare the endovascular virulon of PB and RB
strains, we used multiplex PCR to determine the presence or
absence of 33 such virulence genes, representing a range of
regulatory, adhesin, exotoxin, antiphagocytic, and exoenzyme
genes. The PCR-based assays were performed by use of
previously published techniques .
Rabbit infective endocarditis model. A well-characterized
model of catheter-induced infective endocarditis in rabbits was
used to study the following 4 phasic outcomes of infection due to
PB and RB isolates in vivo [15, 24]: early bacteremia clearance,
initial vegetation colonization, composite virulence, and
responsiveness to vancomycin therapy. Infective endocarditis was
produced by intravenous injection of select PB isolates (from
strain 420) or RB isolates (from strain 1507) 24 h after
catheterization. Strains 420 and 1507 were selected because of their
highly distinguishable in vitro phenotypic and genotypic profiles
(table 2). Phenotypically, the PB strain differed from the RB
strain in several features likely to impact endovascular
pathogenesis: the PB strain has greater resistance to killing by a
representative platelet antimicrobial peptide, tPMP-1 ; greater
adherence to endothelial cells, fibronectin, and fibrinogen in vitro; and
substantially more fluidic cell membranes (table 2).
Genotypically, the PB strain was CC30 and spa type 16 (both of which were
associated with more-severe MRSA clinical infections in other
studies ) and harbored SCCmec type II; in addition, this
strain was agr type III and positive for cna and tst-1. In contrast,
Adhesin, binds collagen
Adhesin, binds fibronectin
Adhesin, binds fibrinogen
Adhesin, binds fibrinogen
Adhesin, binds fibrinogen
Polysaccharide intercellular adhesion (involved in biofilm formation)
Exotoxin, putative virulence factor
Exotoxin, toxic shock syndrome toxin-1
Enzyme; serine protease
Capsular polysaccharide antiphagocytic factor
in life cyclea
a A detailed account of the life cycle of PB isolates is presented in figure 1.
the RB strain was CC8 and spa type 7, contained SCCmec type
IV, and was negative for agr type I, cna, and tst-1 (table 2). Both
strains were pvl negative and had vancomycin MICs of 1 g/mL.
Neither strain exhibited a hetero-VISA phenotype (data not
Each animal was challenged intravenously with 109 cfu of PB
strain 420 or RB strain 1507 24 h after catheterization. Blood
samples were obtained for quantitative culture 1 min and 30 min
after challenge. Bacterial blood density was expressed in mean
log10 cfu/mL (SD). In addition, animals were sacrificed 30 min
after challenge, and all vegetations were removed for
quantitative culture. The in vivo adherence of each strain to vegetations
was expressed in mean log10 cfu per g of vegetation (SD).
The intrinsic virulence of infective endocarditis strains can
be measured as a composite of infective endocarditis
induction rates over an inoculum challenge range and of target
tissue bacterial densities 24 h after receipt of the ID95
inoculum. Thus, 24 h after catheterization, aortic-catheterized
animals were challenged intravenously with 104, 105, or 106 cfu
of PB or RB strains, the inoculum range that encompasses the
ID95 for most S. aureus strains in this model .
Twentyfour hours after inoculation, all animals were euthanized, and
PB strain 420
RB strain 1507
Absent Present 8 Present
NOTE. MLST, multilocus sequence type; SCCmec, staphylococcal cassette chromosome
a Data are percentage of isolates (SD), unless otherwise indicated.
b Data are percentage of isolates that survived exposure to 1 g/mL of tPMP-1.
NOTE. Data are no. (%) of isolates, unless otherwise indicated. The following genes were also
tested, but there were no significant differences between PB and RB isolates: bsaA1, sea-e, sej, sem,
sei, seg, seo, sen, sdrC-E, bbp, ebpS, icaA, efb, sbl, ssp, clfB, splB, fnbA, lukDE, lukM, and EDIN. CC30,
clonal complex 30; MLST, multilocus sequence type; NS, not signficant; SCCmec, staphylococcal
cassette chromosome mec.
a Data are percentage of isolates (SD) that survived exposure to 40 g/mL of hNP-1.
b Data are percentage of isolates (SD) that bound to fibrinogen.
their cardiac vegetations, kidneys, and spleen were removed
and quantitatively cultured.
Because PB isolates were obtained from patients with no
response to vancomycin therapy, we postulated that the in vivo
response of the PB and RB strains to vancomycin therapy would
differ in the infective endocarditis model. Twenty-four hours
after receipt of the infective endocarditis-inducing inoculation
(105 cfu), animals were randomized to receive either no therapy
(controls) or vancomycin (15 mg/kg intravenously twice per day
for 3 days). This vancomycin dose strategy is associated with
slow and incomplete clearance of most susceptible S. aureus
strains from vegetations and other target organs over a 3-day
treatment period . Twenty-four hours after the last
antibiotic dose, all animals were sacrificed, and target tissues were
removed and quantitatively cultured as described above.
