Inoculation of nonhuman primates with the N40 strain of Borrelia burgdorferi leads to a model of lyme neuroborreliosis faithful to the human disease

Lyme borreliosis is a tick-transmitted spirochetal infection with protean clinical manifestations, ^[1] including involvement of the nervous system, ^[2,3] cardiovascular system, skin, ^[4] and joints ^[5]. Neurologic involvement, Lyme neuroborreliosis (LNB), is the most feared of the sequelae of the infection but the one that is least understood from the standpoint of pathogenesis, latency, and therapy.

A better understanding of this infection can be gained from the study of animal models. Inoculation of subprimate animals, such as mice and hamsters, has resulted in systemic infection with the spirochete but an absence of consistent infection and inflammation in the CNS. To test whether a model more faithful to human LNB could be elicited in nonhuman primates (NHPs), we injected rhesus macaques with infectious Borrelia burgdorferi and evaluated the course of the infection.

Methods. Animals. Animals used in this study were housed and cared for in accordance with the Animal Welfare Act and the Guide for the Care and Use of Laboratory Animals (Department of Health and Human Services publication no. 85-23, 1985) in facilities accredited by the American Association for Accreditation of Laboratory Animal Care. Prior to initiation, the study was reviewed and approved by the Institutional Animal Care and Use Committee. Caretakers of the animals, who observed the NHPs daily, were asked to identify any signs of skin lesions, arthritis, or changed behavior. The veterinary staff also observed the animals once a day for any new signs. In addition, at times of anesthesia for CSF and blood collection, the NHPs were examined for skin lesions and arthritis.

The animals in the study consisted of five adult rhesus monkeys (Macaca mulatta) ranging in age from 5 to 12 years; three were female and two were male. Four animals (designated "A," "B," "C," and "D") were untreated prior to inoculation. One female monkey (designated "E") was rendered marginally immunocompromised 1 week prior to inoculation by intramuscular injection with 2 mg/kg/d dexamethasone, which was continued for 1 week after B burgdorferi inoculation. Under ketamine (7 mg/kg)/xylazine (3 mg/kg) anesthesia, whole blood was collected by venipuncture of the femoral vein into evacuated tubes containing either no anticoagulant or sodium heparin. Cisternal punctures were performed to obtain CSF after shaving, prepping, and draping of the high posterior cervical area. All blood samples for culture were obtained using aseptic collection techniques to avoid contamination. CSF and blood were obtained at intervals of approximately 1 week for the first 2 months and once every other week for the next 2 months.

MRIs were also performed using the ketamine/xylazine combination with supplemental dosing as required to maintain anesthesia. To ensure hydration during MRI, a 5% dextrose in 0.9% NaCl intravenous drip was maintained through an indwelling 21-gauge catheter in the saphenous vein. At sacrifice, euthanasia was performed on ketamine/xylazine-anesthetized animals by intravenous administration of an overdose of sodium pentobarbital (65 mg/kg).

Spirochetes. NHPs were injected with N40Br, a previously described substrain of N40 that has been multiply passaged in vivo in mice ^[6-8] and has been demonstrated to induce chronic infection in mice ^[6-8] and rabbits ^[9]. Inoculation of each NHP was with 10 to 15 intradermal injections consisting of a total of 1 million spirochetes in 1 ml along the shaved and cleansed back. NHPs A, B, C, and D were inoculated on the same day with an identical N40Br culture, while a later N40Br culture was used for NHP E. The inoculum was examined for motility after inoculation, and in each case the percent motility of the inoculum (ie, >90%) had not changed from that recorded before inoculation.

