FLI1 polymorphism affects susceptibility to cutaneous leishmaniasis in Brazil

Abstract

Mapping murine genes controlling cutaneous leishmaniasis (CL) identified Fli1 as a candidate influencing resistance to L. major and enhanced wound healing. We examine FLI1 as a gene controlling CL and mucosal leishmaniasis (ML) caused by L. braziliensis in humans. Intron 1 single nucleotide polymorphisms tagging promoter and enhancer elements were analysed in 168 nuclear families (250 CL; 87 ML cases) and replicated in 157 families (402 CL; 39 ML cases). Robust case-pseudocontrol logistic regression analysis showed association between allele C (odds ratio (OR) 1.65; 95% confidence interval 1.18–2.29; P = 0.003) of FLI1_rs7930515 and CL in the primary sample that was confirmed (OR 1.60; 95% confidence interval 1.10–2.33; P = 0.014) in the replication set (combined P = 1.8 × 10⁻⁴). FLI1_rs7930515 is in linkage disequilibrium with the functional GAn microsatellite in the proximal promoter. Haplotype associations extended across the enhancer, which was not polymorphic. ML associated with inverse haplotypes compared with CL. Wound healing is therefore important in CL, providing potential for therapies modulating FLI1.

Leishmania infection is associated with a broad spectrum of clinical phenotypes and many studies have demonstrated that host genetic factors have a part in determining the outcome of infection (reviewed in refs. 1–4). L. braziliensis infection causes cutaneous leishmaniasis (CL) with prolonged time to lesion healing. Pro-inflammatory cytokines, including tumour necrosis factor and interferon-γ, and macrophage activation are important in eventual self-healing, but an exaggerated response is associated with mucosal leishmaniasis (ML).^5,6 A number of studies^7–11 have now been published that report on the role of polymorphisms at candidate genes (TNFA, SLC11A1, CXCR1, IL6, IL10, MCP1) associated with pro- and anti-inflammatory responses in regulating clinical disease outcome in L. braziliensis infection in humans. These studies have generally been carried out using small sample sizes and without replication in a second sample. Nevertheless, they have been underpinned by functional data^9–11 and/or are supported by previous immunological studies^5,6,12–14 demonstrating the importance of these pathways in determining disease outcome. Two of these studies^8,9 were undertaken using the set of families that we use here as a primary sample set, now strengthened by collection of a new replication sample. Thus far no genome-wide approaches to identify novel genes contributing to clinical outcome in L. braziliensis infection have been reported.

One alternative way to identify genes regulating clinical outcome in humans is via a mouse-to-human approach. This has proven quite successful in identifying genes that regulate visceral leishmaniasis in humans.¹ In the mouse, genome-wide linking mapping of genes controlling CL caused by L. major has demonstrated complex genetic control,^3,4,15–18 among which a particular role for wound healing genes has been proposed.¹⁹ Recent fine mapping in the region of chromosome 9 in mice (chromosome 11q24 in humans) has identified Fli1 (Friend leukaemia virus integration 1; FLI1 in humans) as a novel candidate influencing both resistance to L. major and enhanced wound-healing responses.²⁰ We also recently reported⁸ associations between CL caused by L. braziliensis infection in humans and polymorphisms at CXCR1/SLC11A1, which we interpreted in relation to their roles in regulating polymorphonuclear neutrophils, macrophages and/or dendritic cells in wound-healing responses. Here we report on primary and replication family-based genetic association studies that support a role for polymorphism at FLI1 in determining susceptibility to CL caused by L. braziliensis in Brazil.

To undertake our study, we selected four single nucleotide polymorphisms (SNPs; Table 1) that tagged the first two of four major linkage disequilibrium blocks in the large (>50 kb) intron 1 of the FLI1 gene (Figure 1a). The SNPs selected also tagged the proximal promoter region that contains a functional GAn microsatellite,²¹ as well as the CpG island that spans the proximal promoter region and the 5′ region of intron 1 (Figure 1a). We focused on this region extending into intron 1 of FLI1 because a previous study²² had demonstrated a functional ETS/ETS/GATA (E-twenty six family of transcription factors;²³ GATA family of transcription factors that bind to the sequence GATA²⁴) enhancer element residing in this intron (Figure 1a). FLI1 is itself a member of the ERG sub-family of ETS transcription factors.²³ The 4 tag-SNPs (Table 1) had minor allele frequencies >0.1 in the HapMap populations representative of the three major ethnicities (Caucasian, African and Asian) that contribute to this population in Brazil.

Table 1.

