Tropheryma whipplei as a Cause of Epidemic Fever, Senegal, 2010–2012

Findings suggest that the bacterium has role in febrile episodes, is contagious, and has an epidemic character.

D etermining the etiologic causes of febrile illness in tropical settings provides public health and local community benefits. In the context of a decline in malaria cases in many parts of sub-Saharan Africa, the few studies that have been conducted in recent years to analyze the burden of bacterial infections used traditional blood cultures and identified typhoid fever and Streptococcus pneumoniae as the leading documented causes of nonmalarial bloodstream infections (1)(2)(3). However, this method does not enable the identification of intracellular organisms, and most causes of fever remain unknown. In 2008, we initiated a study of the etiologies of fevers of unknown origin in Africa, particularly in Senegal. Our preliminary studies showed the presence of previously known pathogenic microorganisms, such as Borrelia crocidurae, Rickettsia felis, R. conorii, and Coxiella burnetii, and the unexpected presence of Tropheryma whipplei (4)(5)(6)(7)(8)(9).
T. whipplei is highly prevalent in rural Senegal, where carriage rates reach 75% among children <2 years of age, and overall seroprevalence is 72% (21)(22)(23)(24)(25)(26). In our preliminary study in Senegal, which was conducted in 2 villages (Dielmo and Ndiop) during December 2008-July 2009, we detected T. whipplei bacteremia in 6.4% of the analyzed specimens (8). Bacteremia was significantly associated with cough, but no link to feces carriage was observed (8). However, our study had several limitations, such as a small number of febrile patients, no local control group of afebrile persons, and a short study period. In this same area, we recently showed that humans comprise the only source of T. whipplei among the populations in whom the bacterium is highly prevalent. Moreover, our findings showed that limited access to toilets and exposure to human feces was associated with the high prevalence of T. whipplei, suggesting that these conditions may facilitate fecal-oral transmission of the bacterium (31). To better characterize T. whipplei bacteremia, we extended our analysis, beginning in 2010, in this same area of rural Senegal to include the collection of >1,000 blood samples from healthy persons and ambulatory patients with acute fever.

Materials and Methods
We conducted the study during June 2010-March 2012 in Senegal's rural Sine-Saloum area, a dry sahelian ecosystem with 2 typical seasons: dry (November-May) and rainy (June-October). We obtained written consent for every person included in the study. The National Ethics Committee of Senegal approved the study.  (8).
In parallel, we collected blood samples from a control group of 385 healthy, afebrile villagers; 62.5% of these study participants were <15 years of age, and the sex ratio was 1:1.

DNA Extraction
For DNA extraction, we used a BioRobot EZ1 Workstation (QIAGEN, Courtaboeuf, France) according to the manufacturer's instructions. Extraction was performed in Senegal, and specific quantitative real-time PCR (qPCR) was performed in France.

Specific qPCR
We used a 7900HT-thermocycler (Applied Biosystems, Foster City, CA, USA) with the QuantiTect-Probe PCR Kit (QIAGEN) to perform qPCR. First, we analyzed specimens for T. whipplei by using the primer pair Twhi3F (5′-TTG TGT ATT TGG TAT TAG ATG AAA CAG-3′)/Twhi3R (5′-CCC TAC AAT ATG AAA CAG CCT TTG-3′) and the specific Twhi3 probe (6-FAM-GGG ATA GAG CAG GAG GTG TCT GTC TGG-TAMRA). For specimens with positive results, we ran a second, confirmatory qPCR with the Twhi2F (5′-TGA GGA TGT ATC TGT GTA TGG GAC A-3′)/Twhi2R (5′-TCC TGT TAC AAG CAG TAC AAA ACA AA-3′) primer pair and the specific Twhi2 probe (6-FAM-GAG AGA TGG GGT GCA GGA CAG GG-TAMRA) (8,21). To validate the assays, we included positive (T. whipplei) and negative (PCR mix) controls in each run, as previously reported (8,21). We considered samples to be T. whipplei-positive if qPCR results for the 2 specific genes were positive at a log-based fluorescence cycle threshold (C t ) of <38. We used qPCR for the β-actin housekeeping gene, as previously described (7), to check the quality of DNA handling and blood specimen extraction; only positive samples were considered reliable.

