INTRODUCTION

Top Abstract Introduction Case Report Materials & Methods Results Discussion References

The term "small colony variant" (SCV) refers to the phenomenon where certain variants of bacteria grow slowly on routine media and yield unexpectedly small colonies in comparison to the normally growing parent strains (for a review see reference 28). The phenomenon of SCVs has been known since the beginning of this century (17) and has been reported for many genera and species, including the genera Staphylococcus and Pseudomonas and diverse enterobacteriaceae species (1, 5, 7, 9, 18, 31-33, 37). So far, nearly all SCVs isolated from clinical specimens have been identified as Staphylococcus aureus and were most commonly associated with persistent and relapsing infections (1, 16, 29, 35). SCVs of other genera isolated from clinical materials have been reported only rarely and were not well characterized (15, 25, 31). The largest number of studies concerning SCVs are available for S. aureus.

Some characteristic features that are often but not always found in SCVs have been described: (i) auxotrophy (e.g., for hemin, menadione, or thiamine) (1, 28, 33, 38), (ii) decreased respiration (20, 21), (iii) resistance to aminoglycoside antibiotics (3, 27, 36, 38), (iv) limited fermentation of sugars (10, 20, 21, 39), and (v) the ability to revert to normal growth (1, 10, 21, 27). There are no data available about the genetic events responsible for the reappearance of larger colonies. Therefore, in the context of SCV the term "revertants" refers only to changes in the size of the colonies and does not imply actual genetic reversion.

In vitro and in vivo exposure to aminoglycosides at or above the MIC results in selection of SCVs in many genera (1, 18, 26, 27). Slow growth, atypical colony morphology, altered biochemical profile, and morphological instability may lead to failure in identification of SCV strains in clinical laboratories. Inactivation of the heme biosynthetic pathway (Hem) is one possible cause for the appearance of SCVs (7, 21, 24, 34, 38). Hemes are key components of the electron transfer apparatus and the prosthetic groups in different enzymes. The enzymes for heme biosynthesis and their encoding genes (hemA to hemH and hemL) are known in Escherichia coli (6). Mutation in one of these genes resulted in cytochrome oxidase-, catalase-, and nitrate reductase-negative strains which grew slowly under aerobic conditions (21). Most E. coli strains are unable to take up heme (6, 32).

In the present report we characterize a SCV of E. coli isolated from a relapsing prosthetic hip infection. This strain carries a mutation in the heme biosynthetic pathway and accordingly shows multiple phenotypic changes. To our knowledge, this is the first description of a SCV E. coli strain as the etiological agent of a chronic infection.

	CASE REPORT

Top Abstract Introduction Case Report Materials & Methods Results Discussion References

A 62-year-old woman who had undergone two previous arthroplasties of her left hip presented at a surgical outpatient department in May 1997 with a small, red, swollen abscess which had developed over a period of 5 weeks in the scar area of her left hip after implantation of a third endoprosthesis in October 1995. There were no signs of systemic disease. The axillar temperature was 36.9°C. A blood count showed slight leukopenia (3.9 × 10⁹/liter). Upon sonographic examination of her hip, the abscess ruptured and a purulent bloody fluid was spontaneously discharged. On microbiological analysis of the abscess material, a slow-growing gram-negative bacterium was isolated in pure culture (the first isolation of Z-2376). Radiologically, a fistula was shown connecting the abscess with the bone cavity of the implanted prosthesis. After admission, the fistula was surgically revised and drained, since the patient refused implantation of a new prosthesis. Pathological examination of the excised fistula showed a chronic granulomatous and fibrosing inflammation. Two weeks later, the patient was discharged home after clinical improvement following empirical intravenous treatment with cefuroxime, which was later changed to oral ofloxacin. In the following 4 weeks, a new abscess developed, which had to be opened and drained surgically after rehospitalization. Intraoperatively obtained swab material again produced slow-growing gram-negative organisms (the second isolation of Z-2376). After 2 weeks, the patient was discharged with only a slightly secreting fistula, which was treated by draining the wound repeatedly on an outpatient basis. Antibiotic treatment with ofloxacin was discontinued after 6 weeks. Four months later, the fistula showed no signs of infection. Swabs were taken from the fistula, yielding no bacterial growth, and a serum specimen was obtained. The patient's past medical history included two previous implantations of total endoprostheses in 1984 and 1994. Five days after the second implantation, the prosthesis had to be removed due to an infection with Proteus mirabilis diagnosed in a peripheral laboratory. Reimplantation at that time was refused and postponed until October 1995, when the currently infected third prosthesis was implanted. It is noteworthy that gentamicin was not used as pre- or postsurgical prophylaxis in any of the three hip arthroplasties.