Details of the FLI1 intron 1-genotyped SNPs including allele frequencies for Caucasian (CEPH), Asian (HCB) and sub-Saharan (YRI) populations as recorded in Build 37.1 of the NCBI Entrez SNP database³²

Gene/SNP	Physical position (bp)	Allelesa	MAF Brazil	Caucasian	Asianb	Africanb
FLI1_rs7930515	128564794	C>A	0.291	0.259	0.411	0.254
FLI1_rs619521	128578157	A>G	0.422	0.483	0.178	0.342
FLI1_rs590520	128620132	C>T	0.362	0.289	0.344	0.395
FLI1_rs531894	128628756	A>G	0.418	0.375	0.602	0.425

Abbreviations: MAF, minor allele frequency; SNP, single nucleotide polymorphism.

^aMajor>minor alleles for this Brazilian population, which matches major>minor for the Hapmap CEU population.

^bFrequency for the allele that is a minor allele in Hapmap CEU population.

Figure 1.

Gene structure of FLI1 showing position of SNPs used in this study in relation to linkage disequilibrium (LD) blocks and functional regulatory elements. (a) A diagrammatic representation of gene structure, the location of the GAn repeat and the ETS/ETS/GATA enhancer, the positions of SNPs, and the four large LD blocks (as determined for the CEU HapMap sample) that cover the gene. (b) Provides more detail of the first LD block in intron 1 that contains the associated SNP rs7930515 relative to the functional regulatory elements. D′ values and confidence levels (LOD) are represented as black for D′ = 1, LOD>2; shades of grey for high D′< 1, LOD>2; white (none present) for D′< 1, LOD< 2. The numbers within the squares represent the D′ scores for pairwise LD. Where no numbers occur that D′ value is 1 (that is, 100%).

Initially we genotyped these SNPs in 168 nuclear families (Table 2) used in our previous studies^8,9 that contain 250 CL cases and 87 ML cases collected in the area of Corte de Pedra, Bahia, Brazil, where L. braziliensis is endemic. Using robust case-pseudocontrol conditional logistic regression analysis (Table 3), we demonstrated association between the common allele C (odds ratio (OR) 1.64; P = 0.003) at SNP rs7930515 and risk of CL, which was replicated (OR 1.60; P = 0.014) in a second sample (Table 2) of 402 CL cases in 157 nuclear families (combined OR 1.62; P = 1.8 × 10⁻⁴; Table 3) collected in the same region of Corte de Pedra. Application of a Bonferroni correction for four independent SNPs provides a significance cutoff of P ≤ 0.013 (that is, P = 0.05/4), which is achieved at rs7930515 in primary and combined samples. Power to detect association with ML disease in our study is limited by sample size. Single point association analysis undertaken in TRANSMIT (http://www-gene.cimr.cam.ac.uk/clayton/software/transmit.txt) (Figure 2) using the combined primary and replication samples was suggestive of association at this SNP for ML. This showed overtransmission of the minor allele A at SNP rs7930515, consistent with the observation that analysis of CL and ML cases together as leishmaniasis per se showed reduced significance compared with analysis of cases with a long history of CL only disease. Haplotype analysis in TRANSMIT (Figure 2 and Supplementary Figure 1) demonstrated (a) that haplotypes controlling CL extended across the remaining tag-SNPs, strongest over the region containing the enhancer element, and (b) that ML was associated with the inverse 2.1.2 ( = A.A.T) rs793515_rs6199521_rs590520 haplotype compared with CL (1.2.1 = C.G.C; Figure 2). Data for one degree of freedom tests for specific haplotype associations presented in Figure 2 are supported by global test statistics (Supplementary Figure 1) in which the total number of haplotypes is taken into account as indicated by the degrees of freedom. Diminishing strength of the CL haplotype associations (Figure 2; Supplementary Figure 1) moving 3′ across the four SNPs concurs with the observation in this Brazilian population that, although D′ between rs7930515 and rs619521 (D′ = 0.62) and between rs590520 and rs531894 (D′ = 0.86) are relatively high, r² values that take allele frequencies into account are low (0.11 and 0.29, respectively) and below the stringent cutoff (r² = 0.8) we set in choosing tag SNPs. In summary, the association between FLI1 and CL observed in primary and replication samples appears to map within linkage disequilibrium block 1 (Figure 1b) to the region that is most strongly tagged by SNP rs7930515 and closest to the proximal promoter containing the functional GAn microsatellite repeat, with haplotype associations extending across the region of intron 1 containing the functional ETS/ETS/GATA enhancer element proximal of SNP rs619521. Interestingly, no SNPs are observed in public domain databases within the region of the ETS/ETS/GATA enhancer element (Supplementary Figure 2), attesting to its importance as a highly conserved regulatory element.²² The FLI1 promoter is itself upregulated by ETS factors ETS1, ETS2, FLI1 and ELF1, in combination with GATA factors.²⁵ Recent improved motif identification and analysis of ChIP-Seq data demonstrated that FLI1 activates gene transcription when its binding site is located in close proximity to the gene transcription start site (up to ~150 kb), especially when it contains a microsatellite sequence.²⁶ Hence, there may be strong functional significance to the maintenance of linkage disequilibrium across the proximal promoter and intron 1 of FLI1 containing the GAn microsatellite and the ETS/ETS/GATA enhancer.

