HTLV-1 is endemic in Central and West Africa, the Caribbean, and parts of South America, Japan, and Melanesia/Australia. STLV-1 can be found in most African and Asian monkey and ape species. Molecular phylogenetic analyses have shown that HTLV-1 most likely arose as a zoonotic infection through several species crossings from nonhuman primates to humans and that species crossing of STLV-1 continues to occur among nonhuman primates (8, 31, 37, 43, 49, 64, 70, 87). Thus, HTLV-1 and STLV-1 do not belong to independent phylogenetic lineages and are conveniently called PTLV-1.

There are two main subtypes of HTLV-2 (33). Both subtypes are present in intravenous-drug users (IDUs) in North America, Europe, and Asia (21, 34, 65, 66, 77, 92) and have also been found sporadically in Africa (20, 25, 26, 38, 55, 76). HTLV-2a is present in some American Indian tribes of North, Central, and South America, including the Navajo and Pueblo in New Mexico (35) and the Kayapo, Kraho, and Kaxuyana in Brazil (2, 53, 72). The distinct subclustering of the Brazilian Indian HTLV-2a strains, together with the fact that HTLV-2a, except for this Brazilian subcluster, has a Tax open reading frame (ORF) that is truncated with respect to that of HTLV-2b, resulted in the designation of HTLV-2c for these Brazilian strains (13). Most Amerindian HTLV-2 strains cluster within subtype 2b, including those found in the Guaymi in Panama (60), the Wayu and Guahibo in Colombia (56, 72), the Toba and Mataco in Argentina (18), and some Navajo and Pueblo in New Mexico (35). Due to the high incidence of HTLV-2 in isolated Amerindian populations, this virus was originally considered to be of New World origin. The recent discovery of endemic HTLV-2 infections in remote Pygmy populations (20, 25, 28, 29, 32) and the identification of a simian virus closely related to HTLV-2 in bonobos indicate that HTLV-2 seems to have its origin in Africa. The molecular characterization of an HTLV-2b isolate from a Cameroonian Pygmy (25) supports the ancient African origin of HTLV-2, but also raises questions about the extremely close phylogenetic relation with Amerindian HTLV-2b strains (66).

We here report the molecular characterization of a Congolese Efe Pygmy HTLV-2 strain belonging to a potential new subtype, HTLV-2d, that is genetically and possibly also phenotypically different from HTLV-2a, HTLV-2b, or HTLV-2c. The Efe Pygmies belong to the Bambuti (or Mbuti) Pygmies from the Ituri Forest and are the least admixed of all Pygmies. The gene flow (including sexually transmitted viruses) is almost always from Pygmies to other Africans and seldom is reversed. They are generally believed to represent the oldest “Proto-Africans” (4).

Sampling. Twenty-three Pygmies, presenting at the health center of Nduye (Ituri Forest, D.R. Congo) in January and February 1995 for various ailments, agreed to provide blood samples (32). Four spots of capillary blood were taken on a filter paper (Whatman no. 2), and 1 ml of venous blood was mixed with an equal quantity of ethanol (EtOH). After drying of the filter paper sample, the samples were kept in a refrigerator until shipment to Leuven, Belgium.

Serological assays. One filter spot of each sample was eluted in phosphate-buffered saline and screened for HTLV antibodies with a particle agglutination assay (Serodia; Fujirebio). Confirmatory assays were a Western blot, including group- and type-specific recombinant antigens (HTLV blot 2.3; Genelabs, Singapore), an indirect immunofluorescence assay with MT-2 cells (HTLV-1 producing) and clone 19 cells (HTLV-2 producing [22]), and enzyme-linked immunosorbent assays (ELISAs) with type-specific synthetic antigens (Select HTLV; IAF/Biochem, Montreal, Canada).

Extraction of proviral DNA. PCR was performed on proviral DNA isolated from the filter spot and from the EtOH-fixed sample. The filter spot was cut in four pieces, and each part was processed in a separate tube. Hemoglobin was fixed with methanol for 5 to 15 min, and the filter was dried in a vacuum chamber. The EtOH-fixed cells were pelleted (10 min, 2,350 rpm) and washed with phosphate-buffered saline. Both samples were digested overnight at 56°C in a proteinase K solution containing PCR buffer (10 mM Tris-HCl [pH 8.3], 50 mM KCl) (Perkin-Elmer), 2 mM MgCl₂, 0.5% Tween 20, 0.5% Nonidet P40, and 100 μg of proteinase K per ml (Boehringer Mannheim stabilized proteinase K solution). For the four filter spot tubes, the contents of each tube were incubated in 100 μl of proteinase K solution, while the pelleted cells were incubated in 1 ml of proteinase K solution. The DNA was extracted from the lysates with the QIAamp Blood kit (Qiagen, Hilden, Germany) and eluted in Milli-Q water (Millipore, Brussels, Belgium) (50 μl for the pooled lysates of the filter spot and 500 μl for 0.5 ml of lysate of the EtOH-fixed sample). Ten microliters of eluted DNA was used per PCR.