Genotyping
We performed genotyping of T. whipplei as previously described (32). We attempted to amplify and sequence each of 4 multispacer sequences (TW133, ProS, SecA, and Pro184) from positive specimens. When sequences were obtained, we compared them with those available in the GenBank database and our internal laboratory database to determine their corresponding genotype.

Statistical Analyses
We performed statistical analyses by using Epi Info 6 software (http://www.cdc.gov/epiinfo/index.html); results with p<0.05 were considered statistically significant. The corrected χ 2 test or the Fisher exact test was used where indicated.

Prevalence of T. whipplei Bacteremia
A total of 786 febrile patients and 385 healthy controls were included in the study, among whom 36 (4.6%) and 1 (0.25%), respectively, were positive for T. whipplei DNA (p<0.00007). The positive control participant was a 13-year-old boy who had low concentrations of T. whipplei DNA (C t of 36.85 and 37.99). The C t for febrile patients ranged from 26.10 to 36.41 (mean ± SD 33.40 ± 2.53).

Clinical Manifestations
Clinical data were available for 786 febrile patients ( Table

Treatment and Follow-Up
Data about antimicrobial drug therapy was available for 33 patients, 23 of whom benefited from treatment with amoxicillin (18 patients), metronidazole (3 patients), or cotrimoxazole (2 patients). In Dielmo, 24 specimens from 23 patients were positive for T. whipplei; 1 patient was sampled twice 15 days apart, and both specimens were positive. For 17 patients, blood specimens were also sampled during other febrile episodes. Nine specimens from 5 patients were sampled from 15 days to 13 months before the positive sample was detected, and 43 specimens from 17 patients were sampled from 3 weeks to 16 months after the positive sample was detected; all of these samples were negative. Moreover, our previously published data (8) included test results for a 4-year-old boy who was diagnosed with T. whipplei bacteremia in January 2009 (19 months before August 2010). Four other blood specimens from this patient were tested 1 month before (1 sample) or 4, 11, and 15 months after (3 samples) the positive specimen was detected, and all were negative for T. whipplei.
In Ndiop, 12 specimens from 12 patients were positive. For 8 of these patients, blood specimens were sampled during other febrile episodes. The specimen for 1 patient was sampled 1 month before the positive sample, and 9 specimens from 6 patients were sampled from 7 weeks to 18 months after the positive samples; all of these specimens were negative. No data were available for these patients about antibody response against T. whipplei.

Genotyping
Because of the lack of specimens available for genotyping and the low sensitivity of genotyping, we could obtain multispacer sequences for only 8 patients at the time of the 2010 peak in T. whipplei bacteremia cases ( Table 2). The T. whipplei genotype corresponds to the concatenation of the 4 spacers (TW133-ProS-SecA-Pro184); however, TW133 sequencing was not successful, so the corresponding spacer was not available (NA) for any of the patients. ProS sequence was obtained for 5 patients, SecA for 6 patients, and Pro184 for all patients. For 4 patients, 3 spacers were available, enabling the detection of the same multispacer sequence combination (NA-7-2-1) for the 4 patients. For another 4 patients, 2 spacers were available, enabling the detection of the NA-7-NA-1 combination for  2 of the patients and the NA-NA-2-1 combination for the other 2 patients. None of the potential combinations has previously been sequenced in Senegal. Moreover, the NA-7-2-1 combination has also not previously been detected in any other area of the world and is thus a new genotype. Overall, our data suggest that the same genotype was detected in Dielmo and Ndiop during the summer of 2010. However, T. whipplei genotyping was performed (sometimes only partially) for only 8 of 36 patients, so we can only suspect, but not confirm, that an epidemic clone was present and that an outbreak was ongoing at that time.

Affected Households
In 39 had a furnace in which they baked bread that they marketed locally. In December 2010, most of the family left the village and the furnace was shut down; no additional T. whipplei bacteremia cases were subsequently observed.

