Solitary Erythema Migrans in Georgia and South Carolina

Objective  To evaluate the incidence of Borrelia burgdorferi infection in humans with erythema migrans (EM) in 2 southeastern states.

Design  Prospective case series.

Setting  Family medicine practice at academic center.

Patients  Twenty-three patients with solitary EM lesions meeting Centers for Disease Control and Prevention (CDC) criteria for Lyme disease.

Interventions  Patients underwent clinical and serologic evaluation for evidence of B burgdorferi infection. All lesions underwent photography, biopsy, culture and histopathologic and polymerase chain reaction analysis for B burgdorferi infection. Patients were treated with doxycycline hyclate and followed up clinically and serologically.

Main Outcome Measures  Disappearance of EM lesions and associated clinical symptoms in response to antibiotic therapy; short-term and follow-up serologic assays for diagnostic antibody; growth of spirochetes from tissue biopsy specimens in Barbour-Stoenner-Kelly II media; special histopathologic stains of tissue for spirochetes; and polymerase chain reaction assays of tissue biopsy specimens for established DNA sequences of B burgdorferi.

Results  The EM lesions ranged from 5 to 20 cm (average, 9.6 cm). Five patients (22%) had mild systemic symptoms. All lesions and associated symptoms resolved with antibiotic therapy. Overall, 7 patients (30%) had some evidence of B burgdorferi infection. Cultures from 1 patient (4%) yielded spirochetes, characterized as Borrelia garinii, a European strain not known to occur in the United States; 3 patients (13%) demonstrated spirochetallike forms on special histologic stains; 5 patients (22%) had positive polymerase chain reaction findings with primers for flagellin DNA sequences; and 2 patients (9%) were seropositive for B burgdorferi infection using recommended 2-step CDC methods. No late clinical sequelae were observed after treatment.

Conclusions  The EM lesions we observed are consistent with early Lyme disease occurring elsewhere, but laboratory confirmation of B burgdorferi infection is lacking in at least 16 cases (70%) analyzed using available methods. Genetically variable strains of B burgdorferi, alternative Borrelia species, or novel, uncharacterized infectious agents may account for most of the observed EM lesions.

LYME DISEASE (LD) is a complex, multiple-system illness of worldwide importance.^1-5 Cases have been reported on nearly every continent, and from 48 of the 50 United States, but to date, enzootic cycles of Borrelia burgdorferi have been identified only in some temperate zones of North America, Europe, and Asia. Since LD surveillance by the Centers for Disease Control and Prevention (CDC), Atlanta, Ga, began in 1982, more than 101,000 cases have been reported in the United States alone,⁶ triggering comprehensive clinical reviews,⁷ collaborative research,⁸ and public concern.^9,10 However, the precise geographic distribution of LD in the United States remains uncertain and controversial.

In the southeastern United States, reported cases of LD have fluctuated widely in numbers and in focal occurrence patterns.^11-13 Because of lack of definitive evidence, epidemiological interpretation of LD reports has been problematic, and clinical recommendations have been tentative, as in other regions of the United States.^14-17 Although the presence of B burgdorferi in human specimens has been confirmed by culture findings in published reports from at least 6 states,^18-24 there has been no similar confirmation of B burgdorferi infection in patients with tick bite–associated erythema in the southeastern United States. This article reports a prospective clinicopathologic study of 23 patients with tick bite–associated rash consistent with erythema migrans (EM), the best single clinical marker for LD as defined by CDC national surveillance case definition criteria.²⁵

Patients were enrolled from June 4, 1991, through June 7, 1994, based on the following clinical criteria: an expanding erythematous rash exceeding 5 cm, physician diagnosis of EM, and tick bite or tick exposure within the preceding month. Patients were excluded if the rash had resolved by time of enrollment, if antibiotics had been administered in the preceding 2 weeks, or if pregnant, nursing, or younger than 3 years. Enrollment efforts included recruitment letters to practicing physicians within a 200-mile radius of the Medical College of Georgia (MCG), Augusta; posters for display in emergency departments and outdoor (ie, hunting, fishing, camping) establishments; and media announcements in local newspapers in Georgia and South Carolina. All patients were seen at MCG by 3 of us (M.W.F., F.W.C., and D.W.R.). The protocol was reviewed and approved by the MCG Human Assurance Committee.