PCR. Proviral tax/rex DNA was amplified from 10 μl of both the filter and the EtOH-fixed DNA samples by using the HTLV-1/HTLV-2 generic TR101-104 (48) or the PTLV generic and type-specific AV42-83 (86) nested primer sets. The presence of genomic DNA was confirmed by using a globin PCR (primers PC03 and KM38) (83). From the HTLV PCR-positive sample with an HTLV-2 serology, the entire genome was amplified with nested or seminested primers developed by using the Oligo software (Medprobe, Oslo, Norway) and an alignment for all available HTLV and STLV strains (including the new STLV-L PH969 strain and STLV-2 strain PP1664) in that particular gene region. Primer synthesis was done by Perkin-Elmer/Applied Biosystems and by Life Technologies. The sequences and positions of the primers are given in Table 1. Positive controls were 729 cells harboring HTLV-2a Mo (kindly provided by Helen Lee, Abbot Laboratories, North Chicago, Ill.) and Gu cells harboring HTLV-2 Gu (66). The PCR conditions were as follows. For the long terminal repeat (LTR), nested primers AV125-126/AV127-128 (LTR gag) and AV129-130/AV131-132 (tax LTR) were used in a 50-μl reaction volume containing 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 2 mM MgCl₂, 200 μM nucleoside triphosphates, 0.2 μM outer or 0.5 μM inner primers, and 0.025 U of AmpliTaq (Perkin-Elmer) per ml. The cycling conditions were 30 s at 95°C, 30 s at 55°C, and 45 s at 72°C with a 10-min final extension at 72°C on a GeneAmp PCR System 9600 (Perkin-Elmer) for the outer PCR or on a Triothermobloc (Biometra) for the inner PCR. The outer fragment was amplified for 40 cycles, and then 2 μl was transferred to the inner PCR and amplified for 30 cycles. For the gag gene, two overlapping fragments were amplified by using the nested primers AV169-170/AV171-172 and AV181-182/AV183-184 and the same conditions as for the LTR PCRs but with an MgCl₂ concentration of 1 mM for the primers AV169-170, AV181-182, and AV183-184 and an annealing temperature (T_a) of 60°C for the primers AV181-182 and AV183-184. For the pol gene, three outer PCRs, each one in combination with two inner PCRs, were performed with the same conditions as for the LTR PCRs but 1 mM MgCl₂ for all the inner PCRs, as follows: AV147-148 (T_a, 60°C) for the first outer PCR with AV149-150 (T_a, 60°C) and AV151-152 (T_a, 55°C) for inner PCRs, AV153-154 (T_a, 60°C) for the second outer PCR with A155-156 (T_a, 60°C) and AV157-158 (T_a, 55°C) for inner PCRs, and AV159-160 (T_a, 55°C) for the third outer PCR with AV161-162 (T_a, 55°C) and AV163-164 (T_a, 60°C) for inner PCRs. The cycling conditions were as follows: for the outer PCRs, 94°C for 40 s, the appropriate T_a for 40 s, and 72°C for 1 min 40 s for 40 cycles; for the inner PCRs, 1 min at 94°C, the appropriate T_a for 45 s, and 72°C for 1 min for 35 cycles. For the env gene, two outer PCRs were used, the first one in combination with two inner PCRs and the second one with a single inner PCR: AV141-142 for the first outer PCR with the same conditions as for the LTR PCRs and, for inner PCRs with the same conditions as for the LTR PCRs but with a T_a of 45°C, AV143-144 (except that the MgCl₂ concentration was 1 mM) and AV145-142 (seminested); AV173-174 for the second outer PCR with AV175-176 for the inner PCR, using the same conditions as for the LTR PCRs but with a T_a of 60°C. For the beginning of the pX region, the nested primers AV177-178/AV179-180 were used with the same conditions as for LTR PCRs but with a T_a of 60°C. For the tax/rex gene, three seminested PCRs were used, with AV135-138/AV135-136, AV137-140/AV137-138, and AV139-134/AV139-140, all with the same conditions as for the LTR PCRs but with 60°C as the T_a for the outer PCR and 50°C as the T_a for the inner PCR. Amplification products were separated on a 6% polyacrylamide gel and visualized by ethidium bromide staining.

TABLE 1 Primers used for the amplification and sequencing of the Efe2 HTLV-2dgenome

Generation of the nucleotide sequences. Inner PCR fragments were separated on a 1% agarose gel (Seakem; FMC, Life Sciences International, Zellik, Belgium), and the HTLV-2-specific DNA was excised and extracted from the gel by using a Sephaglass bandprep kit (Pharmacia Biotech, Roosendaal, The Netherlands). Direct sequencing of PCR products was performed, resulting in an average PCR population sequence and thus circumventing the problem of Taq polymerase errors. Solid-phase sequencing was performed on an ALF automated sequencer (Pharmacia Biotech) when the M13 universal and reverse sequencing primers (M13USP and M13RSP, respectively, tagged on the inner primers [Table 1]) were used, as described previously (85). On some occasions, sequencing was performed with the M13USP and M13RSP primers with an AutoCycle Sequencing Kit on an ALF automated sequencer or with a Dye Primer Cycle Sequencing Kit with AmpliTaq DNA polymerase FS on an ABI Prism 310 automated sequencer (Perkin-Elmer). For the primer sets that lacked the M13USP or M13RSP tag, sequencing was performed with the Dye Terminator Cycle Sequencing Kit with AmpliTaq DNA polymerase FS on an ABI Prism 310 automated sequencer. Sequences were assembled and aligned with those of all available HTLV and STLV strains from the EMBL database by using the GeneWorks software package (IntelliGenetics, Antwerp, Belgium). Optimal alignments were obtained after minimal manual editing.

