Introduction

Leishmaniasis is a vector-borne disease transmitted by sand flies, caused by a protozoan flagellate of the genus Leishmania. It is endemic in 88 countries on four continents and occurs in at least four major forms: cutaneous, diffuse cutaneous, mucocutaneous, and visceral (Desjeux 2004). The current global estimate of 350 million people at risk of leishmaniasis and of an incidence per year of 500,000 for visceral leishmaniasis (VL) and 1.5 million for cutaneous leishmaniasis (CL) belies higher burdens of disease in endemic foci. In the current context of worldwide (re)emergence and spreading of leishmaniasis, the relevance of species identification further gains importance.

As with many parasites, the taxonomy of the genus Leishmania is very complex because species definitions and boundaries are hard to define. The links between clinical disease, geographic origin, and nomenclature of Leishmania species have long been controversial and confusing. It has been recognized that the gold standard for taxonomy and strain typing of Leishmania based on multilocus enzyme electrophoresis (MLEE, also known as zymodeme typing) has several disadvantages: It requires large culture volumes and specialized equipment, it is costly and laborious, and it has relatively poor discriminatory power (Kuhls et al. 2005; Bañuls et al. 2007). To overcome these problems, DNA-based methods have been widely used for Leishmania spp. identification and/or phylogenetic reconstruction with a variety of targets such as protein-coding genes, non-coding segments, microsatellites, and restriction fragment length polymorphisms (e.g., Piarroux et al. 1995; Noyes et al. 1997; El Tai et al. 2001; Ibrahim and Barker 2001; Kuhls et al. 2005; Asato et al. 2009; Fraga et al. 2010; Montalvo et al. 2010). Notably, Lukeš et al. (2007) went a long way to resolving the issues for the Leishmania donovani complex. They proposed that this species complex consists of only two species, L. donovani sensu stricto (with the synonym Leishmania archibaldi) and L. infantum (with the synonym Leishmania chagasi), by using a combination of widely used DNA analysis techniques and further demonstrated that geographic origin of a strain is a more important predictor of genetic relatedness than the type of disease caused (visceral versus cutaneous leishmaniasis). Few isolates from China, however, were included in these studies, and no work has yet focused on the phylogenetic relationships among different isolates from different epidemic areas in China.

Leishmaniasis remains endemic in China, especially in the west and northwest frontier regions. Both human VL and CL occur in China, with most VL cases, rare CL cases (Guan et al. 1992a; Zheng et al. 2009; Wang et al. 2010). According to different geographical origin, infective agent, and clinical evidences, epidemic foci of VL in China were classified into three types, i.e., plain foci, hill foci, and desert foci (Lu et al. 1994). Different foci and types of leishmaniasis in China have brought forth the difficulty of identifying the strains of Leishmania, as the etiological pathogens cannot be distinguished easily on the basis of morphological characteristics. On the basis of MLEE of six representative isolates from the plain, mountainous, and desert regions, Xu et al. (1984) first identified the causative agents responsible for VL as L. donovani sensu lato and L. infantum. Xu et al. (1984) also provided support for the specific status of Leishmania gerbilli, a nonpathogenic parasite of the great gerbil Rhombomys opimus in the desert areas of Gansu Province and Xinjiang Uygur Autonomous Region in N.W. China (Wang et al. 1964a, b). Interestingly, Xu et al. (1989) further found that two isolates from kala azar patients in Kashi city of Xinjiang could not be designated definitely as L. infantum or L. donovani by isoenzyme electrophoresis. Subsequently, on the basis of zymodeme typing, Guan et al. (1995) first reported that L. turanica had been found in great gerbils from Xiaoguai Farm in Karamay, Xinjiang Uygur Autonomous Region. Although in recent years considerable work has been done to determine the heterogeneity of different isolates from three types of foci in China by analyzing kDNA and/or nDNA (e.g., Lu et al. 1994; Lu et al. 1997, 1998; Lu et al. 2001; Hu et al. 2002), the phylogenetic and taxonomical relationships among isolates covering China are still unclear. On the basis of kDNA and nDNA heterogeneity, Lu et al. (1994) classified 19 Leishmania isolates from epidemiologically different foci in China into five genotypes (groups I–V). They found that members of group II, tentatively designated as L. infantum sensu lato, displayed much heterogeneity in both kDNA and nDNAs. They further inferred that the isolates in China were more heterogeneous than previously thought, necessitating the reassignment of some isolates into different groups. Such a perspective is therefore still far from clearly promoting us to pursue further studies of the molecular phylogenetics of Leishmania.

