Molecular Evidence for Flea-Borne Rickettsiosis in Febrile Patients from Madagascar

Rickettsiae may cause febrile infections in humans in tropical and subtropical regions. From Madagascar, no molecular data on the role of rickettsioses in febrile patients are available. Blood samples from patients presenting with fever in the area of the capital Antananarivo were screened for the presence of rickettsial DNA. EDTA (ethylenediaminetetraacetic acid) blood from 1020 patients presenting with pyrexia > 38.5 °C was analyzed by gltA-specific qPCR. Positive samples were confirmed by ompB-specific qPCR. From confirmed samples, the gltA amplicons were sequenced and subjected to phylogenetic analysis. From five gltA-reactive samples, two were confirmed by ompB-specific qPCR. The gltA sequence in the sample taken from a 38-year-old female showed 100% homology with R. typhi. The other sample taken from a 1.5-year-old infant was 100% homologous to R. felis. Tick-borne rickettsiae were not identified. The overall rate of febrile patients with molecular evidence for a rickettsial infection from the Madagascan study site was 0.2% (2/1020 patients). Flea-borne rickettsiosis is a rare but neglected cause of infection in Madagascar. Accurate diagnosis may prompt adequate antimicrobial treatment.

Some, but not all rickettsiae harbored by arthropods are well-known human pathogens, and their pathogenicity differs from species to species. R. rickettsii, the agent of Rocky Mountain spotted fever (RMSF), is endemic in the Americas and is transmitted by ticks of the genera Dermacentor and Rhipicephalus. RMSF is probably the most lethal rickettsiosis, with a case-fatality rate of >35% when treatment with doxycycline is delayed [3]. In Brazil, mortality of RMSF may exceed 50% in some clusters [4]. On the African continent, entomological studies have confirmed the presence of rickettsial DNA in several different tick and flea species [5][6][7]. R. africae, which infests Amblyomma ticks to a percentage of >90%, is the causative agent of African tick-bite fever (ATBF). ATBF is often reported in travelers to Sub-Saharan Africa and usually manifests as a benign disease [8].
Fleas are typical vectors for R. typhi and R. felis. While R. typhi is mainly found in the rat flea Xenopsylla (X.) cheopis, R. felis has been detected in several flea species, including the cat flea Ctenocephalides felis, X. cheopis and the human flea Pulex irritans [6]. Molecular studies from Northern Africa have reported clinical cases of murine typhus [9,10]. Seroepidemiological reports from Sub-Saharan Africa suggest that up to 11% of the population are exposed to R. typhi, e.g., in Mali [11,12]. The mortality of murine typhus is very low (<1%) [13]. R. felis was identified as a potential cause of febrile illness in 1-15% of patients with fever from some African countries, e.g., Ghana and Senegal [14,15]. However, R. felis has also been found in blood samples of non-febrile control subjects, albeit at a significantly lower rate than in febrile patients [16,17]. Consequently, there is some controversy about the pathophysiology of R. felis infections and its role as an obligatory pathogen.
Clinically, spotted fever group rickettsioses often present with fever and an efflorescence called "eschar", a cutaneous ulcerative lesion at the site of tick bite [18]. Typhus group rickettsioses such as murine typhus (R. typhi) and epidemic typhus (R. prowazekii) are not associated with eschar formation and are thus more difficult to diagnose. Murine typhus presents with a "classical" triad of fever, headache and rash, which is encountered in 33% of patients [19]. R. felis has been associated with similar symptoms as murine typhus [14,20].
Until recently, the importance of rickettsial infections in Madagascar has remained largely unknown. High percentages of Amblyomma variegatum (86.5%) and Amblyomma chabaudii ticks (100%) were found to be infested with Rickettsia (R.) africae [7,21]. Xenopsylla cheopis fleas, collected from small mammals in urban Malagasy areas, tested positive for R. typhi and R. felis, while Pulex irritans fleas from the same area were positive for R. felis [6]. Although it was shown that humans carry antibodies against both typhus group (TG) and spotted fever group (SFG) rickettsiae [6,21], direct molecular evidence for rickettsial infections in humans from Madagascar is so far lacking. A recent pan-Madagascan study that enrolled 682 febrile patients failed to detect rickettsial DNA by applying a novel macroarray system [22]. Here, we present the results of a real-time PCR-based screening for rickettsial DNA in EDTA blood of febrile patients from Madagascar.

