Since young and elder children, males and females were infected during the outbreak, transmission was likely to occur in the village around the huts. This conclusion was further supported by the high number of phlebotomine sand flies that were trapped in the huts. Among them, Phlebotomus orientalis, which represented 5% of the collected sand flies, was likely responsible for the indoor transmission of L. donovani [21]. Since phlebotomine sandflies require a specific environment for their reproduction and nutrition and do not fly far, the distribution of phlebotomines breeding sites could have well influenced KA spreading in the village. It was found that a few environmental variables that could have influenced vector distribution, were associated with KA [22]: these include the presence of Acacia trees (which produce a gum on which sandflies feed), of neems (Azadirachta), which are known as insect repellents, and the presence of domestic animals near the house probably because they attract sandflies and because their faeces support the development of sandfly larvae. Interestingly, these factors have influenced the risk of KA in the first two to three years of the outbreak; later, these environmental factors were no longer associated with KA, probably because the disease had spread throughout the whole village, allowing a high infection rate in sandflies (transmission is mostly anthroponotic). The clustering of the huts had also probably contributed to the homogenization of parasite transmission.

A most interesting observation was made in the analysis of the parasite population that was collected from subjects with active disease and from dogs. Isoenzyme analysis revealed a large heterogeneity in the collected Leishmania strains [23,24]. No less than seven different zymodemes were found in human cases, all being closely related and holding ancestral position on the phylogenetic tree, which places Sudan as a possible original focus of VL, from which the disease has spread throughout the world [24]. Unfortunately, the virulence/pathogenicity of the different isolates was not evaluated; nevertheless this observation indicates that the Leishmania population had evolved during the outbreak, possibly allowing the emergence of parasites with enhanced capabilities to escape immune destruction and to bypass certain host barriers.

The village comprised Aringa (2/3), Fellata (1/6) and Haousa (1/6), who were living close to each others. The risk of KA was found to be 4.7 (1.9–11.7) and 2.2 (0.9–5.5) fold higher in Aringa than in Haoussa and Fellata, respectively [22]. In contrast, the prevalence of infections (with or without overt clinical disease) at the end of the outbreak was comparable in all ethnic groups. There was also a family clustering of KA cases [22] that could not be related to ecological factors (Fig. 2). KA distribution in families was not, however, consistent with a single major gene control and testing the hypothesis of a major locus by segregation analysis yielded negative results [25]. This could have several explanations, including clinical heterogeneity, unappreciated differences in environmental risk factors or a polygenic control.

Nevertheless, familial clustering of KA and differences in disease prevalence in ethnic groups suggested an influence of host genetics on susceptibility to KA. Then linkage analysis was carried out on affected siblings with genetic markers of regions containing obvious candidate genes. Analysis of the data was performed using the Maximum Likelihood Binomial (MLB) model-free method [26,27] that accommodates several cases in the same families. The study was concentrated on the Aringas to reduce the genetic heterogeneity of the study sample.

The genotypes of 172 individuals from 37 families containing two to five affected sibs (97 children) were determined. The mean age of onset was 10 years (3 to 27 years). No evidence of linkage was obtained with the micro satellite markers of the IFNGRI, or HLA/TNF or 5q31–q33 region (LOD scores 0.14). A suggestive LOD-score was obtained with D12S92 (LOD score = 0.94; p=0.018) in the IFN-γ locus, but this finding was not supported by the results of the multipoint analysis with all of the markers typed in this region (maximum multipoint LOD score = 0.58; p=0.05). Significant evidence of linkage was obtained, however, for a 5′(CA)n in the NRAMP1 promoter (LOD score = 1.32; p=0.007), although this polymorphism was not very informative for linkage. Only three alleles were detected at this site (allele 1 = 199 bp; allele 2 = 201 bp; allele 3 = 203 bp), and linkage information could not be obtained from 61% of the families. The information content of the multipoint analysis combining all NRAMP1 markers was higher (75%), and the LOD score was 1.08 (p=0.01).