Phylogenetic analysis. Phylogeny construction and evaluation were done with the Phylip (version 3.56) (17), MacClade (version 3.0) (51), and PAUP (version 5.2) (74) software packages. A chart of nucleotide state changes was made from the alignment with MacClade by using as a guide tree a neighbor-joining (NJ) tree obtained with standard parameters in Phylip. For the analyzed nucleotide alignments, the chart can be described as symmetrical, with a transition-transversion bias of between 1.5 and 5.5 for all alignments used. The ratio was 3.0 for the HTLV-2 LTR alignment, 1.7 for the HTLV-2/STLV-2 LTR alignment, 5.4 for the HTLV-2 env alignment, 1.9 for the HTLV-2/STLV-2 env alignment, and 1.9 for the tax gene alignment. Next, two different methods implemented in the Phylip package were used: the NJ method and the maximum-likelihood (ML) method. The Fitch-and-Wagner parsimony (PARS) method was applied by using PAUP with a heuristic search on 25 replicates of random stepwise-added sequences, with TBR branch swapping and the MULPARS option on (except for bootstrap analysis, where the MULPARS option was off). For all methods, the empirical transition-transversion bias was used. Distances were calculated with the Felsenstein model, which uses the empirical base frequencies. A reverse transcriptase (RT) protein tree with all PTLV types and BLV as the outgroup was also made with the NJ and PARS methods implemented in PAUP. An ML tree for the nucleotide sequence of the RT gene was obtained with Phylip. The NJ and PARS trees were statistically evaluated by using 1,000 bootstrap samples (16). No bootstrapping was done for the ML method, which is itself already a statistical method. The values on the branches represent the percentages of trees for which the sequences at one end of the branch are a monophyletic group.

Strains used in the phylogenetic analyses. The HTLV and STLV strains (accession numbers are in brackets) were those described by Bazarbachi et al. (1) (BOI, France [L36905]), Chou et al. (5) (ATL-YS, United States [U19949]), Coulston et al. (7) (BLV-A1 [D00647]), Digilio et al. (10) (PanP, pygmy chimpanzee, D.R. Congo [U90557]), Eiraku et al. (12, 13) (WY, Wayuu, Colombia; DSA, FH, 408N, and 72969N, Navajo, United States; JD, SC, and AG, Pueblo, United States; DOG, FLN, MIN, SAC, ASB, MER, WEN, CAM, FUC, GAR, PAR, and VIN, New York, United States; MSA1bp and 130P, Pueblo, New Mexico, United States; SP1 to SP7, Sao Paolo, Brazil; RJ-1, Rio de Janeiro, Brazil; and KAY1 and KAY2, Kayapo, Brazil [L37129 to L37146, U25135, and U32886 to U32906]), Evangelista et al. (14) (TSP-1, Japan [M86840]), Ferrer et al. (19) (FH39399, W175, CH610, and W43, Gran-Chaco Amerindians [U46555 to U46558]), Gessain et al. (24, 25) (Mel5, Solomon Islands, Melanesia [L02534]; and PYGCAM-1, Cameroon, Pygmy [Z46888 and Z46889]), Hall et al. (33) (WH6 and WH7, New York, United States [M85226]), Hjelle et al. (35) (CG, LM, DS [Navajo], and JD [Pueblo], New Mexico, United States [M63881 to M63884]), Ibrahim et al. (37) (TE4, Macaca tonkeana from Indonesia [Z46900]), Ishak et al. (39) (KAY1 and KAY2, Kayapo, Brazil [U19109 and U19110]), Lee et al. (45) (NRA, United States [L20734]), Lin et al. (46) (VIET13, VIET19, VIET22, VIET32, and VIET35, Vietnam [U72524 to U72533]), Malik et al. (52) (HS35, Caribbea [D13784]), Mauclère et al. (55) (PH230PCAM, Cameroon [Z46837 and Z46838]), Miura et al. (56) (2C1801, 2C2517, 2C3821, and 2C5505, Guahibo, Colombia; 2C11521, WY018, and WY100, Wayuu, Colombia; and CH13504, Mapuche, Chile [D82952 to D82959]), Mukhopadhyaya and Sadaie (58) (RD-1, Caribbea [L10341]), Pardi et al. (60) (G12, Guaymi Amerindian [L11456]), Ratner et al. (62) (EL, D.R. Congo [S74562]), Sagata et al. (63) (BLV-CG [K02120]), Salemi et al. (65–67) (Va, Bo, and Md, Italy [X80242 to X80244]; Gu, Italy [X89270]; and I-AM, I-EC, I-IT, I-EA, I-GI, I-OG, and I-OV, Italy [Y09149 to Y09155], Seiki et al. (68) (ATK1, Japan [J02029]), Shimotohno et al. (69) (Mo, United States [M10060]), Switzer et al. (71–73) (ATL18, Georgia, United States; BRAZ.A21, Brazil; LA8A, California, United States; NAV.DS, Navajo, New Mexico, United States; NOR2N, IDU, Norway; PUEB.AG and PUEB.RB, Pueblo, New Mexico, United States; ITA47A and ITA50A, Italy; NY185, New York, United States; PENN7A, Pennsylvania, United States; SEM1050 and SEM1051, Seminole, Florida, United States; SPAN129 and SPAN130, Spain; WYU1 and WYU2, Wayuu, Colombia; GHKT, Ghana; KAY73 and KAY139, Kayapo Amerindian; Mexy17, Mexico [L42507 to L42510, U10252 to U10266, and U12792 and U12794]), Takahashi et al. (75) (JG and ED, New York, United States [L06853, L06856, L06857, L06859, and L06860]), Tuppin et al. (76) (JPS, Gabon [Z47788]), Vallejo et al. (78) (RC, BF, DP, AA, JA, JL, JAN, 130, 324, and RVP, Spain [L77235 to L77244]), Van Brussel et al. (80, 81) (PH969, Papio hamadryas, Eritrea [Y07616]; PP1664, Pan paniscus, D.R. Congo [Y14570]), Wang et al. (88) (FLW, United States [S67545]), and Zhao et al. (93) (MT2, Japan [L03561]).