After informed consent, EM lesions were photographed. Under sterile technique, after anesthesia was induced with 1% lidocaine hydrochloride, two 4-mm punch biopsy specimens of these lesions were obtained, one specimen from the expanding erythematous edge and the other from the central area of each lesion. Each biopsy specimen was bisected under sterile conditions. One portion was placed in Barbour-Stoenner-Kelly (BSK) II medium for microbiologic culture, and the other in 10% buffered formalin for histopathologic evaluation and polymerase chain reaction (PCR) assay. Complete blood cell count; urinalysis; sequential multiple analysis for sodium, potassium, carbon dioxide, chloride, glucose, serum urea nitrogen, uric acid, calcium, phosphorus, total protein, albumin, cholesterol, total bilirubin, alkaline, phosphatase, aspartate aminotransferase, and lactate dehydrogenase (SMA-18); erythrocyte sedimentation rate; and LD serologic evaluation also were performed.

All patients were treated by standard protocol with doxycycline hyclate (Vibra-Tabs; Pfizer, Inc, New York, NY), 100 mg twice a day for 21 days.²⁶ Each patient returned for follow-up photography in 7 to 10 days and follow-up serologic assessment 4 to 6 weeks after enrollment. Aliquots of all unused serum samples were stored at −70°C. Telephone or personal contact was maintained with each patient for 3 to 36 months to monitor side effects or sequelae.

One portion of each bisected biopsy specimen was fixed in 10% buffered formalin for 24 hours or less. Fixed tissues routinely were processed into paraffin, sectioned at 5 µm, and stained using hematoxylin and eosin and the Steiner silver impregnation procedure.²⁷ At least 4 replicate sections of each biopsy specimen were stained using each procedure. Sections of tissue processed in a similar manner and containing known spirochetes (Treponema pallidum) were included as positive staining controls.

Skin biopsy specimens were bisected under sterile conditions, and 1 portion was cultured in BSK II medium^28,29 using a modification of previously reported techniques.³⁰ All cultures were incubated at MCG in 5% carbon dioxide at 33°C and examined by one of us (F.W.C.) for spirochetes by darkfield microscopy twice weekly for 3 weeks and, if spirochetes were not demonstrated, weekly thereafter for 4 additional weeks.

The DNA was extracted from formalin-fixed, paraffin-embedded skin biopsy specimens (4 sections from each block) using a procedure modified from previously published techniques.^31,32 Polymerase chain reaction analysis was used to detect known DNA target sequences found in the B burgdorferi B-31 reference strain. One set of primers (supplied by Barbara Johnson, PhD, CDC, Fort Collins, Colo) was used to amplify a 610-base pair (bp) sequence found in the flagellin (fla) gene.³³ An aliquot of the 610-bp product of the first amplification was then amplified using nested fla primers (Table 1) to yield a 390-bp product. A second set of primers (designated 788/944) was used to amplify a 158-bp sequence found in the outer surface protein A (OspA) gene.³⁴ For each PCR assay, 5 ng of pure genomic DNA from B burgdorferi B-31 strain (supplied by Patricia Rosa, PhD, National Institutes of Health, Hamilton, Mont) was used as a positive control, and sterile distilled water was used as a negative control. Amplification of the target sequences was performed in a thermal cycler (Perkin-Elmer Cetus, Norwalk, Conn) (Table 1). The PCR-amplified products and appropriate size markers were electrophoresed in 2% agarose gels and stained with ethidium bromide. Photographs of the UV transilluminated gels were made for permanent documentation using Polaroid film (Polaroid Corp, Cambridge, Mass).

Serologic analysis was first performed on initial and follow-up serum samples by SmithKline Beecham Clinical Laboratories (SKBL), Norristown, Pa, using Western immunoblotting (SKBL-Wb). Frozen serum samples were thawed and retested in 1996 with a 2-step analysis³⁵ using a CDC enhanced fla enzyme immunoassay (CDC-EIA) followed by Western immunoblotting (CDC-Wb). A commercial Western immunoblot (MarDx Inc, Carlsbad, Calif) was used for CDC-Wb, and the manufacturer's protocol was followed. The reference strain of B burgdorferi used in immunoblots at both laboratories was low-passaged B-31. Serum samples from patient 14 were lost in storage, leaving samples from 22 patients for serologic analysis. Samples from 30 healthy blood donors in Georgia and South Carolina were collected and analyzed as control samples.

Twenty-three patients were enrolled during the 3-year study. Fifteen patients (65%) were referred by a physician; 8 (35%) responded to media announcements. Fourteen patients (61%) were men. Ages ranged from 25 to 69 years, with a median of 47 years. Twenty-twopatients were white; patient 22 was black. Tick bite was reported in 21 patients (91%), from 1 to 30 days before study enrollment (median, 11 days).