	MATERIALS AND METHODS

Top Abstract Introduction Case Report Materials & Methods Results Discussion References

Bacteria and culture conditions. E. coli Z-2376 was obtained from two different specimens of the scar area of a patient with a prosthetic hip infection. E. coli DH5 was purchased from Gibco-BRL (Eggenstein, Germany). Unless otherwise stated, bacteria were cultured on Trypticase soy agar (TSA; Oxoid, Unipath Ltd., Basingstoke, England) supplemented with 7% defibrinated sheep blood (Oxoid, Unipath Ltd.) for 48 h at 37°C under aerobic conditions. The sizes of the colonies were determined by plate microscopy. Auxotrophy for hemin or -aminolevulinic acid (-ALA) was tested by incubating bacteria on hemin- or -ALA-supplemented TSA for 48 h at 37°C under aerobic conditions. Hemin (Sigma-Aldrich Chemie GmbH, Deisenhofen, Germany) dissolved in 10% Tween 80 was used at a final concentration of 20 µg/ml, and -ALA (Sigma-Aldrich) was used at a final concentration of 50 µg/ml. In addition, hemin permeability was tested on McConkey agar (Merck, Darmstadt, Germany) at 37°C under aerobic conditions by adding X-factor discs (Oxoid, Unipath Ltd.).

Biochemical reactions and motility testing. The tests for catalase, nitrate reductase, and motility were performed by standard procedures. For the catalase reaction, a colony was dipped on a glass slide and overlaid with 3% H₂O₂. The appearance of bubbles indicates a positive test result. Motility was tested in a hanging-drop chamber and in motility-soft agar. For biochemical characterization, bacteria were grown in brain heart infusion (BHI; Oxoid, Unipath Ltd.) with or without 20 µg of hemin/ml for 24 h. Bacteria were harvested by centrifugation and identified with an API 20 E system (bio-Merieux Ltd., Marcy l'Etoile, France). The benzidine test for detection of peroxidase activity of heme proteins and heme cytochromes was performed with benzidine hydrochloride (Sigma-Aldrich) and H₂O₂ as described previously (12).

Determination of the reversion frequency of the SCV. To determine the reversion frequency, single small colonies grown on blood agar plates for 48 h were resuspended in phosphate-buffered saline (PBS). Serial dilutions of these suspensions were plated on blood agar plates and incubated at 37°C. After 48 h small and large colonies were counted. The test was performed in duplicate and repeated three times.

Gentamicin resistance. The MIC of gentamicin was determined for small and large colonies with E-test strips (VIVA Diagnostika GmbH, Cologne, Germany) on blood agar plates (8). McFarland standard 0.5 inoculum was used. The plates were incubated at 37°C under aerobic conditions for 48 h.

Antibody detection. Cell lysates from bacteria grown under different conditions were separated by sodium dodecyl sulfate-11% polyacrylamide gel electrophoresis, transferred to nitrocellulose sheets (BA85; Schleicher and Schüll, Inc.), and incubated with sera diluted in PBS buffer. The serum of the patient was used at a dilution of 1:500. Pooled sera from 10 healthy volunteers diluted 1:100 served as a control. Subsequently, the binding of immunoglobulin G was visualized by anti-immunoglobulin G-alkaline phosphatase conjugate (Sigma-Aldrich) as described elsewhere (30).

16S rRNA gene sequence analysis. Amplification and direct sequencing of the gene encoding 16S rRNA was done as described previously (13). Universal primers corresponding to the E. coli rRNA gene from bp 8 to 28 and bp 1542 to 1522 were used for PCR amplification. Amplicon contamination controls were performed in parallel. The hypervariable regions V1 and V2 were sequenced with a primer corresponding to bp 361 to 341. For solid-phase DNA sequencing one of the oligonucleotides was biotinylated at the 5' end. Dynabeads (DYNAL GmbH, Hamburg, Germany) were used for the preparation of single-stranded DNA as recommended by the manufacturer. Sequence data were compared with the EMBL GenBank (HUSAR-DKFZ, Heidelberg, Germany).

