Leishmania aethiopica Field Isolates Bearing an Endosymbiontic dsRNA Virus Induce Pro-inflammatory Cytokine Response

Background Infection with Leishmania parasites causes mainly cutaneous lesions at the site of the sand fly bite. Inflammatory metastatic forms have been reported with Leishmania species such as L. braziliensis, guyanensis and aethiopica. Little is known about the factors underlying such exacerbated clinical presentations. Leishmania RNA virus (LRV) is mainly found within South American Leishmania braziliensis and guyanensis. In a mouse model of L. guyanensis infection, its presence is responsible for an hyper-inflammatory response driven by the recognition of the viral dsRNA genome by the host Toll-like Receptor 3 leading to an exacerbation of the disease. In one instance, LRV was reported outside of South America, namely in the L. major ASKH strain from Turkmenistan, suggesting that LRV appeared before the divergence of Leishmania subgenera. LRV presence inside Leishmania parasites could be one of the factors implicated in disease severity, providing rationale for LRV screening in L. aethiopica. Methodology/Principal Findings A new LRV member was identified in four L. aethiopica strains (LRV-Lae). Three LRV-Lae genomes were sequenced and compared to L. guyanensis LRV1 and L. major LRV2. LRV-Lae more closely resembled LRV2. Despite their similar genomic organization, a notable difference was observed in the region where the capsid protein and viral polymerase open reading frames overlap, with a unique −1 situation in LRV-Lae. In vitro infection of murine macrophages showed that LRV-Lae induced a TLR3-dependent inflammatory response as previously observed for LRV1. Conclusions/Significance In this study, we report the presence of an immunogenic dsRNA virus in L. aethiopica human isolates. This is the first observation of LRV in Africa, and together with the unique description of LRV2 in Turkmenistan, it confirmed that LRV was present before the divergence of the L. (Leishmania) and (Viannia) subgenera. The potential implication of LRV-Lae on disease severity due to L. aethiopica infections is discussed.


Introduction
In the highlands of Ethiopia, patients infected with L. aethiopica mostly develop localized cutaneous lesions, which can self heal. There is no accurate national figure for the overall burden of cutaneous leishmaniasis (CL) in Ethiopia, although some estimates suggest annual incidence to range from 20 to 50 thousands [1]. The prevalence varies from location to location, mostly being sporadic or endemic. In some cases, CL tends to persist and to metastasize to other parts of the body causing mucosal leishmaniasis (ML) or diffuse cutaneous leishmaniasis (DCL), two different clinical presentations. DCL starts with a cutaneous lesion that metastasizes to other cutaneous sites. Patients are anergic in response to parasite antigens and poor responders to treatment. ML patients also begin with a single lesion, but in this case infection metastasizes to mucosal tissue causing chronic inflammation and facial disfiguring lesions. The mechanisms underlying the ability of L. aethiopica infections to cause such pathologies are not known. The parasite genetic variability does not account for the variability in clinical presentations [2].
In South America, leishmaniases are major health problems represented by a spectrum of pathological manifestations leading to clinical forms ranging from cutaneous leishmaniasis (CL), visceral leishmaniasis (VL), mucosal (ML) or disseminated leishmaniasis (DL) depending on the infecting species [1]. ML and DL are caused primarily by infections with species of the Leishmania subgenus Viannia (e.g. Leishmania braziliensis, L. panamensis or L. guyanensis). One of the key questions concerning ML and DL is the basis by which some isolates are associated with metastasis in human infections. It is likely that these outcomes reflect polygenetic factors of both the human host and parasite [3][4][5][6][7][8][9][10][11][12][13]. This includes for example parasite resistance to oxidative stress [3], and a variety of host genetic susceptibilities (reviewed in [12]) related to TGF-beta signalling [6], infiltration of phagocytes [7,11] or IL-6 production [8]. In addition, the risk of ML is increased in HIV co-infection [4] and was also found to correlate with the age, gender, nutritional status of the patients as well as duration of the disease [14]. Recently, we reported that, in L. guyanensis mice infection, inflammation and severity of the infection was associated with high burden of a Leishmania RNA virus (LRV) infecting the parasites, suggesting that LRV might also contribute to disease severity in metastasizing leishmanisis [15].