Patients 12 and 20 had extensive tick exposure but no memory of tick bite at the site of EM. All known bites occurred within a 200-mile radius of MCG, except in patient 10, a medical student who was bitten while camping in southern Virginia and had returned to his home in Georgia before the appearance of EM. All patients with known tick bites reported exposure during activities in wooded areas. Unfortunately, ticks were not retrieved by patients for definitive identification. Verbal reports by patients 21 and 22 were suggestive of adult female Amblyomma americanum. Cases occurred from April to October, with peak incidence in May (Figure 1). Patient 12 warrants special description. An enlarging erythematous lesion developed over the left shoulder of this 36-year-old male research acarologist, with an eccentric punctum and final diameter of 20 cm. Although he could not recall a recent tick bite at the punctum site, the patient had removed at least 3 embedded ticks from body sites remote from the lesion while performing field research studies in Georgia and South Carolina during the previous few months. Each specimen was identified as Ixodes scapularis. During the month before presenting with the EM lesion, he vacationed in Holland, Sweden, and Denmark, without known tick bite. He was in the United States for 2 weeks before noticing the EM lesion. At the time of enrollment, the patient was asymptomatic.

All patients in our study presented with EM lesions (Figure 2) that met clinical and histologic criteria previously described.¹ Clinical manifestations were limited to solitary, annular, erythematous skin lesions at the tick bite site and mild systemic complaints. Lesions appeared 1 to 30 days after known tick bites and showed gradual expansion. Skin lesions and systemic complaints resolved completely after institution of antibiotic therapy.

Lesions ranged in size from 5 × 4 cm to 20 × 17 cm, with an average final diameter of 9.6 cm (diameter range, 2-20 cm). Lesions had been observed by the patients for 3 to 21 days (average, 9.3 days). Intensity of erythema at enrollment ranged from pale pink to bright crimson. Lesions occurred on the leg (8 patients), back (6 patients), shoulder (4 patients), abdomen (4 patients), and chest (1 patient). Clinical features are summarized in the following tabulation:

Five patients (22%) reported influenzalike symptoms at or just before the time of study enrollment; associated symptoms in all patients are given in the following tabulation:

Secondary or multiple erythematous lesions, neuropathy, arthritis, and bradycardia were absent.

Hematologic screening tests revealed white blood cell counts ranging from 4.5 to 14.3 × 10⁹/L (mean 7.5 × 10⁹/L), normal differential levels, and normal hemoglobin and platelet values. Erythrocyte sedimentation rates (Westergren) ranged from 0 to 41 mm/h, with a mean of rate of 15 mm/h. Results of urinalyses were normal in all cases. The SMA-18 chemistry panels were unremarkable in all patients except for slight elevations of alanine aminotransferase level at 121 U/L (reference range, 0-47 U/L) and aspartate aminotransferase level at 92 U/L (reference range, 0-36 U/L) in patient 4.

The histopathologic findings were similar in all cases and varied only in the degree of inflammation. These changes consisted of mild to moderate lymphohistiocytic infiltrates around small blood vessels in the papillary and reticular dermis (Figure 3, A). Rare plasma cells and mast cells also were present in the perivascular infiltrates in some cases. The epidermis was of normal thickness and unremarkable. The inflammatory response was similar, regardless of whether the biopsy specimen was taken from near the center or at the periphery of the lesion. However, in biopsy specimens taken near the site of tick attachment, a dispersed inflammatory infiltrate that contained hemorrhagic foci and varying numbers of eosinophils and neutrophils, in addition to mononuclear inflammatory cells, was often present.

Spirochetallike forms were demonstrated in replicate skin sections stained with the Steiner silver impregnation procedure in 3 (13%) of 23 patients (patients 1, 3, and 12) (Figure 3, B-D). At least 4 replicate sections per biopsy specimen were examined carefully and repeatedly by one of us (F.W.C.) and by a dermatopathologist at MCG, using a 63× oil immersion objective. In these cases, the blackened spirochetallike forms were very sparse, often distorted and atypical, and located in the dermis between collagen fibers, around dermal blood vessels, and extracellularly within dispersed inflammatory infiltrates. The length, width, and loose coiling of spirochetallike forms were morphologically suggestive of a Borrelia species or a similar organism, and these forms were more often demonstrated in biopsy specimens taken at or near the site of tick bite.