Statistical analysis. Data were analyzed by Kruskal-Wallis
analysis of variance, with corrections made for multiple
comparisons when appropriate. P values of .05 were considered
In vitro susceptibility to hNP-1. Nearly all strains were
resistant to a 20- g/mL dose of hNP-1. However, at a 40- g/mL dose
of hNP-1, the percentage of isolates that survived was
significantly higher in the PB group, compared with the RB group
(P .04) (table 3).
In vitro adherence to host endovascular ligands (fibrinogen and
fibronectin) and host cells (endothelial cells and platelets). In
general, PB isolates bound better than RB isolates to matrix
ligands and host cells relevant to endovascular infection. For
example, the PB group had substantially greater mean adherence
to fibrinogen and fibronectin than the RB group (figure 2;
P .05 for fibrinogen binding). Results also indicated that PB
isolates selected for infective endocarditis studies bound
substantially better to human endothelial cell monolayers than their
RB counterparts (figure 2). In contrast, there was no difference
in the capacity of the PB and RB strain sets overall to bind to
platelets (figure 2).
S. aureus cell membrane fluidity. The PB strain set
exhibited higher cell membrane fluidity than RB isolates. The mean
polarization indices of membrane fluidity (SD) for PB and RB
isolates were 0.276 0.03 and 0.292 0.02, respectively.
However, these differences did not reach statistical significance
PB and RB isolates did not differ in other relevant
phenotypic characteristics. No PB or RB isolate had a hetero-VISA
phenotype detected by means of a macro E-test.
MLST and CCs. A total of 9 sequence types contained in 5
CCs were represented among the 39 isolates. The most common
CC was CC30, present in 27 (69%) of 39 isolates. PB isolates were
significantly more likely than RB isolates to be CC30 (90% vs.
44%; P .05) (table 3).
Staphylococcal protein A (spa) typing. Consistent with
findings in the recent study by Fowler et al. , there was an
association trend of spa type 16
(W-G-K-A-K-A-O-M-Q-QQ-Q) with the PB strain set. Thus, 57% of PB isolates were spa
type 16, compared with 39% of RB isolates (P .16) (table 3).
SCCmec classification. SCCmec II was the most commonly
observed type (85% of isolates). In addition, 20 PB isolates
(95%) contained the SCCmec II element, compared with 13 RB
isolates (72%; P .05) (table 3).
Multiplex PCR for S. aureus virulence genes. A total of
81% of PB isolates were agr type III, compared with only 44% of
RB isolates (P .017) (table 3). All isolates possessed the fnbA
and clfB surface adhesin genes. Of note, 85% of PB isolates were
cna positive, compared with only 56% of RB isolates (P .07)
(table 3). Also, only 15% of PB isolates carried sdrD and sdrE,
compared with 47% of RB isolates (P .05). Carriage of the
tst-1 endotoxin gene was significantly over-represented among
PB isolates, compared with RB isolates (86% vs. 44%; P .03).
Panton-Valentine leukocidin locus was absent in all strains
(table 3). A total of 86% of PB isolates had the cap type 8 genotype
(cap8); among RB isolates, 39% had the cap8 genotype, and 61%
had the cap5 genotype (table 3). Last, all strains carried the
biofilm-associated ica gene, as well as the V8 protease gene ssp.
There were no significant differences between the PB and RB
strain sets with regard to the presence or absence of other
virulence genes studied (table 3).
Infective endocarditis model. No significant differences
between the rate of early clearance of the PB isolate and the rate
of early clearance of the RB isolate were observed 1 and 30 min
after infection (data not shown). In addition, no significant
differences were observed between the PB and RB strains with
respect to the extent of initial colonization of vegetations (data not
At the 104, 105 and 106 cfu inocula, all catheterized animals
developed infective endocarditis. At inocula of 104 cfu and 105
cfu (table 4), bacterial densities in the 3 target tissues were not
significantly different between animals infected with the RB
isolate and those infected with the PB isolate. The majority of
animals infected with either the PB or RB strain at an inoculum of
106 cfu died 24 h after infection.
As shown in table 4, vancomycin therapy resulted in
significant reductions of all target tissue MRSA densities in animals
Received 15 mg/kg vancomycin
MRSA density, log10 cfu/g of specimen, mean
Kidneys Spleen Group 1 PB
NOTE. The PB isolate was strain 420, and the RB isolate was strain 1507.
a P .005, compared with the density in corresponding specimens from PB controls from group 2.
b P .01, compared with the density in corresponding specimens from RB controls from group 2.
c P .005, compared with the density in corresponding specimens from RB controls in group 2.
infected with either PB or RB isolates, compared with respective
untreated controls (P .005). Of note, for cardiac vegetations,
the residual MRSA density after vancomycin treatment in
animals infected with the PB strain was 2.0 log 10 cfu per g of
vegetation greater than that observed for animals infected with
the RB strain (P .01) (table 4).