ELISA for anti-B burgdorferi antibodies. ELISA of NHP specimens was performed as previously described for human specimens ^[10]. In brief, the antigen used in the ELISAs and Western blots was a sonicate of the strain N40Br. Two hundred microliters of antigen coating solution, carbonate solution at pH 9.6, was added to a microtitration plate (Linbro Scientific, Hamden, CT) at a concentration of 5 micrograms/ml and incubated overnight at 4^degrees C. Plates were washed three times with phosphate-buffered saline (PBS)-0.05% Tween 20 and 200 microliters of the sera at a dilution of 1:5,000 in PBS-Tween 20. Plates were incubated for 2 hours at 37 degrees C and then washed again as described above. Two hundred microliters of horseradish peroxidase-conjugated goat anti-human immunoglobulin, isotypes G, A, and M (Organon Teknika-Cappel, Malvern, PA), was diluted 1:5,000 to 1:10,000 in PBS-Tween 20 and added to each well. Incubation followed for 2 hours at 37 degrees C. Plates were washed, and 200 microliters of TMB Microwell Peroxidase Substrate (Kirkegaard and Perry Laboratories, Gaithersburg, MD) was added to each well, immediately after which 50 microliters of 8% sulfuric acid was added to stop the reaction. The plates were read immediately on an ELISA spectrophotometer (Bio-Rad Laboratories, Richmond, CA) at 450 nm. To make plate-to-plate adjustments for optical density (OD), a standard positive control serum at multiple dilutions was run on each plate. The experimental values on each plate were multiplied by a factor based on the relative OD values of readings in the linear range of the OD-dilution curve of the positive control. Aliquots of serum and CSF samples were frozen after collection. The serum and CSF samples were then thawed and run on the same ELISA to avoid intertest variations.

CSF was treated in an identical manner except that a 1:10 dilution was used. To avoid concerns about contamination of CSF by antibody in blood, those taps in which there were more than 1,000 RBCs were not included.

Western blot. Western blotting was performed as previously described for human specimens ^[10]. In brief, whole-cell sonicate was boiled in loading buffer (1.25 M TRIS-HCl, pH 6.8, with SDS, mercaptoethanol, glycerol, and bromphenol blue) for 5 minutes and allowed to cool at room temperature prior to application to the gel. Twenty micrograms of spirochetal proteins per well were loaded onto a 15% SDS-PAGE vertical apparatus (Owl Scientific, Cambridge, MA) and subjected to 25 mA constant current for approximately 40 minutes (or until the bromphenol blue dye was completely off the gel). Proteins were then transferred to 0.45-microns-pore size nitrocellulose for 2.5 hours at 200 mA constant current in Towbin's TRIS-glycine-methanol buffer on an LKB Midget Multiblot (LKB Products, Bromma, Sweden). Lanes containing molecular weight markers and B burgdorferi sonicate were stained with amido black. The membrane was blocked in a 3% gelatin/500 mM NaCl buffer (pH 7.5). NHP serum samples were diluted 1:1,000 or as specified in 1% gelatin/500 mM NaCl/0.05% Tween buffer (pH 7.5) before incubation with the nitrocellulose. After three TRIS buffer washes, goat anti-human alkaline phosphatase conjugate was used at a dilution of 1:5,000. The second antibody was affinity-purified goat anti-human IgG, IgA, and IgM alkaline phosphatase conjugate (Cappel, West Chester, PA), which was used at a dilution of 1:3,000, and the strips were developed using a substrate consisting of 5-bromo-4-chloro-3-indolyl phosphate p-toluidine salt (BCIP) and p-nitroblue tetrazolium chloride (NBT), available as BCIP/NBT solution (Bio-Rad Laboratories).

Culture. Blood and CSF were cultured as previously described for murine blood and tissues ^[6-8]. Blood was collected from the femoral vein by syringe, and a few drops immediately dripped into a 4-ml culture vial without the addition of anticoagulants. Approximately 0.1 ml of either fluid was placed in 4 ml of BSKII medium. The culture was maintained at 33 degrees C and aliquots observed under darkfield microscopy at 14 and 28 days after culture. Cultures were considered positive if motile spirochetes were seen.

Polymerase chain reaction/hybridization (PCR/H) analysis. PCR/H of blood, CSF, and organs was performed as previously described ^[7-9,11].

