Population Structure of East African Relapsing Fever Borrelia spp.

B. recurrentis may have evolved directly from B. duttonii.

R elapsing fever is a recurrent febrile infection caused by various Borrelia spirochetes transmitted either by lice (epidemic relapsing fever) or by ticks (endemic relapsing fever). Relapsing fever was once a disease of global epidemic importance. However, largely as a result of the demise of the clothing louse Pediculus humanus, it is now restricted to areas where clothing lice are still commonplace, such as Ethiopia (1). In these regions it still has a major impact and was documented in a recent Ethiopian Department of Health report as being the seventh most common cause of hospital admission (2.5% of total; 3,777 cases) and fi fth most common cause of death (0.9%, 42 cases) (2004) (1). The epidemic tendency of this infection is likely to reside in its vector transmission, with waves of clothing lice fl eeing febrile patients and thus facilitating epidemic transmission. The endemic tick-borne relapsing fever spirochetes are transmitted by Ornithodoros ticks; O. sonrai serve as the principle vector for Borrelia crocidurae in West Africa, and O. moubata complex ticks effectively maintain these spirochetes in East Africa (1). Because soft ticks are only associated with hosts for their typically rapid nocturnal feeding, this limits the spread of infection beyond the confi nes of the areas where tick vectors reside.
Although these differing life styles of the vectors can account for the epidemiologic differences between these infections, recent studies have highlighted the similarity between B. recurrentis, the cause of louse-borne relapsing fever, and B. duttonii, the agent of East African of tickborne relapsing fever (2-4). It has been postulated that B. recurrentis is a louse-adapted variant of B. duttonii (5). In these disease-endemic regions, diagnosis is typically achieved through demonstration of spirochetes in stained blood fi lms from patients. However, this technique is unable to discriminate between the different Borrelia spp. that cause relapsing fever.
To provide a method able to reliably identify these spirochetes, we validated the use of sequence analysis of an intragenic spacer (IGS) region for typing these spirochetes (3). This method was used to analyze a noncoding spacer region and proved to be highly discriminatory; it resolved 4 groups among B. duttonii found among isolates and directly in tick vectors. Two groups were found among B. recurrentis isolates and louse vectors (3). Furthermore, a novel Borrelia species detected previously was found, and some sequence types resembled B. crocidurae, previously believed to be only in West Africa (3,6). The surprising fi nding was that 1 phylogenetic group of B. duttonii overlapped with a group of B. recurrentis (3).
This colinearity was further supported by complete genomic sequencing of 1 isolate of B. recurrentis and 1 of B. duttonii, which suggests that B. recurrentis is a decaying genome that evolved from either B. duttonii or a common ancestral strain (5).
To further explore this apparent overlap between B. duttonii and B. recurrentis, we sequenced and compared additional gene targets. However, this study focused upon coding genes that were under different selective pressure and thus were not necessarily comparable to the noncoding IGS previously used. Furthermore, these investigations used only cultivable strains, and thus could represent bias toward those able to grow under axenic conditions. These investigations disclosed a clear demarcation between B. duttonii and B. recurrentis (4).
To resolve this apparent dichotomy, we undertook additional IGS typing directly on serum samples collected from patients with clinical cases, thus removing the selective pressure of cultivation. We report the results of those investigations.

Clinical Samples
Eighty-eight serum samples from patients in Ethiopia with relapsing fever that were blood fi lm-positive for spirochetes were collected and stored frozen at -20°C before testing. Similarly, 23 samples collected from patients in Tanzania were collected and stored frozen. An additional series of 45 samples from family members accompanying patients to hospital as potential blood donors were available for analysis. Ethical approval for their collection and testing for relapsing fever had been granted for earlier studies (Ethiopia [7] and Commission for Science and Technology

DNA Extraction
Serum samples were centrifuged at 13,000 rpm in a microfuge for 30 min and 100 μL of the pellet was used for DNA extraction. After an initial proteinase K digestion in a waterbath at 56°C for 1 h, DNA extraction was conducted by using DNeasy reagents (QIAGEN, Valencia, CA, USA) according to the manufacturer's protocol for the QIAcube robotic platform with a fi nal elution volume of 200 μL.

Preliminary Screening for Borrelia spp.
A real-time TaqMan fl agellin assay was used to screen samples for the presence of borrelial DNA before investigation. The primers are shown in Table 1. Primers were used at a fi nal concentration of 1,000 mmol/L, and the probe was used at a fi nal concentration of 50 mmol/L. A mastermix composed of PCR working concentration buffer without MgCl 2 (Invitrogen, Carlsbad, CA, USA), 0.2 mmol/L of each dNTP, 5 mmol/L MgCl 2 , 0.075 nmol/L Rox, and 0.06 U Taq polymerase (Invitrogen). Mastermix was divided into aliquots of 23 μL to which 2 μL of extracted DNA was added. A MX3000 thermocycler (Stratagene, La Jolla, CA, USA) was used with an initial denaturation of 95°C for 10 min, followed by 50 cycles each at 94°C for 15 s and 60°C for 60 s. Results were read at a Fam wavelength (emission 516) by using Rox as a reference wavelength (emission 610).