Sequence data of the ribosomal RNA (rRNA) gene, in particular the two highly variable internal transcribed spacer regions (ITS1 and ITS2), have been successfully used to resolve taxonomic questions and to determine phylogenetic affinities among closely related Leishmania species (e.g., Dávila and Memen 2000; Schönian et al. 2000; El Tai et al. 2001; Kuhls et al. 2005; Parvizi et al. 2008) and other parasites (e.g., Chen et al. 2007; Lin et al. 2007; Al-Kandari and Al-Bustan 2010). Thus, the starting point of our work is to obtain a better understanding of the status of different isolates from different epidemic areas in China by DNA-based taxonomy. The ITS1–5.8S fragments were sequenced to analyze the phylogenetic relationships of ten isolates collected from different foci in China, in conjunction with seven Chinese isolates retrieved from GenBank. The aims of the present study are (1) to report a set of original ITS1–5.8S sequences for ten Leishmania isolates from different foci in China, (2) to determine the level of variation among ITS1–5.8S sequences among them, and (3) to infer the phylogenetic relationships among isolates in China and explore the taxonomic implications.

Materials and methods

DNA extraction, amplification, and sequencing protocols

The ten Leishmania isolates used in this study were listed in Table 1. Promastigotes were cultivated in medium 199 supplemented with 15% heat-inactivated fetal bovine serum at 25°C. Approximately 1–5 × 109 promastigotes were collected at room temperature by centrifugation at 4,000 rpm for 10 min and washed with distilled water. Total genomic DNA was extracted from the promastigotes using a standard sodium dodecyl sulfate-proteinase K procedure, as described by Sambrook and Russell (2001). The primers of LITSR (5′-CTGGATCATTTTCCGATG-3′) and L5.8S (5′-TGATACCACTTATCGCACTT-3′; El Tai et al. 2000) were used to amplify ITS1–5.8S segments. The PCR protocols were 94°C for 3 min followed by 30 cycles of 94°C for 30 s, 49°C for 30 s, 72°C for 1 min, and then a final elongation step at 72°C for 10 min. The PCR products were purified on a 2.0% agarose gel stained with ethidium bromide, using a commercial DNA purification kit following the manufacturer’s protocol. Sequencing was performed using the same PCR primers with ABI Big Dye Terminator chemistry on an ABI 3730 automated sequencer. The sequences have been deposited in GenBank under accession numbers HM130599–HM130608 (Table 1).

Table 1 List of Leishmania strains, origin, and database accession numbers, including sequences of L. donovani complex retrieved from GenBank
Full size table

Sequence alignment and analyses

A set of ITS1–5.8S sequences of Leishmania were retrieved from GenBank, including nine species of subgenus Leishmania and two species of subgenus Viannia (see Tables 1 and 2). The sequences were first aligned using Clustal X 1.83 (Thompson et al. 1997) with a gap-opening penalty of 5 and gap-extension penalty of 1, following the recommendation (use of small gap costs) of Hickson et al. (2000). The aligned matrix from this procedure was checked by eye, and minor adjustments were made manually with SeaView v.4.2.5 (Gouy et al. 2010). The data matrices are available from the corresponding author.