Nucleotide sequence accession number. The nucleotide sequence reported here has been assigned accession no. Y14365.

Serological and PCR identification of HTLV-2-infected Efe Pygmies. Three of the 23 sera were HTLV screening positive by the particle agglutination test; two of them showed an HTLV-2-like Western blot reactivity (Fig. 1, lanes 4 and 14), and both of these also were HTLV-2 positive in the type-specific ELISA. The sample in lane 4 in Fig. 1 was negative in indirect immunofluorescence; the sample in lane 14 reacted with the HTLV-2 clone 19 cells and also slightly with the HTLV-1 MT-2 cells. Only the sample from lane 14 was PCR positive with the TR101-104 and AV42-46 primer sets and was typed by PCR as HTLV-2 (86). The other HTLV-2-seropositive sample was PCR negative on all occasions, including with the PTLV generic nested primer set (AV42-46 primers) (86). To evaluate the presence of genomic DNA and the absence of PCR inhibitors, the globin PCR was performed and was positive. Therefore, the lack of reactivity in the HTLV PCR is probably not due to the divergence of the strain or to the lack of good genomic DNA but is possibly due to a low proviral load and a low DNA input into the PCR. The third HTLV screening-positive serum was indeterminate on the Western blot but proved to be HTLV-1 by ELISA and by PCR (Fig. 1, lane 12) (32, 86). Nine other screening-negative samples were indeterminate on Western blots but negative with both ELISA and PCR. There is no known familial relationship between the 12 HTLV-reactive Pygmies; the two HTLV-2-positive individuals belonged to two different clans. The only known familial relationships among the 23 Pygmies were between seronegative individuals, and two of the people with sera indeterminate by Western blotting, one PCR-negative woman and one HTLV-1 PCR-positive man, each had a child among the seronegative individuals.

FIG. 1 Western blots of the eluted antibodies from the 23 Efe Pygmy blood spots. MTA1, HTLV-1-specific recombinant gp41 peptide; K55, HTLV-2-specific recombinant gp41 peptide. Lanes: HTLV-1 and HTLV-2, positive controls; 1 to 23, Pygmy blood spot eluates.

Sequence analysis of the Efe2 proviral genome. Due to the limited amount of cellular DNA, each PCR was carefully optimized to allow a maximum amount of amplified proviral DNA from only one infected cell of the control cell lines (containing HTLV-2a Mo or HTLV-2b Gu [see Materials and Methods]). Subsequently, the optimized conditions were used to amplify the Efe2 proviral DNA. By using the PCR sequencing strategy described in Materials and Methods, the sequence of the entire genome (8,971 nucleotides [nt]) could be reconstructed, of which about 15% was sequenced at least twice as a result of overlapping PCR fragments. No discrepancies were found between the two overlapping sequences, ensuring that PCR errors had little effect on the overall sequence of the Efe2 proviral genome. The entire sequence could be unambiguously aligned with those of strains belonging to the subtypes HTLV-2a and HTLV-2b. The Efe2 strain was clearly an HTLV-2 sequence, although it was quite divergent and did not belong to the HTLV-2a or the HTLV-2b subtype. All major genes could be identified based on the homology between the HTLV-2 Efe2 strain and the HTLV-2a Mo and HTLV-2b NRA strains.

The LTR of the Efe2 strain is 769 nt long and very divergent, differing by 10 to 11% from those of HTLV-2a and HTLV-2c and by about 7% from that of HTLV-2b (Table 2). The divergence between the other HTLV-2 subtypes is less than 7.5%. The major functional sites are conserved, including the three 21-bp repeats, the Ets and NF-κB-responsive site, the polyadenylation site, the TATA box, and the major splice donor (Fig. 2A). Also, the sequence of the Rex-responsive element of Efe2 is conserved such that the predicted RNA secondary structure (54) is largely maintained with respect to those of HTLV-2a Mo and HTLV-2b NRA (results not shown). The primer binding site, just downstream of the LTR, is entirely conserved.

