Clinical isolates of Enterococcus faecium exhibit strain-specific collagen binding mediated by Acm, a new member of the MSCRAMM family


  • Sreedhar R. Nallapareddy,

    1. Division of Infectious Diseases, Department of Internal Medicine,
    2. Center for the Study of Emerging and Re-emerging Pathogens,
    Search for more papers by this author
  • George M. Weinstock,

    1. Center for the Study of Emerging and Re-emerging Pathogens,
    2. Department of Microbiology and Molecular Genetics, University of Texas Medical School at Houston, 6431 Fannin Street, Houston, TX 77030, USA.
    3. Human Genome Sequencing Center, Baylor College of Medicine, Houston, TX 77030, USA.
    Search for more papers by this author
  • Barbara E. Murray

    Corresponding author
    1. Division of Infectious Diseases, Department of Internal Medicine,
    2. Center for the Study of Emerging and Re-emerging Pathogens,
    3. Department of Microbiology and Molecular Genetics, University of Texas Medical School at Houston, 6431 Fannin Street, Houston, TX 77030, USA.
    • For correspondence at the second address. E-mail; Tel. (+1) 713 500 6767; Fax (+1) 713 500 5495.

    Search for more papers by this author


A collagen-binding adhesin of Enterococcus faecium, Acm, was identified. Acm shows 62% similarity to the Staphylococcus aureus collagen adhesin Cna over the entire protein and is more similar to Cna (60% and 75% similarity with Cna A and B domains respectively) than to the Enterococcus faecalis collagen-binding adhesin, Ace, which shares homology with Acm only in the A domain. Despite the detection of acm in 32 out of 32 E. faecium isolates, only 11 of these (all clinical isolates, including four vancomycin-resistant endocarditis isolates and seven other isolates) exhibited binding to collagen type I (CI). Although acm from three CI-binding vancomycin-resistant E. faecium clinical isolates showed 100% identity, analysis of acm genes and their promoter regions from six non-CI-binding strains identified deletions or mutations that introduced stop codons and/or IS elements within the gene or the promoter region in five out of six strains, suggesting that the presence of an intact functional acm gene is necessary for binding of E. faecium strains to CI. Recombinant Acm A domain showed specific and concentration-dependent binding to collagen, and this protein competed with E. faecium binding to immobilized CI. Consistent with the adherence phenotype and sequence data, probing with Acm-specific IgGs purified from anti-recombinant Acm A polyclonal rabbit serum confirmed the surface expression of Acm in three out of three collagen-binding clinical isolates of E. faecium tested, but in none of the strains with a non-functional pseudo acm gene. Introduction of a functional acm gene into two non-CI-binding natural acm mutant strains conferred a CI-binding phenotype, further confirming that native Acm is sufficient for the binding of E. faecium to CI. These results demonstrate that acm, which encodes a potential virulence factor, is functional only in certain infection-derived clinical isolates of E. faecium, and suggest that Acm is the primary adhesin responsible for the ability of E. faecium to bind collagen.


Enterococci, members of the normal intestinal flora of man and animals and also used as probiotic bacteria, have long been recognized as a common cause of endocarditis, and are now well recognized as important opportunistic pathogens, ranked third among the organisms isolated from nosocomial infections (Murray, 1990; Tannock and Cook, 2002). Among the enterococcal species described, Enterococcus faecalis and Enterococcus faecium represent over 90% of clinical isolates and cause a wide spectrum of diseases, including intra-abdominal and surgical wound infections, urinary tract infections, catheter-related bactaeremia and central nervous system infections, among others. Many of these infections, as well as endocarditis, are probably dependent upon the ability of the infecting bacterium to adhere to the layer of extracellular matrix (ECM) proteins exposed after tissue damage. Recently, E. faecium infections have become a life-threatening challenge to clinicians because of the accumulation of multiple antibiotic resistances, including ampicillin and vancomycin (Murray, 2000).

Colonization of host tissue is considered a vital step in the bacterial infection process (Patti et al., 1994a; Foster and Hook, 1998). Most pathogenic bacteria have been shown to recognize and adhere to various components of the ECM, thus leading to colonization of the host tissue. ECM is a stable macromolecular structure consisting of a complex mixture of glycoproteins and proteoglycans including collagens, laminin and fibronectin, which provides structural support to epithelial and endothelial cells in addition to other roles (Westerlund and Korhonen, 1993). Any trauma that damages host tissue integrity exposes underlying ECM proteins, thus making them accessible for bacterial adhesion. The major structural protein of the ECM is collagen, the single most abundant protein in vertebrates. There are at least 20 different collagens that are characterized by the formation of triple helices (consisting of repeats of the amino acid sequence Gly-X-Y), in which three polypeptide chains are wound tightly around one another in a rope-like structure. Network-forming collagen type IV (CIV) shows tissue-specific distribution and is found exclusively in basal lamina, whereas fibril-forming collagen type I (CI) shows a broad distribution (Miller and Gay, 1987).

A subfamily of bacterial surface adhesins, collectively known as MSCRAMMs (microbial surface components recognizing adhesive matrix molecules), which specifically bind to host ECM, has been identified in a number of Gram-positive bacteria (Patti and Hook, 1994). MSCRAMMs that adhere to host collagen, fibronectin, fibrinogen and/or laminin have been isolated and the corresponding genes well characterized (Jonsson et al., 1991; Patti et al., 1992; Cheung et al., 1995; Rich et al., 1999; Holmes et al., 2001; Terao et al., 2002). In some cases, the primary binding sites in these MSCRAMMs have been localized to specific domains of the respective adhesins (Patti et al., 1993; Boden and Flock, 1994; Courtney et al., 1996; Hartford et al., 1999; Rich et al., 1999). Similar to other Gram-positive pathogens, opportunistic enterococci probably possess genes that can code for distinct surface proteins to enable the bacteria to colonize, as well as those that promote multiplication within the host and assist in evading host defences, thus leading to infection.

Relatively few reports have evaluated adherence of enterococci to ECM proteins, and these have shown variable adherence phenotypes among different strains of E. faecalis, with little or no significant adherence by E. faecium strains tested (Zareba et al., 1997; Xiao et al., 1998; Shiono and Ike, 1999; Styriak et al., 1999). One of our earlier studies on adherence of clinical isolates of enterococci showed that most E. faecalis isolates, but not E. faecium isolates, displayed conditional (after growth at a stress condition) binding to collagens and laminin (Xiao et al., 1998). Subsequent studies identified a collagen-binding MSCRAMM in E. faecalis, and the corresponding gene (ace) has been cloned and characterized (Rich et al., 1999; Nallapareddy et al. 2000a,b). Ace has structural features that are common to other surface proteins expressed by Gram-positive bacteria, including the staphylococcal collagen adhesin Cna (Fig. 1A).