Table 2 List of the other strains, origin, and database accession numbers retrieved from GenBank
Full size table

Compositional heterogeneity was evaluated using Chi-square (χ 2) tests implemented in PAUP* 4.0b10 (Swofford 2002) and assessed using the software SeqVis v.1.3 (Ho et al. 2006) to visualize and to conduct matched-pairs tests of symmetry (Ababneh et al. 2006). Evidence of evolution under conditions more complex than that assumed by commonly applied models (i.e., stationary, reversible, and homogeneous conditions) was inferred if the scatter of dots in the tetrahedral plots was widely dispersed and if x% of the matched-pairs tests of symmetry produced p values greater than or equal to x; this procedure is consistent with that advocated by Jermiin et al. (2008). Substitution saturation was tested by inspecting a new entropy-based index as implemented in DAMBE (Xia and Xie 2001). For this approach, if I ss (i.e., index of substitution saturation) is not smaller than I ss.c (i.e., critical I ss), then we can conclude that the sequences have experienced severe substitution saturation (Xia et al. 2003; Xia and Lemey 2009). The K80 + G distance matrices (Kimura, 1980) were computed with MEGA v. 4.1 (Tamura et al. 2007), with the gamma shape of 0.5780.

Phylogenetic analyses

Phylogenetic hypotheses of Leishmania were generated with ITS1–5.8S rRNA segments using two types of commonly applied phylogenetic method: heuristic searches using equally weighted maximum parsimony (MP) analyses performed with the program PAUP* and Bayesian inference (BI) with the program MrBayes v.3.2 (Ronquist and Huelsenbeck 2003). In both MP and BI analyses, gaps were treated as missing data.

For heuristic searches under parsimony, invariant characters were removed from the dataset, and all remaining characters were treated as equally weighted. Each search involved ten random addition replicates, one tree held at each step, TBR branch swapping, steepest descent on, and a maximum of 10,000 saved trees; all other search settings were left at default values. Non-parametric bootstrapping was used to generate phylogeny confidence values (Felsensten 1985), with 1,000 pseudoreplicates using a heuristic tree search for each pseudoreplicate. Leishmania panamensis (FJ948422) was used to root the trees. Because intraspecific gene evolution cannot always be represented by a bifurcating tree, haplotype networks may more effectively portray the relationships among haplotypes within species (reviewed by Posada and Crandal (2001)).Therefore, we constructed unrooted parsimony networks of haplotypes for L. donovani complex and Leishmania sp. (see below) using TCS v.1.21 (Clement et al. 2000), with gap treated as a fifth state.

Prior to Bayesian analyses, the best-fit model of evolution, K80 + G, was selected using jModeltest v. 0.1.1 (Posda, 2008) under the Bayesian information criterion (Schwarz 1978), following recent recommendations (Posada and Buckley 2004). We estimated posterior probability distributions by allowing four incrementally heated Markov chains (default heating values) to proceed for four million generations, with samples taken every 200 generations. Analyses were repeated beginning with different starting trees to ensure that our analyses were not restricted from the global optimum (Huelsenbeck et al. 2002). Convergence was first tested by examining the average deviation of the split frequencies of the two runs, in order to determine whether the two runs had converged. MCMC convergence was also explored by examining the potential scale reduction factor (PSRF) convergence diagnostics for all parameters in the model (provided by the sump and sumt commands) and graphically using the cumulative, compare, and absolute difference options of the program AWTY (Nylander et al. 2008). The first one million generations, before this chain reached apparent stationarity, were discarded, and the remaining samples from the independent runs were pooled to obtain the final approximation of the posterior distribution of trees. To yield a single hypothesis of phylogeny, the posterior distribution was summarized as a 50% majority-rule consensus.

In addition, as gap (or “indel”) characters have been widely recognized as a valuable source of data for phylogenetic inference across the tree of life (e.g., Dessimoz and Gil 2010), phylogenetic information from indel events of ITS1 was also included in MP and BI by coding indel events into a separate data matrix with the program SeqState (Müller 2005) using the simple indel coding method (Simmons and Ochoterena 2000). In the latter, all indels are scored as binary characters regardless of their length. In BI, a discrete model employing identical rates of forward and backward transitions (Lewis 2001) was applied to the indel matrix.