TABLE 2 Comparison of the divergence between PTLV strainsa

FIG. 2 (A) LTR nucleotide sequence alignment of the prototype strains of HTLV-2a (Mo, accession no. M1006) (68), HTLV-2b (NRA, accession no. L20734) (45), and the potential new subtype HTLV-2d (Efe2). The functional elements are indicated. (B) Tax amino acid (more ...)

The gag region is somewhat more conserved among HTLV-2 strains, but again, the Efe2 strain is the most divergent one, differing at the nucleotide level by almost 7% from the other subtypes, while HTLV-2a and -2b differ by about 4% (Table 2). The p24 Gag region is one of the major immunodominant epitopes in HTLV-2. It is very conserved among all HTLV-2 subtypes, including the Efe2 strain (98 to 99% amino acid identity), and also between HTLV-1 and HTLV-2 (85 to 87% amino acid identity). p19 is also very conserved within HTLV-2, with the Efe2 strain being the most divergent one, having 96 to 97% amino acid identity to HTLV-2a and HTLV-2b. Between HTLV-1 and HTLV-2, p19 is much less conserved (54 to 55% amino acid identity). The corresponding strong cross-reactivity of HTLV-2 p24 antibodies with HTLV-1 p24 antigens, compared with the weak cross-reactivity for the p19 Gag epitope, is used in Western blotting as an indication of the presence of HTLV-2 (44). Both Pygmy sera cross-reacted with the HTLV-1 p24 epitope, and the serum from the Efe2-infected individual also cross-reacted weakly with the HTLV-1 p19 Gag epitope. These serological data are in accordance with the genotypic data found for the Efe2 strain.

Two ribosomal frameshifts are needed for the translation of the HTLV-2 pol genes (15). These are controlled by a heptanucleotide slippery sequence followed by a stem-loop structure about 7 nt downstream (41). The alignment of the prototype strains of the three HTLV-2 subtypes (Mo for 2a, NRA for 2b, and Efe2 for 2d) shows that the beginning and the end of the protease ORFs are at the homologous position (nt 2085 to 2640 in Efe2) and that the frameshift sequence elements are entirely conserved (around nt 2090 and 2603 in Efe2). The protease gene of Efe2, however, has a 9-nt insertion compared with the two other subtypes, corresponding to an extra 3 amino acids (aa) just upstream of the protease cleavage site at the end of the protease. The pol ORF of Mo, which is shifted −1 with regard to the pro ORF, is extended by 96 nt at the 5′ end compared with those of both NRA and Efe2 (nt 2341 in Efe2), but this does not influence the translated proteins, since this part of the pol ORF is upstream of the ribosomal frameshift site for Pol. The protease cleavage sites around the protease gene, as identified through their homology with the HTLV-1 protease cleavage sites (9), have an entirely conserved amino acid sequence (nt 2149 and 2530 in Efe2), while the protease cleavage site that separates the RT from the integrase is not conserved in Mo compared with NRA and Efe2 (nt 4293 in Efe2). The protease active-site aspartic acid and the amino acid region around it are perfectly conserved (LLDTGA) (50). The three aspartic acids of the RT active site are also entirely conserved (TIDLT and QYMDD). This conservation of essential amino acids in the active sites suggests an active protease and RT. The overall similarity in the pol gene is 4 to 7% among the HTLV-2 subtypes, with the Efe2 strain being the most divergent one (Table 2). All HTLV-2 subtypes are very distantly related to HTLV-1 and STLV-L (34 to 35% nt divergence) and somewhat more closely related to the STLV-2 strain PP1664 (22 to 23% nt divergence).

In the env region, the divergence between the HTLV-2 subtypes is similar to those in the gag and pol regions and lower than that in the LTR region, ranging between 3.9 and 6.9% (Table 2), again with the Efe2 strain being the most divergent one, confirming the divergence of this new HTLV-2d subtype. Only HTLV-2c is much closer to HTLV-2a (1.0% divergence) than to the other two subtypes of HTLV-2. In the 44-aa K55 epitope in the gp46 surface protein that is used to identify HTLV-2 strains in Western blot assays (47), only two amino acids are not identical among the three subtypes HTLV-2a, -2b, and -2d. For one, HTLV-2a is different, for the second amino acid, HTLV-2d is different. This corresponds with the observation that the antiserum from the Pygmy carrying the HTLV-2d strain is reactive with the K55 peptide in the Western blot (Fig. 1). The major splice donor in the LTR and splice acceptor just upstream of the env ORF, which allow splicing to generate the single spliced messenger coding for the Env proteins, are conserved and therefore probably are functional. The second splice donor at the beginning of the env gene to generate the double-spliced messengers encoding Tax and Rex is also conserved and probably is functional.

