and Immunity (2004) 5, 46–57 & 2004 Nature Publishing Group All rights reserved 1466-4879/04 $25.00 www.nature.com/gene Evidence for a cluster of genes on 17q11–q21 controlling susceptibility to tuberculosis and leprosy in Brazilians

SE Jamieson1, EN Miller1, GF Black1, CS Peacock1, HJ Cordell1, JMM Howson1, M-A Shaw2, D Burgner3,6,WXu4, Z Lins-Lainson5, JJ Shaw5,7, F Ramos5, F Silveira5 and JM Blackwell1 1Cambridge Institute for Medical Research, University of Cambridge School of Clinical Medicine, Addenbrookes Hospital, Cambridge, UK; 2Department of Biology, University of Leeds, Leeds, UK; 3The Wellcome Trust Centre for Human Genetics, Oxford, UK; 4Wolfson Institute for Biomedical Research, Rayne Institute, University College, London, UK; 5Instituto Evandro Chagas, Belem, Brazil

The region of conserved synteny on mouse chromosome 11/human 17q11–q21 is known to carry a susceptibility (s) for intramacrophage pathogens. The region is rich in candidates including NOS2A, CCL2/MCP-1, CCL3/MIP-1a, CCL4/MIP-1b, CCL5/RANTES, CCR7, STAT3 and STAT5A/5B. To examine the region in man, we studied 92 multicase tuberculosis (627 individuals) and 72 multicase leprosy (372 individuals) families from Brazil. Multipoint nonparametric analysis (ALLEGRO) using 16 microsatellites shows two peaks of linkage for leprosy at D17S250 (Zlr score 2.34; P¼0.01) and D17S1795 (Zlr 2.67;

P¼0.004) and a single peak for tuberculosis at D17S250 (Zlr 2.04; P¼0.02). Combined analysis shows significant linkage (peak

Zlr 3.38) at D17S250, equivalent to an allele sharing LOD score 2.48 (P¼0.0004). To determine whether one or multiple genes contribute, 49 informative single nucleotide polymorphisms were typed in candidate genes. Family-based allelic association testing that was robust to family clustering demonstrated significant associations with tuberculosis susceptibility at four loci separated by intervals (NOS2A–8.4 Mb–CCL18–32.3 kb–CCL4–6.04 Mb–STAT5B) up to several Mb. Stepwise conditional logistic regression analysis using a case/pseudo-control data set showed that the four genes contributed separate main effects, consistent with a cluster of susceptibility genes across 17q11.2. Genes and Immunity (2004) 5, 46–57. doi:10.1038/sj.gene.6364029

Keywords: chromosome 17q11-q22; tuberculosis; leprosy; genetic susceptibility

Introduction (eg CCL2/SCYA2/MCP-1, CCL3/SCYA3/MIP-1a, CCL4/ SCYA4/MIP-1b, CCL5/SCYA5/RANTES), the gene Many studies support the hypothesis that susceptibility (CCR7) encoding the receptor for CCL19/CCL21 and to Mycobacterium tuberculosis and M. leprae is genetically genes for signal transducers and activators of transcrip- regulated in humans.1,2 One approach to identifying tion STAT3, STAT5A and STAT5B. candidate genes is to study genetic susceptibility in mice The shared intramacrophage niche of leishmanial and and examine the regions of conserved synteny in man. mycobacterial (M. tuberculosis and M. leprae) pathogens, We took this approach to analyse susceptibility to and the similar spectra of disease caused by the two leishmanial infections.3,4 This led to identification of a groups of organisms, means that any gene identified as a region on mouse chromosome 11, syntenic with human susceptibility gene for one group of pathogens becomes a chromosome 17q11.1–q12, which carries susceptibility candidate for the others. A well-characterized example of genes to cutaneous leishmaniasis.5,6 This region is rich in this is the gene (Nramp1/NRAMP1) located on mouse candidate susceptibility genes. These include the gene chromosome 1/human 2q35 that encodes the natural (NOS2A) encoding the inducible form of nitric oxide resistance associated 1, now rede- synthase (iNOS), genes encoding members of the family signated as solute carrier family 11a member 1 (Slc11a1 in of small inducible gene (designated, for mice; SLC11A1 in man). This gene controls innate example, chemokine, CC motif, ligand 1 or CCL1) cluster resistance to Leishmania donovani,7–10 M. bovis BCG,11 M. lepraemurium12,13 and M. intracellulare14 in the mouse, 15–22 Correspondence: Dr JM Blackwell, Cambridge Institute for Medical and contributes to susceptibility to M. tuberculosis, Research, Wellcome Trust/MRC Building, University of Cambridge School M. leprae23 and L. donovani24,25 in man. of Clinical Medicine, Addenbrookes Hospital, Hills Road, Cambridge CB2 The successful mouse-to-man precedent set with 2XY, UK. E-mail: [email protected] Slc11a1/SLC11A1 encouraged us to take a similar 6Present address: Princess Margaret Hospital, Subiaco, Western approach in studying the region of human chromosome Australia, Australia. 17q syntenic with murine chromosome 11 in relation to 7Present address: Department of Parasitology, Institute of Bio- medical Sciences, Sao Paulo University, 05508-900, Brazil. genetic susceptibility to leprosy and tuberculosis in man. Received 17 April 2003; revised 14 August 2003; accepted 19 August Here we present linkage and allelic association data for 2003 polymorphic loci at 17q11–q21 from 92 multicase families Genetic susceptibility to tuberculosis and leprosy SE Jamieson et al 47 Table 1 Family structures for the 92 tuberculosis and 72 leprosy families collected from Belem, Brazil

Number

Families structure TB Leprosy per se Leprosy LL subtype Leprosy TT subtype

No. of families 92 72 72 72 No. of nuclear families 118 87 56a 48a Nuclear families with 1 affected sib 58 19 27 22 Nuclear families with 2 affected sibs 37 50 25 17 Nuclear families with 3 affected sibs 17 7 3 7 Nuclear families with 4 affected sibs 4 5 1 1 Nuclear families with 5 affected sibs 2 5 0 0 Nuclear families with 6 affected sibs 0 0 0 1 Nuclear families with 7 affected sibs 0 1 0 0 No. of affected offspring 209 192 90 87 No. of affected parents 71 41 25 10 Total no. of affected individualsb 280 208c 109 93 Total no. of individuals 627 372 372 372

Nuclear families with a single affected offspring were always part of an extended multicase pedigree. aIn all, 17 nuclear families contain both LL and TT affected individuals; therefore, the numbers of LL+TT nuclear families do not add up to leprosy per se. bDue to pedigree structure, some individuals are classed as both sibs and parents in different nuclear families; therefore, the total number of affected is not the sibs+parents. cSix leprosy per se individuals not classified as LL or TT; therefore, the numbers of affected children and parents for LL+TT do not add up to leprosy per se.