Bayesian hypothesis testing

We used Bayes factors to compare our preferred Bayesian tree topology (see below) to Bayesian trees with constraint. This method differs from traditional hypothesis testing because it does not offer a criterion for absolute rejection of a null hypothesis but instead an evaluation of the evidence in favor of the null hypothesis (Kass and Raftery 1995). The phylogeny inferred from the ITS1–5.8S data set was constrained to alternative hypotheses. Constraint analyses were conducted in MrBayes v.3.2 using the command prset topologypr = constraint. All analyses consisted of two simultaneous runs each with an abbreviated three MCMC chains run for four million generations or more (as necessary). The Bayes factor was determined by calculating the marginal likelihood for both unconstrained and constraint analyses using Tracer v.1.5 (Rambaut and Drummond 2009). The difference in these ln-transformed marginal likelihoods was compared to the table provided by Jeffreys (1935, 1961) and further modified by Raftery (1996). Based on these tables, we consider a 2ln Bayes factor ≥10 as significant evidence for a hypothesis (Kass and Raftery 1995).

Results

Base composition and nucleotide substitution patterns

The newly determined ITS1–5.8S fragments ranged in size from 297 bp for isolate IPHL/CN/77/XJ771, 311 bp for isolate MHOM/CN/84/JS1, and 313 bp for the remaining isolates with exception of isolate MRHO/CN/88/KXG-2 (320 bp). Specifically, we found that the isolate MRHO/CN/88/KXG-2 shared the same ITS1 sequence with six strains of Leishmania turanica retrieved from GenBank, as listed in Table 2. The 5.8S rRNA segment was 69 bp in length. The alignment of the Leishmania taxa required accommodation of 85–112 gaps in the ITS1 region per sequence. Indels (insertion/deletion events) represented between 20.7% and 27.4% of the aligned sequence length. Most indels were 3–4 bp in length, and the maximum indel length was 19 bp. Of the 409 aligned characters, 111 were variable, with 89 parsimony-informative. Percentage base compositions were A, 31.76; C, 20.01; G, 22.54; T, 25.69. The average maximum likelihood estimated Ti/Tv ratio was 1.45.

A base stationarity test showed insignificant differences among taxa in base composition bias in the data (χ 2 = 22.34, df = 171, p = 1.00). Figure 1 presented the tetrahedral plot from the ITS1–5.8S rRNA. Clearly, there was no conspicuous compositional heterogeneity in the alignment. The implication of this plot was that these sites were likely to have evolved under the same stationary, reversible, and homogeneous conditions. To corroborate whether this was the case, the matched-pairs test of symmetry was used in conjunction with the alignment. Table 3 summarized the distribution of p values. The distribution of p values clearly showed that the evolutionary process was likely to have been stationary, reversible, and homogeneous, implying that it would be wise to analyze this data using a phylogenetic approach that assumes a stationary, reversible, and homogeneous evolutionary process. The observed I ss value of 0.764 was not significantly different from the I ss.c value of 0.692 for a symmetrical topology (p = 0.4757, two-tailed test) and was significantly greater than the I ss.c value of 0.362 for an asymmetrical topology (p = 0.0001, two-tailed test), suggesting that the ITS1–5.8S might have experienced substitution saturation.

Fig. 1
figure 1

Tetrahedral plots for ITS1–5.8S dataset, which were obtained using the Select Sites command from the View menu in the program SeqVis (Ho et al. 2006)

Full size image
Table 3 Summary of results from matched-pairs tests of symmetry
Full size table

K80 distances among the Leishmania species except Leishmania sp. ranged from near zero (between Leishmania braziliensis and L. panamensis) to 0.224 (between Leishmania mexicana and L. panamensis). Most pairwise comparisons mentioned above had divergence values of less than 0.224, with 0.102 on average. Meanwhile, the divergence between Leishmania sp. and other species ranged from 0.104 (Leishmania sp. versus Leishmania aethiopica) to 0.231 (Leishmania sp. vs. L. panamensis), with an average of 0.147 (Table 4).