Although first described more than two decades ago [16,17], it was only very recently reported that high burden of LRV, a member of the Totiviridae family, is present in parasites metastatic in hamsters and isolated from secondary lesions of ML patients [15,[18][19][20]. Thus far, these viruses have been described only in L. (Viannia) braziliensis and L. (V.) guyanensis, with the exception of one L. major strain from Turkmenistan (Lmj ASKH) [21,22]. On the basis of sequence divergence and organization of the open reading frames, the L. (Viannia) and L. major viruses have been named LRV1 and LRV2 respectively. The viral particles are composed of a capsid protein (CP), an RNA-dependent RNA polymerase (RdRp) and a 5.3 kb double stranded RNA (dsRNA) genome. In most Totiviridae, such as in yeast, RdRp is not expressed as a single polypeptide, but as a fusion protein with CP [23]. Although the presence of a fusion protein was not directly proved for LRV in vivo, in vitro translation experiments conducted on LRV1 from Lg M4147 shows that the overlapping region of the two ORFs promotes translational frameshifting [24]. Another hypothesis is that an individual RdRp protein, as observed using a specific antibody, is produced after clevage of the potential fusion protein [25]. As demonstrated in the Totiviridae from Helminthosporium, RdRp could also be produced as a single protein after a termination/reinitiation mechanism [26].
In the L. guyanensis infection model, the dsRNA genome is recognized by the host endosomal Toll-like receptor 3 (TLR3) to induce pro-inflammatory cytokines and chemokines [15]. These TLR3-mediated immune responses render mice more susceptible to infection, and the animals develop an increased footpad swelling and parasitemia. Thus, LRV1 in L. guyanensis parasites subverts the host immune response to Leishmania and promotes parasite persistence.
In this study, we hypothesized that, because of the severity of the clinical presentations observed in L. aethiopica infections, a Leishmania RNA virus (LRV) could be present in some L. aethiopica strains. We showed that a virus related to L. major LRV2 was widespread and can evoke cytokine responses similar to those seen previously with LRV1 from L. (Viannia) [15]. This sets the stage for future studies looking at the role of LRV2 in the severity and nature of human leishmaniasis.

Parasite strains and cultures
Two L. guyanensis clones, designated here as Lg M4147 LRV1+ and Lg M4147 LRV12, and which were previously shown to be highly-and non-infected by LRV respectively (designated as LRV high and LRV neg in [27]), were used as reference parasites.
Eight strains of L. aethiopica parasites were freshly isolated from infected patients who contracted leishmaniasis in Ethiopia (Table 1, fresh isolates). In addition, three L. aethiopica lines from Leishmania species reference centers were also used ( Table 1, cryobank lines), and kindly provided by Charles Jaffe and Lee Schnur (Jerusalem, Israel).
All parasite strains were cultivated as promastigotes at 26uC in Schneider's insect medium (Sigma) supplemented with fetal bovine serum, Hepes, penicillin/streptomycin, biopterin and hemin as described before [28].

Detection of dsRNA by dot blot and immunofluorescence microscopy
Viral dsRNA genome was detected using the J2 monoclonal mouse antibody (English & Scientific Consulting) as described before [28]. Briefly, approximately 5610 5 stationary phase promastigotes (2 mg of total proteins by BCA quantification) were directly spotted on a nitrocellulose membrane for dot blot analysis. After blocking with milk, the membrane was incubated with the J2 antibody (1:1000) that was finally recognized by an anti-mouse IgG antibody coupled to peroxidase (Promega).
For immunofluorescence microscopy (IFM), stationary phase promastigotes were fixed in formaldehyde before being attached to poly-lysine coated slides. After permeabilization and blocking steps, slides were incubated with the J2 antibody (1:800), which was then visualized using a goat anti-mouse IgG coupled to AlexaFluor 488 (1:600, Invitrogen).