IGS Typing
Samples positive for Borrelia DNA by the fl agellin real-time assay were further subjected to IGS typing. A nested PCR was used to amplify the IGS, with outer primers anchored within the 3′ end of the rrs ribosomal gene and the 5′ end within the ileT gene as used previously (3). Primer sequences are shown in Table 1. Conventional PCR was performed by using 25-μL reaction volumes in My Cycler thermocylers (Bio-Rad, Hercules, CA, USA) with an initial denaturation at 94°C for 3 min, followed by 35 cycles at 94°C for 30 s, 55°C for 30 s, and 74°C for 60 s, and a fi nal extension at 72°C for 7 min. A 2-μL volume from the fi rst-round reaction was added to 23 μL of fresh second-round mastermix. This mixture was heated at 94°C for 3 min, followed by 35 cycles at 94°C for 30 s, 60°C for 30 s, and 74°C for 60 s, and a fi nal extension at 72°C for 7 min. Resulting amplicons were resolved by electrophoresis on 1% agarose gels stained with ethidium bromide (0.5 μg/ mL). Different manipulations of the PCR were conducted in different laboratory areas to avoid contamination, and nontemplate controls were included at a ratio of 1 for every 15 samples tested. Resulting amplicons were subjected to sequence analysis on an ABI 3700 automated DNA sequencer (Applied Biosystems, Foster City, CA, USA) at the Genome Centre Queen Mary's University, London. The inner nested primers were additionally used for sequencing primers at a concentration of 10 pmol/μL. Sequence results were aligned by using ClustalW (www.ebi.ac.uk/clustalw) trimmed and a neighbor-joining phylogenetic tree was constructed by using MEGA4 (8).

Results
Most samples collected from patients with either louseborne (all 88 serum samples) or tick-borne relapsing fever (21/23 samples), yielded a positive PCR result upon initial screening and subsequently when tested for IGS. The 2 samples positive by initial microscopy, but negative by PCR, may represent DNA degradation over the prolonged period of frozen storage before testing. Additionally, 15 of 45 blood donors who accompanied patients in Tanzania had positive test results upon screening; samples from 9 of these also produced IGS amplicons. Nontemplate controls showed negative results.
IGS sequence data were obtained for 90% of the 124 PCR screen-positive samples (18 from patients with tickborne relapsing fever, 85 from patients with louse-borne relapsing fever, and 9 from blood donors in Tanzania) and used for phylogenetic analysis. Only 5 of the 34 identical B. recurrentis type I sequences (represented by B. recurrentis A1 DQ000277) and 7 of the 43 B. recurrentis type II sequences (represented by B. recurrentis A11 DQ000278) are shown in Figure. Similar to results found when looking at cultivable strains and IGS data in ticks or lice (3), we were able to show 4 clades among B. duttonii and 2 for B. recurrentis (Figure). Two of the B. duttonii clades contained only sequence types from patients in Tanzania or blood donors. Notably, all of the cultivable strains of B. duttonii resembled sequence type I. The other previously designated type IV (3), represented by B. duttonii WM (DQ000282) that had been represented by a single patient sequence, was found in 2 additional patient samples and in 1 serum sample from a blood donor. The remaining 2 B. duttonii clades overlapped with the 2 B. recurrentis clades. The accession numbers for the sequences types shown in Figure are deposited under GenBank nos. GQ401243-GQ401278 (Table 2).