of tuberculosis and 72 multicase families of leprosy from D17S33 27.9Mb NOS2A CCL2 Belem, Brazil that support the hypothesis that a cluster of 5.7cM XulNOS 28.6Mb TNFAIP1 CCL7 4.2cM 31.0Mb CRLF3 CCL11 genes in this region regulate mycobacterial infections in man. CCL8 3.7cM D17S798 CCL13 CCL1 5.8cM D17S1293 34.6Mb CCL Cluster 8.0cM } D17S927 36.4Mb Results 4.8cM D17S250 40.1Mb CSF3 D17S1814 THRA1 MMP28 41.1Mb Genetic and physical maps for chromosome 5.8cM D17S1299 CCR7 CCL5 42.5Mb STAT5A & 5B CCL16 CCL14 17q11.1–q21.31 7.8cM STAT3 3.8cM D17S930 44.9Mb CCL15 1.4cM CCL23 A total of 627 individuals from 98 multicase tuberculosis 2.3cM CCL18 CCL3 families and 372 individuals from 72 multicase leprosy 3.9cM CCL4 D17S2015 48.5Mb SP2 families (Table 1) were initially genotyped for 16 2.6cM D17S1785 49.6 4.2cM D17S1868 D17S1795 microsatellite markers across the chromosome 17q11.1– D17S1869 q21.31 region. Marker–marker linkage data from the 6.8cM D17S1877 combined set of families were used to generate a genetic D17S787 55.7Mb 26 27 map (Figure 1) using CRI-Map and MAP-O-MAT. Figure 1 Line diagram comparing genetic map (Kosambi cM) This is compared (Figure 1) to the current physical map generated using CRI-Map and MAP-O-MAT with physical map order for the same markers obtained from the UCSC (Mb) based on information from the November 2002 freeze of the Human Project Working Draft, November 2002 working draft of the at (Electronic Database freeze (NCBI Build 31). The only inconsistency between Information 5). the genetic and physical map order is for the marker D17S2015. This discrepancy in position could be due contributed by the TT leprosy families (Zlr 2.58; P¼0.005), either to assembly errors in the physical map or to a lack whereas LL families appear to make some contribution to of informative meioses in the genetic mapping. Since the the leprosy per se peak at D17S1975. Combined linkage genetic map provides a more accurate representation of analysis for all leprosy and tuberculosis families gave a recombination events across the region, we used the peak multipoint Zlr¼3.38 (P¼0.0004). Again, the TT leprosy genetic map generated from the Brazilian families in subtype appeared to account for all of the combined multipoint linkage analyses presented below. linkage (multipoint Zlr 3.56; P¼0.0002). No evidence for linkage is observed for either leprosy or tuberculosis at Linking disease susceptibility to NOS2A, which lies immediately proximal to the region of Results of multipoint nonparametric linkage analyses linkage. Multipoint information content remained constant carried out in ALLEGRO are presented in Figure 2. For (mean7s.d.) across the region (tuberculosis families: leprosy per se, a broad region of linkage was observed, 0.5370.05; leprosy families: 0.5170.05; LL subtype: 7 7 with an initial peak (Zlr 2.34; P¼0.01) at D17S250 and a 0.71 0.05; TT subtype: 0.81 0.09). second peak (Zlr 2.67; P¼0.004) at D17S1795. For tuberculosis, a single peak (Zlr 2.04; P¼0.02) suggestive Family-based transmission disequilibrium testing of linkage was observed at D17S250, with no linkage at (TDT) D17S1795. Analysis by disease subtype (Figure 2) shows To determine whether one or multiple candidate genes that all of the linkage to leprosy at the D17S250 peak is might contribute to mycobacterial disease susceptibility

Genes and Immunity Genetic susceptibility to tuberculosis and leprosy SE Jamieson et al 48 on 17q11.2, an additional 69 single nucleotide (Table 2). From proximal to distal 17q11.2, these were polymorphisms (SNPs) were typed in candidate genes. as follows: These included 10 novel SNPs in the promoter of NOS2A and, where possible, at least two SNPs (except 1. Significant allele-wise TDT associations were ob- served for SNPs NOS2A-1026 (Wald w2¼4.26, STAT5B and CCR7 where only one SNP was available) in 2 each of the candidates (in order across the region) P¼0.039) and NOS2A-2447 (Wald w ¼4.8, P¼0.029) TNFAIP1, CRLF3, CCL2,11,8,13,1,5,16,15,14,23, 18,3,4, in the promoter of NOS2A. 2. Significant allele-wise TDT associations were ob- CSF3, THRA, CCR7, STAT5A, STAT5B and STAT3. 2 Details of the SNPs, including primer sequences, served for SNPs CLL18-rs2015086 (Wald w ¼6.25, P¼0.012), CCL18-rs2015070 (Wald w2¼4.9, P¼0.027) nucleotide transitions and allele frequencies, appear 2 in Web-Tables 8 and 9. Of these SNPs, 49 were and CCL18-rs14304 (Wald w ¼7.35, P¼0.007). X 3. Significant allele-wise TDT association was observed informative (variant allele frequency 0.1) for allelic 2 association testing in the tuberculosis families. Analysis for SNP CCL4-rs1719144 (Wald w ¼7.35, P¼0.007). of these data using TDT28 with a robust sandwich 4. Significant allele-wise TDT association was observed estimator for the variance and a Wald w2 test to allow for SNP STAT5B-rs2230097 without robust variance estimates (w2¼4.26, df¼1, P¼0.04), marginal after for family clustering provides evidence for four 2 regions contributing to susceptibility to tuberculosis correction for pedigree clustering (Wald w ¼3.24, df¼1, P¼0.07). CCL18, CCL4 and STAT5B all fall under Pp0.05 area of the first linkage peak on 17q11.2. NOS2A lies immedi- CCR7 ately proximal to the first linkage peak (Figure 2), but D17S250 STATs outside the Pp0.05 area of linkage. No other significant 4 D17S1795 allelic associations were observed for candidate genes under this linkage peak for tuberculosis. 3 Lep CCLs TT Lep Case/pseudo-control single marker and intralocus 2 LL Lep stepwise logistic regression analyses NOS2A TB Family-based allelic association testing using TDT that

Zlr Score 1 TB + Lep was robust to pedigree clustering suggested that poly- morphism at multiple candidate genes might contribute TB + TTLep 0 to the disease associations with tuberculosis. SNPs at multiple candidate genes, some separated by intervals 020406080 (NOS2A–8.4 Mb–CCL18–32.3 kb–CCL4–6.04 Mb–STAT5B) Distance (cM) -1 of several Mb, all indicated positive allelic associations. Conditional logistic regression analysis (Table 3) that Figure 2 Multipoint nonparametric Zlr scores for linkage between microsatellite markers across chromosome 17q11.2–q23.2 and used a case/pseudo-control data set derived from the susceptibility to tuberculosis, leprosy per se, leprosy LL vs TT families and was robust to nuclear family clustering subtypes, and for combined analysis for mycobacterial suscept- essentially confirmed the associations (cf below) ob- ibility genes controlling tuberculosis and leprosy per se or served using robust TDT. In comparing allele-wise and tuberculosis plus TT leprosy. Linkage scores are plotted as a genotype-wise associations in the case/pseudo-control function of marker location in centimorgans (cM) according to the genetic linkage map generated using data for all tuberculosis and analysis, no dominance effects (defined as departure leprosy families (see Figure 1). Graphs are annotated with from the multiplicative model for the effects of alleles on approximate positions for candidate genes across the region. Dotted the overall genotype relative risks) were observed at any lines indicate the levels of significance associated with Zlr scores. of the candidate gene loci (data not shown). To determine

Table 2 Allele-wise robust TDT statistics for association between chromosome 17q11.2 markers and tuberculosis

Marker SNP/MIC N NT NIT w2 df P

NOS2A À277 214 76 52 1.78 1 0.182 À1026 185 50 17 4.26 1 0.039 À1659 200 60 26 0.8 1 0.371 À2447 185 56 34 4.8 1 0.029 XuiNOS 203 118 138 13.65 7 0.058 CCL18 rs2015086 199 64 48 6.25 1 0.012 rs2015070 203 69 40 4.91 1 0.027 rs712043 202 75 43 2.69 1 0.101 rs14304 192 61 29 7.35 1 0.007 CCL4 rs1634514 214 65 36 3.27 1 0.070 rs1719144 214 74 27 7.35 1 0.007 rs1719147 205 64 25 0.03 1 0.866 STAT5B rs2230097 219 73 19 3.24 1 0.072

N: number of affected offspring typed; NT: number of trios analysed; NIT: number of informative transmissions; w2: chi squared; df: degrees of freedom; P: probability. XuiNOS is a microsatellite, all other markers are SNPs. Bold indicates significant associations ( Pp0.05).

Genes and Immunity Genetic susceptibility to tuberculosis and leprosy SE Jamieson et al 49 Table 3 Robust case/pseudo-control allelic association testing for chromosome 17q11.2 markers and tuberculosis

2 Marker SNP SNP transition Frequency of the common allele w df P Pc

NOS2A À277 A-G 0.64 1.67 1 0.196 — À1026 G-T 0.72 5.37 1 0.021 1.0 À1659 C-T 0.80 0.93 1 0.335 — À2447 C-G 0.65 3.63 1 0.057 — CCL18 rs2015086 T-C 0.68 10.19 1 0.001 0.049 rs2015070 G-A 0.73 4.56 1 0.033 1.0 rs712043 C-T 0.75 2.56 1 0.110 — rs14304 G-A 0.74 7.80 1 0.005 0.245 CCL4 rs1634514 T-A 0.83 3.95 1 0.047 1.0 rs1719144 G-A 0.83 9.35 1 0.002 0.098 rs1719147 G-A 0.79 0.03 1 0.864 — STAT5B rs2230097 T-C 0.89 4.29 1 0.038 1.0

SNP transition indicates the sequence change from common to variant allele. Robust variance estimates and a Wald w2 test were used throughout to control for family clustering. Pc¼P-value with Bonferroni correction (multiplied by 49 markers used). Bold indicates significant associations ( Pp0.05).