Results of repeated PCR assays of multiple skin sections (4 from each biopsy specimen) for B burgdorferi DNA sequences were positive in 5 (22%) of the patients (Figure 4). In these 5 patients (patients 2, 3, 4, 5, and 12), specific DNA sequences were amplified and detected only after several attempts (2-9 assays per biopsy specimen; Table 2). The fla primers amplified target DNA sequences in these 5 patients; assays using the OspA primers 788/944 consistently yielded negative findings, as they did for the spirochetal isolate from patient 12. Complete PCR results for each case are summarized in Table 2. Spirochetal forms were directly demonstrated in replicate, Steiner-stained skin sections from 2 (patients 3 and 12) of the 5 patients with positive PCR findings, and spirochetes were isolated in culture from patient 12. Only 3 of 9 PCR assays using fla primers yielded positive findings in patient 12, despite positive findings of Steiner stain (Figure 3, D) and serologic and culture assessments.

One (4%) of 23 EM biopsy specimens yielded positive culture results. In patient 12, spirochetes were isolated on day 13 of incubation in BSK II medium from skin biopsy specimens obtained from the erythematous lesion edge. The spirochetal isolate from patient 12 (designated HK-1) demonstrated morphology consistent with B burgdorferi on darkfield microscopy. Immunofluorescence assay reactivities to monoclonal antibodies against B burgdorferi were demonstrated to H5332/OspA but not to H3TS/OspA, H5TS/OspB, or H6831/OspB (Robert S. Lane, PhD, and J.H.O., written communication, September 1998). The OspA gene was not amplified by PCR using primers 788/944. Subsequent analysis in 1997 of the highly conserved 330-bp product of the OspA gene and pulsed field gel electrophoresis of the isolate (David H. Persing, MD, PhD, written communication, October 1997) demonstrated a close phylogenetic relationship between HK-1 and IP-90, a tick isolate from Russia recognized as Borrelia garinii, not known to exist in the United States.³⁶

The serologic reactivities in 22 of the 23 patients are listed in Table 2. Numerous immunoblot reactive bands were documented using SKBL-Wb, but only in patient 12 were criteria met for serodiagnosis as defined by Dressler et al³⁷ in 1993. In 1995, a 2-tier test algorithm was recommended by expert consensus for use in the United States.³⁸ Using this algorithm, serum samples from patients 12 and 21 were positive for B burgdorferi when tested at CDC. Positive immunoblots for both were limited to acute sample IgM reactivity against the 41-kd and OspC antigens. An additional 7 patients (patients 11, 13, 15, 18, 19, 20, and 23) displayed equivocal results of CDC-EIAs, but these findings were negative on CDC-Wb. Hence, a total of 2 patients (9%) were seropositive by the 2-tier CDC method. Requisite IgG bands were not detected by SKBL-Wb or CDC-Wb in any patient. Of the control serum samples from healthy blood donors, results in 4 of 30 were equivocal on CDC-EIA, and 1 (3%) of these yielded positive findings on IgM CDC-Wb to the 41-kd and OspC antigens. Findings for all control serum samples were negative on CDC-Wb for IgG. Comparison of case-patient seropositivity with that of control samples revealed no statistically significant difference (Fisher exact test, P = .57).

In summary, a total of 7 (30%) of 23 patients had evidence of B burgdorferi infection. Patient 12 had positive results on culture, silver staining, PCR analysis, and serologic assessment; patient 3, silver staining and PCR analysis; patient 1, silver staining; patients 2, 4, and 5, PCR analysis; and patient 21, serologic assessment.

Initial clinical follow-up for all 23 patients occurred within 5 to 10 days. Patients 1 and 8 reported transient fever (>38°C) and arthralgias within 48 hours of initiation of antibiotic therapy. The EM lesions resolved within 7 days in 22 patients. However, patient 22, the only black patient, had persistent, 16 × 6-cm pruritic erythema and purpura for more than 14 days despite compliance to doxycycline therapy. All 23 patients successfully completed 21 days of antibiotic therapy. No known sequelae suggestive of ongoing or disseminated LD have occurred in any patient during follow-up evaluation.