The tax/rex overlapping ORF is very conserved among all PTLVs (Table 2), with a maximum divergence of about 23%. Again, Efe2 is the most divergent HTLV-2 subtype, equally different from HTLV-2a as from HTLV-2b (about 3%). The HTLV-2c strain KAY1 is again very close to HTLV-2a. The lengths of the Tax proteins are not conserved among the PTLV strains. HTLV-2a has the shortest Tax protein, and STLV-2 PanP has the longest Tax protein (Fig. 2B). The Tax proteins of the other strains have intermediate lengths. The HTLV-2d Tax protein is 344 aa long, which is between the lengths of the HTLV-2a Tax (331 aa) and the HTLV-2b and -2c Tax (356 aa), and at the same homologous site as for STLV-2 PP1664. This might reflect the position of the stop codon for Tax in the common ancestor of PTLVs (Fig. 2B) and could imply that HTLV-2d represents an older lineage of HTLV-2 than does HTLV-2a or -2b. Thus, the phenotype of HTLV-2d Tax is potentially different from those for the other HTLV-2 subtypes. The part of the PTLV Tax proteins that is homologous to the shortest Tax of HTLV-2a has a very high similarity among all PTLV strains (Table 2 and Fig. 2B). In the part of the Tax proteins that is extended with respect to the HTLV-2a Tax, the amino acid sequences are much more divergent. Tax and Rex are translated from double-spliced messengers, using the major splice donor in the LTR, a splice acceptor just upstream of the env region, a second splice donor at the beginning of the env gene, and a second splice acceptor just upstream of the second ORF of the tax/rex gene. These are all conserved, and correct splicing of double-spliced messengers can be assumed. In the proximal pX region, preceding the second tax/rex ORF, two alternative splice sites have been identified for HTLV-2a Mo that allow translation from additional ORFs in this region (6). Both are conserved in HTLV-2b but only the first of these two alternative splice sites (nt 6821) is conserved in HTLV-2d.

Phylogenetic analysis of the Efe2 LTR. The similarity between the HTLV-1 and HTLV-2 LTRs is so low (57%) (Table 2) that only a few small stretches of nucleotide sequence can be aligned unambiguously. Therefore, the HTLV-1 LTR is unsuitable as an outgroup to root the HTLV-2 LTR tree (67, 82). Instead, we have used the bonobo PP1664 LTR sequence (81) (EMBL accession no. Y14570), which has a similarity of about 70% with all HTLV-2 subtypes (Table 2), as an outgroup. Both the rooted and unrooted trees (Fig. 3) are drawn in a star-like format usually employed for unrooted trees, because this format better illustrates the real proportions of the branches and clusters. Details of the analysis are described in Materials and Methods. Many sequences in the LTR are available, with a large amount sequenced only in the R and part of the U5 region. Since the U3 region is the most divergent part of the LTR, valuable phylogenetic information would be discarded by trying to include all strains. We have performed an analysis with all available strains by using the R and part of the U5 region and another analysis with a limited number of strains by using almost the entire LTR. The trees in Fig. 3 represent the analysis of almost the entire LTR. The unrooted tree (NJ tree [inset in Fig. 3]) shows a stable clustering of all HTLV-2a (including HTLV-2c) strains (>99% by the NJ and PARS methods; P < 0.01 in the ML tree) and HTLV-2b strains (>93% by NJ and PARS; P < 0.01 by ML), both different from the new Efe2 strain, which clearly represents a new subtype, HTLV-2d. When the two STLV-2 strains, PP1664 and PanP, are added as an outgroup (main tree in Fig. 3), the Efe2 strain stably clusters with the outgroup, showing that it diverged from the bonobo virus earlier than the separation between the two other subtypes. All three HTLV-2 subtypes are very different from the STLV-2 strains, which have a real branch length that is about 10 times larger than the truncated branch shown in Fig. 3. Both HTLV-2a (including HTLV-2c) and HTLV-2b remain a stable cluster, although with a somewhat lower support for the HTLV-2b subtype. It is also clear from this tree that HTLV-2c is not a separate subtype but rather a cluster within HTLV-2a, which is stably supported by the PARS and ML methods but not by the NJ method. The previously reported Cameroonian Pygmy strain (PYGCAM-1) (25) is very different from the Efe2 strain and belongs to the Amerindian cluster within HTLV-2b. Two other African strains, PH230PCAM (Cameroon) (55) and GHKT (Ghana) (38, 73), belong to the HTLV-2a subtype. When all available strains with the smaller LTR fragment are included, the same clusters appear within HTLV-2a but with lower bootstrap support, except for the HTLV-2c cluster, where additional strains result in an increased support of this cluster by the NJ method (74.3% of bootstrap replicates). Within HTLV-2b, additional Amerindian strains cluster with the WYU1/G12 clade and the Cameroon Pygmy/Amerindian clade. The WY100 strain sequenced from a Wayuu Amerindian in Colombia had a sequence identical to that of the PYGCAM-1 strain from a Cameroonian Pygmy. An extra Amerindian clade emerges, consisting solely of strains from Colombian Guahibo Indians (56). Several additional IDU strains (including Vietnamese strains [46]) cluster within the IDU clade of HTLV-2b but with a lower bootstrap support (around 40% by NJ).

FIG. 3 Phylogenetic analysis of the entire LTR sequence (homologous to nt 1 to 769 of HTLV-2 Efe2) by the NJ approach described in Materials and Methods. The values on the branches represent the percentages of trees for which the sequences at one end of the (more ...)