CCL18 CCL3 CCL4 distal end of CCL18. SNPs rs2015086 and rs2015070 are in 123321123strong linkage disequilibrium (Table 4) with each other (D0¼0.85), but not with SNP rs14304 (D0¼0.22 and

16.8Kb 13.6Kb 0.18, respectively). Stepwise logistic regression analysis (Table 5) demonstrated that rs2015086 and rs2015070 did not add significant main effects to each other, but both had significant main effects when compared to rs14304

CCL3+459 CCL3+113 alone. Similarly, rs14304 adds significant main effects to both rs2015070 alone and rs2015086 alone, and the two CCL18/rs14304

CCL18/rs712043 CCL4/rs1634514 CCL4/rs1719144 CCL4/rs1719147 together. This suggests that there are separate functional CCL18/rs2015086 CCL18/rs2015070 polymorphisms in linkage disequilibrium with the Figure 3 Line diagram showing physical relationship between markers at the 50 and 30 ends of CCL18 that are CCL18-CCL3-CCL4 on chromosome 17q11.2. contributing to disease association with tuberculosis. Case/pseudo-control analysis confirmed significant associations between tuberculosis and the SNPs whether each of these loci contributed significant main rs1634514 and rs1719144 at CCL4 (Table 3). The allele- effects to the chromosome 17q11.2 associations observed, wise relative risks associated with allele rs1634514T we carried out tests to determine main effects at (common allele) was 2.00 (95% CI 1.01–3.96; P¼0.047) each in a forward stepwise logistic regression and with allele rs1719144A (variant allele) was 0.35 (95% procedure.29 To achieve this, we first carried out a CI 0.18–0.68; P¼0.002). SNP rs1634514 lies in the 50 forward stepwise logistic regression analysis for the flanking region of CCL4 (Figure 3). CCL4 rs1719144 lies in multiple markers within each candidate gene locus to 1 485 bp distal to rs1634514. CCL4 rs1634514 and determine which markers contributed significant main rs1719144 are in strong linkage disequilibrium (Table 4) effects. All comparisons used logistic regression analysis with each other (D0¼0.94). The stepwise analysis with robust variance estimates and a Wald w2 test to (Table 5) demonstrates that the associations at CCL4 allow for family clustering. rs1634514 and rs1719144 can each be accounted for by Case/pseudo-control analysis confirmed significant association at the other, that is, neither adds separate associations between tuberculosis and the SNP at main effects to the other. position –1026 in the promoter of NOS2A (Table 3). The Case/pseudo-control analysis confirmed significant allele-wise relative risk associated with allele À1026G associations between tuberculosis and SNP rs2230097 at was 3.25 (95% CI 1.19–8.81; P¼0.021), that is, the common STAT5B (Table 3). The allele-wise relative risks associated allele was disease associated. Since only one SNP with allele rs2230097C (variant allele) was 0.36 (95% CI showed significant association, no intralocus stepwise 0.13–0.95; P¼0.038). Since only one SNP was available, analysis was performed. no intralocus stepwise analysis was possible. Case/pseudo-control analysis confirmed significant The only single marker association to remain signifi- associations between tuberculosis and the SNPs cant after application of a strict Bonferroni correction30 rs2015086, rs2015070 and rs14304 at CCL18 (Table 3). (multiplication by 49 informative SNP types) for multiple The allele-wise relative risk associated with allele testing was CCL18 rs2015086 (Table 3). rs2015086C was 0.41 (95% CI 0.24–0.71; P¼0.001), with Linkage disequilibrium occurred between markers allele rs2015070A was 0.35 (95% CI 0.18–0.68; P¼0.002) within the three regions (see Figure 1) carrying the four and with allele rs14304A was 0.38 (95% CI 0.19–0.75; candidate genes (Tables 4 and 6), but no extended P¼0.005), that is, the variant allele at each SNP was linkage disequilbrium was observed across the three disease protective. SNPs rs2015086 and rs2015070 lie regions (D0p0.5; data not shown). Interestingly, no 166 bp apart in the 50 flanking region and intron 1 at the associations were seen between tuberculosis and mar- proximal end of CCL18 (Figure 3). SNP rs14304 lies 6.7 kb kers for candidate genes (CCL2,11,8,13,1) in the more distal to rs2015086 and rs2015070 in the 30UTR at the proximal group of chemokine genes (see Figure 1). Intra-

Genes and Immunity Genetic susceptibility to tuberculosis and leprosy SE Jamieson et al 50 and interlocus linkage disequilibrium was observed for

1 CCL2/11 and CCL13/1 within this group of markers (Table 6b).

Case/pseudo-control interlocus stepwise logistic

0.84 regression analyses To determine whether all four candidate genes for which significant associations had been observed contributed independent main effects to the association between 1 17q11.2 and tuberculosis, stepwise interlocus logistic regression analysis was performed (Table 7). Since CCL18 rs2015086 and rs2015070 did not show separate main

1 effects from each other, only the more significant CCL18 rs2015086 was used in the interlocus stepwise analysis. Similarly, since CCL4 1634514 and rs1719144 did not 459 rs1634514 rs1719144 rs1719147 rs1634516

þ show separate main effects, only the more significant CCL4 rs1719144 was used in the interlocus stepwise analysis. All comparisons used logistic regression analy- 0.5). 1 2

4 sis with robust variance estimates and a Wald w test to 0

D allow for family clustering. In all of these analyses, a significant Wald w2 indicates that addition of the marker 0.88 0.740.81 0.67 0.73 0.94 0.92 0.94 in bold under the alternative model adds a significant separate main effect to the marker considered under the null hypothesis.

1 0.45 0.41 Table 7 demonstrates that the À1026 bp SNP in the promoter of NOS2A contributed separate main effects to all other markers for which significant associations were observed across 17q11.2. While there was strong linkage 0.35 0.130.58 0.61 1 disequilibrium between markers in the NOS2A promoter (Table 6a), there was no extended linkage disequilibrium with SNPs at candidate genes (TNFAIP1 and CRLF3)

1 0.790.73 0.85 immediately distal to NOS2A (Table 6a) or any of the other markers across 17q11.2 (data not shown). Taken together, these data suggest that polymorphism at NOS2A itself is associated with susceptibility to tubercu- 1 0.270.32 0.03 0.03 losis in this population. Similarly, the 50 SNP rs2015086 at CCL18 contributed separate main effects to all other markers for which 0.58 0.88 0.55 0.52 1 significant associations were observed across the region (Table 7), indicating that polymorphism at CCL18 may also contribute independently to susceptibility to tuber-

0.58 culosis in this population. Interestingly, SNP rs1719144 at CCL4 contributed separate main effects to rs2015086 but not rs14304 at CCL18. Similarly, SNP rs14304 at CCL18 contributed significant main effects when compared to 0.52 0.82 0.65 rs2015086 at CCL18 (Table 5), but not when compared to SNP rs1719144 at CCL4. SNP rs1719144 lies in intron 1 of CCL4 32.756 kb distal to CCL18 rs14304 (Figure 3), with CCL3 encoded within this interval. CCL4 rs1719144 is in strong linkage disequilibrium (Table 4) with the 30 CCL18 SNP rs14304 (D0¼0.73), which explains why these two

) between markers in the more distal chemokine gene cluster on 17q11.2 SNPs do not contribute separate main effects to each 0

D other. These results suggest that there may be more than one functional polymorphism across the region CCL18- 403 rs2063979 rs854680 rs854684 rs712048 rs854625 rs1003645 rs1719204 rs2015086 rs2015070 rs712043 rs14304 1

À CCL3-CCL4 that contribute separately to susceptibility to tuberculosis in this population, one in linkage disequili- 0 CCL5 CCL16 CCL15 CCL23brium CCL18with SNP rs2015086 CCL3 at the 5 end CCL4 of CCL18 and a

0.86 second in linkage disequilibrium with SNPs across In1.1 TC 0 À 3 CCL18-CCL3-CCL4. SNP rs2230097 in the 50UTR of STAT5B, 6.04 Mb distal to CCL4, contributed significant main effects when Linkage disequilibrium ( 459 0.01 0.03 0.05 0.09 0.09 0.05 0.35 0.08 0.04 0.01 0.35 0.48 0.42 1 In1.1TC403 1 compared to CCL4 rs1719144 but not to more proximal rs1719204rs2015086 0.13rs2015070 0.09rs712043 0.09 0.01rs14304 0.07À 0.04 0.29 0.02rs1634514 0.04 0.33rs1719144 0.04 0.30 0.31rs1719147 0.05 0.04 0.36rs1634516 0.24 0.13 0.44 0.22 0.20 0.30 0.44 0.17 0.22 0.04 0.05 0.03 0.10 0.07 0.01 0.28 0.18 0.06 0.16 0.07 0.01 0.04 0.10 0.14 0.02 0.13 0.13 0.06 0.13 0.25 0.09 0.31 0.32 0.25 0.01 0.02 0.30 0.12 0.24 0.31 0.08 0.01 0.03 0.20 0.42 0.01 0.29 0.11 0.11 1 0.08 0.01 0.06 0.22 0.29 0.06 0.18 0.17 0.30 0.29 0.34 0.07 0.14 0.08 0.13 0.48 0.35 0.37 0.19 0.43 0.24 1 À À rs2063979rs854680 0.38rs854684rs712048 0.07 0.29rs854625 0.15rs1003645 0.24 0.01 1 0.40 0.07 0.23 0.09 0.08 0.07 0.04 0.07 0.39 1 0.12 0.28 0.16 0.19 0.15 1 0.22 0.46 0.35 1 candidate genes at 17q11.2. STAT5B is not in significant linkage disequilibrium with markers at any of the CCL18 CCL3 CCL4 CCL5 CCL16 CCL15 CCL23 Table 4 See Figure 1 for the position of markers in theproximal distal chemokine cluster. Bold indicates moderate to strong linkage disequilibrium ( candidate genes for which significant associa-