Figure 1.

Structural organization of collagen-binding proteins.

A. Domain organization of E. faecium (Acm), S. aureus (Cna) and E. faecalis (Ace). S, signal sequence; A domain, unique collagen-binding domain; B domain, repeat region (B domain of acm in some isolates consists of 187-aa region followed by a partial repeat of 97 aa); W, cell wall domain containing LPKTS motif in Acm and LPXTG motif in Cna and Ace; M, membrane-spanning domain; and C, charged C-terminal region. Collagen-binding domain is represented by CBD (hatched region). The region of the expressed recombinant protein (rAcm) is also shown.

B. Alignment of the predicted minimal binding region of E. faecium Acm (151–320 aa), with the experimentally proven minimal binding region of S. aureus Cna (151–318 aa) and the suggested binding region of E. faecalis, Ace (174–319 aa). Identical amino acids are shaded. The five amino acids known to be critical for collagen binding by Cna of S. aureus are marked with arrows. Acm versus Cna, 56% similarity and 48% identity; Acm versus Ace, 43% similarity and 34% identity; Ace versus Cna, 41% similarity and 33% identity.

In the current investigation, we have studied the collagen-binding capacity of 32 diverse E. faecium isolates and identified 11 isolates of E. faecium that bind to collagen. In this study, we report the identification and characterization of a collagen-binding adhesin from E. faecium that appears to be expressed and functional only in some clinical isolates.


The E. faecium adhesin, Acm, shares sequence homology and structural organization with Gram-positive collagen-binding adhesins

We searched the E. faecium strain DO partial genome database from our group (http:www.hgsc.bcm.tmc.edumicrobialefaecium) for adhesin homologues using the collagen-binding domain sequence of Ace of E. faecalis and identified a 2166 bp gene that predicted a 721-amino-acid (aa) open reading frame (ORF). This protein was subsequently designated Acm (adhesin of collagen from E. faecium). Although the similarity (47%) between Acm and Ace is confined to the collagen-binding A domain, to our surprise, Acm shows 60% and 75% similarity with A and B domains, respectively, of the Staphylococcus aureus collagen-binding protein, Cna (62% overall similarity). The predicted collagen-binding A domain of Acm is almost the same size (501 aa) as that of the A domain of Cna (503 aa), but is 166 amino acids larger than that of the collagen-binding A domain of Ace. Within the primary sequence of the A domain, we identified a 170-aa putative collagen-binding region (aa 151–320), which shows 48% identity and 56% similarity with the 168-aa minimal binding region (aa 151–318) of S. aureus Cna (Patti et al., 1993). The same region shows 34% identity and 43% similarity with a 146-aa central region of E. faecalis Ace A domain, previously implicated as being the putative collagen-binding trench (Fig. 1B) (Rich et al., 1999). Structural modelling carried out using ExPASy SWISS-Model automated protein modelling server using the previously determined structures for Ace and Cna revealed a trench-shaped binding site in the putative minimal binding region of Acm, which is characteristic of collagen-binding MSCRAMMs. A homology search of the NCBI database with the amino acid sequence of Acm using blastp software revealed that Acm also shows 46% and 40% similarity to a region of fibronectin-binding MSCRAMMs of Streptococcus equisimilis and Streptococcus dysgalactiae. However, this sequence similarity was localized to the N-terminal region (aa 5–647 of S. equisimilis and 5–672 of S. dysgalactiae) not involved in fibronectin binding.

The predicted Acm structure has features typical of other cell wall-anchored proteins of Gram-positive bacteria (Fig. 1A), including a putative 28-aa signal sequence (region S) involved in the transport of proteins across the cytoplasmic membrane, followed by a 501-aa non-repetitive A domain located at the N-terminus. After the A domain, a 97-aa B domain (repeat region in some strains; see below), showing significant identity with B repeats of S. aureus Cna, was identified. The C-terminus amino acid sequence has an LPKTS motif, which is a potential target of sortase (this enzyme has been shown to anchor proteins covalently to the peptidoglycan cell wall in other bacteria), a stretch of hydrophobic residues (18 aa, region M) thought to span the cell membrane, followed by a short cytoplasmic charged tail.

A putative ribosome binding site was located 8 nucleotides 5′ of the predicted Acm start codon. Upstream of this start site, we identified a TTGAAG-20 bp-TATAGT sequence that exhibits reasonable homology to the consensus −10 and −35 (TTGACA-17 bp-TATAAT) sequences of σA. The coding sequence is followed by inverted repeats of 15 bp, possibly forming a hairpin structure of a transcriptional terminator. A systematic search for possible regulatory elements in the 5′ end of the acm gene identified two perfect 11 bp inverted repeats (IR) and three 8 bp direct repeats (DR) in the 400 bp region upstream to the start site. IR and DR sequences in the upstream regions of several genes from Gram-positive bacteria are known to be involved in both positive and negative regulation (Chen et al., 1998; Eder et al., 1999; Ouyang et al., 1999; Saxild et al., 2001; Vats and Lee, 2001)

Detection of the acm gene in E. faecium isolates

Polymerase chain reaction (PCR) was used to amplify the region coding for the A domain of Acm from 32 E. faecium isolates from different sources and laboratory strains. Six of these were from our previous study that showed no appreciable binding among any of 13 E. faecium isolates tested, and 26 additional strains were chosen from our collection for their temporal or isolation site or geographical diversity. A DNA fragment of ≈ 1500 bp was detected in all 32 isolates. However, PCR amplification with the A domain reverse primer (AcmR1; see Table 1) and a primer 3′ to the putative −35 promoter region (AcmF1) showed two different band sizes, ≈ 1650 bp and ≈ 2700 bp, suggesting a possible insertion upstream of the Acm A domain coding region and promoter sequences in some isolates. PCR with both outside primers (AcmF1 and Acm R3; Table 1) showed bands of six different sizes, subsequently shown to result from the presence or absence of the insertion sequences and/or B repeats (see below).

Table 1. . Oligonucleotide primers used in this study.
OligonucleotideForward primer sequences (5′−3′)Locationa
OligonucleotideReverse primer sequences (5′−3′)Locationa, complementary strand
  • a

    . Acm primer locations are relative to the ATG start codon of E. faecium strain DO (TX0016).