Isolation and purification of template DNA from blood and CSF. Techniques were identical to those previously described ^[8,9,11]. In brief, 1 ml of blood was brought to a final volume of 5 ml with 1 x PBS. Next, 0.5 ml of 10% SDS (Sigma Chemical, St. Louis, MO) stock solution was added, followed by 55 microliters of proteinase K (Boehringer Mannheim Biochemicals, Indianapolis, IN) from a stock solution of 10 mg/ml. After an overnight incubation at 37 degrees C, the blood was extracted twice with a 1:1 mixture of phenol and chloroform. A final extraction with chloroform was performed to remove all phenol. One-tenth volume of 3 M sodium acetate (Sigma Chemical), pH 5.4 to 6.0, was then added to the aqueous mixture, followed by the addition of 2 volumes of ice-cold 100% ethanol (Warner-Graham, Cockeysville, MD). DNA was precipitated overnight at -20 degrees C and then spun at 14,000 rpm in a Sorval SS-34 rotor for 20 minutes. The pellet was dried and resuspended in 0.2 ml TE buffer (10 mM TRIS, 1 mM EDTA; pH 7.6). Proteinase K was then inactivated by boiling the samples for 5 minutes. Samples were read on a spectrophotometer at OD 260 and 280, and reextraction was performed if the OD 260:OD 280 ratio was less than 2. DNA concentration was calculated based on 1 OD at 260 representing a concentration of 50 micrograms/ml.

For CSF specimen preparation, 400 microliters of CSF was boiled in 1.8-ml microfuge tubes for 10 minutes. Three hundred twenty microliters of double-distilled water was then placed in the bottom of Ultrafree-MC-30,000 NMWL filter units (Millipore, Bedford, MA) and the boiled CSF placed in the top of the tube. The tube was spun at 10,000 rpm for 10 minutes, and 10 microliters of the remaining fluid in the top chamber was used in the PCR.

PCR protocol. Two genes of B burgdorferi were chosen as targets for amplification. The first was a genomic sequence first described by Rosa and Schwan ^[12,13] and used in the PCR previously described for the murine model ^[8]. The second was the OspA sequence first described by Bergstrom et al ^[14]. Each PCR run and the subsequent hybridization of PCR products included a range of concentrations from 1 to 30 fg of B burgdorferi DNA. PCR/H results were excluded from the data unless the sensitivity for B burgdorferi DNA controls was at least 10 fg. Also, one-eighth of all samples performed were either saline controls or DNA extracted from the brains of uninfected mice. Data from any PCR/H run yielding a positive on a negative control was excluded; this occurred in fewer than 3% of all PCR/H runs.

Genomic primers. The following primers were used for the genomic PCR:

- genomic-1: 5'-CGAAGATACTAAATCTGT

- genomic-2: 5'-CAGGAGCCTATGGAAACG

- genomic-3: 5'-AATCAGTTCCCATTTGCA

- genomic-4: 5'-TGTTAGGATCTGAGGGTG

- genomic-5: 5'-GATCAAATATTTCAGCTT.

Each PCR was run in 100-microliters volumes with a final magnesium concentration of 1.5 mM. The primers were prepared in the Lombardi Cancer Center Oligonucleotide Synthesis Facility at Georgetown University with a SAM One DNA Synthesizer (Biosearch, San Raphael, CA). The initial primer pair used consisted of genomic-1 and genomic-5. Forty-four picomoles of each primer was mixed with 44 nmol of each nucleotide (Boehringer Mannheim Biochemicals) and 2.5 U Taq polymerase (Cetus, Emoryville, CA). Five hundred nanograms of extracted DNA was then added, 50 microliters of mineral oil layered on top, and the tube placed in an automated DNA thermal cycler (Perkin Elmer-Cetus, Norwalk, CT). Denaturing was performed at 94 degrees C for 1 minute, annealing at 37 degrees C for 2 minutes, and extension at 72 degrees C for 2 minutes. The PCR product was then separated from the residual primers, desalted, and concentrated by spin dialysis in a UMC Filter (Millipore). Ten microliters of this solution was then added to a new solution of nucleotides, buffer, and Taq polymerase, and the internal set of primers, genomic-2 and genomic-4, was then used for the second round. The same conditions for the concentrations of reagents and thermal cycler as for the first round were used in the second round. Eight and one-half microliters of the PCR product was then added to 3 microliters of gel loading buffer and run on 1.5% agarose at 160 V. Gel loading buffer consisted of 0.04% bromphenol blue, 0.04% xylene cyanol, and 5% glycerol.

Each PCR run consisted of 32 samples: 24 experimental samples, four positive controls (consisting of 10, 3, 1, and 0.3 fg B burgdorferi (N40)-extracted DNA), and four negative controls (consisting of two saline samples and two tubes of DNA extracted from the brain of an uninfected mouse). A positive result was considered a clear band at 294 bp.