Genes and Immunity Genetic susceptibility to tuberculosis and leprosy SE Jamieson et al 51 Table 5 Intralocus forward stepwise regression analysis

Test statistic

Null model Alternative model w2 df P

(a) CCL18 rs2015086 rs2015086+rs2015070 2.94 1 0.086 rs2015070 rs2015070+rs2015086 0.63 1 0.428 rs2015086 rs2015086+rs14304 9.25 1 0.003 rs14304 rs14304+rs2015086 4.51 1 0.034 rs2015070 rs2015070+rs14304 5.51 1 0.019 rs14304 rs14304+rs2015070 16.45 1 4.99E-05 rs2015086+rs2015070 rs2015086+rs2015070+rs14304 5.31 1 0.021 rs2015070+rs14304 rs2015070+rs14304+rs2015086 0.56 1 0.453 rs2015086+rs14304 rs2015086+rs14304+rs2015070 6.76 1 0.009 (b) CCL4 rs1634514 rs1634514+rs1719144 0.44 1 0.507 rs1719144 rs1719144+rs1634514 0.05 1 0.818

Test to determine whether SNPs showing significant allelic associations within each candidate gene locus contribute separate main effects. A significant Wald w2 test comparing null and alternative models indicates that the marker added (bold) under the alternative model is contributing a separate main effect from the marker(s) considered under the null hypothesis. Robust variance estimates to control for family clustering were used throughout. Bold indicates significant ( Pp0.05) main effects added under the alternative model.

Table 6 Linkage disequilibrium (D0) between markers in different regions of 17q11.2

NOS2A TNFAIP1 CRLF3

(a) À277 À1026 À1659 À2447 rs1007398 rs733914 rs999798 rs999797

NOS2A À277 1 À1026 0.77 1 À1659 0.86 0.70 1 À2447 0.90 0.77 0.79 1 TNFAIP1 rs1007398 0.03 0.08 0.06 0.01 1 rs733914 0.03 0.05 0.01 0.06 0.72 1 CRLF3 rs999798 0.02 0.01 0.04 0.10 0.08 0.05 1 rs999797 0.09 0.30 0.02 0.06 0.01 0.01 0.67 1 CCL2 CCL11 CCL8 CCL13 CCL1

(b) À2518 rs1024610 rs1860184 rs1019109 rs1233650 rs1431991 rs159313 rs159314 rs159315 rs210839

CCL2 À2518 1 rs1024610 0.74 1 CCL11 rs1860184 0.72 0.74 1 rs1019109 0.24 1.00 0.45 1 CCL8 rs1233650 0.21 0.35 0.22 0.16 1 CCL13 rs1431991 0.22 0.04 0.28 0.19 0.09 1 rs159313 0.28 0.26 0.30 0.16 0.13 0.53 1 rs159314 0.36 0.05 0.22 0.08 0.03 0.51 0.33 1 rs159315 0.44 0.29 0.33 0.35 0.07 0.67 0.23 0.70 1 CCL1 rs210839 0.20 0.02 0.30 0.10 0.17 0.29 0.69 0.51 0.20 1 CSF3 THRA CCR7 STAT5B STAT3

(c) rs2227319 rs2227322 rs25645 rs2827 rs3471 rs1467258 rs2002724 rs2230097 rs744166 rs1026916

CSF3 rs2227319 1 rs2227322 0.76 1 rs25645 0.90 0.71 1 rs2827 0.79 0.61 0.89 1 THRA rs3471 0.11 0.22 0.09 0.54 1 rs1467258 0.35 0.42 0.37 0.11 0.42 1 CCR7 rs2002724 0.37 0.53 0.56 0.07 0.26 0.03 1 STAT5B rs2230097 0.72 0.06 0.85 0.39 0.15 0.14 0.25 1 STAT3 rs744166 0.24 0.11 0.26 0.20 0.08 0.12 0.12 0.37 1 rs1026916 0.18 0.02 0.26 0.05 0.18 0.01 0.15 0.36 0.74 1

Markers around (a) NOS2A, (b) the proximal chemokine gene cluster, and (c) STAT5B (see Figure 1). Bold indicates moderate to strong linkage disequilibrium (D040.5).

Genes and Immunity Genetic susceptibility to tuberculosis and leprosy SE Jamieson et al 52 Table 7 Interlocus forward stepwise regression analysis

Test statistic

Null model Alternative model w2 df P

(a) Adding SNP at NOS2A CCL18 rs2015086 CCL18 rs2015086+NOS2A À1026 4.8- 1 0.028 CCL18 rs14304 CCL18 rs14304+NOS2A À1026 3.71 1 0.054 CCL4 rs1719144 CCL4 rs1719144+NOS2A À1026 4.21 1 0.040 STAT5B rs2230097 STAT5B rs2230097+NOS2A À1026 5.12 1 0.024 (b) Adding SNP at CCL18 NOS2A –1026 NOS2A À1026+CCL18 rs2015086 8.25 1 0.0041 NOS2A À1026 NOS2A À1026+CCL18 rs14304 2.13 1 0.145 CCL4 rs1719144 CCL4 rs1719144+CCL18 rs2015086 10.69 1 0.001 CCL4 rs1719144 CCL4 rs1719144+CCL18 rs14304 2.23 1 0.135 STAT5B rs2230097 STAT5B rs2230097+CCL18 rs2015086 7.81 1 0.005 STAT5B rs2230097 STAT5B rs2230097+CCL18 rs14304 8.08 1 0.005 (c) Adding SNP at CCL4 NOS2A À1026 NOS2A À1026+CCL4 rs1719144 0.41 1 0.524 CCL18 rs2015086 CCL18 rs2015086+CCL4 rs1719144 7.12 1 0.008 CCL18 rs14304 CCL18 rs14304+CCL4 rs1719144 1.38 1 0.241 STAT5B rs2230097 STAT5B rs2230097+CCL4 rs1719144 10.55 1 0.001 (d) Adding SNP at STAT5B NOS2A À1026 NOS2A À1026+STAT5B rs2230097 0.47 1 0.495 CCL18 rs2015086 CCL18 rs2015086+STAT5B rs2230097 2.81 1 0.094 CCL18 rs14304 CCL18 rs14304+STAT5B rs2230097 1.07 1 0.300 CCL4 rs1719144 CCL4 rs1719144+STAT5B rs2230097 4.25 1 0.039

Test to determine whether SNPs at different loci across 17q11.2 contribute separate main effects. A significant Wald w2 test comparing null and alternative models indicates that the marker added (bold) under the alternative model is contributing a separate main effect from the marker considered under the null hypothesis. Robust variance estimates to control for family clustering were used throughout. Bold indicates significant ( Pp0.05) main effects added under the alternative model.