Table 4 Pairwise genetic distances for ITS1–5.8S segments among Leishmania species in this study
Full size table

Phylogenetic relationships

The heuristic search of the ITS1–5.8S matrix resulted in 10,000 equally parsimonious trees of 149 steps, with high values of CI (0.8456) and RI (0.9599). In the strict consensus phylogram (Fig. 2), eight isolates in China formed a strongly supported clade (clade A; Leishmania sp.; BP = 100%) that was sister to the remaining members of subgenus Leishmania (BP = 100%). Leishmania amazonensis and L. mexicana formed a robust clade (BP = 100%) that was basal to all remaining subgenus Leishmania species (BP = 100%). Within the other members of subgenus Leishmania, L. donovani complex clustered with Leishmania tropica (BP = 66%), next joined by L. turanica plus L. gerbilli (BP = 52%), L. major (BP = 100%), and finally by L. aethiopica (BP = 100%). Nevertheless, the monophyly of L. tropica was not supported. Unexpectedly, the isolate MRHO/CN/88/KXG-2, identified as L. turanica by MLEE, did not cluster where expected, which appeared in the Leishmania sp. branch. When indels of the ITS1 were treated as additional characters, the heuristic search yielded 10,000 equally parsimonious trees of 298 steps, with high values of CI (0.7651) and RI (0.9435). As shown in Fig. 3, the consensus tree was similar to Fig. 2 with respect to the placements of clades A and B. The placements of other species were incongruent with those in Fig. 2.

Fig. 2
figure 2

Maximum parsimony consensus tree from 1,000 bootstrap replicates of ITS1–5.8S dataset by using PAUP*. Numbers above the branch represent percent recovery in bootstrap analysis (1,000 pseudoreplicates). `Tree length = 149, CI = 0.8456, RI = 0.9599

Full size image
Fig. 3
figure 3

Maximum parsimony consensus tree from 1,000 bootstrap replicates of ITS1–5.8S with indel coding by using PAUP*. Numbers above the branch represent percent recovery in bootstrap analysis (1,000 pseudoreplicates). Tree length = 298, CI = 0.7651, RI = 0.9435

Full size image

For the BI analyses, the likelihood value of the 50% majority consensus tree (Fig. 4) was ln L = −1,541.09. The average PSRF was 1.001. Overall, as well with maximum parsimony analyses, Leishmania sp., consisting of eight isolates from China, was sister to the subgenus species (PP = 0.76). Similarly, L. amazonensis and L. mexicana formed a robust clade (PP = 1.00) that was sister to all remaining subgenus Leishmania species (PP = 0.96). The relationships within the remaining species were similar to those of Fig. 3 except for recognizing the monophyly of L. tropica (PP = 0.56) instead of L. donovani complex. When the ITS1 indels were incorporated as additional characters, the resultant 50% majority consensus tree was shown as Fig. 5, with ln L of −2,167.78 and the average PSRF of 1.001. In this context, the topology is similar to Fig. 4 except that the monophyly of L. donovani complex was recovered with moderate posterior probability (PP = 0.88) instead of L. tropica.

Fig. 4
figure 4

The 50% majority-rule consensus tree inferred from Bayesian inference of ITS1–5.8S dataset by using MrBayes v. 3.2. Numbers at nodes represent Bayesian posterior probabilities

Full size image
Fig. 5
figure 5

The 50% majority-rule consensus tree inferred from Bayesian inference of ITS1–5.8S plus indel coding by using MrBayes v. 3.2, with indels treated in a manner similar to the simple gap coding outlined by Simmons and Ochoterena (2000). Numbers at nodes represent Bayesian posterior probabilities

Full size image

To get additional insight into the relationships among the L. donovani complex strains, we analyzed our data set, using the coalescent-based statistical parsimony network approach. The network of 13 haplotypes was shown as Fig. 6. H9 and H12 seemed to be central haplotypes, and the haplotype diversity was highest in Sudan. L. donovani revealed much more polymorphism than L. infantum despite a wider geographical distribution for the latter (see H3 in Fig. 6). L. infantum (H3) was most closely related to the H1 of L. donovani, with one mutational step. H2, shared by five strains from China, was also most closely related to H1. Having an advantage over the bifurcating tree in detail at the intraspecific level, the haplotype network could intuitively reflect the genetically greater distances between the singleton (H11) and one central haplotype (H12; five mutational steps, see Fig. 6). As shown in Fig. 7, the haplotype shared by GSH2 and GS2 (i.e., HM130602 and HM130603) was the interior haplotype of Leishmania sp. and may be older than any other haplotypes. There were three tip haplotypes harbored by GS6/GS1, SC10H2, and JS1, respectively. Similarly, this network reflected a greater distance between JS1 and the interior haplotype as four mutational steps.