Table 2.

Characteristics of the primary and replication samples

	Primary sample			Replication sample
	CL	ML	Leishmaniasis per se	CL	ML	Leishmaniasis per se
No cases	250	87	337	402	39	441
Males	128	60	188	219	24	243
Females	122	27	149	183	15	198
Age at disease
Mean	19.1	30.3	22.4	21.5	26.6	21.9
95% CI	17.1–21.2	25.8–34.7	20.3–24.4	20.1–22.9	20.7–32.4	20.6–23.3
No nuclear families	168	—	168	157	—	157
Total N in families/trios	767	—	767	764	—	764

Abbreviations: CI, confidence interval; CL, cutaneous leishmaniasis; ML, mucosal leishmaniasis.

The study was conducted in the area of Corte de Pedra, Bahia, Brazil, where L. braziliensis is endemic. Corte de Pedra is in a region of rural rain forest, where agriculture underpins the local economy. Around 3300 subjects were interviewed during 4 years to select the study population. The primary sample was collected during the period 2000–2004; the replication sample during 2008–2010. For the primary sample, index cases of ML were ascertained from medical records of the Corte de Pedra Public Health Post, families visited and blood samples for DNA collected from all ML and CL cases, as well as available unaffected family members. For the replication sample, case samples and attending parents were sampled at the health post, and additional family members collected in follow-up visits to the household. The case definition of ML is a characteristic mucosal lesion with either parasitological confirmation or two of the three following criteria: positive delayed-type hypersensitivity test, positive leishmania serology and a histopathology suggestive of leishmaniasis. All cases in the current study also responded to antileishmanial therapy. CL is defined as the presence of a single chronic ulcerative lesion at a skin site without evidence of mucosal involvement and without evidence of dissemination to 10 or more sites (disseminated leishmaniasis), also confirmed by detection of parasites or two of the three criteria listed above. Past cases that have been treated in the health post of Corte de Pedra have their diagnoses confirmed by medical records using the same criteria defined above and all cases were examined for detection of a characteristic well-delimited scar. Informed consent was obtained from all the participants and the research was approved by the ethical committee of the Hospital Universitário Professor Edgard Santos, Salvador, Brazil. Transmission disequilibrium test (TDT) power approximations³³ show that 250 primary, 402 replication and 652 combined CL trios had ≥53%, ≥76%, and ≥94% power, respectively, to detect an odds ratio ≥1.5 at P = 0.01 for markers with minor allele frequency ≥0.2. For leishmaniasis per se, 337 primary, 441 replication, and 778 combined CL trios had ≥81%, ≥81% and ≥97% power, respectively, to detect an odds ratio ≥1.5 at P = 0.01 for markers with minor allele frequency ≥0.2. In contrast, primary, replication and combined ML trios all had ≤20% power for the same parameters. As markers used for association analysis were not known to be causal variants, or to be in complete LD with a functional variant, robust association tests were performed to take account of multiple trios within a pedigree. It is possible that some CL cases in our study could progress to ML disease at a later date. Epidemiological studies show that this will affect < 4% of CL patients,³⁴ thus representing a small reduction in the power of our study to detect CL-specific genetic effects. Demographical, epidemiological and phenotype characteristics of these subjects were previously described in full.⁹

Table 3.

Results of robust CPC analysis for FLI1 tag SNP rs7930515 (see Table 1 and Figure 1) for transmission of alleles CPC analysis for transmission of alleles from heterozygous parents to CL, ML and L. braziliensis per se (CL and ML) individuals in families

Gene/marker	Sample set	Phenotype	Allele	# case/pseudo-control sets	OR	95% CI	P-value
FLI1_rs7930515	Primary	CL	C	78	1.65	1.18–2.29	0.003
	Replication	CL	C	131	1.60	1.10–2.33	0.014
	All	CL	C	209	1.62	1.26–2.09	1.8 × 10−4
	Primary	ML	C	23	1.25	0.64–2.42	0.44
	Replication	ML	C	16	0.44	0.16–1.24	0.12
	All	ML	C	39	0.90	0.52–1.58	0.73
	Primary	CL+ML	C	101	1.54	1.09–2.18	0.014
	Replication	CL+ML	C	146	1.36	0.96–1.94	0.083
	All	CL+ML	C	247	1.45	1.13–1.85	0.003

Abbreviations: # case/pseudo-control sets, number of informative transmissions; CI, confidence interval; CL, cutaneous leishmaniasis; CPC, case/pseudo-control; ML, mucosal leishmaniasis; OR, odds ratio; SNP, single nucleotide polymorphism.