Results
The patient population, a subset from the Typhoid Fever Surveillance in Africa Program (TSAP) study, was previously characterized in detail by Boone et al. [23]. Briefly, 1020 samples from patients living in a rather rural environment (Imerintsiatosika) or in urban slums (Isotry, central Antananarivo) and presenting with pyrexia ≥38.5 • C were selected for the present study. From 1,020 EDTA samples screened for rickettsial DNA by gltA-specific qPCR, five samples showed amplification in at least one of two replicate runs. Of those, 2/5 samples (#4133, #4196) were confirmed by repeated testing using the same target.
To identify the rickettsial species, gltA amplicons were sequenced applying Sanger sequencing. Excluding primer regions, a sequence of 116 bp was analyzed (for gltA sequences, see Tables A1 and A2). The two positive samples differed in nine positions. Compared to published sequences in the NCBI database, the gltA amplicon of sample #4133 showed 100% homology to several reference strains of R. typhi, while sample #4196 showed 100% homology to R. felis. The phylogenetic analysis, which included other closely related Rickettsia species, confirmed this identification ( Figure 1). To identify the rickettsial species, gltA amplicons were sequenced applying Sanger sequencing. Excluding primer regions, a sequence of 116 bp was analyzed (for gltA sequences, see Tables A1, A2). The two positive samples differed in nine positions. Compared to published sequences in the NCBI database, the gltA amplicon of sample #4133 showed 100% homology to several reference strains of R. typhi, while sample #4196 showed 100% homology to R. felis. The phylogenetic analysis, which included other closely related Rickettsia species, confirmed this identification ( Figure 1). The Ct values of > 35 from the two blood samples suggested a low concentration of bacterial target DNA (Table 1). Thus, the molecular typing and confirmation approaches were different in both cases. The amplification of a larger partial ompB sequence (amplicon size: 855-861 bp) [24] was successful only for sample #4133 (in one sequencing direction). Sanger sequencing of the amplicon (GenBank OL310470) showed 100% homology with R. typhi, thereby confirming the gltA sequence result. The Ct values of >35 from the two blood samples suggested a low concentration of bacterial target DNA (Table 1). Thus, the molecular typing and confirmation approaches were different in both cases. The amplification of a larger partial ompB sequence (amplicon size: 855-861 bp) [24] was successful only for sample #4133 (in one sequencing direction). Sanger sequencing of the amplicon (GenBank OL310470) showed 100% homology with R. typhi, thereby confirming the gltA sequence result.
A larger ompB amplicon could not be obtained from sample #4196. Instead, we utilized the intergenic spacer-typing approach by Fournier et al. to confirm the gltA sequencing result [25]. The two intergenic spacers dksA-xerC (amplicon size: 164 bp, see Table A2) and rpmE-tRNAfMet (amplicon size: 351 bp, GenBank OL310471) were successfully sequenced from sample #4196. Both showed 100% homology with R. felis. The identification of the rickettsial species in sample #4196 as R. felis was further confirmed by a R. felis-specific ompB qPCR [26] (Table 1).
Both patients with molecular confirmation of Rickettsia spp. in their blood samples were from the urban study site in Antananarivo. The R. typhi patient was a 38-year-old female from the Manarintsoa-Isotry district in central Antananarivo, while the R. felis patient was a 17-month-old infant from Antohomadinika Sud, another central area in Antananarivo. In both individuals, a fever of > 39 • C and a cough were noted, but no rash (Table 1). These results provide a molecular confirmation of human-pathogenic rickettsiae for the first time in febrile patients from Madagascar. Only the flea-borne species R. typhi and R. felis were found; no tick-borne rickettsiae were identified. Overall, the positive rate for Rickettsia spp. DNA in febrile patients was 0.2% (2/1020 patients) in this study.