Strain-dependent differences in in vitro adherence of E. faecium isolates to collagen type I

Adherence to CI and CIV was tested for the 32 E. faecium isolates described in Experimental procedures. Using a cut-off of 5% cells bound, results for binding of five out of six E. faecium isolates used in our earlier study confirmed the lack of binding to CI and CIV, as reported previously (Xiao et al., 1998); however, for one catheter isolate (labelled as isolate 14 in Fig. 2), we identified borderline binding (7.5 ± 1.5%) to collagen type I. Including an additional 26 isolates, four out of six endocarditis isolates (all four were vancomycin resistant) and seven out of 14 other clinical isolates (three blood isolates, one catheter isolate, one urine isolate and two wound isolates) showed adherence to collagen type I. The percentage of cells that bound to CI varied in different binding isolates, ranging from 7.3% to 32% (Fig. 2). The binding of these isolates to CIV was variable (data not shown). None of the 12 non-clinical strains (10 community-derived isolates from faeces of healthy volunteers, a recipient ATCC strain and a laboratory recipient strain) included in this study adhered to CI or CIV. There was a statistically significant difference (P < 0.005 by Fisher's exact test) in the collagen-binding phenotype of clinical versus non-clinical isolates. The CI binding phenotype in different stages of the growth cycle was also examined for strains TX2555 (CI binder) and GE-1 (non-binder). No significant differences were observed in CI binding of TX2555 from exponential or stationary phase cells (data not shown), and GE-1 was also consistently a non-binder.

Figure 2.

Adherence of 32 E. faecium isolates to immobilized collagen type I (CI). Adherence was tested in wells coated with 1 µg of ECM proteins (see text). Bars represent the means of percentage of cells bound ± standard deviation for six wells. Results are representative of at least three independent experiments. BSA was used as a negative control. Isolates 1–6 are endocarditis isolates (bars shaded in light grey); 7–20 are other clinical isolates (bars shaded in dark grey); and 21–30 are faecal isolates from healthy community-based volunteers (P < 0.005 for clinical versus non-clinical). The nine E. faecium isolates that were studied further (see also Table 4) are labelled as TX0016, #1; TX0054, #2; TX2535, #6; TX2555, #17; TX2405, #19; TX2442, #20; TX1330, #28; D344S (a laboratory recipient strain), #31; and GE-1 (a recipient strain), #32.

Some non-adhering E. faecium isolates have a pseudo acm gene as a result of disruption or mutation in the acm locus

The complete acm gene was sequenced from three clinical E. faecium isolates (TX0054, TX2535, TX2555), which showed 15%, 32% and 20% of cells adhering to CI respectively (labelled as isolates 2, 6 and 17 in Fig. 2). These sequences were compared with that from the non-adhering (2.3% cells bound CI; labelled as isolate 1 in Fig. 2) E. faecium strain TX0016, which revealed that the acm genes of these four clinical isolates of E. faecium are 2166 bp in length, with 100% identity.

The complete acm gene was also sequenced from five other strains showing < 5% of cells adhering to CI (non-binders). Sequence analysis identified four different conditions that should result in lack of functional Acm expression (Table 2). In one strain, a single basepair deletion in the 8 bp homopolymeric ‘A’ stretch of the acm gene of TX1330 (at 192 bp) and a 4 bp deletion in the cell wall-anchoring region caused frameshifts resulting in the introduction of stop codons. In GE-1 (a commonly used recipient strain), the acm gene was interrupted by the insertion sequence (IS6770-like), commonly found in enterococci, in the middle of the signal sequence. In TX2405, IS6770 was found between the 3′ end of the promoter region and the ribosomal binding site, which should prevent transcription. For strains TX2442 and D344S, a mutation-introduced stop codon was identified at amino acid 224 (in the middle of the binding domain), in addition to the identification of insertion sequence IS6770 (in between the 3′ end of promoter region and the ribosomal binding site). The B-domain regions of TX2405, TX2442 and D344S pseudo acm genes were 852 bp in length coding for a 187-aa repeat followed by a partial repeat of 97 aa, whereas the B domains of the other two isolates were found to be 97 aa in length (Table 2). Further estimation of B repeat numbers by PCR indicated the occurrence of 0.5, 1.5, 2.5 or 3.5 repeats among other strains that we have not sequenced.

Table 2.  Relationship between in vitro collagen-binding phenotype and characteristics of acm based on sequence analysis as well as its expression on the surface of nine different E. faecium strains.
Adherence phenotypeaSource of isolate(name of isolate)Characteristics of acm gene(acm size in bp)Number ofB repeatsSurface expressionb
  • a

    .+ and – denote binding (+ defined as> 5% bacteria bound) or lack of binding (≤ 5%) to collagen type I.

  • b

    . Surface expression in E. faecium strains detected by whole-cell ELISA; +/– denotes expression or lack of expression based on OD450 measurements. The OD450 value of negative control E. coli (sum of average and twice the standard deviation) was used as the cut-off value for determining positive expression (see also Fig. 6).

  • c

    . 97-amino-acid B-domain region. The strains containing 1.5 B repeats have a 187-aa region (showing extensive similarity with B-domain repeats of S. aureus, Cna) followed by a 97-aa partial (0.5) repeat.

  • d

    . These strains have been studied previously, and results of binding of these E. faecium strains to ECM proteins were consistent with our previous report (Xiao et al., 1998).

+Endocarditis isolates (TX0054 and TX2535) acm gene codes for 721-aa ORF (2166 bp)0.5c+
dEndocarditis isolate (TX0016) acm gene codes for 721-aa ORF (2166 bp)0.5
+Perineal wound isolate (TX2555) acm gene codes for 721-aa ORF (2166 bp)0.5+
Clinical isolate (TX2405)IS6770 insertion in the promoter of acm gene1.5
Clinical isolate (TX2442) and
laboratory recipient strain
IS6770 insertion in the promoter of acm gene and
stop codon in the A domain
dCommunity isolate (TX1330)Deletion of a single base in the A domain and 4
bases in the cell wall domain of the acm gene
resulted in stop codons
dRecipient strain (GE-1)IS6770-like insertion in the signal sequence of

Specific binding of recombinant Acm A domain (rAcmA) to collagens and the influence of rAcmA on adherence of E. faecium to CI

As the experimental data from the two well-studied collagen-binding proteins Ace and Cna have proved that the binding domain is contained in the unique A domain, we overexpressed His-tagged rAcmA in Escherichia coli. The purified rAcmA migrated as two bands in reducing PAGE gels and as a single band in native gels with a molecular mass of ≈ 65 kDa, which is higher than the predicted molecular mass (60.749 kDa). The molecular mass calculated using mass spectroscopy showed 60.797 kDa. The acidic nature of the polypeptide (pI of 4.72) may explain the slower migration of this protein in PAGE gels. On storage at 4°C, a small amount of this protein was found to degrade to a protein of ≈ 30 kDa.