OspA primers. The OspA protocol was a single-cycle PCR identical to that performed by Persing et al ^[15,16]. The following primers were used:

- OspA149: 5'-TTATGAAAAAATATTTATTGGGAAT

- OspA319: 5'-CTTTAAGCTCAAGCTTGTCTACTGT

- Probe: 5'-ATTGGGAATAGGTCTAATATTACGCT.

A master mix was prepared containing 10 mM TRIS-HCl (pH 8.3), 50 mM KCl, 2.5 mM MgCl₂, 200 micromolars each deoxynucleoside triphosphate, 50 pmol each primer, 10% glycerol, and 1.25 U AmpliTaq DNA polymerase. Fifty microliters was used in 5 microliters of DNA template; the reaction mixture was overlaid with two drops of mineral oil. DNA was amplified using 50 cycles of 94 degrees C for 45 seconds, 55 degrees C for 45 seconds, and 72 degrees C for 45 seconds. The PCR product was then run on a 2% NuSieve/1% SeaKem GTG agarose gel.

Hybridization by slot blotting. This was performed as previously described ^[8,11]. Ten microliters of PCR product was diluted in 200 microliters of buffer. The slot-blotting apparatus (Schleicher & Schuell, Keene, NH) was set up and membrane prepared (Zeta-Probe, Bio-Rad Laboratories). After a rinse with the SSPE buffer, the 200 microliters of sample was added to the wells and vacuum was applied. After another rinse with buffer, wells were dried by prolonged application of vacuum.

The DNA was denatured by incubation in 0.5N NaOH at room temperature for 10 minutes. After washing with SSPE buffer and drying, the membrane was exposed to ultraviolet light for 5 minutes. Oligonucleotides labeled as "genomic-3" or "probe" for the genomic and OspA PCRs, respectively, were labeled with digoxigenin by terminal transferase using the Lumiphos protocol (Boehringer Mannheim Biochemicals). The membrane was exposed to a hybridization solution containing 2.5 pmol oligonucleotide per 10 ml and then incubated for 100 minutes at 45 degrees C. The membrane was then washed in sequential 10- and 20-minute washes at room temperature and 64 degrees C. After the final wash, blocking solution was added for 30 minutes and then washed off. The anti-digoxigenin antibody-alkaline phosphatase conjugate (Boehringer Mannheim Biochemicals) was then added and incubated for 30 minutes at room temperature. Lumiphos-530 was then added and exposed to film.

For each slot-blot experiment, positive controls (ie, B burgdorferi DNA and lambda DNA labeled with digoxigenin) and negative controls (unlabeled lambda DNA and DNA from hearts and brains of uninfected mice) were included. All specimens underwent both PCR and hybridization unless otherwise indicated.

CSF analysis. Determination of cell counts and protein in the CSF was performed by Maryland Medical Laboratories using standard techniques. CSF was delivered by messenger to this laboratory within 12 hours of collection and was analyzed within 24 hours of collection. In no cisternal tap were there more than 10,000 RBCs. Of the 65 cisternal taps performed, there were five that contained between 1,000 and 10,000 RBCs, and in these taps, the "cell count" recorded was the actual number of WBCs counted minus the total RBCs divided by 1,000.

Histopathology of skin. Four-millimeter punch biopsy samples of skin were obtained from the erythema migrans lesions. Biopsy samples were fixed in 10% formalin followed by a change of formalin at 24 hours. They were then processed by dehydration in graded alcohols, embedded in paraffin, cut in 6-microns sections, and stained with hematoxylin and eosin. Stained tissue sections of all the collected samples were examined using standard light microscopy.

