PCR assays were performed as previously reported 16 17 18 19 20 . Two hundred μl cerebrospinal fluid specimens were submitted for DNA extraction using the QIAamp DNA MiniKit (Qiagen, Hilden, German) according to the manufacturer's recommendations. PCR mixes were prepared using a Fast-Start DNA Master SYBR Green kit (Roche, Mannheim, Germany) following the manufacturer's instructions. Quantitative PCR (qPCR) was performed in a LightCycler thermocycler (Roche biochemicals, Mannheim, Germany). For each assay, positive and negative controls were used. At intervals of 5 samples, negative controls (water, PCR mix and human samples) were evaluated. A tenfold dilution of a standard suspension of 10⁶ T. whipplei strain Marseille-Twist was used as a positive control and for quantification, as previously reported 21 . The quality of DNA extraction from samples was estimated by PCR targeting a housekeeping gene coding β-actin. If a first PCR assay was positive, the result was systematically confirmed by a second PCR assay using a second set of primer pairs. In the event of discrepancies between the two PCR assays or incorrect controls, samples were submitted to new DNA extraction and/or new qPCR assays.

The PCR primers that we used have evolved with improved knowledge about T. whipplei. Between January 2001 and February 2004, all samples were tested using our regular qPCR, targeting a 489-bp fragment of the 16S-23S rRNA gene intergenic spacer using the primers tws3f (5'-CCGGTGACTTAACCTTTTTGGAGA) and tws4r (5'-TCCCGAGGCTTATCGCAGATTG), as previously reported 17 . If this assay was positive, a 650-bp fragment of the rpoB gene using the primers TWRPOB.F (5'-TTTTTCCGGCGTGCGCTCAA) and TWRPOB.R (5'-TTTCTCCGAGGTTGCGGTGC) was performed 17 . For all of the assays, when an amplified product was detected, T. whipplei was systematically confirmed by sequencing, as previously reported 19 . Since October 2003, the availability of the T. whipplei genome has offered the option of using repeated sequences of T. whipplei for highly sensitive and specific PCR assays 16 . Subsequently, a PCR targeting repeated sequences of T. whipplei was developed that, when an amplified product was detected, confirmed the identification of T. whipplei by sequencing (between October 2003 and March 2004) or by using specific oligonucleotide Taqman* probes (since April 2004). From October 2003 to March 2004, T. whipplei qPCR targeting a 164-bp sequence of the bacterium incorporated the primer pairs 53.3F (5-AGAGAGATGGGGTGCAGGAC) and 53.3R (5'-AGCCTTTGCCAGACAGACAC) into the reaction mix. If this assay was positive, the result was confirmed by a second assay using a second set of primer pairs, 342F (5'-AGATGATGGATCTGCTTTCTTATCTG) and 492R (5'-AACCCTGTCCTGCACCCC), which targeted a different DNA sequence. Since April 2004, the reaction mix has included T. whipplei qPCR targeting a 155-bp sequence of the bacterium and incorporating the primer pair TW27F (5'-TGTTTTGTACTGCTTGTAACAGGATCT) and TW182R (5'-TCCTGCTCTATCCCTCCTATCATC) and a Taqman* probe (27F-182R, 6-FAM-AGAGATACATTTGTGTTAGTTGTTACA-TAMRA). If this assay was positive, the result was confirmed by a second assay using a second set of primer pairs TW13F (5'-TGAGTGATGGTAGTCTGAGAGATATGT) and TW163R (5'-TCCATAACAAAGACAACAACCAATC) and a Taqman* probe (13F-163R, 6-FAM-AGAAGAAGATGTTACGGGTTG-TAMRA) targeting a different 150-bp sequence 20 21 22 . Finally, all the cerebrospinal fluid specimens prior to October 2003 were retrospectively tested using PCR targeting repeated sequences, and the specimens sampled between October 2003 and February 2004 were tested using both regular and repeat PCR.

Serologic studies were performed as previously reported 23 . Cerebrospinal fluid specimens were cultivated using both cell-culture and cell-free culture media, as previously described 24 25 .