Viral dsRNA extraction from nucleic acids
Total nucleic acids from stationary phase promastigotes were obtained either by standard Trizol (Invitrogen) protocol or after lysis in sarkosyl followed by proteinase K and ssRNase incubation as described before [28]. After phenol-chloroform extraction and precipitation of total nucleic acids (containing genomic parasitic DNA and LRV dsRNA), parasite DNA was eliminated by a RQ-DNase treatment (Promega or Invitrogen) and LRV dsRNA was visualized on 0.8% agarose gel by staining with ethidium bromide, and purified from the gel for further cDNA preparation (below). To decrease intensity of ethidium bromide stained rRNA, and thus better visualize LRV dsRNA from the L. aethiopica L494 strain, total cellular RNA (after Trizol extraction) was treated with two volumes of FFS buffer (7.5M formaldehyde, 20% formamide, 0.2M sodium chloride, 30% glycerol and 0.02% bromphenol blue) at 37uC for 15 min.

Author Summary
Leishmania RNA virus (LRV) has been detected in Leishmania (Viannia) braziliensis and guyanensis species, parasites causing not only cutaneous but also mucosal and disseminated leishmaniases. In a mouse model, the viral dsRNA genome within L. guyanensis parasites is recognized by host Toll-like receptor 3 (TLR3) and induces proinflammatory cytokines and chemokines, typically IL-6 and TNF-a, which are hallmarks of human mucosal leishmaniasis. Metastatisic complications such as mucosal and diffuse cutaneous leishmaniasis have also been described in other parts of the world, e.g. in Ethiopia. We detected LRV within L. aethiopica human isolates. Sequencing of three L. aethiopica LRVs (LRV-Lae) genomes confirmed that LRV-Lae belongs to the same Totiviridae family of LRVs found in South American species (LRV1) and present in a single L. major isolate from Turkmenistan (LRV2). LRV-Lae genomic organization is similar but not identical to the other LRVs, with a unique 21 frameshift situation in the overlapping region of the capsid protein/polymerase genes. Finally and similarly to L. guyanensis LRV1, LRV-Lae induced a TLR3-dependent inflammatory response in infected macrophages. The presence of LRV and its detection could be a crucial step towards better understanding the pathology spectrum of L. aethiopica infections.

Viral genomes sequencing and comparison
The LRV-Lae L494 sequence was obtained by combination of total small RNA sequencing and specific-primer sequencing from cDNA. A library of small RNAs (,42 nt) was generated from total RNA (purified with Trizol, described above) using the method described by Atayde and co-workers [29]. Recent genome sequence data of the Lae L147 line reveals an absence of genes required for RNAi, consistent with the evolutionary position of L. aethiopica within the Leishmania clade shown previously to lack RNAi [27], and thus the small RNAs represent primarily degradation products (unpublished data), as seen previously in similar studies in Leishmania tarentolae which also lacks RNAi [27,30]. The library was sequenced using Illumina HiSeq2000 technology, yielding 35.9 million reads. The 59 and 39 adapters were trimmed from the data and then mapped to the sequence of the Lae L494 PCR products described above and/or the L. major LRV2. From this analysis, three large contigs were obtained, and the remaining regions of the LRV2 were obtained by PCR amplification across the gaps and sequencing. This strategy allowed us to get the complete 5193 bp LRV-Lae L494 genome sequence (GenBank accession number: KF757256).
After purification of LRV genomic dsRNA from the infected L. aethiopica 303 and 327 strains (see previous section), it was quantified and used for cDNA synthesis as described before [28]. Different overlapping PCR fragments were progressively amplified from the viral cDNA and further sequenced (by Fasteris, Switzerland) to finally obtain most of the viral genomic sequence (5048 bp), including the complete open reading frames encoding the capsid protein (CP) and the RdRp (GenBank accession numbers: KF256264 and KF256265). PCR was performed as previously described [28] with an annealing temperature adapted to each set of oligonucleotides that were used (generally about 2uC below the lowest melting temperature of both primers used). All primer sets (Microsynth, Switzerland) that were used for LRV sequencing, detection and cDNA are listed in Table S1.