Brain MRI. MRI was performed using a 1.5-tesla General Electric (Milwaukee, WI) Signa MR unit and a 5-inch receive-only surface coil. Animals were positioned prone in a stereotaxic head frame with the surface coil located in close proximity to the top of the skull. Coronal and sagittal localizing images were obtained at an echo time (TE) of 12 msec and a repetition time (TR) of 400 msec, with a 5-mm slice thickness in a 16-mm field of view, to establish values for the remaining series of scans. A T (_1-weighted) inversion recovery scan was obtained using an inversion (TI) of 600 msec, a TR of 200 msec, and a TE of 25 msec in the anatomic coronal plane. T_1-weighted images were also obtained in the axial plane, using spin-echo sequence (TR 600 msec/TE 20 msec) with an in-plane resolution of 0.79 to 0.39 mm. T_2-weighted images (TR 2,000 msec/TE 80 msec) and proton density images (TR 2,000 msec/TE 20 msec) were obtained in the anatomic axial plane. Gadopentetate dimeglumine (Magnevist, Berlex Laboratories, Cedar Knolls, NJ) was administered intravenously at a dose of 0.4 mmol/kg body weight and postcontrast T_1-weighted images were obtained in the axial and coronal planes. All scans were done with a 3-mm slice thickness and interleaved in a field of view of 10 cm. All images were acquired with 128 phase encode steps, processed with 2-D Fourier transform technique, and displayed on a 256 x 256 matrix. MRIs were performed at irregular intervals beginning after the first month, due to the lack of predictability of MRI scanner availability.

Results. Clinical signs. Ten days to 2 weeks after inoculation, all four of the immunocompetent NHPs (A, B, C, and D) displayed characteristic erythema migrans. The skin lesions began as mild erythematous areas near the inoculation sites. These areas of erythema became confluent and spread outside the area of injection. The lesions were very erythematous at their peak by 3 weeks after infection and clearly discernible along the shaved back. After the first few weeks of the lesions, they began to fade somewhat and had disappeared by 6 weeks after inoculation.

Careful daily monitoring by the staff revealed no sign of meningitis, encephalitis, changed behavior, or arthritis in any animal.

Enzyme-linked immunosorbent assay. Serum samples were obtained before inoculation and at intervals after inoculation. Data are shown in Figure 1A.

The results are consistent with three phases of antibody response. The first phase, which constitutes the first week after inoculation, is a short latency phase in which there is little or no detectable antibody response. The second phase is the progressive increase in specific antibody over weeks 2 to 6. Following this is the third phase, a period of antibody plateau in which high-titered antibody is present.

CSF antibody testing revealed a pattern of rising antibody reactivity in NHPs A, B, and E but lower reactivity in the other two NHPs Figure 1B.

Western blot. The Western blot data on NHP C, presented in Figure 2 and representative of the data for all immunocompetent NHPs, show increasing numbers of antigens recognized by the humoral response with increasing duration of infection. The predominant antigens were of 31, 34, 41, 22, and 16 kd (in order of appearance). Thirty-one-kilodalton and 34-kd bands in Western blots using N40Br represent antibodies to OspA and OspB, respectively. There were some variations in the Western blots between animals. For instance, NHP A had much less of an anti-22 kd response than did the other animals, while the anti-41 kd response of NHP B was much less prominent. The first bands visualized were 41 kd for NHPs C and D, 31 kd for NHP A, and 34 kd for NHP B. A 16-kd band was a late-appearing band in all animals.

Culture. None of 57 blood cultures from the infected animals was positive. Two CSF cultures were positive, one 29 days after inoculation in NHP E and one 36 days after inoculation in NHP A. These cultures were active and were grown to larger volumes.

Thirty-two CSF cultures obtained before, and 39 CSF cultures obtained after, the appearance of pleocytosis or positive PCRs were negative.

Polymerase chain reaction/hybridization. PCR/H was demonstrated to be more sensitive than culture in this study, as it was in previously studied animal models ^[13,15]. All samples were tested by genomic PCR, but because of a lack of adequate volumes, only selected samples were tested by OspA PCR.

Blood. PCR/H positivity of the blood was present on a single occasion in three NHPs--A (on postinoculation day 13), B (on day 16), and D (on day 22).

Cerebrospinal fluid. Positivity of the CSF PCR is demonstrated in the Table 1. CSF PCR positivity appeared first on day 29 in one NHP and on day 36 in the four others. Of 41 CSF specimens after the appearance of pleocytosis in which both cell count and PCR were tested, the PCR correlated with the presence of CSF pleocytosis in 27 (66%): CSF PCR was negative in 10 cases when no cells were present and positive in 17 cases when a pleocytosis was present. In four instances the pleocytosis was not present while the CSF PCR was positive, and in 10 instances the pleocytosis was present with a negative CSF PCR.