Fig. 6
figure 6

Statistical parsimony network showing genetic relationships and distance among 13 haplotypes of L. donovani complex from different countries. Numbers of haplotypes correspond to Table 1. In the network, solid circles indicate sampled haplotypes; small hollow circles indicate unsampled or extinct haplotypes. Each mutation step is shown as either a short or longer line connecting neighboring haplotypes (including observed and unobserved one). The size of the solid circles roughly represents the numbers of strains carrying the haplotype, with the scale given beside the network; different filled patterns represent the corresponding geographical origin from which the haplotype was sampled

Full size image
Fig. 7
figure 7

Statistical parsimony network showing genetic relationships and distance among six haplotypes of Leishmania sp. from different sites in China. In the network, solid circles indicate sampled haplotypes; small hollow circles indicate unsampled or extinct haplotypes. Each mutation step is shown as either a short or longer line connecting neighboring haplotypes (including observed and unobserved one)

Full size image

Bayesian hypothesis testing

Bayes factor comparisons were summarized in Table 5. The analyses conducted reflect our primary interests of evaluating the inclusion of Leishmania sp. in L. donovani complex. As mentioned above, analyses of the ITS1 data resulted in a Chinese Leishmania clade that excluded L. turanica and L. donovani complex. Bayes factor analyses of the ITS1–5.8S incorporating indels coding were conducted to compare topologies with constraints to the optimal tree topology. In all cases, there was very strong (2ln Bayes factor >10) evidence against the constrained topologies.

Table 5 Summary of 2ln Bayes factor comparisons of alternative phylogenetic hypotheses
Full size table

Discussion

Probabilistic methods, namely maximum likelihood and BI, have progressively supplanted the MP method for inferring phylogenetic trees. One of the major reasons for this shift is that MP is much more sensitive to the Long Branch Attraction artifact than are probabilistic methods. Based on simulation studies, Philippe et al. (2005) found that MP can be affected by heterotachy and that it is much less efficient than probabilistic methods in dealing with all other evolutionary heterogeneities. Thus, in support of several recent studies (e.g., Gadagkar and Kumar 2005; Gaucher and Miyamoto 2005; Spencer et al. 2005), Philippe et al. (2005) strongly urged the continued preference of probabilistic methods for inferring phylogenies from real sequences. In our study, there is no significant compositional heterogeneity in ITS1 sequences, which are likely to have evolved under the same stationary, reversible, and homogeneous conditions. The MP analysis incorporating indels coding resulted in a topology that is congruent with the Bayesian trees. However, Bayesian support values in the BI tree (Fig. 5) were found to be comparatively higher than bootstrap values for the clades in that MP tree (Fig. 3), suggesting that the Bayesian inference could be properly applied to the phylogenetic analyses of subgenus Leishmania. Considering congruent with the recent studies on the interrelationships of subgenus Leishmania (Asato et al. 2009; Fraga et al. 2010), we tentatively support the relationships inferred from BI of the dataset incorporating indels coding (Fig. 5) as the preferred phylogeny.

As expected, the isolate MRHO/CN/88/KXG-2, previously identified as L. turanica by MLEE (Guan et al. 1995), clusters with L. gerbilli. This result is in congruent with the taxonomic scheme published by the World Health Organization (WHO 1990). Interestingly, a common allele was shared by the isolate MRHO/CN/88/KXG-2 with six other strains of L. turanica from Central Asia, as shown in Table 2. This further lends support that the MRHO/CN/88/KXG-2 belongs to L. turanica. On the other hand, as highlighted by Guan et al. (1992b), Leishmania parasite of the Karamay great gerbils (including isolate MRHO/CN/88/KXG-2) was pathogenic to monkey and man, causing cutaneous leishmaniasis. This medical characteristics, however, is different from that of L. turanica, being nonpathogenic to humans, as described by Strelkova et al. (1990). Thus, more isolates of L. turanica from different geographical areas and multiple loci are required for phylogeographic studies in order to clarify the intra-species genetic diversity and complex phylogeographic pattern.