A number of interesting observations emerged from the present
investigations. Phenotypically, the current data extend upon past
findings regarding antimicrobial peptide resistance; PB strains tend
to be more resistant to key innate cationic host defense molecules
from both neutrophils (e.g., hNP-1) and platelets (e.g., tPMP-1)
. These findings suggest that PB isolates have an intrinsic capacity
to survive interactions with 2 predominant host defense cells early
in the course of bloodstream invasion (figure 1, phase 1
bacteremia). The ability of PB strains to circumvent such innate immune
defenses likely enhance their subsequent pathogenic potential [15,
28]. Furthermore, since these 2 innate defense molecules may be
important in multiple stages of endovascular pathogenesis (e.g.,
tPMP-1resistance fostering infective endocarditis progression)
, the ability of a strain to resist their microbicidal actions
probably contributes to PB.
The present findings also demonstrated that PB isolates adhere
better than RB isolates to host cells (i.e., endothelium) and matrix
ligands relevant to endovascular pathogenesis (i.e., fibrinogen and
fibronectin). Such capabilities may facilitate the colonization phases
of PB infection (phase 2 colonization) (figure 1). Also, since
fibronectin and fibrinogen binding are now considered integral to
endothelial cell and vegetation persistence in experimental infective
endocarditis , increased binding of PB isolates to these ligands,
compared with binding of RB isolates, would theoretically provide
an advantage for PB pathogenesis.
Importantly, PB isolates exhibited substantially more fluidic
membranes than RB isolates. A fluidic membrane phenotype has
previously been linked to cationic antimicrobial peptide
resistance in S. aureus laboratory strains . This characteristic may
facilitate persistent and progressive infective endocarditis due to
PB isolates . The mechanism (or mechanisms) by which
enhanced fluidity causes increased resistance to such peptides is
not well understood but is postulated to be associated with
reduced membrane binding or intramembrane organization of
these cationic molecules.
Genotypically, a greater percentage of PB isolates were
associated with SCCmec II (95%, compared with 72% of RB isolates),
CC30 (90% vs. 44%), and spa type 16 (57% vs. 39%). This
observation is in line with recent findings by Fowler et al. , who
demonstrated a significant trend toward more frequent
hematogenous complications in strains exhibiting these genotypic
profiles. Multiplex PCR revealed that PB isolates differed from
RB isolates in terms of overrepresentation of capsule type 8 and
cna and tst-1 gene carriage. The increased presence of the
adhesin gene, cna, would theoretically allow PB strains to exploit
specific anatomic targets, such as bones, joints, or endothelial
substrata . This role could potentially contribute to
enhancement of the colonization and/or persistence phases in the
life cycle of PB isolates. In addition, overrepresentation of the
tst-1 gene could ostensibly increase the incidence of cytokine
stormassociated sepsis syndromes, leading to worse clinical
outcomes for patients with PB .
The net intrinsic virulence properties of these 2 strains were
not different in the context of infective endocarditis induction or
progression. In contrast, vancomycin therapy clearly divulged
significant outcome differences between groups. Thus, isolates
from the PB strain set were able to persist within cardiac
vegetations to a greater extent than those from the RB strain set during
vancomycin therapy. This result occurred despite identical
vancomycin MICs and no hetero-VISA subpopulations in the PB
and RB strain sets. A similar observation by Fowler et al.  and
Hawkins et al.  demonstrated that vancomycin susceptibility
was not decreased among PB isolates. However, these findings
contrast with those of other studies in which the PB phenotype
has been associated with reduced susceptibility to vancomycin
[33, 34]. These in vivo data regarding vancomycin-induced
disclosure of the PB outcome suggest several interesting
possibilities. For example, PB and RB isolates may differ in other features
not assessed in our profiling that may impact net responsiveness
to vancomycin (e.g., cell wall perturbations, cell surface charge,
and global metabolic pathway abnormalities) . Also, it is
possible that vancomycin may itself differentially impact virulence
pathways in PB isolates, compared with RB isolates.
Our study has several potential limitations. First, all strains
were obtained from Duke University Medical Center, raising the
possibility of single-center and/or geographic bias. Second, these
strains emanated from 1994 1999 and therefore do not
represent recent shifts. Third, we compared only the initial PB and RB
isolates and did not screen for virulence signatures that may have
adaptively evolved during treatment among follow-up blood
isolates. Finally, we only examined a single PB-RB strain pair in
vivo. Current studies are being designed to address these
In summary, the present data support our hypothesis that
there are significant phenotypic and genotypic profiles that can
distinguish PB isolates from RB isolates. Characterization of PB
isolates may afford breakthrough discoveries in the treatment of
life-threatening MRSA infections.