Phylogenetic analysis of Efe2 env. The env sequence of the HTLV-2d Efe2 strain was aligned with the available HTLV-2 env sequences from the EMBL and GenBank databases. As for the LTR region, the analysis was done with (Fig. 4) or without (Fig. 4, inset) the STLV-2 strains as an outgroup. Again, the Efe2 strain clusters with the outgroup, clearly separated from HTLV-2a and HTLV-2b (85.8% bootstrap value by NJ and 87.6% by PARS; P < 0.01 by ML). In the analysis with the STLV-2 outgroup, both HTLV-2a and -2b are stable clusters (>91% of bootstrap replicates in all cases; P < 0.01 by ML), while without the outgroup, HTLV-2a is unstable by the NJ method (63.9% by NJ and 84.0% by PARS; P < 0.01 by ML). Adding STLV-2 as an outgroup in the analysis has shifted the branch leading to Efe2 and STLV-2 towards the HTLV-2b subtype. Within HTLV-2a, the Brazilian Kayapo cluster, referred to as HTLV-2c by Eiraku et al. (13), is seen as a stable cluster, different from all other HTLV-2a strains. Within HTLV-2b, the Cameroonian Pygmy strain PYGCAM-1 clusters with another African strain, JPS (Gabon) (76) and with strains from Amerindians and IDUs. The other African strain from Cameroon, PH230PCAM (55), belongs to HTLV-2a.

FIG. 4 Phylogenetic analysis of the entire gp21 gene in the env region (homologous to nt 6119 to 6655 of HTLV-2 Efe2) by the NJ approach described in Materials and Methods. The values on the branches represent the percentages of trees for which the sequences (more ...)

Phylogenetic analysis of the Efe2 RT and pX genes and proteins. The similarity in the RT protein is high enough to allow an unambiguous alignment of all PTLV RTs with BLV RT. The three PTLV types can be clearly distinguished, with PTLV-1 containing HTLV-1 and STLV-1 strains, PTLV-2 containing the three HTLV-2 subtypes (a, b, and d) and the bonobo STLV-2 viruses, and PTLV-L containing the only strain of this type known so far, STLV-L PH969. The BLV outgroup branches off close to the center of the tree but on the branch leading towards PTLV-1. The real branch length of the outgroup is almost five times larger than that of the truncated branch in Fig. 5. Within PTLV-1, Asian strains branch off closer to the other types than African strains, but the relative branching order among these Asian strains is dependent on the method. Within the PTLV-2 clade, there is an early split between the bonobo STLV-2 strains and the HTLV-2 strains. In the evolution of HTLV-2, HTLV-2d first diverged from HTLV-2a and HTLV-2b, and later HTLV-2a and HTLV-2b diverged from each other. All of these HTLV-2 clades are reliably separated from each other by all three methods used.

FIG. 5 Phylogenetic analysis of the RT protein (homologous to nt 2534 to 4293 of HTLV-2 Efe2) by the NJ approach described in Materials and Methods. The values on the branches represent the percentages of trees for which the sequences at one end of the branch (more ...)

The similarity in the tax gene is also high enough to allow an unambiguous alignment of all PTLV tax genes (not shown). The picture is very similar to that for the RT analysis. Again, Efe2 is clearly different from HTLV-2a and -2b. In our analyses, HTLV-2a is still a stable phylogroup (>98% by all methods), but HTLV-2b is no longer a stable cluster (50 to 60% of bootstrap replicates by the NJ and PARS methods; P < 0.01 by ML). As in the RT tree, HTLV-2d is the first subtype of HTLV-2 to branch off. The three HTLV-2 subtypes (2a, 2b, and the new 2d) again form a close stable cluster (100% in all analyses) different from STLV-2, STLV-L, and PTLV-1. When the translated Tax and Rex protein alignment is used, none of the three HTLV-2 subtypes is consistently well supported, and the branching order among the subtypes is dependent on the translated reading frame (Tax or Rex) and the method used. Still, in all analyses, HTLV-2 remains separated from STLV-2 with highly significant bootstrap values (>98%).

We have described here the identification and genomic characterization of a highly divergent new HTLV subtype, HTLV-2d, found among Efe Pygmies in D.R. Congo. Serologically, this virus can be clearly typed as HTLV-2, and its genome can be amplified with HTLV-2 type-specific PCR primers. Sequence analysis showed that although it is different from the known HTLV-2a and -2b subtypes, this new virus is clearly an HTLV-2 virus, different from the recently discovered closest simian relative of HTLV-2, the Congolese bonobo virus STLV-2 (27, 48, 84). In all gene regions analyzed, this new Pygmy Efe2 strain is the most divergent HTLV-2 strain. Sequence analysis implied that major functional elements seemed to be conserved, except for the lack of one of the alternative splice acceptors in the proximal pX region, used by HTLV-2a Mo to generate the accessory protein p28XII (6), and a different location of the Tax stop codon. The similarity between the HTLV-2d Gag and Env proteins and the corresponding HTLV-2a and -2b proteins is consistent with the observed serological reactivity. An interesting feature of this potential new HTLV-2 subtype is the length of the Tax protein, as inferred from the Tax ORF. The HTLV-2b Tax protein is 25 aa longer than the HTLV-2a Tax protein. The HTLV-2c Tax protein, which is phylogenetically indistinguishable from that of HTLV-2a, is extended to a length similar to that of HTLV-2b. The longer Tax protein of HTLV-2b and HTLV-2c seems to be linked to a stronger transactivation activity of the viral LTR, as was shown by Eiraku et al. (13) in transient-expression assays. The inferred HTLV-2d Tax protein has a length that is between those of the HTLV-2a Tax protein and the HTLV-2b and -2c Tax proteins. The length of the Tax protein is not conserved among the different PTLV types, due to the introduction or disappearance of a stop codon along different phylogenetic lineages of PTLV (Fig. 2B). The longest Tax protein is the one from STLV-2 PanP (400 aa), the shortest one is from HTLV-2a Mo (331 aa), and those from HTLV-1 (353 aa), STLV-L (350 aa), HTLV-2b (356 aa), HTLV-2c (356 aa), and HTLV-2d (344 aa) are of intermediate lengths. Interestingly, the position of the Tax stop codon for HTLV-2d is homologous to the position of the Tax stop codon for STLV-2 PP1664. It would be interesting to investigate the transactivation potency of the HTLV-2d Tax protein in comparison with those of the other PTLV Tax proteins. Together with the lack of one of the alternative splice acceptor sites in the pX region, the length of the Tax protein suggests that HTLV-2d has a phenotype different from that of HTLV-2a or HTLV-2b.