Figure 3 illustrates the results from an enzyme-linked immunosorbent assay (ELISA) showing that rAcmA binds in vitro to immobilized CI and CIV, but not to laminin (Ln), fibronectin (Fn), fibrinogen (Fg) or bovine serum albumin (BSA). Binding of rAcmA to CI and CIV was found to be concentration dependent and exhibited saturation kinetics with half maximal binding concentrations (KD) of 3.8 µM and 12.8 µM respectively. As is also evident from Fig. 3, binding of rAcmA to CI was greater than binding to CIV at all the concentrations tested. Treatment of collagen-precoated wells with collagenase VII completely eliminated the binding of rAcmA to CI and CIV, confirming the specific binding of rAcmA to collagens rather than to any contaminants in collagen preparations (data not shown).

Figure 3.

Binding of recombinant Acm A to immobilized collagen types I and IV (1 µg) as a function of concentration of Acm A. Binding of recombinant Acm A up to 10 µM concentration has been expanded and is shown in the upper left corner. BSA was used as a negative control. All OD450 values were corrected for the response of penta-His monoclonal antibodies with the respective ECM proteins. Data points represent the means of OD450 values ± standard deviation for six wells representing three independent experiments. (CI, collagen type I; CIV, collagen type IV; Fn, fibronectin; Fg, fibrinogen; Ln, laminin; and BSA, bovine serum albumin). Half-maximal binding of recombinant Acm A with CI and CIV was calculated to be 3.8 and 12.8 µM respectively.

Preincubation of CI-coated wells with different concentrations of rAcmA (5 µM to 100 µM) before the addition of labelled cells resulted in concentration-dependent partial reduction in adherence of a vancomycin-resistant endocarditis isolate TX2535 and a vancomycin-resistant perineum wound isolate TX2555 to CI. Partial inhibition was achieved only at concentrations greater than the saturated binding of rAcmA to CI, whereas rAcmA has no effect on adherence of these strains to fibrinogen over the range of concentrations tested (Fig. 4).

Figure 4.

Effect of different concentrations of recombinant Acm A domain on adherence of two E. faecium strains to collagen type I and fibrinogen (A) TX2555, perineal wound isolate; and (B) TX2535, endocarditis isolate. Adherence was tested in wells coated with 1 µg of ECM proteins (see text). ECM proteins in wells were preincubated with recombinant Acm A for 2 h at 37°C before the addition of 35S-labelled bacteria. Data points represent the mean ± standard deviation for eight wells. Results are representative of four independent experiments. CI, collagen type I; Fg, fibrinogen.

Acm is detected exclusively in E. faecium isolates showing collagen binding

Polyclonal immune rabbit serum raised against rAcmA reacted with a single ≈ 86 kDa protein band in mutanolysin-phenylmethylsulphonyl fluoride (PMSF) cell wall extracts of CI-adhering isolates TX2555, TX2535, TX0054 (Fig. 5). The apparent observed molecular weight is higher than the predicted molecular weight (75.09 kDa after signal peptide and post-translational processing), perhaps because of association of Acm with peptidoglycan. A second possibility is that the difference in migration might result from the slightly acidic nature (pI of 5.18) of the Acm protein. In highly concentrated (5× those from binding strains) mutanolysin-PMSF extracts of strain TX0016 (which has an intact acm gene, but does not adhere to collagen), a very faint immunoreactive protein band was observed. No bands were detected from mutanolysin-PMSF extracts of the pseudo acm gene containing strains TX1330 and D344S (Fig. 5). Probing of mutanolysin extracts of all the above isolates with preimmune rabbit serum showed no bands.

Figure 5.

Western blots of mutanolysis extracts of E. faecium isolates. Lanes 1–3, TX1330, TX2555 and TX2535 probed with preimmune rabbit serum; lane 4, molecular weight standards; lanes 5–7, TX1330, TX2555 and TX2535 probed with anti-Acm A polyclonal serum.

Antiserum to purified Acm A failed to detect the recombinant A domains of either Ace and Cna. Similarly, anti-Ace A polyclonal serum reacted only with the recombinant A domain of Ace, but not with either of the recombinant A domains of Cna or Acm.

Complementation of the naturally occurring mutation in the acm gene of two E. faecium strains conferred collagen type I binding capacity

We were unsuccessful in several attempts to create a disruption in the acm gene of adhering clinical strains, largely because of the low transformation capability of E. faecium clinical strains (data not shown). As our three known moderately transformable E. faecium strains each contains a naturally interrupted pseudo acm gene, we introduced a functional acm gene into two of these strains using the low-copy-number shuttle vector pWM401. Owing to the apparent instability of pWM401 and, hence, the construct pTEX5343 (pWM401::acm), in the absence of chloramphenicol, we tested adherence after growing cultures in the presence of chloramphenicol. The acm complemented strain, TX5345 (TX1330 with pTEX5343), showed an 11-fold increase in the percentage of cells binding to CI (from 1.6% to 17.6%) compared with TX5101 (TX1330 with pWM401) (Table 3). Similarly, for the complemented strain TX5346 (D344S with pTEX5343), the percentage of binding to CI increased by threefold (3.1% to 10%) compared with D344S containing pWM401. The increased adherence of complemented strains of E. faecium was specific for collagen and did not extend to other ECM proteins (Table 3).

Table 3.  Adherence characteristics of two natural acm mutant E. faecium strains when complemented with a functional acm gene from a collagen-adhering E. faecium isolate.
Strain% of cells bound to ECM proteinsa
  • a

    . ECM, extracellular matrix proteins; CI, collagen type I; CIV, collagen type IV; Fn, fibronectin; Fg, fibrinogen; Ln, laminin; BSA, bovine serum albumin.

  • b

    . Values are means of percentage of cells bound ± standard deviation for eight wells. Results are representative of four independent experiments.