Macrophage in vitro infection and cytokine production analysis by ELISA
Bone marrow derived macrophages (BMM) were obtained from C57BL/6 and TLR3 knock-out mice, and infected by Leishmania promastigotes as described before [15]. Culture supernatants were collected 24 h post-infection and analyzed for IL-6 and TNF-a cytokine production. For this purpose, ELISA kits were purchased from eBioscience and read on a Biotek Synergy HT spectrophotometer. Cytokine production was quantified relatively to purified mouse IL-6 and TNF-a standards. The number of parasites per macrophage was counted after fixation with formalhehyde followed by DAPI staining (as described in [28]). Four different pictures from each experiment were used for counting (at least 90 macrophages).

Detection of LRV in L. aethiopica human fresh isolates
We surveyed eight strains of L. aethiopica freshly isolated from infected patients who contracted leishmaniasis in Ethiopia, six exhibiting typical CL and two with typical DCL pathologies ( Table 1, fresh isolates). LRV presence was first assessed using a dot blot technique based on a monoclonal antibody (J2), which recognizes specifically dsRNA irrespective of the nucleic acid sequence [31,32]. Here, LRV can be easily detected in minute quantities of whole parasites or from lesion biopsies (e.g. less than 100 parasites or 100 ng RNA extract in highly infected strains) [28]. Briefly, the eight freshly isolated L. aethiopica (Lae) parasites were cultured and then spotted on a nitrocellulose membrane followed by an immunoblotting assay using the J2 antibody. As positive and negative controls, we used two clonal lines shown previously to bear LRV1 (LgM4147 LRV1+) or selected for loss of LRV1 (LgM4147 LRV12) [27,28].
Out of the eight L. aethiopica fresh isolates, four strains showed a detectable level of dsRNA, while all others were found negative ( Table 1). The three strains that showed the strongest dsRNA reactivity (Lae 303, 316 and 327) as well as one negative strain (Lae 372) were selected for further analysis. Figure 1A shows the dsRNA detection by dot blot using the J2 antibody of these four L. aethiopica strains, in comparison to Lg M4147 reference clones. We concluded that Lae 303 and 327 had a higher level of dsRNA than Lae 316 (although weaker than the Lg positive control), while it was undetectable in Lae 372 similarly to the Lg negative control.
In order to demonstrate that the dsRNA detected by dot blot in these three L. aethiopica strains was accompanied by the presence of a dsRNA viral genome, nucleic acids were extracted and treated with DNase and ssRNase, allowing the visualization of viral dsRNA on agarose gels. As expected in the case of Leishmania RNA virus (LRV) infections, dsRNA molecules of approximately 5.3 kb were identified in LRV-positive Lae strains, only in those lines positive in anti-dsRNA tests ( Figure 1B). As previously observed for other infected L. guyanensis and L. braziliensis strains [28], the LRV genome sometimes appears as a doublet in non-denaturing high-resolution gels such as that presented here. Considering the homogeneity between the viral sequences that we obtained so far (including the ones presented below for L. aethiopica), the doublet is unlikely to represent a mixed population of LRV. Instead, we propose that this migration pattern is due to differences in their secondary structures. Currently, it remains an open question, which deserves further analysis. The Lae 303 and 327 parasites harboring high dsRNA levels, as well as the LRV-negative Lae 372 and the Lg controls, were subjected to immunofluorescence microscopy to study LRV localization. As shown in Figure 2, dsRNA was detected as clusters throughout most of the cytosol in Lae 303 (as well as in Lae 327, data not shown), similar to what was seen previously with the Lg M4147 LRV1+ control [28]. In agreement with the prior results, dsRNA reactivity was weaker in Lae relative to Lg M4147 LRV1+ ( Figure 1A), and no reactivity was visible in the LRV-negative Lae 372 promastigotes.