PCR amplification of hemB and hemD and generation of digoxigenin-labelled DNA probes. For amplification of the E. coli hemB gene (GenBank accession no. L44595), we used four primers corresponding to the indicated base pairs of the hemB gene: hemB1, 5'-GGCAGACCATGACAGACTTAAT-3' (bp 8 to 14); hemB2, 5'-ACCTGCAGCAGCTGCAACCA-3' (bp 565 to 546); hemB3, 5'-ACGGCCAGGTACAGGCGATT-3' (bp 597 to 616); and hemB4, 5'-CGCAGAATCTTCTTCTCAGCCA-3' (bp 1062 to 1041). For amplification of the E. coli hemD gene (GenBank accession no. X12614), primer hemD1, 5'-CCGCTGGAGAAGAGTTAGTGA-3', corresponding to bp 38 to 59, and primer hemD2, 5'-GACGACCAATAGTCGACAGTG-3', corresponding to bp 642 to 621 of the gene, were used. The oligonucleotides were synthesized by Roth (Karlsruhe, Germany). PCR buffer, Taq DNA polymerase, and the model 2400 DNA thermal cycler were obtained from Perkin-Elmer Cetus (Foster City, Calif.); deoxynucleoside triphosphates were purchased from Pharmacia LKB (Uppsala, Sweden). DNAs from different E. coli strains grown for 24 h in BHI were released by repeated freezing and thawing and used as templates. The PCR reactions were performed for 30 cycles with a profile of 94°C for 10 s, 58°C for 60 s, and 72°C for 180 s. The PCR products were analyzed on a 1.5% ethidium bromide-stained agarose gel.

Preparation of genomic DNA, Southern blotting, and filter hybridization. Preparation of genomic DNA, Southern blotting, and filter hybridization were done by standard procedures as described previously (2). Small- and large-colony types of Z-2376 and E. coli DH5 were each grown overnight in 10 ml of BHI. Genomic DNAs were extracted with proteinase K (Boehringer, Mannheim, Germany) and phenol (Roth). DNA (5 µg) from each strain was digested with EcoRI (Boehringer). In duplicates, the DNA fragments were separated on 0.7% agarose gels and transferred to nylon membranes (Zeta probe; Bio-Rad, Richmond, Calif.) by vacuum blotting (Vacu gene; Pharmacia). One nylon membrane was hybridized against the hemB probe, and the other was hybridized against the hemD probe at 68°C. Detection of hybrid DNA was carried out by an enzyme immunoassay with anti-digoxigenin antibodies conjugated to alkaline phosphatase (Boehringer) according to the manufacturer's protocol.

	RESULTS

Top Abstract Introduction Case Report Materials & Methods Results Discussion References

Isolation and identification of a SCV of E. coli. Two specimens from the wound area of a patient suffering from chronic prosthetic hip infection were sent to our laboratory. From abscess material (first specimen) and a swab taken intraoperatively 2 months later (second specimen) we isolated a slow-growing gram-negative bacterium in pure culture. All tests performed for growth behavior and differentiation showed the same results for both isolates, and the bacterium was denoted with the laboratory accession number, Z-2376. On blood agar plates the diameters of the colonies of Z-2376 measured less than 0.1 mm after 24 h of aerobic incubation. After 48 h the colony size increased to about 0.2 mm, and some larger colonies 2 mm in diameter appeared. After 48 h of anaerobic incubation both colony types grew with identical morphologies as normal-sized colonies 1.2 mm in diameter. Z-2376 was cytochrome c oxidase negative and glucose fermentation positive. Repeated subcultures on different agar plates revealed that the larger colonies were revertants of the small ones and, remarkably, the larger colonies were only present on culture media containing blood. The biochemical characterization of the small colonies and the revertants with the API system did not facilitate an acceptable identification (low discrimination between E. coli and Enterobacter spp.). Sequencing of 16S rRNA was performed with the two colony types of both isolates of Z-2376, and in all four cases it showed identical sequences matching 100% with the 16S rRNA gene of E. coli. Thus, Z-2376 represented E. coli growing as a SCV.

Instability of the SCV. To determine the reversion frequency of the SCV, we resuspended 48-h-old small colonies in PBS and plated serial dilutions on blood agar plates. After 48 h of incubation large and small colonies were counted. The reversion frequency was 2 × 10³ (standard deviation, ±0.5).

Characterization of E. coli Z-2376 as a Hem mutant. Normal-sized colonies after anaerobic incubation of SCVs resembled respiration-deficient (Res) bacteria (21). Therefore, we tested the small colonies and the revertants for other characteristics of Res mutants, such as (i) aminoglycoside resistance, (ii) negative benzidine reaction, (iii) negative catalase reaction, and (iv) negative nitrate reductase reaction. The small colonies were resistant to gentamicin (MIC, 16 µg/ml), and the reactions for benzidine, catalase, and nitrate reductase were all negative. In contrast, the revertants were susceptible to gentamicin (MIC, 1.2 µg/ml), and the benzidine, catalase, and nitrate reductase reactions were all positive.