Among Pygmy populations, the Bambuti Pygmies (to whom the Efe belong) from the Ituri Forest in D.R. Congo are considered the oldest Proto-Africans (4). In the occasional contacts between Pygmies and Bantus, the gene flow is almost exclusively from Pygmies to Bantus. Among Bambuti Pygmies from eastern Congo, HTLV-2 is highly endemic, with a seroprevalence of 14%. HTLV-2 is not found among Biaka Pygmies from the Central African Republic, but the most western group of Cameroon Pygmies have a prevalence of about 2.3% (29). For both infected Pygmy groups, the seroprevalence of HTLV-2 is higher among the Pygmy population than among the surrounding black population, where it is only found sporadically. These serological data, together with the divergence of the HTLV-2 Efe2 strain found in Pygmies as reported here, strongly suggest that this virus has a long separate history in this African Pygmy population, which is entirely in agreement with the anthropological data.

The consensus picture from the phylogenetic analyses shows that this new HTLV-2 Efe2 strain is a potential new HTLV subtype, 2d, that diverged earlier than the other two subtypes, 2a and 2b. Since the closest related simian virus is the African bonobo STLV-2, this clearly favors an African origin for HTLV-2. The phylogenetic analyses conform with the view that after the separation of the bonobo and the human viruses, the potential HTLV-2d subtype was the one that remained in Africa, with its high divergence showing its long independent evolution. It has been suggested that since HTLV-2a and -2b are present in isolated Amerindian tribes, it is probable that both HTLV-2 subtypes found their way to the Americas with the human migration (2).

The PYGCAM-1 strain found in a Cameroonian Pygmy has an LTR sequence that is identical to that of the WY100 strain found in a Colombian Wayuu Amerindian. The evolutionary rate (or fixation rate) of cellular genes is estimated to be 10⁻⁸ to 10⁻⁹ nucleotide substitutions per site per year (3, 90). HTLV-2 in IDUs has a fixation rate of about 10⁻⁴ to 10⁻⁵ nucleotide substitutions per site per year (67), which is one of the lowest rates among RNA viruses (11). While HTLV-2 probably evolves at an even lower rate in populations in which the virus is endemic than in IDUs (67), where 1 nucleotide change is expected every 20 years in the 500-nt LTR sequence studied here, this virus is evolving much faster than the human genome, where 1 nucleotide change is expected every 200,000 years in this 500-nt LTR sequence. The extremely close relationship of the Pygmy strain PYGCAM-1 and Amerindian strains does not conform with a long independent evolution. It is therefore unlikely that there was an ancient separation between the Pygmy strain PYGCAM-1 and the Amerindian strains as suggested by Gessain et al. (25). Further sampling of the Cameroonian Pygmy population might be necessary to resolve this contradiction. The presence of HTLV-2a and HTLV-2b in Africa can then be explained rather by a recent reintroduction of these strains on the African continent. In this view, the potential new HTLV-2d subtype, which is not found among Amerindians, might be a genuine African subtype.

Detailed analyses of the evolutionary rates within the HTLV-2 and STLV-2 lineages would be necessary to judge whether the separation of the potential new subtype HTLV-2d from the subtypes HTLV-2a and -2b dates back to the time of the exodus of modern humans from Africa over 100,000 years ago (4) and whether HTLV-2a and -2b entered the Americas by two separate human migrations possibly coinciding with two of the three waves of migration over the Bering strait (4). Similarly evolutionary rate analyses are required to judge whether the separation between the human and simian viruses within PTLV-2 dates back to the speciation of humans and chimpanzees or whether interspecies transmission between bonobos and humans directly or via another primate species is involved.

In this paper we have characterized a potential new HTLV-2d subtype among Bambuti Efe Pygmies that is genetically and probably phenotypically different from the other three subtypes. Together with the presence of a divergent STLV-2 in African bonobos, these findings strongly suggest an African origin for HTLV-2.

We thank Martine Michiels and Martin Reynders for excellent technical assistance, and Ria Swinnen for fine editorial help.

Marco Salemi is the recipient of a Training and Mobility of Researchers (TMR) Marie Curie fellowship from the European Commission. The Rega Institute is part of the HERN concerted action supported by the Biomed Programme of the European Commission. This work was supported in part by the Belgian Fonds voor Geneeskundig Wetenschappelijk Onderzoek (Krediet no. 3.0098.94).