LRV detection in a catalogued L. aethiopica strain Although never described in this Leishmania species neither on the African continent before, LRV was strikingly present in half of the recently isolated parasite strains tested. We therefore wondered if it was a particularity of this sampling, the region where samples were isolated or a general phenomenon. To this purpose, we tested already described L. aethiopica lines from reference centers ( Table 1, cryobank lines). Three strains, isolated more than twenty years ago, were screened for LRV presence by PCR using primers specific for regions conserved across known LRV1 and LRV2 genomes (see Methods and Table S1); one of these is a WHO reference line (Lae L147). A specific fragment was amplified from total cDNA from one of these strains, namely Lae L494 ( Figure 3A and Table 1), and correspondingly, this strain alone exhibited a 5.3 kb dsRNA LRV genome ( Figure 3B). Thus the presence of LRVs was not uncommon in L. aethiopica.

LRV-Lae sequence comparison to LRV1 and LRV2
We determined the dsRNA sequence of the virus infecting L. aethiopica L494 by a combination of random small and specific primer-based RNA (cDNA) sequencing (see Methods and Table  S1). This yielded a complete 5193 nucleotide genome (GenBank accession number: KF757256), designated LRV2-Lae L494 by the revised taxonomy proposed for LRVs as discussed below. This primary sequence was then used to design primers for sequencing most of the dsRNA genome from cDNA of two additional LRV-Lae isolated from Lae 303 and 327 (5048 bp), including the complete open reading frames for the capsid protein (CP) and the viral polymerase (RdRp) (GenBank accession numbers: KF256264 and KF256265) ( Figure 4A).
The LRV-Lae nucleotide sequences were then aligned to those of the three complete LRV genomes in GenBank: LRV1-Lg CUMC1 (formerly LRV1-1), LRV1-Lg M4147 (formerly LRV1-4) and LRV2-Lmj ASKH (formerly LRV2-1) (GenBank accession numbers NC002063, NC003601 and NC002064 respectively). The overall nucleotide sequence identity amongst the three L. aethiopica LRVs ranged from 77 to 85%, while it was 68% with L. major LRV2 and 52-58% with the two L. guyanensis LRV1s ( Table 2, genome column). Thus L. aethiopica LRVs were most closely related to L. major LRV2 at the overall nucleotide and, even more at the amino acid level as discussed below. This close relationship of LRV-Lae with LRV2-Lmj was clearly illustrated by a phylogenetic analysis ( Figure 4B). Significantly, it mirrored the evolutionary relationship of Lae and Lmj parasites [33], consistent with the prevailing view that most members of the Totiviridae are thought to be inherited vertically and that Leishmania RNA viruses show relationships similar to their host [22]. For these and reasons evident below, we have assigned the L. aethiopica LRVs as new members of the LRV2 species of Leishmania RNA virus (LRV2-Lae).
Conceptual translation of the LRV2-Lae genomes revealed the presence of two long and overlapping open reading frames (ORFs) coding for a capsid protein (CP) and an RNA-dependent RNA polymerase (RdRp) similarly to previously described LRVs ( Figure 4A). The position of these ORFs in the viral genome and the size of the encoded proteins are strikingly similar to what was observed in LRV2-Lmj, if we admit that LRV capsid starts at an internal AUG (position 341, Figure 4A) and not at an upstream in-frame AUG, which is present in two of the LRV2-Lae but not in LRV2-Lmj (position 248). Analysis of LRV2-Lae303 sequence further supported our hypothesis that the AUG at nucleotide 248 is unlikely to be used, since it is followed by an in-frame stop codon upstream of the AUG at nucleotide 341 (GenBank KF256264). Therefore, in the discussion below, we took this shorter predicted protein as the LRV2-Lae CP.
Similarity between the LRV2 genomes also applied to an additional short ORF located upstream of the CP gene, that potentially encoded a 39 amino acid peptide highly conserved in both LRV2-Lae and LRV2-Lmj (85% identity). In contrast, the upstream short ORFs that were described in LRV1 genomes were not conserved ( Figure 4A). Whether such ORFs are translated into protein is still unknown.