View larger version (83K): [in this window] [in a new window]   FIG. 1.   After 24 h of incubation at 37°C the large-colony type of E. coli Z-2376 showed hemin-dependent growth on McConkey agar only surrounding an X-factor disc (X DD; 5 µg of hemin).

Detection of a mutated hemB gene in E. coli Z-2376. Lewis et al. reported that in Res E. coli strains the hemB gene is a hot spot for spontaneous mutations (21). We first tested the integrity of the hemB gene in both colony types of Z-2376 by PCR amplification of (i) the 5' half of the hemB gene (primers hemB1 and -2), (ii) the 3' half of the hemB gene (primers hemB3 and -4), and (iii) a part of the hemD gene (primers hemD1 and -2) as a control. All three PCRs resulted in PCR products of the correct size for E. coli DH5. The small-colony type and the revertant of Z-2376, however, showed correct PCR products only for the hemD PCR. The PCRs for the hemB gene resulted in nonspecific or no amplification (data not shown). For further characterization we generated two digoxigenin-labelled DNA probes, one for the hemB gene of E. coli and the other for the hemD gene of E. coli. Both probes were hybridized against identical nylon membranes carrying immobilized genomic DNAs of both colony types of Z-2376 and E. coli DH5. The Southern blot is shown in Fig. 2. In contrast to the hemD gene, which was present in all three strains, hemB-homologous sequences could not be detected in either colony type of Z-2376. This result demonstrates that the hemB gene in Z-2376 is inactivated by a deletion of part of the gene or the whole gene.

View larger version (88K): [in this window] [in a new window]   FIG. 2.   Southern hybridization. Two identical nylon membranes carrying immobilized, EcoRI-digested, genomic DNA were hybridized with the hemB probe (A) and the hemD probe (B). Lanes 1, DH5; lanes 2: small-colony type of Z-2376; lanes 3, large-colony type of Z-2376; lanes M, marker (HindIII-digested  DNA). The localization of the different fragments at 23.7, 9.46, 6.66, 4.2, 2.25, and 1.96 kb (top to bottom) are indicated by lines.

Demonstration of E. coli-specific antibodies. Persistent infections should result in the production of specific antibodies against the bacterial agents. We tested the serum of the patient by immunoblotting for the presence of antibodies against Z-2376. To induce different antigens, small and large colonies of Z-2376 were incubated under aerobic and anaerobic conditions on TSA-blood agar plates and whole-cell lysates for immunoblotting were prepared. Pooled sera from healthy volunteers served as a negative control. A strong antibody response against Z-2376 could be detected in the serum of the patient but not in the control sera (Fig. 3).

View larger version (67K): [in this window] [in a new window]   FIG. 3.   Immunoblotting. Cell lysates of E. coli DH5 and the two colony types of Z-2376 grown under aerobic or anaerobic conditions on TSA-blood agar plates were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, transferred to nitrocellulose sheets, and tested against antisera. (A) Antiserum from the patient diluted 1:500; (B) pooled antisera from 10 healthy volunteers diluted 1:100. Lanes 1, DH5; lanes 2, large-colony type of Z-2376 grown aerobically; lanes 3, small-colony type of Z-2376 grown aerobically; lanes 4, large-colony type of Z-2376 grown anaerobically; lanes 5, small-colony type of Z-2376 grown anaerobically; lanes 6, prestained low-molecular-weight marker.

	DISCUSSION

Top Abstract Introduction Case Report Materials & Methods Results Discussion References

Persistent bacterial infections are typically due to slow-growing microorganisms, e.g., Mycobacteria, Borrelia, or Helicobacter spp., whereas fast-growing bacteria like S. aureus or E. coli most often cause acute, purulent infections. Nevertheless, chronic infections have also been reported for S. aureus which could often be correlated with the appearance of SCVs (1, 16, 29, 35). Mutations in biosynthetic pathways resulting in a decreased growth rate of the pathogen obviously could influence the course of infections. This phenomenon, described for S. aureus, might occur in other bacteria as well.

In the present study, we describe a SCV of E. coli as the etiological agent of a recurrent prosthetic hip infection. Our patient had a history of three prosthetic hip replacements in the last 15 years, with one episode of acute prosthetic hip infection 3 years ago. From two different specimens of the infected hip taken at an interval of 2 months we isolated a slow-growing gram-negative bacterium in pure culture that was identified as E. coli by 16S rRNA sequencing. As a consequence of the chronic infection, the patient had developed large amounts of antibodies against the E. coli strain, as demonstrated by immunoblotting. This immune response, however, was obviously not effective in eliminating the pathogen.