Brain magnetic resonance examinations were performed on a 1.5 T Magnetom Vision Plus system (Siemens, Erlangen, Germany). The imaging protocol included one axial FLAIR sequence (TR = 8000 ms, TI = 180 ms, TE = 110 ms) and one axial T1-weighted sequence (TR = 644, ms TE = 15 ms) performed before and after contrast agent administration (0.2 ml/kg of Gadolinium-DTPA, Guerbet, Paris, France). Magnetisation transfer imaging was performed using two axial proton-density weighted FLASH sequences (TR = 500 ms, TE = 4.7 ms, 21 slices, thickness = 5 mm, flip angle = 30°, without and with MT saturation: 1.5 kHz off resonance, 500°, FOV = 240 mm, matrix size = 256 × 256). Diffusion-weighted imaging was performed using a single-shot EPI sequence (b = 0, 500, 1000 s/mm²) applied in the x, y and z directions (19 slices, thickness = 5 mm, matrix size = 128 × 128, FOV = 256 × 256 mm²). Apparent diffusion coefficient (ADC) maps were reconstructed using this sequence. Single voxel proton magnetic resonance spectroscopy (SVS) included STEAM acquisition at an echo time of 20 ms. TR was 1500 ms. One voxel was centred on the medulla oblongata lesion, while the reference voxel was located on a contiguous area (the pons) that appeared normal on all MR sequences. The voxel volume was adapted to the size of the lesion. Several normal brain metabolites are detected using the STEAM 20 ms sequence: N-acetylaspartate (NAA) is an amino acid that is present almost exclusively in neurons; reduced levels may correspond to neuronal death or injury. Myoinositol (MI) is a sugar that is only present in glia. It reduces when these cells experience specific damage and increases in cases of glial activation or proliferation (in glial tumours or reactive gliosis) 26 . Other molecules are detected under pathological conditions. Lactate and lipids or macromolecules may be detected when there is a local ischemia that eventually leads to necrosis (as with ischemic stroke and some tumours). Lactate may also be detected in relation to a macrophage infiltration. High lactate and lipid levels are also detected in cases of abscess. Other molecules such as succinate, acetate, alanine and other amino acids can be detected using in vivo magnetic resonance spectroscopy in cases of brain abscess, but only when the infectious lesion is cystic. Creatine/phosphocreatine (tCr) is a ubiquitous metabolite involved in energy storage. It is a marker of overall cellular density. Choline (Cho) is mostly involved in cell membrane metabolism, and it is thus a membrane turnover marker. It increases whenever there is cellular proliferation (as with brain tumours), inflammation or membrane catabolism (as with multiple sclerosis).

Data were analysed using EpiInfo software, Version 3.4.1 (Centers for Disease Control and Prevention, Atlanta, GA, USA). P < 0.05 was considered statistically significant.