TX1330 1.4 ± 1.1b0.9 ± 0.31.1 ± 0.71.0 ± 0.61.2 ± 1.11.3 ± 1.2
TX5101 (TX1330 with pWM401) 1.6 ± 0.81.1 ± 0.51.4 ± 0.71.2 ± 0.51.1 ± 0.90.8 ± 0.5
TX5345 (TX1330 with pTEX5343)17.6 ± 3.14.1 ± 2.41.1 ± 0.41.0 ± 0.33.3 ± 0.91.0 ± 0.5
D344S 2.4 ± 1.01.0 ± 0.82.4 ± 0.61.8 ± 0.51.2 ± 0.50.9 ± 0.7
D344S with pWM401 3.1 ± 1.01.6 ± 0.92.7 ± 1.12.0 ± 0.41.8 ± 0.90.7 ± 0.4
TX5346 (D344S with pTEX5343)10.0 ± 2.41.9 ± 0.43.1 ± 2.11.8 ± 0 61.7 ± 1.90.6 ± 0.3

Cell surface expression of Acm

Using Acm-specific antibodies eluted from rAcmA, we assessed the surface expression of Acm by nine E. faecium strains from which we had sequenced acm, as well as by the complemented strains, using a whole-cell ELISA. Antiserum raised against formalin-killed TX0016 whole cells was used as a positive control. Only the three CI-adhering clinical isolates (Table 2) and the CI-adhering complemented strains were positive in this assay (Fig. 6). TX0016 (the strain with an intact acm locus that is unable to bind CI) whole cells did not show a positive titre with IgGs against rAcmA. No significant difference was observed in exponential and stationary growth phase cells of TX2555 and TX0016 (data not shown).

Figure 6.

Comparison of Acm expression on cell surface using whole-cell ELISA. The strains used include: one collagen non-binding endocarditis isolate with intact acm gene (TX0016); two collagen-binding E. faecium clinical isolates (TX2535 and TX2555); one collagen non-binding community isolate with pseudo acm gene (TX1330) before and after complementation with shuttle vector (TX5101) or a functional acm gene (TX5345). Bars represent the means of OD450 values ± standard deviation for six wells. Results are representative of three independent experiments. PRS, IgGs from preimmune rabbit serum; anti-Acm, Acm-specific IgGs.


Our interest in E. faecium virulence factors, especially those mediating direct interaction with the host, is related to our concern about the disproportionate increase in multidrug-resistant E. faecium infections over the last decade and the fact that very little is known about virulence factors in this species. Although we and others have shown the presence of homologues of some E. faecalis virulence factors among E. faecium strains, including efaAfm, espfm and gene(s) coding for aggregation substance (Singh et al., 1998; Elsner et al., 2000; Eaton and Gasson, 2001; Willems et al., 2001), none of the genetic determinants has been well characterized in E. faecium, in large part because of difficulties in creating isogenic mutants of clinical isolates of this species.

Enterococcus faecium, a normal inhabitant of the human gut, can cause infection when the organism spreads beyond its original niche into the bloodstream. Thus, adherence to colonic epithelium, to urinary epithelium or heart valves by enterococci is likely to be a key to subsequent infections, and each of these processes may involve some kind of adhesin for ECM proteins. Although two recent studies demonstrated significant binding of plasma proteins including vitronectin, lactoferrin and thrombospondin to E. faecium strains (Zareba et al., 1997; Styriak et al., 1999), no appreciable adherence of E. faecium strains to ECM proteins has been detected in these studies or in our earlier study (Xiao et al., 1998). In contrast to these studies, we describe here the identification of collagen-binding capacity in certain E. faecium strains. The disagreement between the published binding data and the results from the present study may result from our inclusion of more diverse E. faecium isolates, including some from severe infections such as endocarditis.

After failing to identify a homologue of the E. faecalis collagen-binding adhesin Ace in E. faecium strains using antibodies raised against Ace or by screening with an ace probe using low-stringency hybridization (Nallapareddy et al., 2000b; Duh et al., 2001), we were successful in identifying a 721-aa homologue of Cna, subsequently named Acm, in the database sequence of the human endocarditis E. faecium isolate TX0016. Of the five amino acids that are critical for collagen binding by Cna of S. aureus (Patti et al., 1995; Symersky et al., 1997), four (tyrosine, arginine, phenylalanine and asparagines at positions 176, 190, 192 and 194 of Acm) were present, and the fifth critical residue (asparagine in Cna) was conserved as lysine (at position 234 of Acm), the same as in Ace. Furthermore, the domain organization of these three collagen-binding proteins is similar (Fig. 1), suggesting common characteristics for collagen-binding proteins. Interestingly, there is less similarity to Ace, a collagen-binding MSCRAMM of the more genetically related species, E. faecalis.

We initially tested for the presence of acm in 32 E. faecium isolates (including six E. faecium isolates used in our earlier study that showed no binding to CI) by PCR and identified amplification products from all, albeit of different sizes. We then retested the collagen-binding capacity of the six E. faecium isolates from our previous study (Xiao et al., 1998) and identified borderline CI binding (7.5% cells bound versus our cut-off of 5% cells) for one catheter isolate (labelled as isolate 14 in Fig. 2). However, among the total 32 E. faecium strains included in this study, four out of six endocarditis isolates and seven out of 14 other clinical isolates, but none of the non-clinical isolates, showed adherence to CI. To address the size difference in the PCR products of acm from different E. faecium isolates and also to investigate the sequence differences between the binding (to CI) versus non-binding strains, we sequenced the acm locus from eight strains and compared these with the database TX0016 sequence.

Comparison of acm sequences from three CI- binding clinical E. faecium isolates showed that the acm genes of these three clinical isolates are identical. In contrast, sequence of the acm locus from six non-CI-binding E. faecium strains (including the database sequenced strain) identified four different situations leading to non-functional acm (Table 2), which can explain the non-binding phenotype. Thus, the absence of collagen-binding capacity in these five strains was perfectly correlated with the presence of the pseudo acm gene, because of the presence of an insertion sequence and/or stop codons. Similar inactivated pseudogenes have been reported previously in the genome sequences of Mycobacterium leprae, Yersinia pestis, Neisseria gonorrhoeae and Rickettsia sp. (Simonet et al., 1996; Cole, 1998; Zhu et al., 1999; Andersson and Andersson, 2001). In fact, the data from two recent reports demonstrated that expression of the siaA gene of Neisseria meningitidis and the ica locus of Staphylococcus epidermidis is modulated by alternating insertion and excision of insertion sequence (Hammerschmidt et al., 1996; Ziebuhr et al., 1999). Similarly, a more recent study showed, in an experimental human carriage study with pneumococcus, that an in vivo frameshift mutation in pspA is compensated by a second mutation, further demonstrating the naturally occurring variability of expression of virulence factors in pathogenic bacteria (McCool et al., 2002). Taking these results together, it appears that a functional acm is present only in certain clinical isolates, even though all 32 E. faecium isolates tested have the acm locus. This is in contrast to the staphylococcal homologue, cna, which is present in only 38–56% of S. aureus strains (Switalski et al., 1993; Ryding et al., 1997; Smeltzer et al., 1997). The similarity of Acm and Cna, the presence of acm in 32 out of 32 E. faecium isolates tested, but the presence of cna in only some isolates, raises the possibility that E. faecium may be the remote origin of the cna gene of S. aureus. Even though it is tempting to speculate that the Acm adhesin plays a role during infections, based on its highly statistically significant presence in functional form only in clinical isolates, but not in the community faecal or laboratory isolates, further study will be necessary to determine whether this is the case.