Comparison of the two PCR methods. The OspA PCR had a higher rate of positivity than did the genomic PCR for CSF specimens, despite equal sensitivity of all three techniques for DNA extracted from cultured B burgdorferi. In 48 specimens in which both targets were used, both were negative in 29, both were positive in five, OspA was positive and genomic negative in 11, and OspA was negative and genomic positive in three. In the data shown in the table, PCR is labeled as positive if either PCR method yielded a positive result. In none of the PCR runs was there positivity in water controls, which constituted one-eighth of all PCR specimens. All specimens positive by PCR were confirmed by slot-blot hybridization.

CSF analysis. Cell count. As shown in the table, CSF pleocytosis was first observed at approximately the same time at which the CSF became positive by culture, by PCR/H, or by both. The pleocytosis within the second month disappeared by the end of 2 months after inoculation but then reappeared in the fourth month in all four of the NHPs maintained that long.

Protein. CSF protein was normal, ie, 7 to 24 mg/dl, in all samples tested in four of the five NHPs. The protein level in NHP B, which was high in the preinoculation CSF, did not vary significantly during the course of the infection. The reason for the high preinoculation CSF protein in this animal is unknown.

Histopathology. Microscopic changes observed in the skin rash tissue (2/2 NHPs biopsied) consisted of acanthosis, hyperkeratosis, and focally extensive infiltration of the upper dermis and parafollicular structures with lymphocytic cells and eosinophils. In some areas, focal accumulations of lymphocytes and eosinophils formed small nodules or surrounded dermal vessels.

Brain MRI. Brain MRI without and with gadolinium contrast was negative on all animals except NHP E. Twelve negative studies, spread evenly over 3 months, were performed on these animals--on postinoculation days 22, 49, 75, and 91 in NHP A, on days 34, 65, and 89 in NHP B, on day 36 in NHP C, and on days 22, 49, 63, and 84 in NHP D.

The gadolinium-enhanced MRI on NHP E revealed enhancement of the meninges at the base of the temporal lobe Figure 3 and was present on days 32, 46, and 53 after inoculation.

Discussion. We demonstrated that the NHP model of LNB is remarkably similar to the human disease. In both human and nonhuman primates, erythema migrans precedes the development of neurologic involvement. In both, culture positivity of the CSF is a rare event and tends to occur very early in the course of neurologic involvement ^[17]. In both, CSF pleocytosis occurs within the second month after infection and persists. Other similarities include the pace and complexity of the humoral immune response, the superiority of PCR over culture in detecting the presence of the spirochete in CSF, and the occurrence of meningeal inflammation as demonstrated by MRI.

In the NHP model, as in human disease, neurologic involvement is associated with an anti-B burgdorferi antibody response, as demonstrated by ELISA and reactivity of serum antibody to multiple antigens as determined by Western blotting. A lower antibody response is likely to result from a smaller inoculum, ^[7,8] but on the basis of the present study it appears unlikely that immunocompetent NHPs can be infected and have a complete lack of an antibody response.

In both NHPs and humans, B burgdorferi is readily detected in the CSF by PCR. Two separate targets, genomic and plasmid sequences, were employed for two reasons. First, they were confirmatory for each other. Second, we wished to test whether OspA targets were more sensitive for infection in the CNS than other targets. This study confirmed in the CNS what had previously been observed in joints--namely, that OspA targets were more sensitive than other targets ^[15,16] for infected body fluids and tissues, despite equivalent sensitivities for DNA extracted from cultured B burgdorferi. The cause of this "target imbalance" is unknown, but it may represent duplication of plasmid DNA in vivo or greater accessibility of DNA polymerase to plasmid targets than to genomic ones.

In both NHP and human LNB, brain MRI in the first months of neurologic involvement occasionally reveals uptake of gadolinium in the dura consistent with meningeal inflammation. This finding is consistent with the clinical findings in human LNB of radiculitis and cranial neuritis, as well as with the prominent CSF pleocytosis in both humans and NHPs.

PCR analysis of the CNS in our study revealed that spirochetal DNA was persistently present in the CSF. In the mouse model of Lyme disease, ^[7,8,18] infection persisted in the heart and bladder, often indefinitely, while brain infection occurred in only a small minority of animals. In the rabbit, in contrast, intradermal inoculation resulted in spinal cord and brainstem positivity by PCR months after inoculation in all animals injected; however, CSF was consistently negative for both PCR and culture. The NHP model of LNB is the first model in which CSF findings show many features consistent with human infection: pleocytosis and positivity by PCR in all animals and positivity by culture in some. As in human disease, the CSF is not always positive by PCR. The PCR data are also consistent with the persistence of infection in the CNS in untreated LNB.