tions were observed across 17q11.2 (data not shown). case/pseudo-control data set derived from the larger set However, some extended linkage disequilibrium (Table of tuberculosis families confirmed that at least four 7c) was observed between STAT5B and CSF3, and separate candidate genes, NOS2A, CCL18, CCL4 and between CSF3 and CCR7. Although association between STAT5B, or genes in linkage disequilibrium with them, disease susceptibility and the particular SNPs examined may contribute to the region of linkage on 17q11.2. at these candidate loci was not observed, it is possible Robust tests that controlled for pedigree clustering that the association with STAT5B could result from confirmed allelic association in the presence of linkage. linkage disequilibrium with a functional polymorphism Although these results do not provide definitive evi- at these adjacent candidate genes. dence that these are the aetiological loci regulating disease susceptibility, some discussion of their likely candidacy in relation to recent genetic and functional Discussion studies is relevant. The first gene contributing separate main effects to A common theme emerging from studies of complex allelic associations across the region was NOS2A, which traits such as infectious and autoimmune diseases is that lies immediately proximal to the region of linkage at multiple genes frequently contribute to regions of Po0.05 on chromosome 17q11.2. The SNPs used in this linkage to disease susceptibility. In mice, this has been study were identified as part of an earlier infectious highlighted by analysis of congenic mouse strains where disease study in which a single haplotype, uniquely progressive refinement of the congenic interval has defined by the NOS2A-1659T allele, was associated with identified multiple genes contributing to a region of cerebral malaria in The Gambia.35 In Brazil, no associa- linkage.31–34 In humans this is more difficult to demon- tion with NOS2A-1659 was observed, even though all strate as definitively as in the mouse, but the application four promoter SNPs were in linkage disequilibrium with of stepwise logistic regression procedures29 provides a each other. Our sample size was not sufficient to statistical approach that allows the relative contribution undertake a detailed analysis of haplotypes associated of candidate genes within a linked region to be with tuberculosis in this Brazilian population. From determined from case–control or family data. This is studies in Tanzania and Kenya, Hobbs et al36 also particularly appropriate in the analysis of regions of the identified a novel SNP, À1173C-T, associated with genome that contain clusters of immune response, such protection from symptomatic malaria and severe malar- as HLA, chromosome 5q23–q33, and the region of ial anaemia. Functionally, NOS2A provides a credible chromosome 17q11–q21 that we have studied here. candidate for tuberculosis susceptibility. Choi et al37 Linkage analysis performed here using 16 microsatellite examined resected from eight tuberculosis patients markers provided evidence for linkage between this for expression of iNOS and nitrotyrosine, a marker of region and susceptibility to leprosy and tuberculosis. nitric oxide production. Immunohistochemical analysis Stepwise conditional logistic regression analysis using a revealed that NOS2A was expressed in CD68-positive

Genes and Immunity Genetic susceptibility to tuberculosis and leprosy SE Jamieson et al 53 epithelioid and giant cells in the inflam- and CCL4 were produced within 8 h after infection of matory zone of granulomas of patients, but not in dendritic cells with M. tuberculosis. Interestingly, their histologically normal tissue obtained from control sub- work also showed that maturing M. tuberculosis-infected jects. Studies in vitro demonstrated that macrophages dendritic cells express high levels of CCR7 and become from active tuberculosis patients produced less nitric responsive to its ligand CCL21. Linkage disequilibrium oxide in response to lipopolysaccharide stimulation than across the region means that we cannot preclude controls,38 suggesting that downregulation of nitric oxide polymorphism at CCR7 as contributing to the fourth production may be counterprotective. This is consistent component of susceptibility to tuberculosis measured with the demonstration that nitric oxide production here using SNPs at STAT5B. Sallusto et al50 demonstrated correlates with inhibition of bacterial growth in alveolar that expression of CCR7 divides human memory T cells macrophages in vitro.39 Other work40 has also demon- into two functionally distinct subsets. CCR7-negative strated high and specific expression of iNOS in granu- memory T cells display immediate effector function. lomas surrounding and infiltrating dermal nerves in CCR7-positive memory cells provide long-term central borderline leprosy lesions, with and without reversal memory and efficiently stimulate dendritic cells and reaction, suggesting that regulation of iNOS expression differentiate into CCR7-negative effector cells upon might also be involved in leprosy susceptibility and/or secondary stimulation. The work of Lande et al49 high- in the nerve damage following reversal reaction in lights the complex interplay between and leprosy. Although our sample size for leprosy precluded receptors that occurs with M. tuberculosis infection, which confirmation of a role for polymorphism at NOS2A in they postulate orchestrates the recruitment and selective contributing to susceptibility to leprosy in this popula- homing of activated/effector cells known to accumulate tion, further studies in Brazil to obtain larger samples are at the site of M. tuberculosis infection and take part in the in progress. formation of the granulomas. It is possible that long- The second region that contributed separate main range regulatory elements on 17q11.2 could coordinate effects to tuberculosis susceptibility contained genes expression in response to infection and may contribute to encoding chemokines CCL18, CCL3 and CCL4. Strong the complex pattern of disease susceptibility genes and linkage disequilibrium across the region encoding these linkage disequilibrium observed across the region. three genes (Table 4) means that we cannot be certain as In our study we have accessed high-throughput to which genes contain the functional polymorphisms genotyping technology that has allowed us to undertake influencing tuberculosis, although it seems likely that an extensive analysis of SNPs across the region of more than one candidate gene across the region may be chromosome 17q11.2. Since we used families with involved. Some evidence for extended linkage disequili- multiple cases, we were careful to employ tests for allelic brium to other chemokine genes (CCL16,15,23 but not association that were robust to pedigree clustering. CCL5 encoding RANTES, Table 4) in this cluster means While the robust stepwise conditional logistic regression that these may also be involved. Nevertheless, the analysis allowed us to demonstrate that different genes stepwise logistic regression analysis suggested an in- across the region contribute separate main effects, the dependent role for polymorphism at the 50 and 30 ends of ability to type many markers has provided an additional CCL18, the latter in linkage disequilibrium with SNPs in problem in deciding how to correct for multiple testing. CCL4 also contributing separate main effects. CCL18 is a The only single marker test statistic to remain significant small inducible chemokine that was formerly known as after application of a strict Bonferroni correction30 (ie pulmonary and activation-regulated chemokine or multiplying by the number of informative SNPs typed PARC. Northern blot analysis shows that PARC is across the region) was CCL18 rs2015086. The issue of expressed at high levels in lungs, and at lower levels in appropriate correction for multiple testing is complex some lymphoid tissues,41 making it an alluring candidate and one that has preoccupied statisticians in general30,51 for a tuberculosis susceptibility gene. Independently, and genetic statisticians in particular. In our case, Adema et al42 identified the PARC gene, which they application of a strict Bonferroni correction is over- called DCCK1, as a gene expressed in dendritic cells with conservative, since we know that the 49 SNPs are not all high levels of expression dependent on the presence of independent of one another. A less conservative ap- 4. DCCK1 acted preferentially as a chemo- proach is to carry out permutation tests. We calculated kine for naı¨ve T cells, suggesting that it is important in permutation P-values for the case/pseudo-control single the initiation of an immune response. Other workers43 marker test statistics (data not shown) by carrying out also identified DCCK1 as a chemokine induced in 1000 replications randomly permuting the ‘case’ and macrophages by alternative activation with interleukin ‘control’ labels within each set of case and pseudo- 4 and glucocorticoids, a factor that could be important in controls and repeating the regression analysis. However, the light of studies demonstrating that activation of the we did not improve on a Bonferroni correction, possibly hypothalamus–pituitary–adrenal axis affects antimyco- due to loss of power when generating a case/pseudo- bacterial activity of macrophages in vivo.44,45 control data set as it was not always possible to infer CCL3 and CCL4 encode macrophage inflammatory phase in the parents when using all the markers at once. (MIP) MIP1a and MIP1b that act as chemoat- For this analysis, it was necessary to generate the case/ tractants for polymorphonuclear leucocytes and mono- pseudo-control data set using all markers simultaneously cytes. Many studies have documented the importance of to take into account linkage disequilibrium between CC chemokines in pathology,46 with MIP-1a and/or markers for the permutation procedure. Our experience MIP-1b specifically identified as chemokines expressed highlights important issues relating to sample size that in response to M. tuberculosis stimulation in macrophage need to be addressed in planning genetic studies of cell lines47 and alveolar macrophages.48 In a recent study, complex traits. When we collected our families in Brazil Lande et al49 also demonstrated that high levels of CCL3 between 1991 and 1993, our efforts were focused on