The only exception to the above-mentioned correlation between the acm sequence and the collagen-binding capacity is the presence of an intact acm in the non-adhering (2.3% cells bound CI) endocarditis strain E. faecium TX0016. At least two possibilities might account for this apparent discrepancy of having intact acm but absence of the binding phenotype. (i) The very low amount of Acm detected in concentrated mutanolysin preparations might result from the regulation of this gene in this strain. This possibility is suggested by the presence of putative regulatable regions (both direct and indirect repeats) in the upstream promoter region of acm. Smeltzer's group (Gillaspy et al., 1998; Blevins et al., 1999; 2002) have shown that SarA (staphylococcal accessory regulator) represses expression of the S. aureus collagen adhesion gene (cna) by binding to cis elements associated with the cna promoter. Our ongoing studies will investigate a possible role for these putative regulatory elements in the differential expression of acm in strains possessing an intact acm. (ii) In addition, the non-detection of Acm on the cell surface (as measured by whole-cell ELISA) suggests the possibility that, besides the low levels (detected in concentrated mutanolysin extracts) of Acm being produced, Acm may be being masked by polysaccharide or other surface components under our experimental conditions. Previous studies from our laboratory have identified small electron-dense clumps of polysaccharide adjacent to the cell wall, resembling a capsule, in electron micrographs of TX0016 (Arduino et al., 1994; Rakita et al., 2000). Similar observation of low collagen binding by some cna-positive S. aureus strains was shown to result from either masking by heavy encapsulation or failure to transcribe cna (Gillaspy et al., 1997; 1998; Blevins et al., 1999).

Consistent with the domain organization, the in vitro binding characteristics of rAcmA more closely resembled that of the recombinant A domain of S. aureus Cna and differs from E. faecalis recombinant Ace A by: (i) binding exclusively to collagens, but not to laminin; (ii) the concentration at which it binds; and (iii) having higher affinity for CI than for CIV. It is also noteworthy that rAcmA had a partial inhibitory effect on adherence (to CI) of two out of two clinical E. faecium tested, suggesting that native Acm has higher affinity for immobilized CI compared with rAcmA. Of note, preincubation with concentrations of CI as low as 0.5 µg ml−1 resulted in clumping of cells (data not shown), suggesting that the binding of CI to Acm is at multiple sites.

Confirming that the E. faecium strains with the pseudo acm gene were in fact deficient in Acm expression, polyclonal rabbit serum raised against rAcmA failed to detect Acm in extracts of E. faecium strains with a pseudo acm gene, whereas it readily detected a band in mutanolysin cell surface extracts of CI-binding E. faecium strains. The eluted rAcmA-specific IgGs also failed to detect the Acm on the surface of the pseudogene containing E. faecium strains in whole-cell ELISA. These results demonstrated unambiguously that the natural acm mutant strains of E. faecium lacked Acm on the cell surface.

Although we were unable to generate an acm knock-out mutant in clinical E. faecium isolates with collagen-binding capacity, introduction of the acm locus into two natural acm mutant E. faecium strains that lacked CI-binding capacity resulted in up to 11-fold enhancement in their capacity to bind CI. However, there was variation in the level of complementation of these natural acm mutant strains in terms of the CI-binding phenotype (Table 3). Although surface detection of Acm in the complemented strains, as judged by ELISA readings, was indistinguishable from that of TX2555 (the source of the complemented acm locus), nothing is known about the surface distribution of Acm in these strains. Thus, the variations in spatial distribution may have affected the adhesive properties of these strains.

In summary, the data from this study show that acm encodes an MSCRAMM, a potential virulence factor of E. faecium, but is functional only in certain E. faecium isolates, all 11 of which were infection derived. In vitro production of Acm by nine isolates studied was perfectly correlated with the binding phenotype observed for E. faecium strains, suggesting that the expression of Acm is necessary to mediate the attachment of E. faecium strains to collagen type I. Other corroborating data confirming that the acm gene encodes a primary collagen binding adhesin were derived from complementation experiments. In view of the observations that the presence of the collagen-binding adhesin (Cna) and its binding to collagen correlate with the pathogenicity of S. aureus in infections including endocarditis (Patti et al., 1994b; Hienz et al., 1996; Montanaro et al., 1999; Smeltzer and Gillaspy, 2000; Johansson et al., 2001; Elasri et al., 2002), further investigation of a possible role for Acm in E. faecium pathogenesis will be the subject of our future studies.

Experimental procedures

Bacterial strains and culture conditions

A total of 32 E. faecium isolates from different geographic locations was selected from our laboratory collection (obtained over a 12 year period from 12 cities in the USA and Belgium), including six isolates from severe clinical infections, such as endocarditis. Fourteen non-endocarditis clinical isolates were obtained from wounds, urine, bile, catheter and blood. The 10 community-derived isolates were obtained from faeces of healthy volunteers. Nine selected E. faecium strains including two laboratory/recipient strains that were widely used in this study are listed in Table 4. Two strains, TX0016 and TX0054, isolated from the same hospital in Houston, were previously shown to have distinct pulsed field gel electrophoresis (PFGE) patterns; the other isolates were from different hospitals, cities or states. Escherichia coli cells were grown in Luria–Bertani (LB) broth or on LB agar with appropriate antibiotics overnight at 37°C. Enterococci were grown in either brain–heart infusion (BHI) broth/agar or Todd–Hewitt broth/agar (Difco Laboratories) overnight at 37°C. Antibiotics were used at the following concentrations: chloramphenicol at 20 µg ml−1 and ampicillin at 50–100 µg ml−1 for E. coli; chloramphenicol at 8–16 µg ml−1 for the E. faecium complemented stains. All constructs were given TX numbers as shown in Table 4. Plasmids from these constructs were assigned respective pTEX numbers. All plasmids used in this study are also listed in Table 4.