The amino acid sequences of the CP and RdRp from the three LRV2-Lae were then compared to the three available LRVs. As expected, and even more strikingly than the genome analysis, both LRV proteins from the L. aethiopica strains were clearly more homologous to their counterpart in L. major (sharing 80% and 60-61% identical residues for CP and RdRp respectively), than to the South American LRV1s (with only 36-41% of the residues being conserved for both proteins) ( Table 2, CP and RdRp columns). These genome analysis also revealed that RdRp showed more diversity than CP, even between closely related strains such as the   (Table S1). As a cDNA quality control, a 372 bp fragment of the beta-tubulin gene was also amplified. B. LRV dsRNA visualization. Total RNA was analyzed on agarose gel. Ribosomal RNA (rRNA) and the complete 5.3 kb LRV genomic dsRNA are indicated. doi:10.1371/journal.pntd.0002836.g003 three LRV2-Lae, with the exception of certain highly conserved central domains (Figures S1 and S2). These include six regions that were previously reported to be conserved among various Totiviridae, three of them having been directly shown as critical for polymerase activity from the S. cerevisiae L-A totivirus [34][35][36] ( Figure S3). From the CP and RdRp alignments, phylogenetic trees were constructed, again clearly dividing the LRVs into two separated groups, the New and Old World species accordingly to their parasite hosts ( Figure 4C-D).

A unique CP/RdRp frameshift in LRV2-Lae
In LRV1-Lg M4147 and LRV1-Lg CUMC1, the open reading frames (ORFs) for CP and RdRp overlap over 71 nucleotides with a 21 frameshift ( Figure 5). A similar organization was seen in LRV1s isolated from other Viannia subgenus species (manuscript in preparation). In contrast, LRV2-Lmj CP and RdRp are encoded by non-overlapping ORFs that are in-frame [21]. LRV2-Lae showed a third pattern, where the reading frames overlaped by 46 nucleotides, but now with a +1 frameshift ( Figure 5). Potentially, LRVs are particularly diverse in the mechanisms used despite their close relationships. If RdRp is produced as a fusion protein with CP (as in yeast [23]), this would occur through a nonconserved mechanism of either translational frameshifting (+1 for LRV1 or 21 for LRV2-Lae), or via ribosomal hopping as in the case of in-frame ORFs of LRV2-Lmj. However it is important to note that no evidence has been provided yet establishing the existence of LRV CP-RdRp fusion proteins in vivo (unpublished data and [21,24,25]). Alternatively, RdRp could be produced as a single protein, as has been observed in Helminthosporium virus, which is undertaken by a termination/ reinitiation mechanism [26].

LRV2-bearing L. aethiopica induces TLR3-dependent cytokine induction in macrophage infections
Previously, we showed that LRV1-bearing L. guyanensis induced a TLR3-dependent hyperinflammatory response in in vitro infected macrophages, characterized by elevated expression of a suite of cytokines [15]. We performed similar studies here, focusing on two representative important inflammatory cytokines, IL-6 and TNFa, by measuring their release into supernatants of Lae-infected macrophages. Similar to L. guyanensis, the infection with the two L. aethiopica strains harboring the highest levels of LRV (Lae 303 and 327) yielded significantly elevated levels of both cytokines. This was dependent on TLR3 signalling, as shown by infection of macrophages from TLR3 knock-out mice ( Figure 6). Only background levels of IL-6 and TNF-a were seen with infections by the LRV-negative Lae 372 parasites, as it was observed with the Lg M4147 LRV1-negative control clone and non-infected macrophages. Interestingly, the Lae 316 strain, that had a low LRV load, behaved identically to a LRV-negative strain, suggesting that a minimum amount of virus was required to drive the TLR3-dependent production of IL-6 and TNF-a. The differences in cytokine levels observed with the four Lae strains were not due to differences in parasite uptake or survival, since a similar number of amastigotes per macrophage were quantified 24 hours post-infection with all Lae strains ( Figure S4). IL-6 and TNF-a production upon Lae 303 and 327 infection was less than that observed for L. guyanensis, which might be attributable to the lower LRV load in the L. aethiopica strains (Figures 1A and 2) and/or because of a higher parasite survival rate in the case of L. aethiopica. This hypothesis is supported by the observation that L. aethiopica survive significantly better than L. guyanensis parasites after macrophage infection ( Figure S4). Since the activation of TLR3 by the viral dsRNA requires parasite killing and release of the virus in the phagolysosome, the different survival rates (in addition to the lower LRV load) might therefore explain the lower cytokine production observed with L. aethiopica LRVinfected strains in comparison to L. guyanensis.