This series of 23 patients with EM is, to our knowledge, the first reported from Georgia and South Carolina. Although clinically consistent with early LD, these EM lesions were not associated with currently defined laboratory evidence of B burgdorferi infection in most (70%) of the patients. Only 1 patient had unequivocal cultural evidence of infection. Overall, 5 patients (22%) had positive PCR findings; 3 (13%), positive silver staining of histologic sections; 2 (9%), positive serologic findings; and 1 (4%), positive culture yield. The culture-positive patient had positive findings on all 4 laboratory tests; 1 patient, on 2 tests; and 5 patients, on single tests. These data imply that EM in our region may represent B burgdorferi infection in some cases, but is of an origin yet to be determined in most patients. The relative lack of laboratory confirmation contrasts sharply with results from larger series of EM cases characterized in areas of highly established endemicity for B burgdorferi infection.^39,40

The culture-positive patient (patient 12) was exposed in Georgia and South Carolina to I scapularis, the vector of LD in the eastern United States and a species from which B burgdorferi has been isolated in this region.⁴¹ Genomic analysis of isolate HK-1 from patient 12, however, was consistent with B garinii of European origin.^36,42 Further characterization of this isolate using additional OspA primers more recently available to us has revealed DNA sequences in HK-1 comparable to those found in isolates of B burgdorferi sensu lato (various genospecies other than the typical B-31 strain) recently obtained from ticks in Georgia, South Carolina, and Missouri (J.H.O. and F.W.C., unpublished data, October 1998). Hence, the molecular and epidemiologic data on this human isolate are not fully convergent, and it is uncertain whether this patient acquired his infection in the United States or Europe.

Borrelia burgdorferi sensu stricto recovered in the northeastern and north central regions of the United States has been the only reported human isolate and is genetically more homogeneous than B burgdorferi sensu lato isolated from ticks and mammals in the southern and western United States.^36,43,44 Although published studies^23,39 demonstrate that B burgdorferi sensu stricto is cultivable from biopsy specimens of up to 86% of EM cases acquired in northern, endemic regions, this laboratory method is not widely practiced or available. However, failure to culture B burgdorferi in BSK II medium from EM cases in the south does not prove its absence. A recent study revealed great diversity of genotypes among B burgdorferi sensu stricto circulating in an enzootic cycle in Colorado, and that diversity was greatly reduced in cultured spirochetes.⁴⁵ Current laboratory methods for B burgdorferi culture in BSK II medium may isolate successfully only a limited number of the genotypes circulating in a population and may have limited our ability to isolate spirochetes from skin biopsy specimens in our series.

FIVE PATIENTS IN our series had skin biopsy specimens that yielded positive findings for specific B burgdorferi fla DNA sequences on PCR analysis. The PCR results were inconsistent, however, in that some assays of tissue samples from these 5 patients had positive findings and some others had negative findings when identical amplification techniques were used (Table 2). Even in the 1 culture-proven case, PCR findings were positive in only 3 of 9 assays (4 tissue sections per assay). These results suggest that the negative PCR findings in some of our specimens may have been false-negative. Sources for false-positivity and false-negativity were considered when interpreting PCR results in our cases. False-positive PCR results may have derived from laboratory contamination or the presence of some other organism with B burgdorferi–conserved fla gene sequences. False-negative PCR results could reflect sampling variability among tissues with sparse spirochetal distribution or the presence of B burgdorferi strains with heterologous OspA sequences.^36,43 Isolation of B burgdorferi from ticks and mammals in the southeastern United States has been documented recently.^41,46 To date, more than 200 isolates have been obtained from ticks, birds, and mammals in this region (J.H.O., unpublished data, August 1998). Most of these isolates had positive PCR findings for the fla primers and 1 or more OspA primers specific for B burgdorferi. Some of the isolates yielded positive findings when 4 different OspA primers thus far tested were used, and some did not. Unfortunately, biopsy specimens from patients in our study were analyzed using the only OspA primer (788 of 944) available at that time. Most of the isolates from nature in this region also yielded positive findings with primers to a conserved chromosomal sequence,⁴⁷ whereas some did not (J.H.O. and F.W.C., unpublished data, August 1998). These findings underscore the genetic diversity among nonhuman isolates from this region, as reported by Mathiesen et al.³⁶ The possibility that patients in our study were infected with B burgdorferi strains bearing variant target DNA sequences, leading to our inability to amplify specific genes using PCR, remains to be demonstrated.