This finding was the first observation suggesting that NRAMP1 alleles were involved in the control of human KA. The different alleles at this polymorphic site have different effects on the transcription of NRAMP1 in vitro [28]. In our family-based association analysis, no distinct NRAMP1 polymorphisms were associated with an increased risk of VL. A much larger sample would have been necessary to reveal such an association. Together with the results observed in human tuberculosis [29,30] leprosy [31,32] and HIV [33], our data support the view that NRAMP1 is implicated in the control of intracellular pathogens in humans and in mice. A recent study in Sudan further supports our conclusions: that study has analyzed the genetic control of KA in a village that is close to our study population. This study replicated the linkage with polymorphisms in NRAMP1 and demonstrated that the main effects was contributed by a SNP in intron 4 [34]. These polymorphisms had been previously associated with susceptibility to tuberculosis [29]. Since the extent of linkage disequilibrium upstream the promoter region has not been determined, it cannot be totally excluded that a polymorphism outside the tested region could be responsible for the observed association. This is however unlikely, given the large body of literature on the role of NRAMP1 in both mouse and human susceptibility to intra cellular pathogens including L. donovani. These results altogether indicate a contribution of NRAMP1 to the susceptibility of Sudanese (Aringa and Massalit, which are close ethnic groups) subjects to KA.

A two-step genome-wide linkage study was carried out on the same affected siblings. In a first step, 38 families including two or more siblings affected with KA were genotyped for 367 micro satellite markers, covering the whole genome and providing a 10–15-cM primary genetic map. Regions located on chromosomes 22q12, 2q24 and 2q35 yielded maximum multipoint LOD scores of 0.83 (p0.025): LOD score = 2.9; 1.45 and 0.97, respectively. Two additional regions (14q13–14q23 and 3q27–3qter) provided LOD scores that were slightly lower than 0.83. All other regions provided LOD scores of 0.73 (p0.03).

In the second step of the genetic analysis, the three regions that presented suggestive linkage with KA were further investigated in 25 additional families and typing 13 additional markers at the 22q12, 2q23–2q24 and 2q31 (NRAMP1) loci. The multipoint linkage analysis of chromosomes 2q23–q24 and 2q35 yielded similar maximum LOD scores (1.53 and 0.79, respectively) as in the first step. In contrast, LOD scores in region 22q12 increased to 3.50 (P=3×10−5) at marker D22S280 (Fig. 3). The linkage of KA with that locus is significant on a genome-wide basis [35].

We repeated the analysis only on early KA cases, because we had observed that early and late KA cases might be associated with different immunological responses. KA recorded early in the outbreak had occurred rapidly after primo-infection, whereas later in the outbreak, KA developed in subjects who had seroconverted several months before. Furthermore, adult KA was more frequent late in the outbreak, whereas children were affected at early times in the outbreak (Fig. 4). In this analysis, 52 out of the 63 families remained informative (with at least two affected children). An increased maximum LOD score of 3.90 (P=10−5) was obtained for a position 0.5 cM away from the peak obtained when considering all VL cases (Fig. 3).

We then investigated possible gene–gene interactions between the major 22q12 locus and the other two regions 2q23–q24 and 2q35. A conditional analysis showed no significant positive gene–gene interactions. When looking for negative gene–gene interactions; the single point LOD score increased markedly at the two adjacent 2q23 markers D2S141 and D2S142 (LOD scores = 3.74 and 3.84, respectively), using the 30 families displaying negative LOD scores at D22S280. However, the multipoint analysis yielded rather uniform LOD scores between 1.75 (p=0.003) and 2.29 (p= 0.0006) across a 30-cM region extending from D2S151 to D2S2330. The conditional analysis moderately increased the multipoint LOD scores from 0.79 to 1.00 (p=0.015) in the 2q35 region and the maximum LOD score was observed with NRAMP1 genetic markers.