A 39-year-old male (Patient 1) was hospitalised for cognitive impairment and choreiform movements. His medical history included a nearly 3-week-long period of unexplained gastroenteritis 4 years earlier. Over a 3-year period, the patient developed hepatitis and obesity (a 25-kilogram weight gain) associated with hypercholesterolemia, hypertriglyceridemia and Type II diabetes. Hormonal evaluation showed peripheral hypoandrogenia with elevated levels of follicle-stimulating hormone. Brain magnetic resonance imaging (MRI) did not show any abnormalities. One year prior to hospitalisation, he developed dementia, dysarthria, and apathy, followed by involuntary and almost permanent choreiform movements of the left hemibody and abnormal buccofacial movements. No blood cell count abnormalities or C-reactive protein increases were observed. An increase in hepatic transaminases was identified. A lumbar puncture showed a high protein level (4 g/l) with a normal glucose level. Examinations with brain computer tomography (CT) and brain MRI and spectroscopy were abnormal (Figure 1 and additional file 1, Figure S1). The MRI disclosed multiple unusual lesions corresponding to multiple brain lesions located in the medulla oblongata, hypothalamus, internal side of the right temporal lobe, fornix, lenticular nuclei and head of the caudate nuclei. These lesions showed high signal intensity on FLAIR images and low signal intensity on MT images, and they were slightly enhanced on T1-weighted images after gadolinium administration. Spectroscopy was characterised by a decrease in the NAA and MI peaks (consistent with neuronal and glial injury) and an increase in lipids and macromolecules, whereas lactate and other amino acids suggestive of an abscess were not detected. Finally, the absence of an increase in the choline-creatine ratio ruled out a demyelinating lesion or neoplasm. A brain biopsy was taken from the right striatum; it revealed astrocytic gliosis and vascular proliferation. No granulomas were observed. PAS staining and T. whipplei immunohistochemistry were negative. Regular PCR targeting the intertransgenic spacer (ITS) was negative for this brain biopsy; however, repeat PCR was positive for both the biopsy specimen and cerebrospinal fluids. One month after beginning specific treatment (doxycycline 200 mg, hydroxychloroquine 200 mg 3 times and trimethoprim-sulfamethoxazole 320 + 1,600 mg 3 times each per day), significant improvements were observed; however, cerebrospinal fluid showed persistent high protein levels, and repeat-PCR was still positive. After 2 months, the clinical examination was normal and, surprisingly, the patient had lost 17 kg. One year after the start of the treatment, the patient's obesity had completely regressed, his hepatic tests were normal and he presented no evidence of diabetes; furthermore, brain MRI revealed that the majority of the lesions had disappeared. After 18 months of treatment, brain MRI and spectroscopy results were normal. The repeat PCR for cerebrospinal fluids was negative. The treatment was stopped. Fourteen months later, he experienced a clinical relapse, with weight gain (10 kilograms), the reappearance of diabetes, abnormal choreic movements, headache, difficulty concentrating and increased hepatic transaminases. Brain MRI and spectrometry showed signs of relapse, but T. whipplei repeat PCR for CSF was negative. A treatment based on doxycycline, hydroxychloroquine and sulfadiazine (1,500 mg four times per day) was started. All clinical abnormalities resolved, including obesity, similar to the patient's response to the first treatment. After 17 months of treatment, the patient's brain MRI was normal and spectroscopy was clearly improved. After 52 months, the treatment was stopped. Seventeen months after the completion of treatment, the patient showed no sequelae and felt well.

The characteristics of 4 other patients with T. whipplei encephalitis (Patients 2, 3 and 4), seen by one of the authors (DR) in consultation, and 15 patients from the literature (8 with confirmed T. whipplei encephalitis and 7 with a possible diagnosis) are summarised in an additional file 2, Table S1 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 . Seventy-four patients with Whipple's disease with neurologic manifestations and sufficient clinical information were also identified from the literature review 15 27 44 45 46 47 48 49 50 51 52 53 54 . Comparisons among the different groups are summarised in Tables 2 and 3.

T. whipplei encephalitis has two consistent findings: progressive cognitive impairment and the eventual onset of obesity or ataxia. For our 5 patients, the only diagnostic finding was a positive PCR of for cerebrospinal fluid and brain biopsies. This disease arises from the isolated involvement of T. whipplei localised to the central nervous system, which responded to antibiotic treatments. Among our patients and those in the literature, all 3 who did not receive antibiotics (because T. whipplei infection was diagnosed after post-mortem examination) died, compared with only 3 deaths among those who received antibiotics. This difference was statistically significant (p = 0.01). No specific brain MRI abnormalities were observed. Six out of 18 patients had a normal brain MRI.

Among the 824 cerebrospinal fluid specimens, 7 (0.85%) tested positive for T. whipplei DNA with PCR; 2 out of 16 brain biopsies were PCR positive, including a sample from one patient for whom cerebrospinal fluid was negative. Negative and positive controls were correct, including the amplification of the gene coding β-actin. All of our attempts to cultivate T. whipplei from the cerebrospinal fluid specimens of our 6 patients with T. whipplei encephalitis and positive PCR samples were unsuccessful; in contrast, we established 9 strains of T. whipplei from the cerebrospinal fluid specimens of patients with classic Whipple's disease with associated neurologic manifestations (6/8), asymptomatic neurologic involvement (2/4), or neurologic relapse (1/3) (published and unpublished data) 1 . This difference was statistically significant (p < 0.017). We hypothesised that this difference might be linked to a lower amount of bacteria in the cerebrospinal fluid specimens of patients with T. whipplei encephalitis. Serum sample serologic tests were available from 5 patients with T. whipplei encephalitis. Of these, 2 showed no reaction and 3 showed low-level but specific responses to the deglycosylated proteins of T. whipplei, as has been observed in patients with classic Whipple's disease 23 .