The serologic data on our patients failed to demonstrate the pattern of immunoreactivity typically seen in culture-positive early LD cases.⁴⁸ Based on antigens derived from the B-31 index strain from the northern United States, only 1 sample (patient 12) from Georgia and South Carolina patients had positive CDC-EIA findings. Although patients 12 and 21 had positive CDC immunoblot findings for IgM at the time of presentation, we did not observe broadening of the immune response to include additional recognized B burgdorferi antigens over time, nor did we demonstrate IgM to IgG switching. This pattern of reactivity restricted to IgM seropositivity resembles that seen in the control serum samples tested from our region. It is possible, however, that prompt antibiotic treatment of our patients precluded a serologic response in convalescent samples.

We speculate that the illness in most of our patients may be caused by genetically variant strains of B burgdorferi, as observed in nature in our region,^36,41,46 by an unculturable Borrelia, or by some other tick-borne agent. Campbell et al⁴⁹ described 45 patients from Missouri with an EM-associated illness without confirmatory cultural, molecular, or serologic proof of B burgdorferi infection and concluded that the illness was caused by some agent other than B burgdorferi. However, the conclusions of that study have been controversial.⁵⁰ A series of 14 patients with EM with minor systemic symptoms suspected to be associated with bites of A americanum ticks was recently analyzed at a North Carolina summer camp, with no evidence of B burgdorferi infection by culture or serologic assessment.⁵¹ Furthermore, the presence of an unculturable Borrelia (provisionally called Borrelia lonestari) in A americanum ticks recovered from Missouri, New Jersey, New York, North Carolina, and Texas was recently reported,⁵² in which borreliae were observed in tick midgut contents on immunofluorescence and PCR results. Whether this spirochete is the cause of EM in some of our patients or in others elsewhere in the United States has not yet been determined. Borrelia burgdorferi sensu lato has been isolated recently from ticks in Missouri,⁵³ suggesting that a spectrum of closely related borreliae may be enzootic in the south.

Although 90% of our patients recalled a tick bite at the site of the skin lesion, no tick specimens were available for examination. The predominant species parasitizing humans in the study area is A americanum.⁵⁴ Based on descriptions by 2 of our patients, A americanum may be the vector of the pathogen causing the illness we observed, in contrast to I scapularis⁵⁵ and Ixodes pacificus⁵⁶ transmission of LD in the northeastern and far western United States, respectively. Convincing linkage between bites of A americanum and culture-negative EM lesions has been established in a series of 17 patients undergoing evaluation in Missouri.⁵⁷

Because tick bite–associated EM in Georgia and South Carolina has not been studied previously, its natural history is not fully characterized. In our patients, rapid resolution of lesions and associated symptoms corresponded with doxycycline therapy, suggesting an infectious cause. Until definitive characterization of an infectious agent, we recommend clinical management with doxycycline based on the favorable outcome observed in our patients. Certain noninfectious etiologies of expanding erythemas, such as arthropod assault or granuloma annulare, also need to be considered, but seem unlikely to us.

Recognition of this illness highlights the dilemma that although EM is a characteristic skin lesion of early LD and a widely applied criterion for clinical diagnosis, it may not be pathognomonic for B burgdorferi sensu lato infection. If LD is to be defined by the range of clinical and laboratory manifestations associated with B burgdorferi sensu stricto infection in endemic areas, then EM alone may not be a sufficient clinical criterion for documenting such infection in the southern United States. This observation has important implications for the current national surveillance case definition for LD and generates uncertainties about reported LD cases diagnosed solely on the basis of an expanding erythematous or "bullseye" rash. Efforts need to be continued to cultivate B burgdorferi or other as yet unknown pathogens from tick vectors, mammalian reservoirs, and human tissues in regions that are not endemic for LD, such as the southeastern United States, where cases of EM are documented, disturbing, and awaiting explanation.^58,59

Accepted for publication June 7, 1999.

This publication was supported in part by cooperative agreements U50/CCU 40661 and U50/CCU 410281 from the Centers for Disease Control and Prevention, Atlanta, Ga, and by grant AI 24899 from the National Institute of Allergy and Infectious Diseases, Bethesda, Md.

We thank Lee O. Huey and Byron S. McGuire for performing polymerase chain reaction assays; Elizabeth A. Crans and Leslie Z. Carrow for assistance in manuscript preparation; Phillip M. Jones for preparation of photographs; David H. Persing, MD, PhD, for molecular analysis of the isolate; Robert S. Lane, PhD, for performance of monoclonal assays; and David T. Dennis, MD, for critical comment and support to the work since its inception.

Corresponding author: Michael W. Felz, MD, Department of Family Medicine, Medical College of Georgia, 1467 Harper St, Augusta, GA 30912-3500 (e-mail: mfelz@mail.mcg.edu).