In vivo infection experiments were also conducted in mice using these L. aethiopica parasites. In accordance to previous reports on such species [37][38][39], no significant footpad swelling, weight loss or any other sign of infection was measurable with any strain (data not shown). No swelling of the nose, even faint as was described in [38], was observed. This lack of clinical disease prevented the determination of the parasite load. Therefore, it was not possible to test if the presence of LRV in L. aethiopica and the consecutive inflammatory cytokine production had any effect in mice infection as it is observed with L. guyanensis parasites [15].

Discussion
Using several approaches to detect Leishmania RNA virus (LRV) in whole parasite isolates [28], we showed that some strains freshly isolated from human patients infected with L. aethiopica contained a dsRNA virus (designated LRV-Lae). A complete LRV-Lae genome was obtained for one isolate, and more than 97% of it, including the CP-RdRp regions, for two others. These data allowed us to definitively classify the Lae dsRNA virus as relatives of the Leishmaniaviruses LRV1 and LRV2 within the viral family Totiviridae, on the basis of nucleotide comparisons across the LRV genomes, the organization of the ORFs, and amino acid comparisons of the CP and RdRp proteins ( Table 2, Figures 4, S1 and S2). These comparisons further showed that the L. aethiopica LRVs were much more closely related to L. major LRV2 than to the Viannia LRV1s, and were therefore designated as LRV2-Lae. In combination with the prior report of LRV2 in L. major, these data confirmed the presence of LRV outside of South America  and the likelihood that it was present prior to the divergence of the two L. (Viannia) and L. (Leishmania) subgenera.
Despite being close to LRV2-Lmj, L. aethiopica viruses showed a striking difference in the region surrounding the CP and RdRp open reading frames. Similar to LRV1s and unlike LRV2, the reading frames from LRV2-Lae overlaped, but with a 21 frameshift rather than the +1 seen in all LRV1s. In contrast, the LRV2-Lmj CP and RdRp ORFs are non overlapping and in frame ( Figure 5). Viruses are known to exhibit many forms of translational frameshifting and/or hopping [40]. Amongst Totiviridae, S. cerevisiae L-A virus and Trichomonasvirus TvV use frameshifting [41,42] while others such as Helminthosporium virus HvV use stop/restart mechanisms [26]. LRVs are potentially diverse in the mechanisms used despite their relatively close relationships. Whether RdRp is produced as a fusion protein with CP by ribosomal frameshift/hopping or as a separate protein is still an open question that will be addressed in future studies.
Importantly, LRV2-Lae mirrored our previous findings with LRV1-Lg, and further supported the hypothesis that LRV dsRNA was a major innate immunogen as measured by the TLR3dependent production of two key pro-inflammatory cytokines following infection of macrophages in vitro ( Figure 6). In addition, there was likely a correlation between the viral load and the inflammatory response. It also showed for the first time that these cytokine productions were not restricted to L. guyanensis LRV1 but were indeed probably a general feature of LRV-infected strains.