The growth characteristics of E. coli Z-2376 under aerobic and anaerobic conditions as well as its gentamicin resistance resembled those of a SCV. Further characterization of the clinical isolate demonstrated the phenotypic alterations of Hem E. coli strains. Heme is a prosthetic group of respiratory cytochromes and several other enzymes involved in energy metabolism and oxidative catalysis (6). Mutations in the biosynthesis of heme result in cytochrome oxidase-negative, catalase- and peroxidase-negative, and nitrate reductase-negative strains. As a consequence, these strains are slow growing under aerobic conditions and appear as SCVs (7, 21, 24, 32).

Lewis et al. demonstrated by in vitro selection for respiratory-deficient E. coli strains that 80% of these mutants carried the defect in the hemB gene (21). Deletions or insertions of the transposable element IS2 were responsible for the inactivation (19). Using PCR and Southern hybridization, we tested the integrity of the hemB gene in Z-2376. Both techniques were unable to detect hemB-homologous sequences in Z-2376, demonstrating that a deletion of part or all of the hemB gene is responsible for the Hem phenotype of Z-2376. hemB mutants do not respond to -ALA, the precursor of heme biosynthesis. Likewise, the growth of E. coli Z-2376 could not be stimulated with -ALA.

It is well known that E. coli is usually impermeable for hemin and that Hem E. coli strains cannot utilize supplemented hemin (6, 34). In Hem E. coli strains, however, revertants have been recognized that respond to hemin supplementation and grow as normal-sized colonies on hemin-containing agar plates (7, 23). E. coli Z-2376 showed exactly the same phenomenon. With a remarkably constant frequency of 2 × 10³, hemin-permeable revertants grew as large colonies on blood- or hemin-containing agar plates. As shown by PCR and Southern hybridization, these revertants carried the same deletion in the hemB gene as did the small colonies of Z-2376.

Based on the data presented above, there is evidence for at least two independent mutations in E. coli Z-2376: (i) the first mutation is located in the heme biosynthesis genes (deletion of hemB), resulting in slow growth and the appearance of SCVs; (ii) the second event occurs in the SCVs with a frequency of 2 × 10³ and results in hemin permeability of the SCV. Correlation of the two different mutations with the course of the infection is hypothetical; the first mutation slowed the growth rate of the bacterium, resulting in SCVs and persistance, while the second mutation accelerated the growth rate of the SCVs and resulted in reactivation of the infection.

Three years ago, the patient suffered from an infection of her prosthetic hip following immediately after the second implantation. A sole microorganism, P. mirabilis, had been isolated. After explantation, the scar area was rinsed with a disinfectant (Lavasept) and the patient was treated with cefuroxime. P. mirabilis is a rare and unusual cause of wound infection after surgical intervention, particularly in joints. However, it might be that P. mirabilis was not the only gram-negative rod that gained access to the wound area. The presence of concomitant bacteria, such as E. coli, could have been masked in the diagnostic laboratory by the swarming motility of Proteus spp. On the other hand, we cannot exclude the possibility that E. coli might have infected the joint on a different occasion or from the bloodstream as a result of bacteremia.

Antibiotic treatment with aminoglycosides, which are known to select SCVs, was not reported in the patient's history as pre- or postsurgical prophylaxis in any of the three hip arthroplasties. Therefore, other, yet-unknown factors might have been involved in generation of the SCVs.

How should infections with the E. coli SCV be treated with antibiotics? In general, standardized antimicrobial susceptibility testing and correlation between such testing and successful treatment are not known for infections with the E. coli SCV. However, the presumed pathogenesis of infections with SCVs should be considered in treatment. SCVs are thought to persist intracellularly in a quiescent metabolic state inside the host (29). For eradication of persistent infections only bactericidal antibiotics are promising. Eng et al. (14) demonstrated that fluoroquinolones exhibit the highest activity against slow-growing and nongrowing bacteria. Moreover, fluoroquinolones show high intracellular activity (4, 11). The patient described in this study was treated with a fluoroquinolone derivate for 6 weeks. Six months later the fistula still persisted. Swabs taken from deep within the fistula yielded no bacterial growth. However, it remains questionable whether the patient had overcome the infection with the SCV of E. coli.

In routine laboratory testing, the presence of SCV in clinical specimens is difficult to verify. Colony formation by SCVs can often be detected only after more than 48 h of culture during primary isolation. Moreover, atypical growth behavior and unusual biochemical reactions (in this case, no growth on McConkey agar and negative catalase, nitrate reductase, and indole reactions, respectively) may result in misidentification of the microorganism. Therefore, it is most important to take SCVs into account as a possible cause of persistent infectious diseases, particularly when no bacteria or unusual bacteria are found in materials obtained in such cases.