Table 4. . Selected bacterial strains and plasmids used in this study.
Strains/plasmidsRelevant characteristics; origin; year of isolationReference
  1. ApR, ampicillin resistant; CmR, chloramphenicol resistant; TcR, tetracycline resistant; VmR, vancomycin resistant.

 TX0016 (TEX16, DO)Endocarditis isolate; Houston, TX, 1992 Arduino et al. (1994); Xiao et al. (1998)
 TX0054Endocarditis isolate, VmR, ApR; Houston, TX, 1994This study
 TX2535Endocarditis isolate, VmR, ApR; Houston, TX, 1995This study
 TX2555Perineal wound isolate, VmR; Lubbock, TX, 1995This study
 TX2405Clinical isolate, VmR; Cleveland, OH, 1991This study
 TX2442 (VDA)Clinical isolate, VmR; Belgium, 1993 Van der Auwera et al. (1996)
 TX1330 (SE34)Faecal isolate from community volunteer; Houston, TX, 1994 Coque et al. (1995); Xiao et al. (1998)
 D344SLaboratory recipient isolate; spontaneous mutant of D344 Williamson et al. (1985); Sifaoui et al. (2001)
 GE-1Recipient isolate (ATCC51558); 1988 Eliopoulos et al. (1988)
 TX5101TX1330 containing pWM401, CmRThis study
 TX5345TX1330 containing pTEX5343, CmRThis study
 TX5346D344S containing pTEX5343, CmRThis study
 DH5α E. coli host strain used for routine cloningStratagene
 LMG194 E. coli strain for expression of recombinant proteinsInvitrogen
 TX5343 E. coli DH5α (pTEX5343), CmRThis study
 TX5330 E. coli LMG194 (pTEX5330), ApRThis study
 pBAD/HisAExpression vectorInvitrogen
 pWM401Shuttle vector, CmR, also TcR in E. coli Wirth et al. (1987)
 pTEX53301503 bp TX2555 acm (coding for Acm A domain) cloned into  pBAD/HisA expression vectorThis study
 pTEX53433322 bp complete acm locus with its promoter from TX2555
cloned into pWM401; CmR
This study

General techniques

Collagen types I and IV were purchased from Sigma Chemical. Mouse laminin (isolated from the EHS-sarcoma) was purchased from Invitrogen. Tran 35S label and BSA were purchased from ICN Biomedicals. Oligonucleotide primers were purchased from Invitrogen. All other chemicals used in the investigation were of molecular biology grade. General recombinant DNA techniques such as restriction digestion, agarose gel electrophoresis and ligation were performed using standard methods (Sambrook et al., 1989). Chromosomal DNA from E. faecium isolates were prepared according to the method described earlier (Wilson, 1994). Electroporation of E. coli and E. faecium was carried out using a Bio-Rad gene pulser as described previously (Li et al., 1995). DNA sequencing reactions were performed by the Taq dye-deoxy terminator method on an automated ABI Prism sequencer (Applied Biosystems). The primers used in the sequencing are listed in Table 1.

Adherence assay

Adherence to ECM proteins was tested by a previously described assay (Nallapareddy et al., 2000a). Briefly, bacteria grown overnight in 5 ml of BHI broth containing 10 µCi ml−1 Tran 35S label were harvested, washed three times in phosphate-buffered saline (PBS) and resuspended in 0.1% Tween 80, 0.1% BSA in PBS. The ECM-coated (1 µg of ECM per well, at 4°C overnight) wells were blocked with 200 µl of 0.2% BSA for 2 h with gentle shaking. Labelled bacteria (50 µl adjusted to an OD600 of 0.2) were added to wells and allowed to incubate at room temperature for 2 h with gentle shaking. Wells were then washed with 0.1% Tween 80, 0.1% BSA in PBS three times. The radioactivity of bound cells in each detachable well was counted in a liquid scintillation counter LS6500 (Beckman Coulter). Adherence percentage was calculated using the formula (radioactivity of bound cells/radioactivity of total cells added) × 100. The assays were performed in duplicate. Isolates were considered to adhere to ECM proteins if> 5% of total labelled cells bound to the well (Xiao et al., 1998).

Cloning, expression and purification of recombinant Acm A from TX2555

A 1503 bp DNA fragment coding for the complete A domain was amplified from TX2555 using AcmFE (5′-TTTCTCGAGGATGCAGGCAGAGATATCAGCAG-3′) and AcmRE (5′-TTGGTACCTTATTTATTCTCATTTGTAACGAC TAGC-3′) primers (introduced restriction sites are underlined), cloned into pBAD/HisA expression vector (Invitrogen) followed by electroporation into the E. coli host LMG194, and one of the resulting colonies (designated as TX5330) was verified by sequencing. After electrophoresis of lysates on 10% NuPAGE Bis-Tris gels (Invitrogen), Western transfer was carried out according to the protocol supplied, and the His-tagged recombinant protein was detected with anti-His (penta) antibodies (Qiagen).

Recombinant Ace A domain was overexpressed by inoculating 5 l of LB with 50 ml of overnight culture of TX5330. After 2.5 h of growth at 37°C, arabinose was added to a final concentration of 0.4% to induce protein expression, and incubation was continued for an additional 6 h. Bacteria were harvested by centrifugation, the supernatant was decanted, and the cell pellet was washed and resuspended in 50 ml of buffer A (0.1 M NaCl and 10 mM Tris-HCl, pH 8.0) before being stored at −80°C until use. The suspension was later thawed, and the cells were lysed using a French press. Insoluble cell debris was removed by centrifugation at 28 000 g for 20 min, followed by filtration through a 0.22 µm membrane. Recombinant Acm A domain containing the N-terminal His tag was initially purified by immobilized metal chelate affinity chromatography. A 5 ml HiTrap chelating column (Amersham Pharmacia Biotech) was connected to a fast performance liquid chromatography system (FPLC) and charged with 87.5 mM Ni2+ and equilibrated with buffer A. The filtered bacterial cell lysate was applied to the column, and the column was washed with 15 bed volumes of buffer A containing 5 mM imidazole. The bound protein was eluted with a continuous linear gradient of imidazole (5–200 mM; total volume of 200 ml) in buffer A at a flow rate of 5 ml min−1. Fractions containing recombinant protein, as determined by absorbance at 280 nm and SDS-PAGE, were pooled, dialysed against 25 mM Tris-HCl, pH 8.0, and applied to a 5 ml HiTrap Q-Sepharose column (Amersham Pharmacia Biotech) that had been equilibrated with the same buffer. Bound protein was eluted with a continuous linear gradient of NaCl (0–0.5 M; total volume of 160 ml) in 25 mM Tris-HCl, pH 8.0. Eluted fractions containing purified Acm A domain were identified by SDS-PAGE and estimated to be> 95% pure.