The species concept has long been a matter of debate which is far more resolved (De Meeûs et al. 2003; de Queiroz 2007). As noted by Bañuls et al. (2002), any new species of Leishmania should be based on the clearly distinct phylogenetic approach of Tibayrenc’s discrete typing unit (Tibayrenc 1998). Eight new sequences reported in this work formed a robust clade (clade A; Leishmania sp.) that is sister to the remaining members of subgenus Leishmania. When we constrained all the isolates in China to form a monophyletic group, the tree obtained differs significantly from the BI tree based on 2ln Bayes factor comparison (55.856 > 10; see Table 5). Thus, we reject the monophyletic origin of Chinese Leishmania isolates and exclude the possibility that all the isolates in this study only belong to L. donovani complex. In addition, the mean genetic divergence (K80 distance) between Leishmania sp. and other species is 0.147, which is higher than that among several other species (mean 0.102). We further confirm that Leishmania sp. is an undescribed pathogenic species endemic in China, comprising isolates from hill foci, desert foci, and plain foci (see Fig. 7).

Judging from the kinetoplast and nuclear DNA heterogeneity, the isolate IPHL/CN/77/XJ771 was tentatively designated as L. infantum sensu lato by Lu et al. (1994). This hypothesis, however, is challenged by our molecular data. There is complete identity of the ITS1 sequence of the isolate IPHL/CN/77/XJ771 and that of a strain of L. donovani (MHOM/CN/00/Wangjie1). They share a common haplotype H2 with three isolates from eastern Jiashi County. As shown in the haplotype network (Fig. 6), H2 is most closely related to H1, which is shared by four strains from Sudan, one from India, and one from Ethiopia. Consequently, we argue that the isolate IPHL/CN/77/XJ771 is L. donovani instead of L. infantum.

The three isolates from eastern Jiashi County, causing desert type of zoonotic visceral leishmaniasis, were designated as L. infantum based on the genetic analysis of the ITS1 sequence (Wang et al. 2010). This conclusion, however, should be interpreted with caution, since it was not deduced from a robust phylogenetic tree, and the MP tree was misleading (see Fig. 3 in Wang et al. 2010). In contrast, as mentioned above, the coalescent-based statistical parsimony network approach provides additional insight into the relationships among the L. donovani complex. We infer that the three isolates should belong to L. donovani. As can be seen from Fig. 6, the isolate MHOM/CN/80/801 from Kashi city shares H6 with several isolates from India, Kenya, and Sri Lanka, and there is only one mutational step between H6 and a central haplotype H9 from Sudan. In combination with the results of Lukeš et al. (2007), the demonstration that the isolate MHOM/CN/80/801 belongs to L. donovani is corroborated, in accord with Wang et al. (2010).

In conclusion, phylogenetic analyses suggested that Chinese Leishmania isolates do not form a monophyletic group, but among which eight newly determined isolates form a monophyletic group, being sister to other members of subgenus Leishmania. The genetic distance analysis further provides evidence of the occurrence of an undescribed species of Leishmania. Our results also suggest that the isolate IPHL/CN/77/XJ771 is L. donovani; the three isolates from eastern Jiashi County, Xinjiang Uygur Autonomous Region belong to L. donovani instead of Leishmania infantum. In addition, the results of this study make an important contribution to understanding the heterogeneity and relationships of Chinese Leishmania isolates, further indicating that the isolates from China may have had a more complex evolutionary history than expected. However, more samples from different geographical areas and multiple independent evolving loci are required for phylogenetic studies in order to clarify the evolutionary history among Chinese Leishmania isolates. It further might be useful in understanding the links between clinical disease, geographic origin, and nomenclature of Leishmania species.