Discussion
The high endemicity of plague in Madagascar indicates that the Malagasy population is exceptionally exposed to flea-borne infections [27,28]. However, little is known about the regional epidemiology of flea-borne rickettsioses. Apart from a single case of an R. typhi infection in a French traveler returning from Madagascar [29], there has only been indirect serological evidence that humans may become infected [6]. Molecular data from fleas that showed infection with rickettsiae suggested that there might be a risk of transmission to humans [6]. The present study shows for the first time, based on a real-time PCR screening for rickettsial DNA in febrile patients, that human-pathogenic, flea-borne rickettsiae are transmitted to humans in Madagascar.
Overall, the detection rate of rickettsial DNA in whole blood samples was low in this study (2/1020 patients). While we cannot exclude that storage and transport conditions may have affected the stability of target DNA in our study, previous investigations showed that Rickettsia-specific PCRs from blood samples may yield false-negative results in serologically confirmed cases [30,31]. A systematic review reported a blood DNA median sensitivity of 18% (6-27%) for pan-Rickettsia PCRs (SFG: 42% [6-69%]; TG: 3% [1-18%]) [32]. Thus, we may have missed some rickettsial infections by our approach. Importantly, the high Ct values identified here should not be interpreted as a sign of low clinical relevance, since they reflect the transient nature of these pathogens in circulating blood [33]. In sum, it is likely that the actual number of rickettsial infections among febrile patients may be several-fold higher than that detected by PCR.
With 1/1020 samples (0.1%) testing positive for R. felis, the detection rate in the urban setting of the Malagasy capital is lower than previously reported from continental African countries, where frequencies from 1.5% PCR-positive samples in Ghana to 15% in rural Senegal have been found [14,15]. As in our 1-year-and 5-month-old patients, R. felis is thought to be a common cause of febrile disease in children [14]. It can be detected significantly more often in febrile compared to afebrile patients, but its identification in afebrile pediatric control groups [16,34] has questioned its role as an obligatory pathogen. Another conundrum is the significance of Plasmodium falciparum/R. felis co-infections in regions with high malaria endemicity [14,15]. Notably, in our Malagasy study population, malaria was found only in a very low percentage of febrile patients (0-2%) [35].
R. typhi causes murine typhus, an infection with low mortality, and is transmitted worldwide, mainly in suburban areas [36]. Epidemiological studies indicate its transmission to humans in African countries [11,12,37], and direct molecular evidence from fever studies has primarily been provided from the northern African countries Algeria and Tunisia [38].
Infections with R. typhi and R. felis are associated with similar clinical presentations. Distinguishing them is possible by molecular diagnostics [26,39], but difficult by serology.
In both infections, typical symptoms include fever and cough, as in our patients, and the absence of rashes often complicates the clinical diagnosis [14,36].
The present study found no evidence for spotted fever group rickettsiae in febrile patients. While according to seroprevalence studies, transmission of SFG rickettsiosis occurs in Madagascar [21], these infections are likely to only rarely present as bloodstream infections and may therefore be difficult to diagnose by applying molecular diagnostic approaches with EDTA blood samples.
While it was not determined in the present study whether fleas from the specific patients' areas of residence carry R. felis or R. typhi, a similar study was performed in urban areas of the Tsiroanomandidy district, about 200 km west of Antananarivo. Here, 26/117 (24%) Xenopsylla cheopis fleas collected from small mammals tested positive for R. typhi, and 2/117 (2%) were positive for R. felis. In 8/26 (31%) Pulex irritans fleas, R. felis was detected. No co-infections with both species in the same flea were reported. It is highly likely that rodents in Antananarivo carry fleas with a similarly high proportion of Rickettsia infection, thus posing a risk for transmission of R. felis and R. typhi to humans.
In future research, a prospective study design may be appropriate when investigating rickettsial infections in Madagascar. A comprehensive approach that takes into account clinical clues for rickettsial infections (e.g., eschars or rashes), serological assays (e.g., IgM/IgG against several rickettsial antigens) and immediate sample processing for PCR (e.g., from buffy coat) may provide a more complete picture of spotted and typhus fever rickettsioses.

Human Blood Sample Collection
EDTA blood samples were collected within the framework of the Typhoid Fever Surveillance in Africa Program (TSAP) [40]. A detailed description of how samples were collected is provided elsewhere [23]. In brief, 4500 EDTA blood specimens were collected from febrile patients in Madagascar between 2011 and 2013 continuously throughout the year. Of those, a total of 1,020 samples from individuals presenting with a body temperature ≥38.5 • C were screened for the presence of rickettsial DNA. Samples were previously tested for the presence of Burkholderia pseudomallei, Coxiella burnetii, Brucella spp., Leptospira spp. and Borrelia spp. DNA, with a detection rate of 1.5% for Brucella spp. and otherwise negative results [23,41].

DNA Extraction and Real-Time PCR
Total DNA was extracted from 1 ml of human EDTA blood samples using the Flexigene DNA Kit (Qiagen, Hilden, Germany), following the manufacturer's protocol. Rickettsial DNA was screened for by gltA-specific real-time PCR, using a modification of a previously published protocol based on the primers RKND03_F (5 -GTGAATGAAAGATTACACTATTTAT-3 ) and RKND03_R (5 -GTATCTTAGCAATCATTCTAATAGC-3 ), which yields an amplicon of 166 bp (base pairs) [42,43]. In a total of 10 µL, reactions contained 1x reaction buffer (Invitrogen, Darmstadt, Germany), 2.5 mM magnesium chloride, 0. When a typical amplification curve was found, the result was confirmed in two independent, ompB-specific real-time PCRs for R. typhi and R. felis that allow detection of each species by a molecular beacon [26]. Samples showing amplification in both gltA and ompB assays were interpreted as positive for Rickettsia spp.

Conclusions
Taken together, the present study provides direct molecular evidence for both R. typhi and R. felis in blood samples of febrile patients from Madagascar. We suggest that flea-borne rickettsial infections are a relevant differential diagnosis of fevers of unknown etiology in a small number of patients and should be considered when deciding on empirical antimicrobial therapy by healthcare professionals in Madagascar.

Funding:
The TSAP project was based on research funded by the Bill & Melinda Gates Foundation (OPPGH5231) [40]. In addition, this research received no external funding.

Institutional Review Board Statement:
The study was conducted according to the guidelines of the Declaration of Helsinki. The study protocol was approved by the Malagasy Ethical Committee (no. 088 MSANP/CE from 2014/09/25, amendment to approval 045/MSANP/CE from 2011/09/11) and the IVI Institutional Review Board (no. IVI IRB #2011-001).
Informed Consent Statement: Written, informed consent was obtained from all patients involved in the study.
Data Availability Statement: All relevant data are provided in the manuscript. The ompB sequence have been deposited at NCBI GenBank under accession number OL310470, the rpmE-tRNAfMet sequence is available under accession number OL310471.