Genes and Immunity Genetic susceptibility to tuberculosis and leprosy SE Jamieson et al 54 obtaining large multicase families that would be power- genetic analysis of multicase pedigrees. Families came ful for linkage analysis. With the development of family- from an area of high tuberculosis prevalence. The based allelic association testing, a better strategy now is incidence rate for tuberculosis in Belem City at the to collect a separate sample of case/parent trios to beginning (1991) of the study was 62 per 100 000 facilitate independent allelic association testing, and to habitants. The prevalence rate for tuberculosis in the ensure that this sample has sufficient power to detect whole of Brazil is 68 cases per 100 000 habitants small effect sizes (odds ratios) for variant alleles at low (Electronic Database Information 1). frequency and to correct for multiple testing. To plan our Leprosy diagnosis was made following clinical exam- future studies, we calculated the theoretical power to ination for anaesthetic skin lesions, results from slit skin detect allelic association with 200, 400, 600 and 800 case/ smear testing for acid-fast bacilli, and, in some health parent trios using the method of Knapp52 and assuming a centres, histological analysis. Patients were categorized multiplicative model (see Supplementary Information- into disease subtype groups according to the Ridley–Jo- Table 10). To achieve P¼0.001 (to allow for correction for pling histological scale53 and/or bacterial load. Subtypes 50 markers typed) for odds ratios of the order of recorded were lepromatous (LL), borderline lepromatous magnitude (o0.5 or 42.0) observed here for variant (BL), borderline (BB) or borderline tuberculoid/tuber- alleles at frequencies o0.2, we need a minimum of 200– culoid (BT/TT). For genetic analyses in the present study, 400 case/parent trios. Interestingly, we would never gain lepromatous patients comprised LL plus BL subtypes 445% power to achieve P¼0.001 for an odds ratio o2 (referred to hereafter as LL) and tuberculoid patients and a variant allele at frequency 0.1 even with 800 comprised BB plus BT/TT subtypes (referred to hereafter parent/child trios (¼2400 total samples). Hence, there as TT). The incidence rate for leprosy in Belem City at the has to be a balance between sample size and return for beginning of the study was 45 per 100 000 population. effort. New studies in Brazil designed to improve our The prevalence rate for leprosy in the whole of Brazil is power are underway, as are studies designed to replicate 45 cases per 100 000 habitants (Electronic Database our observation for tuberculosis in other populations and Information 2). to determine whether the same or different genes across All families were of equivalent socioeconomic status. this region regulate leprosy per se or LL and TT disease Attempts to determine BCG vaccination status on all subtypes. For these, we have set a target of 400 case/ members of the pedigrees were either incomplete (as parent trios as a realistic compromise between sample determined by vaccine records) or inconclusive (as availability and power to detect moderately strong determined by recording of scar status at interview), associations. and were not included in the analyses. Blood was Overall, our study follows the trend that clusters of collected by venepuncture for all available members of genes under broad linkage peaks contribute to complex the families. Epstein–Barr virus (EBV)-transformed B disease inheritance. Our results provide encouragement cells were prepared and cryopreserved for transport to that continued investigation of this region of chromo- UK. In Cambridge, EBV cells were expanded and DNA some 17q11.2 may provide valuable insight into the prepared for genetic analysis. A total of 92 multicase genes and mechanisms that contribute to the variable families (627 individuals) for tuberculosis and 72 multi- disease and pathology associated with tuberculosis and case families (372 individuals) for leprosy were available leprosy in Brazil. for analysis after checking for genetic integrity within all families. Table 1 provides a breakdown of the family structures for each disease. Study population and methods Genotyping Ascertainment of families A total of 16 microsatellite markers spanning the Leprosy and tuberculosis families were ascertained candidate region were used in the analysis, 14 chosen through medical records at local Ministerio de Sanidad from the Genetic Location Database maintained by the health centres in Belem, Para, Brazil. Families were Human Genetics Division of the University of South- pursued when there was indication on the medical ampton (Electronic Database Information 3), a micro- record that additional family members had been or were satellite designated here as XuINOS within the NOS2A currently affected with leprosy or tuberculosis. All health promoter,54 and D17S250.55 Primer sequences were centres were staffed by clinicians highly experienced in obtained from the relevant publications54,55 or from the the diagnosis and treatment of leprosy and tuberculosis. Whitehead Institute (Electronic Database Information 4). The study was performed with the approval of the Fluorescently tagged primer pairs were ordered from PE ethical review committee of the Institute Evandro Applied Biosystems (Warrington, UK) or Invitrogen Chagas, Belem, Para, Brazil. (Paisley, UK). All microsatellites were analysed by For tuberculosis families, diagnosis was made on the electrophoresis on 6% polyacrylamide gels using an basis of a sputum test positive for acid-fast bacilli and/or automated sequencer (model ABI-Prism 377, PE Applied chest X-ray. Sputum tests were carried out at the centres, Biosystems) and data collected and analysed using stained (Ziehl–Nielson) and read for acid-fast bacilli by GENESCANt and GENOTYPERt software (PE Applied qualified laboratory technicians. If tuberculosis symp- Biosystems). toms persisted following two negative sputum results, A total of 69 SNPs within or adjacent to candidate patients were referred to the central tuberculosis hospital genes across the region were identified from the Human in Belem, where all X-rays were read by experienced Genome Working Draft available at UCSC (Electronic specialist clinicians. Cases of pulmonary tuberculosis Database Information 5), the Ensembl Genome Browser only were included in this study. Other clinical forms of (Electronic Database Information 6) or publications: tuberculosis were at too low a frequency to permit CCL5-403;56 CCL5-In1.1t/c;57 CCL3 þ 113 and þ 459;58

Genes and Immunity Genetic susceptibility to tuberculosis and leprosy SE Jamieson et al 55 NOS2A promoter SNPs;35 MCP1-2518.59 High-through- had 98–100% power to detect linkage (LOD score 43.95; put genotyping was performed using uniplex or biplex Po0.00001) for a recessive gene at 5, 50 or 99% Invader assays (Third Wave Technologies, Madison, WI, penetrance, with reduced power to detect linkage up to

USA) or TaqMan Assay-by-Design (PE Applied Bio- a critical value (Zlr¼3.27) equivalent to an allele sharing systems), apart from three SNPs where the introduction/ LOD score 43.00 (Po0.0001) of 92–99% for a dominant abolition of a restriction fragment length polymorphism gene at 99% penetrance, 82–84% at 50% penetrance and recognition site was used. All fluorescent assays were 53–55% at 5% penetrance. performed in 384-well plates and were set up using a BiomekFX robotics system (Beckman, High Wycombe, Allelic association testing UK). Following incubation, Invader assays were ana- Family-based allelic association tests were performed lysed using a Wallac Victor2 (PE Applied Biosystems) using the TDT.28 Since all case–parent trios in the families fluorescence plate reader and TaqMan assays analysed were used in the analysis, we corrected for clustering at using an ABI Prism 7900HT Sequence Detecetion System the nuclear family level and nonindependence between (PE Applied Biosystems). Invader and TaqMan assays sibs using a robust sandwich estimator for the variance were analysed for genotype clustering using locally and a Wald w2 test. Only the tuberculosis families had available software. Full details of primer sequences and sufficient numbers of parent/child trios to carry out assays are available (see Supplementary Information- allelic association testing: 92% power to detect an effect Table 8). (odds ratio¼2; P¼0.05) for SNP with a variant allele Marker allele frequencies were determined from frequency of 0.5; 79% power for SNP with a variant allele unrelated individuals in the families using the computer frequency of 0.15; o50% power for SNP with a variant programme SPLINK.60 allele frequency of p0.1. Only SNPs where the variant allele was at frequency X0.1 were included in the Physical and genetic maps for chromosome analyses (see Supplementary Information-Table 9). Alle- 17q11.1–q21.31 lic associations, and relative risk estimates, were Physical positions for markers and candidate genes were obtained by creating a case/pseudo-control study where obtained from the most recent golden path assembly the cases comprise the genotypes of the affected off- (November 2002) of the Human Genome Working Draft spring, and the controls are the one to three other available at UCSC (Electronic Database Information 5). genotypes (depending on whether phase is known or Genetic map distances were calculated using both CRI- inferred) which the affected offspring might have Map26 (available at Electronic Database Information 7) received from the parents.29 The relative risks were and MAP-O-MAT27 (available at Electronic Database estimated using conditional logistic regression analysis, Information 8). The genetic map order was determined again employing robust variance estimates to control for using a likelihood-ratio criterion of X3 in both CRI-Map pedigree clustering and a Wald w2 test to indicate overall and MAP-O-MAT. significance for allelic associations. A stepwise logistic- regression procedure29 was used to evaluate the relative Linkage analysis importance of variants within and between candidate Multipoint nonparametric linkage analysis was per- gene loci. Wald w2 tests were used to compare models in formed in ALLEGRO61 using the Spairs scoring function which the main effects for both loci are modelled with and the exponential allele-sharing model62 with all one in which the main effects at the primary locus only families weighted equally. ALLEGRO reports allele are included. Robust TDT and case/pseudo-control sharing LOD scores and maximum Z scores for the statistical tests implemented within Stata were devel- likelihood ratio (Zlr). One-sided P-values associated with oped by Heather Cordell and David Clayton at the