Discussion
The fi ndings of this study build upon those previously reported (3). As before, on the basis of IGS sequence types, we identifi ed 4 distinct clusters of B. duttonii. Thus, the fi ndings of this study, which analyzed sequence types in clinical samples, generally mirror results found when in-vestigating sequence types of bacteria in arthropod vectors (3). The notable exception to this was the absence of the novel Borrelia sp. previously detected by Scott et al. in ticks (3) and by others in ticks and human samples (6,9). Whether detection of this species in humans with fever was directly related to the fever or incidental remains to be assessed through properly structured epidemiologic investigations. We did not detect this species among samples collected from patients with clinical relapsing fever in our study population. In consequence, our fi ndings do not support the conclusion that this species has the potential to cause human infection.
Our samples from Tanzania were collected from villagers living in traditional mud-built dwellings; these dwellings usually have >80% infestation rates with O. moubata ticks. Regular nocturnal feeding by ticks could account for effi cient transmission of these spirochetes to humans. Consequently, it is not surprising that samples from these persons, or from the asymptomatic blood donors who lived  in similar conditions, were positive for B. duttonii. The absence of evidence of cross-contamination of the assays further substantiates these fi ndings. Others have reported microscopy positive samples and PCR positive results from asymptomatic persons in Tanzania (6), which lends further support to this fi nding. Whether these observations represent persistent infection, as seen in the related spirochete that causes Lyme borreliosis, or refl ects continued exposure by feeding ticks, remains to be elucidated.
Conversely, in Ethiopia, data are not available concerning the prevalence of soft tick species; however, infestation with clothing lice is commonplace (10). Based on the vector prevalence, in Ethiopia, relapsing fever has always been presumed to result from B. recurrentis infection. A subset of serum samples included in the current study were found to harbor a premature stop codon in their recA gene (S.J. Cutler et al., unpub. data), a feature disclosed in B. recurrentis, but not B. duttonii, in the recently completed whole genome sequencing of these spirochetes (5). This fi nding provides further evidence for the identifi cation of infecting spirochetes among these samples.
The distribution of IGS profi les found among Tanzanian blood donors did not appear to differ from that found among persons with clinical cases in this region. The apparent absence of associated clinical signs is likely to correlate with a degree of local immunity. This fi nding is the basis of much local folk lore on the need to have regular tick exposure to boost immunity (11). These spirochetes undergo multiphasic antigenic variation (12), and whether they would continue to cause asymptomatic infection if the spirochetes underwent antigenic variation remains speculative. Furthermore, these organisms are able to evade key components of the innate immune response (13). This evasion might help account for their presence in asymptomatic persons.
Strikingly, the IGS sequence type for the cultivable isolates obtained for B. duttonii all fell into a single clade in which only B. duttonii sequence types were represented. This fi nding would explain the distinct clustering previously reported that separated B. duttonii from B. recurrentis when a multigene sequence-based approach was used (4). It may be that this particular cluster of B. duttonii is more readily cultivable that those belonging to other clades, thus highlighting the potential for misinterpretation of population structure analysis that is based only on cultivable isolates.
We had reported that B. recurrentis fell into 2 clades, which we can reconfi rm with our current in situ fi ndings (3). Of the samples tested from patients in Ethiopia with louse-borne relapsing fever, 44% fell into the type I clade and 56% were type II. Interestingly, our earlier work showed that B. duttonii type II represented by tick sample MA/15 (DQ000280) had a IGS sequence identical to those found for B. recurrentis type I from Ethiopia. Our current fi ndings have provided further support for this observation by showing that another B. duttonii type III represented by tick sample MA/18 (DQ000281) aligns more closely with B. recurrentis IGS types. Further examples of IGS sequence homology were found between spirochetes that cause louse-borne relapsing fever and those that cause tickborne relapsing fever, in samples from 2 patients from Tanzania with tick-borne relapsing fever (Bd4 and Bd14) and also from 2 blood donors from Tanzania (Tz blood donor 6 and Tz blood donor 19); the isolates clustered among B. recurrentis type II. The vectorial specifi city of these spirochetes has been a source of curiosity for many years. Early investigations questioned host and vector specifi cities of relapsing fever spirochetes; studies were carried out that enabled ticks to feed on patients with the louse-borne form of the disease and subsequently letting these ticks feed upon nonfebrile persons. These studies were unable to conclusively establish whether alternative arthropod vectors could transmit these spirochetes. Haberkorn demonstrated that B. crocidurae isolated from O. erraticus successfully multiplied in lice (14). The investigations by Heisch and Garnham initially claimed successful transmission of B. duttonii by body lice (15). However, in a later study Heisch concluded that B. duttonii from humans is unlikely to be naturally transmitted by lice (16). Many of these early investigations were hampered by the inability to reliably identify the organisms used, lack of cultivable strains, and having no means of assessing the immune status of recipient hosts. Speciation of the causative relapsing fever spirochete was based on the 1 vector-1 species criteria that relied on geographic region where infection occurred and identifi cation of the vector responsible for transmission.
Collectively, these fi ndings support the increasing evidence that B. recurrentis has evolved either directly from B. duttonii or from a common ancestral strain. This raises the question of whether these spirochetes should be considered as separate species or as ecotypes of the same species? With the advent of molecular typing tools, should the traditional classifi cation of relapsing fever spirochetes be readdressed? A precedent for this approach already exists with other highly related organisms such as Bacillus anthracis, B. cereus, and B. thuringiensis (17). Although IGS typing provides a highly discriminatory tool to type these spirochetes, these observations should be underpinned by sequence analysis of multiple gene targets to provide a robust phylogenetic analysis.