The 22q12 region contains several genes whose alleles could modulate susceptibility to KA including the IL-2 receptor β chain gene (IL-2RB), which is located in the 22q12 chromosomal region close to D22S280. IL2RB is a good candidate for the control of KA. IL-2 is critical for T-cell proliferation and differentiation. T-cell responses and IL-2 production are strongly suppressed in VL patients [36,37] and high levels of circulating soluble IL-2 receptor have been observed during the acute phase of the disease [38]. Studies on other protozoan parasites such as Trypanosoma cruzi also support the hypothesis that some parasites evade the immune system by inhibiting the transcription of the IL-2R γ and β chain genes [39]. Interestingly, two other genes, NMI (N-myc interactor) and STAM2 (Signal-Transducing Adapter Molecule 2), located in the 2q22–q23 region, encode proteins involved in the IL-2 signalling transduction pathway.

This led us to sequence the genes in several affected and non-affected cases. The association of the most frequent (frequency 20%) was tested. Only a weak association was observed. Work is ongoing to evaluate further this gene – and other candidate ones – in a larger population.

Thus, susceptibility to KA in Barbar's population is determined by a major gene on the chromosome 22q12 and by mutations in NRAMP1. The data also suggest a control by gene(s) in the 2q22–q23 region. The proportion of alleles at the 22q12 locus, shared by affected siblings at the peak LOD score in region 22q12 was 0.63 (all KA cases) and 0.66 (KA before 1999). The locus-specific sibling recurrence-risk ratio (λs) was estimated by using the commonly used relationship between λs and the proportion of affected sib-pairs sharing zero alleles [40]: λs was estimated to be 1.83 (all KA) and 2.11 (VL before 1999), which may indicate a strong genetic effect. λs values for many infectious diseases are not expected to be very high, as λs cannot be greater than 1/disease prevalence (1/0.3=3.33 in this outbreak). λs was 1.66 for a major susceptibility locus in leprosy [41]. The data also indicate that the genetic contribution of the 22q12 locus was important to explain cases of KA that occurred early in the outbreak, as the 30 patients who developed KA in 1999 and 2000 did not contribute much to the LOD score observed on chromosome 22q12. This confirms that the KA phenotype is heterogeneous and that the outbreak has progressively affected different subgroups of subjects, as illustrated in Fig. 4. The observation that a second locus on 2q23–q24 is likely involved in the control of KA confirms this view. Moreover, an other study carried out in a village of the same region of Sudan has reported an association of KA with polymorphisms in or close to IL4 [42]. However, attempts to show associations with genes in the HLA–TNF region has failed in L. chagasi [43], L. infantum [44] and L. donovani [45,46] infections.

Such genetic heterogeneity might be a characteristic of Leishmania infections by comparison with Schistosoma infections, which are influenced by a few major loci confirmed in several populations [47]. Several factors might have contributed to such heterogeneity, first the heterogeneity of the studied phenotype (KA is a broad phenotype and alterations in more than one pathways can cause KA); second, the Leishmania population in Barbar was heterogeneous and has evolved with time: various zymodemes were identified during the outbreak and the composition of the parasite population was different at the end as compared to the start of the outbreak [24,48]. Thus, Leishmania virulence may evolve with time, enabling the parasite to infect subjects who resisted in the early stage of the epidemics. So far, however, there has been no clear demonstration of such an evolution of parasite virulence. Studies in experimental models using strains isolated from patients with cutaneous leishmaniasis have shown, however, that a Leishmania isolate collected from patients with a poor response to treatment escape immunological mechanisms that control less pathogenic isolates [49]. No such observations were made on schistosomes and there is no evidence for different pathogenicity of isolates from subjects (from the same region) with different clinical diseases; most studies rather support the view that the schistosoma genome, unlike the leishmania genome, do not evolve rapidly in order to escape immunological or chemotherapeutical pressures.