LOD and Zlr scores are used throughout. For multipoint Cambridge Institute for Medical Research and are analysis, genetic distances between markers were en- available at Electronic Database Information 9. Correc- tered in cM from the genetic map generated using the tion for multiple testing was carried out using a strict Brazilian families. The fraction of the total inheritance Bonferroni correction30 multiplying by the number (49) information extracted by the available marker data is of informative markers typed. indicated by the information content. To find evidence for loci that might affect both tuberculosis and leprosy Linkage disequilibrium (or subtype), linkage analyses were performed using Linkage disequilibrium between pairs of markers across appropriate combined family data sets. Simulations 17q11–q21 was determined using Hedrick’s definition of performed in ALLEGRO using observed allele frequen- Lewontin’s D0 statistic.63 Initially, haplotype phase in a cies for the 16 microsatellite markers (10–18 alleles; set of unrelated individuals was determined using the average heterozygosity 0.63) demonstrated that the 92 PHASE program,64 and linkage disequilibrium statistics tuberculosis families had 98–100% power to detect were then calculated using HaploXT.65 Both programs linkage up to a critical value (Zlr¼4.27) equivalent to an are available at gMap.net based at the Wellcome Trust allele sharing LOD score 43.95 (Po0.00001) for a Centre for Human Genetics, Oxford (Electronic Database recessive or dominant susceptibility gene with pene- Information 10). trances 5, 50 or 99%. Similarly, the 72 leprosy families had 94–100% power to detect linkage (LOD score 43.95; Po0.00001; 5, 50 or 99% penetrance for recessive gene; 50 Acknowledgements and 99% penetrance for dominant gene) for a gene controlling susceptibility to leprosy per se, with reduced This work was funded by grants from the Wellcome power (48%) at low (5%) penetrance for a dominant Trust. SJ held an MRC research studentship. We thank gene. For leprosy subtypes, the LL and TT families each the people of Belem for their cooperation in this study.

Genes and Immunity Genetic susceptibility to tuberculosis and leprosy SE Jamieson et al 56 Electronic Database Information 15 Shaw M-A, Collins A, Peacock CS et al. Evidence that genetic susceptibility to Mycobacterium tuberculosis in a Brazilian population is under oligogenic control: linkage study of the 1. WHO site for tuberculosis prevalence including candidate genes NRAMP1 and TNFA. Tuberc Lung Dis 1997; Brazil http://www.who.int/gtb/publications/glo- 78: 35–45. brep02/PDF/Regional_Profile-Americas.pdf 16 Bellamy R, Ruwende C, Corrah T, McAdam KPWJ, Whittle 2. WHO site for leprosy prevalence including Brazil HC, Hill AVS. Variation in the NRAMP1 gene is associated http://www.who.int/lep/stat2002/Americas02.htm with susceptibility to tuberculosis in West Africans. N Engl J 3. Genetic Location Database http://cedar.genetics.so- Med 1998; 338: 640–644. ton.ac.uk/pub/chrom17/map 17 Greenwood CM, Fujiwara TM, Boothroyd LJ et al. Linkage of 4. Whitehead Institute http://www.genome.wi.mit. tuberculosis to chromosome 2q35 loci, including NRAMP1, in edu a large aboriginal Canadian family. Am J Hum Genet 2000; 67: 5. UCSC Genome Browser http://genome.ucsc.edu/ 405–416. 6. Ensembl Genome Browser http://www.ensembl.org/ 18 Gao PS, Fujishima S, Mao XQ et al. Genetic variants of 7. CRI-Map http://www.hgmp.mrc.ac.uk NRAMP1 and active tuberculosis in Japanese populations. International Tuberculosis Genetics Team. Clin Genet 2000; 58: 8. MAP-O-MAT http://compgen.rutgers.edu/mapo- 74–76. mat 19 Cervino AC, Lakiss S, Sow O, Hill AV. Allelic association 9. Cambridge Institute for Medical Research—David between the NRAMP1 gene and susceptibility to tuberculosis Clayton software http://www-gene.cimr.cam.ac.uk/ in Guinea-Conakry. Ann Hum Genet 2000; 64: 507–512. clayton/software/ 20 Ryu S, Park YK, Bai GH, Kim SJ, Park SN, Kang S. 30UTR 10. Wellcome Trust Human Genetics Centre, Oxford polymorphisms in the NRAMP1 gene are associated with http://www.gmap.net/analysis.htm susceptibility to tuberculosis in Koreans. Int J Tuberc Lung Dis 2000; 4: 577–580. 21 Awomoyi AA, Marchant A, Howson JM, McAdam KP, Blackwell JM, Newport MJ. Interleukin-10, polymorphism in References SLC11A1 (formerly NRAMP1), and susceptibility to tubercu- losis. J Infect Dis 2002; 186: 1804–1814. 1 Fine PEM. Immunogenetics of susceptibility to leprosy, 22 Liaw YS, Tsai-Wu JJ, Wu CH et al. Variations in the NRAMP1 tuberculosis, and leishmaniasis. An epidemiological perspec- gene and susceptibility of tuberculosis in Taiwanese. Int J tive. Int J Leprosy 1981; 49: 437–454. Tuberc Lung Dis 2002; 6: 454–460. 2 Newport MJ, Blackwell JM. Genetic susceptibility to tubercu- 23 Abel L, Sanchez FO, Oberti J et al. Susceptibility to leprosy losis. Baillier’s Clin Infect Dis Mycobact Dis 1997; 4: 207–230. is linked to the human NRAMP1 gene. J Infect Dis 1998; 177: 3 Blackwell JM. Genetic susceptibility to leishmanial infections: 133–145. studies in mice and man. Parasitology 1996; 112: S67–S74. 24 Mohamed HS, Ibrahim ME, Miller EN et al. SLC11A1 4 Blackwell JM, Black GF, Peacock CS et al. Immunogenetics of (formerly NRAMP1) and susceptibility to visceral leishma- leishmanial and mycobacterial infections: the Belem Family niasis in The Sudan. Eur J Hum Genet 2003 in press. Study. Phil Trans R Soc B 1997; 352: 1331–1345. 25 Bucheton B, Abel L, Kheir MM et al. Genetic control of visceral 5 Roberts M, Mock BA, Blackwell JM. Mapping of genes leishmaniasis in a Sudanese population: candidate gene controlling Leishmania major infection in CXS recombinant testing indicates a linkage to the NRAMP1 region. Genes inbred mice. Eur J Immunogenet 1993; 20: 349–362. Immun 2003; 4: 104–109. 6 Mock B, Blackwell J, Hilgers J, Potter M, Nacy C. Genetic 26 Lander ES, Green P. Construction of multilocus genetic control of Leishmania major infection in congenic, recombinant linkage maps in humans. Proc Natl Acad Sci USA 1987; 84: inbred and F2 populations of mice. Eur J Immunogenet 1993; 2363–2367. 20: 335–348. 27 Matise TC, Gitlin JA. MAP-O-MAT: marker-based linkage 7 Bradley DJ. Genetic control of natural resistance to Leishmania mapping on the World Wide Web. Am J Hum Genet 1999; 65: donovani. Nature 1974; 250: 353–354. A435. 8 Bradley DJ, Kirkley J. Regulation of Leishmania populations 28 Spielman RS, McGinnis RE, Ewens WJ. Transmission test for within the host. I. The variable course of Leishmania donovani linkage disequilibrium: the insulin gene region and insulin- infections in mice. Clin Exp Immunol 1977; 30: 119–129. dependent diabetes mellitus (IDDM). Am J Hum Genet 1993; 9 Bradley DJ. Regulation of Leishmania populations within the 52: 506–516. host. II. Genetic control of acute susceptibility of mice 29 Cordell HJ, Clayton DG. A unified stepwise regression to Leishmania donovani infection. Clin Exp Immunol 1977; procedure for evaluating the relative effects of polymorphisms 30: 130–140. within a gene using case/control or family data: application 10 Bradley DJ, Taylor BA, Blackwell J, Evans EP, Freeman J. to HLA in type 1 diabetes. Am J Hum Genet 2002; 70: Regulation of Leishmania populations within the host. III. 124–141. Mapping of the locus controlling susceptibility to visceral 30 Bland JM, Altman DG. Multiple significance tests: the leishmaniasis in the mouse. Clin Exp Immunol 1979; 37: 7–14. Bonferroni method. BMJ 1995; 310: 170. 11 Gros P, Skamene E, Forget A. Genetic control of natural 31 Wicker LS, Todd JA, Prins JB, Podolin PL, Renjilian RJ, resistance to Mycobacterium bovis (BCG) in mice. J Immunol Peterson LB. Resistance alleles at two non-major histocompat- 1981; 127: 2417–2421. ibility complex-linked insulin-dependent diabetes loci on 12 Brown IN, Glynn AA, Plant J. Inbred mouse strain resistance chromosome 3, Idd3 and Idd10, protect nonobese diabetic to Mycobacterium lepraemurium follows the Ity/Lsh pattern. mice from diabetes. J Exp Med 1994; 180: 1705–1713. Immunology 1982; 47: 149–156. 32 Podolin PL, Denny P, Lord CJ et al. Congenic mapping of the 13 Skamene E, Gros P, Forget A, Patel PJ, Nesbitt MN. Regulation insulin-dependent diabetes (Idd) gene, Idd10, localizes two of resistance to leprosy by chromosome 1 locus in the mouse. genes mediating the Idd10 effect and eliminates the candidate Immunogenetics 1984; 19: 117–124. Fcgr1. J Immunol 1997; 159: 1835–1843. 14 Goto Y, Buschman E, Skamene E. Regulation of host resistance 33 Serreze DV, Bridgett M, Chapman HD, Chen E, Richard SD, to Mycobacterium intracellulare in vivo and in vitro by the Bcg Leiter EH. Subcongenic analysis of the Idd13 locus in NOD/Lt gene. Immunogenetics 1989; 30: 218–221. mice: evidence for several susceptibility genes including a