The reappearance of CSF pleocytosis in all animals after a period in which the CSF was clear indicates a relapsing-remitting character in the inflammatory process in the CNS. This is consistent with human disease, especially with Lyme arthritis, in that joint inflammation occurs episodically without clear precipitating events. The cause of the variability in the serum antibody response or CSF pleocytosis is unclear, but it may be related to variations in the growth rate of the spirochete or the appearance of new genetic variants in the nervous system ^[19,20]. Other possible causes for variation in CSF pleocytosis need to be considered, such as sampling error, lab error, or delay in transport of some CSF samples to the laboratory. However, none of these causes was clearly present, and we believe that the observed fluctuation represents variation in the degree of subarachnoid inflammation.

The fact that 100% of the infected animals developed neurologic infection indicates a higher incidence of neurotropism than is generally assumed to occur in naturally infected humans. Studies in humans have estimated the occurrence of neurologic involvement in Lyme borreliosis at 15 to 62% ^[21,22] of all cases. The higher figures are generally reported from Europe, where in a recent study ^[22] 42% of Lyme borreliosis cases were classified as LNB in Austria, 57% in southern Germany, 41% in northern Germany, 57% in France, and 62% in Switzerland. A number of factors might account for the variability in incidence, such as variations in the detection of subclinical infection, in the aggressiveness of early treatment of erythema migrans, and in the number of spirochetes injected by infected ticks. Another factor may be variations in the degree of neurotropisms among strains, since strains have been shown to vary considerably genetically ^[23]. Our studies indicate that if an inoculum adequate in terms of spirochete numbers and strain is injected intradermally, CNS infection is virtually assured.

Some clinicians use CSF antibody as their primary test for determination of active infection. In our study there were three NHPs with strongly positive CSF antibody responses (A, B, and E) and two NHPs in which antibody levels in CSF were quite low (C and D). These animals were not distinguishable on the basis of CSF pleocytosis or CSF PCR. The possibility that the observed differences in CSF antibody might simply be secondary to differences in serum antibody appears unlikely, since all animals mounted serum anti-B burgdorferi antibody titers of similarly high amplitude. However, necropsy studies in progress may identify differences in persistence of the spirochete in the CNS between the two groups.

Our study contrasts with a previous study of experimental Lyme borreliosis in rhesus macaques ^[24]. The previous study documented the presence of CSF pleocytosis in two animals, each on one occasion, but concluded that CNS infection in the animals was not present because culture results were negative; PCR of blood and CSF was not performed. Our study, which focused much more intensively on neurologic involvement, has demonstrated in NHPs a large part of the spectrum of human LNB not detected in the previous study, including persistent CSF pleocytosis, ^[25] CSF culture positivity, ^[17] positive CSF PCRs, ^[11] progressively complex Western blots of serum antibody, ^[10] and positive MRIs. The two studies did have a similarity outside the CNS in that erythema migrans with typical histopathology was observed in both.

Our understanding of human LNB has previously come exclusively from clinical studies in which patients presented with neurologic symptoms. In many of these patients the clinical diagnosis of LNB was confirmed exclusively on the basis of serology, which frequently can be unreliable. Thus, the natural history of infection within the nervous system in Lyme borreliosis is unknown, since only symptomatic individuals are encountered and since some patients diagnosed with LNB may have diseases other than LNB. We do know that subclinical infection of the nervous system can occur and that, in Lyme borreliosis, absence of neurologic symptoms cannot be equated with absence of neurologic involvement ^[11,26]. The data presented in our study are consistent with the hypothesis that subclinical infection is common, since the spirochete was present in the CNS in all animals without focal neurologic deficits, gross change in behavioral patterns, or large lesions demonstrable on MRI.

Acknowledgments

The authors thank Dr. Lucien Levy for his assistance in reading the MRIs, Dr. Joseph Frank for his aid in obtaining the MRIs, and Dr. Venita Thornton and Lisa Berrey for their help in management of and specimen collection from the nonhuman primates.