Remarkably, LRV2-Lae was found in nearly half (5/11) of the L. aethiopica strains tested, in both 'recent' and archival strains (Table 1). This suggests that LRV is frequently found in Leishmania strains from Ethiopia, although most of the patients develop cutaneous lesions that often self-heal. As a general and primary consequence, it is unlikely that the presence of LRV2-Lae would by itself be sufficient to explain ML and DCL complications. Other aggravating factors may obviously combine to lead to such pathologies, as previously suggested with L. (Viannia) species (described in the introduction [3][4][5][6][7][8][9][10][11][12][13][14]). Unfortunately, our collection did not include any ML patients, therefore no conclusion could be drawn for this clinical presentation. Three samples from DCL patients were included, none of which bore LRV (or any other virus detectable by dsRNA antibody or eletrophoretic profile). While the numbers were small and as yet inconclusive, they did not point to a strong relationship between a 'digital' classification of human disease pathology (CL/ DCL) and LRV2-Lae presence. A similar conclusion was reached in studies of LRV1 in South America recently [43]. This may in part reflect the difficulty of assessing human disease, as potentially CL patients may have progress to more severe disease at the time of diagnosis, parasite isolation and/or treatment. It would be of special interest to follow the CL cases from which LRV-positive parasites were isolated, and establish if the risk of ML complication is increased by the virus presence in such patients. Similarly, given the spectral nature of leishmaniasis, the range of disease severity is unlikely to be fully captured by 'digitization' into CL/ML. In general, a large survey of geographically and comprehensively clinically catalogued patients (CL/DCL/DL/ML) infected with LRV1 and LRV2-bearing parasites is needed to fully assess the contribution of these viruses to human pathology.
Unfortunately, animal models for L. aethiopica are not well developed and thus do not allow tests of the relationship between LRV and in vivo disease as was possible with L. guyanensis [15]. There, the data pointed strongly to a causal association of LRVs with disease severity and metastasis. Thus, another priority in the future is to explore and develop better models to facilitate testing of the pathogenic consequences of LRV in leishmaniasis caused by L. aethiopica. Figure S1 Capsid protein (CP) alignment of L. aethiopica LRVs with LRVs from L. major ASKH (A) and L. guyanensis M4147/CUMC1 (B). Identical residues are highlighted in grey. (JPG) Figure 6. The presence of LRV2-Lae leads to TLR3-dependent production of pro-inflammatory cytokines by in vitro infected macrophages. C57BL/6 (in black) and TLR3 knock-out (in grey) murine bone marrow derived macrophages were infected by Leishmania promastigotes (parasite/macrophage ratio 10:1), and the level of IL-6 (A) and TNF-a (B) in culture supernatants was measured by ELISA 24 hours postinfection. Non inf.: non-infected macrophages. The cut-off line was calculated as 3 standard deviations (SD) above the mean absorbance of the uninfected macrophage control. Average values presented were obtained from two independent experiments performed in duplicates. doi:10.1371/journal.pntd.0002836.g006 Figure S2 RNA-dependent RNA polymerase (RdRp) alignment of L. aethiopica LRVs with LRVs from L. major ASKH (A) and L. guyanensis M4147/CUMC1 (B). Identical residues are highlighted in grey. (JPG) Figure S3 Conserved domains in the RNA-dependent RNA polymerase (RdRp) from LRVs and other Totiviridae. RdRp sequences from Lae L494/Lmj ASKH/Lg M4147 LRVs (indicated simply as Lae, Lmj and Lg) were aligned to their homologues from S. cerevisiae L-A virus (Sc) and T. vaginalis virus (Tv) using ClustalW2. The only six regions that shared at least 50% of identical residues over 5 or more consecutive amino acids are shown for each virus. Identical and similar residues are indicated by asterisks and double dots respectively. All six domains were already described as conserved among similar viral RdRp, and the third, fourth and fifth domains were directly shown to be crucial for polymerase activity [34][35][36]. (TIF) Figure S4 In vitro infectivity of the L. aethiopica strains analyzed is not affected by their LRV load after 24 hours. Average number of parasites per bone marrow derived macrophage (BMM) were counted 24 hours post-infection from two independent experiments (same as used for cytokine measures presented in Figure 6).

(TIF)
Table S1 Primers used for LRV-Lae sequencing and LRV detection by PCR. Forward (f) and reverse (r) primer sequences were designed from the sequence of LRVs infecting the following strains (as abbreviated in the ''strain'' column): L. major ASKH, L. guyanensis CUMC-1/M4147 and L. aethiopica L494/ 303/327. Primer position is indicated relative to the complete L. aethiopica L494 LRV sequence (5193 bp). bTUBf/r primers amplify a fragment of the beta-tubulin locus of all Leishmania species (sequence based on LmjF33.0798 gene). It was used as a quality control (QC) for cDNA preparations. (PDF)