Genes and Immunity Genetic susceptibility to tuberculosis and leprosy SE Jamieson et al 57 possible diabetogenic role for beta 2-microglobulin. J Immunol 48 Sadek MI, Sada E, Toossi Z, Schwander SK, Rich EA. 1998; 160: 1472–1478. Chemokines induced by infection of mononuclear phagocytes 34 Morel L, Blenman KR, Croker BP, Wakeland EK. The major with mycobacteria and present in lung alveoli during active murine systemic lupus erythematosus susceptibility locus, pulmonary tuberculosis. Am J Respir Cell Mol Biol 1998; 19: Sle1, is a cluster of functionally related genes. Proc Natl Acad 513–521. Sci USA 2001; 98: 1787–1792. 49 Lande R, Giacomini E, Grassi T et al. IFN-alphabeta released 35 Burgner D, Usen S, Rockett K et al. Nucleotide and haplotypic by Mycobacterium tuberculosis-infected human dendritic cells diversity of the NOS2A promoter region and its relationship induces the expression of CXCL10: selective recruitment of to cerebral malaria. Hum Genet 2003; 112: 379–386. NK and activated T cells. J Immunol 2003; 170: 1174–1182. 36 Hobbs MR, Udhayakumar V, Levesque MC et al. A new NOS2 50 Sallusto F, Lenig D, Forster R, Lipp M, Lanzavecchia A. Two promoter polymorphism associated with increased nitric subsets of memory T with distinct homing oxide production and protection from severe malaria in potentials and effector functions. Nature 1999; 401: 708–712. Tanzanian and Kenyan children. Lancet 2002; 360: 1468–1475. 51 Aickin M, Gensler H. Adjusting for multiple testing when 37 Choi HS, Rai PR, Chu HW, Cool C, Chan ED. Analysis of nitric reporting research results: the Bonferroni vs Holm methods oxide synthase and nitrotyrosine expression in human [see comments] 1996; 86: 726–728. pulmonary tuberculosis. Am J Respir Crit Care Med 2002; 166: 52 Knapp M. A note on power approximations for the transmis- 178–186. sion/disequilibrium test. Am J Hum Genet 1999; 64: 1177–1185. 38 Tunctan B, Okur H, Calisir CH et al. Comparison of nitric 53 Ridley DS, Jopling WH. Classification of leprosy according to oxide production by /macrophages in healthy immunity. A five-group system. Int J Lepr Mycobact Dis 1966; subjects and patients with active pulmonary tuberculosis. 34: 255–273. Pharmacol Res 1998; 37: 219–226. 54 Xu W, Liu L, Emson PC, Harrington CR, Charles IG. Evolution 39 Rich EA, Torres M, Sada E, Finegan CK, Hamilton BD, Toossi of a homopurine–homopyrimidine pentanucleotide repeat Z. Mycobacterium tuberculosis (MTB)-stimulated production of sequence upstream of the human inducible nitric oxide nitric oxide by human alveolar macrophages and relationship synthase gene. Gene 1997; 204: 165–170. of nitric oxide production to growth inhibition of MTB. Tuberc 55 Weber JL, Kwitek AE, May PE, Wallace MR, Collins FS, Lung Dis 1997; 78: 247–255. Ledbetter DH. Dinucleotide repeat polymorphisms at the 40 Schon T, Hernandez-Pando RH, Negesse Y, Leekassa R, D17S250 and D17S261 loci. Nucleic Acids Res 1990; 18: 4640. Sundqvist T, Britton S. Expression of inducible nitric oxide 56 Liu H, Chao D, Nakayama EE et al. Polymorphism in RANTES synthase and nitrotyrosine in borderline leprosy lesions. Br J chemokine promoter affects HIV-1 disease progression. Proc Dermatol 2001; 145: 809–815. Natl Acad Sci USA 1999; 96: 4581–4585. 41 Hieshima K, Imai T, Baba M et al. A novel human CC 57 An P, Nelson GW, Wang L et al. Modulating influence on HIV/ chemokine PARC that is most homologous to macrophage- AIDS by interacting RANTES gene variants. Proc Natl Acad Sci inflammatory protein-1 alpha/LD78 alpha and chemotactic USA 2002; 99: 10002–10007. for T lymphocytes, but not for . J Immunol 1997; 159: 58 Gonzalez E, Dhanda R, Bamshad M et al. Global survey of 1140–1149. genetic variation in CCR5, RANTES, and MIP-1alpha: impact 42 Adema GJ, Hartgers F, Verstraten R et al. A dendritic-cell- on the epidemiology of the HIV-1 pandemic. Proc Natl Acad Sci derived C-C chemokine that preferentially attracts naive T USA 2001; 98: 5199–5204. cells. Nature 1997; 387: 713–717. 59 Rovin BH, Lu L, Saxena R. A novel polymorphism in the 43 Kodelja V, Muller C, Politz O, Hakij N, Orfanos CE, Goerdt S. MCP-1 gene regulatory region that influences MCP-1 expres- Alternative macrophage activation-associated CC-chemokine- sion. Biochem Biophys Res Commun 1999; 259: 344–348. 1, a novel structural homologue of macrophage inflammatory 60 Holmans P, Clayton D. Efficiency of typing unaffected protein-1 alpha with a Th2-associated expression pattern. J relatives in an affected sib-pair linkage study with single Immunol 1998; 160: 1411–1418. locus and multiple tightly-linked markers. Am J Hum Genet 44 Brown DH, Zwilling BS. Activation of the hypothalamic–pi- 1995; 37: 1221–1232. tuitary–adrenal axis differentially affects the anti-mycobacter- 61 Gudbjartsson DF, Jonasson K, Frigge ML, Kong A. Allegro, a ial activity of macrophages from BCG-resistant and new computer program for multipoint linkage analysis. Nat susceptible mice. J Neuroimmunol 1994; 53: 181–187. Genet 2000; 25: 12–13. 45 Brown DH, Sheridan J, Pearl D, Zwilling BS. Regulation of 62 Kong A, Cox NJ. Allele-sharing models: LOD scores and mycobacterial growth by the hypothalamus–pituitary–adrenal accurate linkage tests. Am J Hum Genet 1997; 61: 1179–1188. axis: differential responses of Mycobacterium bovis 63 Hedrick PW. Gametic disequilibrium measures: proceed with BCG-resistant and -susceptible mice. Infect Immun 1993; 61: caution. Genetics 1987; 117: 331–341. 4793–4800. 64 Stephens M, Smith NJ, Donnelly P. A new statistical method 46 Matsukawa A, Lukacs NW, Hogaboam CM, Chensue SW, for haplotype reconstruction from population data. Am J Hum Kunkel III SL. Chemokines and other mediators, 8. Chemo- Genet 2001; 68: 978–989. kines and their receptors in cell-mediated immune responses 65 Abecasis GR, Cookson WO. GOLD—graphical overview of in the lung. Microsc Res Tech 2001; 53: 298–306. linkage disequilibrium. Bioinformatics 2000; 16: 182–183. 47 Ragno S, Romano M, Howell S, Pappin DJ, Jenner PJ, Colston 66 Kun JF, Mordmuller B, Lell B, Lehman LG, Luckner D, MJ. Changes in in macrophages infected with Kremsner PG. Polymorphism in promoter region of inducible Mycobacterium tuberculosis: a combined transcriptomic and nitric oxide synthase gene and protection against malaria. proteomic approach. Immunology 2001; 104: 99–108. Lancet 1998; 351: 265–266.

Genes and Immunity