Concentration determination and verification of protein identity

The concentration of recombinant Acm protein was determined spectrophotometrically on a Beckman DU-70 UV-visible spectrophotometer using a 1.0 cm path length cuvette. Molar extinction coefficients were calculated by the method of Gill and von Hippel (1989). To verify the protein identity, mass spectrometry measurements were performed at the instrumentation facility of the Department of Chemistry, Texas A and M University, College Station, TX, USA.

Production of rabbit polyclonal serum

After verifying the His-tagged recombinant Acm A on a Western blot probed with anti-His (penta) antibodies (Qiagen), this protein was used to raise polyclonal antibodies by immunization of rabbits at Bethyl Laboratories and stored at −70°C. Antibody titres of sera were determined by ELISA with preimmune serum as a control.

Binding of recombinant Acm A to ECM proteins

An ELISA was used to analyse the binding ability of recombinant Acm A to immobilized ECM proteins. Microtitre plates were coated with 1 µg of ECM proteins or BSA in 100 µl of PBS and allowed to incubate overnight at 4°C. Wells were washed five times with PBST (PBS with 0.01% Tween 20). After blocking wells with 5% BSA, wells were washed again. Varying concentrations of recombinant Acm A (0.1–50 µM) in PBS with 0.1% BSA were added to the wells and incubated at 37°C. After 4 h, unbound protein was removed by washing with PBST. Bound proteins were detected by penta-His monoclonal antibodies (Qiagen) that recognize the His tag of the recombinant Ace A protein, followed by horseradish peroxidase-conjugated goat anti-mouse IgG antibodies (Invitrogen). Relative binding was measured by monitoring absorbance at 450 nm after the addition of 3,3′,5,5′,-tetramethyl benzidine and H2O2. In a parallel set of experiments, collagen-precoated wells were digested with collagenase VII (Sigma Chemical) before incubation with recombinant Acm.

Inhibition of adherence

The wells coated with collagen type I or fibrinogen were incubated with varying concentrations (5–100 µM) of recombinant Acm A for 2 h at 37°C. The wells were then washed in PBS containing 0.1% Tween 80 and 0.1% BSA to remove excess unbound recombinant Acm A, before the addition of labelled cells in the adherence assay described in an earlier section.

Cell wall-associated protein extraction and Western blotting

Protein extracts from E. faecium strains were prepared using mutanolysin (Sigma Chemical) as described previously (Nallapareddy et al., 2000a). Five units of mutanolysin was used for each one OD600 cells. Protein concentrations were estimated by BCA assay (Pierce). Mutanolysin extracts were electrophoresed on 4–12% NuPAGE Bis-Tris gels (Invitrogen) under reducing conditions in MOPS buffer, and transferred to a polyvinylidene difluoride (PVDF) membrane. Electrophoresis and transfer were carried out according to the protocol supplied by Invitrogen. Membranes were then incubated with either anti-Acm A polyclonal antiserum or preimmune serum (antibody I) followed by Protein A horseradish peroxidase conjugate (antibody II), and developed with 4-chloronaphthol in the presence of H2O2.

Complementation of the acm locus in E. faecium strains with a naturally occurring mutant acm gene

To complement the acm gene in trans, a 3.3 kb PCR fragment containing the acm gene and the promoter sequences was amplified using AcmExp1 (5′-TTTGAATTCTTGTTTATA GAATGTAGTGGGAAG-3′) and AcmExp2 (5′-TTTGAAT TCTTGTTTACTTAGAACATCTGCCAG-3′) primers, digested with EcoRI, cloned into a shuttle vector, pWM401 (Cmr) (Wirth et al., 1987) and transformed into E. coli DH5α resulting in pTEX5343. After confirming the correct sequence, pTEX5343 was transformed into two natural acm mutants (TX1330 and D344S) (Table 4). The CmR colonies were screened for plasmid content by a miniprep procedure described previously (Woodford et al., 1993). The adherence of the complemented E. faecium strains to ECM proteins was carried out by the assay described above.

Whole-cell ELISA to detect expression of Acm on the cell surface

Overnight-grown E. faecium cells were washed and resuspended in 50 mM carbonate buffer, pH 9.6, and then 107 cfu in 100 µl were used to coat microtitre plate wells (in triplicate). Bacteria were allowed to bind at 4°C overnight. Wells were washed five times with PBS containing 0.05% Tween 20 (PBST) to remove the non-adherent cells, and the remaining areas of the wells were blocked with 2% BSA in PBST. One microgram of Acm-specific IgGs (eluted from Western blots of rAcmA by the method described earlier; Nallapareddy et al., 2000a) dissolved in 100 µl of 1% BSA in PBS was added to each well and incubated at 37°C for 2 h. The wells were washed five times with PBST, followed by the addition of 100 µl of a 1:3000 dilution of goat anti-rabbit IgG conjugated to horseradish peroxidase. After 2 h incubation at 37°C, the wells were washed five times with PBST. Chromogenic slow TMB solution (100 µl; Pierce) was added, and the reaction was stopped with 35 µl of 2 M H2SO4 after 15 min incubation at room temperature. The absorbance of each well was read at 450 nm. To verify that whole cells were able to bind to microtitre plates, we used antibodies raised against formalin-killed whole cells of TX0016 as a positive control (Rakita et al., 2000).


Fisher's exact test was used to determine the statistical significance of the differences in in vitro collagen-binding properties of strains from clinical versus non-clinical groups.

Nucleotide sequence accession number

The acm nucleotide sequence from TX2555 was submitted to GenBank under accession number AY135217.


We would like to thank Xiaowen Liang from Magnus Hook's laboratory for her initial technical help in FPLC purification. We are grateful to Magnus Hook for allowing us to use FPLC and for valuable discussions. This work was supported by NIH grant AI42399 from NIAID, the Division of Microbiology and Infectious Diseases to B. E. Murray.