SNPs rs619521, rs590520 and rs531894 (see Table 1 and Figure 1) did not show significant single point associations, as demonstrated in Figure 2 for the combined analysis of families in TRANSMIT. Genotyping was performed in Cambridge using Taqman technology (Applied Biosystems, CA, USA) for SNPs presented in Table 1. All SNPs were in Hardy–Weinberg equilibrium in genetically unrelated founders of the families (data not shown). PEDCHECK³⁵ was used to determine Mendelian inconsistencies within families. Missingness (that is failure to score on Taqman assays) ranged from 4.6% (43/935 individuals genotyped) to 10.3% (97/935) across the four SNPs. Association analyses were performed under an additive model using family-based CPC analysis,³⁶ where each affected offspring is matched to one to three pseudo-controls that derive from the remaining possible genotypes of the parental mating. OR, 95% CI and P-values are calculated using robust conditional logistic regression (rCLOGIT) models employing a robust sandwich estimator of variance and a Wald χ²-test statistic to control for clustering of trios within pedigrees. CPC was implemented in STATA v8.0 (http://www.stata.com/).

Bold entries indicate P-values significant at P< 0.05.

Figure 2.

Results of the single point (global P-value) and haplotype (1 degree of freedom tests for individual haplotypes) analyses performed in TRANSMIT for combined families from primary and replication samples. TRANSMIT analyses, which use the EM algorithm to take account of missing parental data, were performed using the robust-ro flag to take account of family clustering.³¹ Results are shown for over and undertransmitted haplotypes with estimated frequencies >0.1 for (a) CL and (b) ML. The single point analysis concurs with the CPC analysis presented in the main text (P-values differ because TRANSMIT infers data for missing parents), with only SNP rs7930515 showing significant single point association with CL. Only haplotype associations with nominal P≤0.05 are shown. The overtransmitted haplotypes (1.1 and 1.2) across the two 5′ SNPs rs793515 and rs619521 show that the association maps to rs793515 for CL, but there is evidence of haplotype associations that extend across the region from the GAn repeat to the enhancer element (see Figure 1).

Here we have demonstrated association (combined OR 1.62; 95% confidence interval 1.26–2.09; P = 1.8 × 10⁻⁴) between FLI1 and susceptibility to CL caused by L. braziliensis. This follows on from genetic and functional mapping of the lmr2 gene controlling cutaneous lesions caused by L. major on murine chromosome 9 to the Fli1 gene.²⁰ Resistance to L. major correlated with a wound-healing response that presented in congenic resistant mice as a large population of fibroblasts and an organized and abundant deposition of collagen bundles in the absence of inflammatory cells. Recent studies have shown an association between enhanced type I collagen expression and epigenetic repression of the FLI1 gene.²⁷ In congenic susceptible mice, response to wounding was associated with a larger population of acute inflammatory cells with sparse and disorganized collagen bundles. In a previous study⁹ we found an association between CL caused by L. braziliensis and a regulatory polymorphism in the promoter of the gene IL6 encoding interleukin 6. Homocysteine-dependent stimulation of interleukin 6 has recently been reported²⁸ to upregulate genes essential for epigenetic DNA methylation (DNA (cytosine-5-)-methyl-transferases (Dnmts) via expression of FLI1. Homocysteine increases the CpG methylation status (and hence represses gene expression) of the CpG-rich proximal promoter of the lysyl oxidase (LOX) gene,²⁸ an extra-cellular copper enzyme that initiates the cross-linking of collagens and elastins. Inhibition of interleukin 6 reverses this repression. Regulation of collagen expression and organization may thus involve epigenetic regulation at both FLI1 and LOX genes, consistent with the presence of the CpG motif across the region of the functional FLI1 promoter elements. This suggests that, although there are many immune-related functions for both interleukin 6 and FLI1 that could account for association with CL caused by L. braziliensis, there may be a direct functional link between these two genes that mediates resistance or susceptibility to infection through the wound-healing response. This, in turn, might provide a novel therapeutic opportunity. For example, the use of imatinib mesylate has been proposed for treatment of systemic sclerosis,^29,30 an autoimmune disorder similarly resulting from immune activation, fibrosis development and damage of small blood vessels, in which FLI1 is downregulated through an epigenetic mechanism.³⁰ Imatinib mesylate reverses the expression levels of FLI1. Further work will be required to analyse expression levels of FLI1 in tissue biopsies from L. braziliensis patients to determine its potential as a therapeutic target. Overall our results strengthen the potential role of genes/mechanisms associated with wound healing in CL, and suggest that a broader analysis of pathways involved in wound-healing responses may contribute to a better understanding of the pathogenesis of disease.