Discussion
We report the detection of T. whipplei DNA in the blood of patients in Dielmo and Ndiop, Senegal. The validity of our data is based on strict experimental procedures and controls, including rigorous positive and negative controls, used to validate test results. In addition, we confirmed each positive PCR result by the successful amplification of an additional specific DNA sequence, and we performed T. whipplei genotyping on several specimens. We also showed that the presence of T. whipplei in blood is significantly linked to the presence of fever; T. whipplei DNA was detected (at a low level) in the blood of only 1 afebrile person in the study area. Moreover, we included a control group of afebrile persons from the same area, thereby reinforcing the validity of our data. Indeed, several well-known pathogens have been detected in recently analyzed specimens from healthy persons. For example, Plasmodium falciparum has been detected in 32% of blood specimens from healthy, afebrile persons in Senegal (33); respiratory viruses, including influenza virus, have been detected in 12% of nasopharyngeal samples from symptom-free Hajj pilgrims (34); and S. pneumoniae has been detected in 6.3% of blood specimens from afebrile children in Tanzania (35). Thus, because of the significantly higher prevalence of T. whipplei among febrile patients compared with healthy controls, we suspect that this microorganism is a pathogenic agent. The overall prevalence of T. whipplei bacteremia is 4.6%. However, in August 2010, we observed a peak in T. whipplei bacteremia cases in Dielmo and Ndiop, where T. whipplei was involved in more than half of the observed cases of fever. This peak corresponds to a short outbreak of T. whipplei bacteremia with 1 potential genotype. A similar new genotype was observed for the patients from Dielmo and Ndiop for whom genotyping was available at the time of the outbreak. To date, 35 different T. whipplei genotypes have been detected in Senegal, but only 1 common genotype has been detected in Dielmo and Ndiop, even though the villages are 5 km apart (25). All of the other genotypes detected in the Sine-Saloum area were specific to each village, including the 2 that were more prevalent: genotype 52 was detected in 54% of feces samples in Dielmo, and genotype 49 was detected in 28% of feces samples from Ndiop (25).
Several familial cases also occurred during this outbreak. The family in household no. 39 in Dielmo was 1 of the most affected families: 3 of 4 persons living in the home had fever and T. whipplei bacteremia. Genotyping was available for 2 of these patients, both of whom exhibited the same potential genotype. The family in household no. 39 was involved in the management of a traditional oven for preparing bread, which was thoroughly cooked and sold directly to other residents. Since the departure of the baker and his family, no other outbreaks have been observed, and the prevalence of T. whipplei bacteremia has dramatically decreased. Thus, this family may have contributed to spread of the outbreak on a daily basis in Dielmo and possibly on a weekly basis at traditional markets, which served as the main contact between villagers from Dielmo  and Ndiop. Also of note, no toilet facilities were present in household no. 39, and a link between a lack of toilet facilities and the high detection of T. whipplei, mainly in feces, has previously been reported (31). Thus, we hypothesize that T. whipplei was transmitted to customers who bought bread contaminated with infectious feces (31). Overall, all of our data confirm human-to-human transmission of the bacterium (22,23,26,31). One of the main symptoms among febrile patients with T. whipplei bacteremia is cough (36.1%). In our preliminary study of T. whipplei bacteremia, cough was also the main manifestation observed (36). Thus, T. whipplei could be involved in respiratory infections (13,14,36,37). However, the presence of cough in ≈36% of febrile patients who were either T. whipplei-positive or -negative may also suggest that this symptom was poorly specific.
Of note, a 4-year-old patient had 2 febrile episodes associated with T. whipplei bacteremia 18 months apart (8); however, it was not possible to make a distinction between relapse and reinfection because genotyping was not available (38). Blood specimens from this patient that we tested for T. whipplei before and after the last febrile episode were negative, confirming that the infection was acute. Thus, these data suggest that some patients may have several febrile episodes linked to T. whipplei.
T. whipplei bacteremia cannot be diagnosed in tropical regions that lack the proper laboratory facilities or in industrialized countries that lack or do not routinely perform molecular biology-based diagnostics due to the specific training, expensive reagents, and excessive time required to perform such tests. Moreover, even recent studies that have looked for causes of nonmalarial fevers, including by performing molecular detection in blood for intracellular bacteria, such as R. felis, have not included the molecular detection of T. whipplei (39). Thus, it is currently difficult to estimate the prevalence of T. whipplei bacteremia. In conclusion, the results of our large-scale study clearly confirm the role of T. whipplei in febrile episodes as well as its contagiousness and epidemic character.