Overlapping BXSB Congenic Intervals, in Combination with Microarray Gene Expression, Reveal Novel Lupus Candidate Genes

ORIGINAL ARTICLE Overlapping BXSB congenic intervals, in combination with microarray gene expression, reveal novel lupus candidate genes

MEK Haywood1, SJ Rose1, S Horswell2, MJ Lees1,GFu3, MJ Walport4 and BJ Morley1 1Rheumatology Section, Division of Medicine, Imperial College Faculty of Medicine, London, UK; 2Medical Research Council, Clinical Sciences Centre, Hammersmith Hospital, London, UK; 3Oxitec Limited, Oxford, UK and 4The Wellcome Trust, London, UK

The BXSB mouse strain is an important model of glomerulonephritis observed in systemic lupus erythematosus (SLE). Linkage studies have successfully identified disease-susceptibility intervals; however, extracting the identity of the susceptibility gene(s) in such regions is the crucial next step. Congenic mouse strains present a defined genetic resource that is highly amenable to microarray analysis. We have performed microarray analysis using a series of chromosome 1 BXSB congenic mice with partially overlapping disease-susceptibility intervals. Simultaneous comparison of the four congenic lines allowed the identification of expression differences associated with both the initiation and progression of disease. Thus, we have identified a number of novel SLE disease gene candidates and have confirmed the identity of Ifi202 as a disease candidate in the BXSB strain. Sequencing of the promoter regions of Gas5 has revealed polymorphisms in the BXSB strain, which may account for the differential expression profile. Furthermore, the combination of the microarray results with the different phenotypes of these mice has allowed the identification of a number of expression differences that do not necessarily map to the congenic interval, but may be implicated in disease pathways. Genes and Immunity (2006) 7, 250–263. doi:10.1038/sj.gene.6364294; published online 16 March 2006

Keywords: autoimmunity; systemic lupus erythematosus; rodent; microarray; candidate genes

Introduction consistently linked to SLE in different models, particularly distal chromosome 1.5–7 However, it is also clear The BXSB mouse is a recombinant inbred strain1 that that several other intervals, unique to each mouse model, provides a model of the proliferative glomerulonephritis play crucial roles in the development of disease. For observed in human systemic lupus erythematosus (SLE). example, in BXSB mice, Bxs1 on chromosome 1 is now This strain develops a range of autoantibodies directed at known to be an important locus for the development of nuclear components, particularly anti-DNA and anti- nephritis,8,9 but this interval is not linked to disease in 2 10,11 chromatin. Alongside this spectrum of autoantibodies, either the (NZB Â NZW)F1 or MRL models. Identifi- more general indicators of immune system dysregulation cation of the genetic basis of these disease-susceptibility can be observed, specifically splenomegaly and lympha- intervals is fundamental to developing our understand- denopathy.1 The culmination of disease occurs in the ing of this prototypic autoimmune disease. kidneys, where mesangial matrix increase and cellular In BXSB mice, in addition to the autosomal loci, the Y- infiltration are observed as characteristic features of the chromosome carries an autoimmune accelerator gene glomerulonephritis in the BXSB strain.2 (Yaa).12 Though the identity of this gene is still unclear, it Genetic studies in both mouse models and human has been demonstrated to accelerate disease observed in populations have indicated that SLE is a complex genetic BXSB male mice in comparison to their female counter- disease; thus, multiple loci contribute to disease suscept- parts.12 Transfer of the BXSB Y-chromosome by breeding ibility, with no absolute requirement for any single onto other autoimmune prone strains of mice, such as locus.3 Comparison of the many linkage studies per- NZB and NZW, has been shown to accelerate disease formed in the lupus prone mouse strains, MRL, BXSB progression.13 It has only recently become clear that Yaa and (NZB Â NZW)F1, amply illustrates the complexity of is capable of breaking tolerance in a non-autoimmune this disease.4 Some chromosomal intervals have been prone strain such as C57BL/10 (B10).8 Whereas linkage studies are invaluable for identifying chromosomal intervals in which disease-susceptibility Correspondence: Dr B Morley, Rheumatology Section, Division of loci are located, they are limited in their resolution. The Medicine, Imperial College, Hammersmith Campus, Du Cane Road, resultant disease-associated interval may contain several London W12 0NN, UK. E-mail: [email protected] hundred positional candidate genes, of which several Received 9 September 2005; revised 2 February 2006; accepted 2 may present as reasonable functional candidates. In February 2006; published online 16 March 2006 order to confirm our linkage data, we generated congenic Microarray analysis of BXSB congenic mice MEK Haywood et al 251 mice, in which the disease-associated intervals of potential candidates, even within the congenic intervals, chromosome 1 were bred onto a non-autoimmune strain microarrays offer a rapid way of screening for expression (B10) containing the Yaa gene.8 Thus, the individual and level differences. Although expression profiling may not isolated contribution of each linkage region to the overall identify functional differences underlying the disease, it phenotype could be interrogated. These novel mouse may detect perturbations in downstream expression strains present a restricted and defined genetic resource, levels secondary to the primary genetic defects on with minimal genetic variation from the non-affected B10 chromosome 1. Such a ‘disease pathway’ may be parental strain, and each interval contributes different identified by examining gene expression in those strains aspects of the overall phenotype8 (Figure 1). with related phenotypes and determining any common The next stage to dissecting the genetic basis of SLE is differentially expressed genes. the identification of candidate genes. One approach is to A major problem in the expression profiling approach select genes that are differentially expressed between is the large number of candidates that may be identified. congenic and control strains. Given the large number of These congenic strains provide an ideal resource to

Figure 1 Schematic representation of chromosome 1 BXSB congenic mouse strains. The microsatellite markers used to direct the breeding for each strain are shown on the left-hand side (scale in Mb). Disease-susceptibility linkage intervals are indicated by grey boxes and named according to Hogarth et al.7 and Haywood et al.9 The congenic interval is indicated by the dark grey bar, and the pale grey bars indicate regions in which the location of genetic recombination remains to be determined. Thus, the B10.Yaa.Bxs1 strain contains the BXSB-derived interval from 46.3 to 89.2 Mb, B10.Yaa.Bxs1/4 spans 20.0–65.9 Mb, B10.Yaa.Bxs1/2 encompasses 64.4–159.0 Mb and B10.Yaa.Bxs2/3, 105.4– 189.0 Mb. The phenotypes associated with each interval,8 or overlapping interval, are indicated on the right-hand side.

Genes and Immunity Microarray analysis of BXSB congenic mice MEK Haywood et al 252 reduce background in such an analysis.14,15 Each con- tosis in B10.Yaa.Bxs2/3.8 However, the samples used for genic interval is relatively small in comparison to the the gene expression analysis presented here were whole genome, so the confounding influence of natural collected prior to this change and to the development variations between strains can be minimised. In addition, of disease. each congenic interval partially overlaps with at least one of the other strains. Thus, the validity of candidate genes Overlapping congenic intervals aid identification of novel can, in part, be determined by the expression profile candidate genes across the strains. By definition, a differentially ex- Over- and underexpressed probes for each congenic pressed candidate gene should be expressed at a similar strain were identified using the arbitrary level of twofold level in all strains containing the BXSB allele of that gene, change in expression levels with respect to B10.16,17 The and must be differentially expressed in all strains results for each congenic strain were overlapped with carrying the B10 allele. To this end, we have carried those for both BXSB and B10.Yaa mice. This allowed us to out an extensive microarray analysis on all four congenic identify the primary genetic causes of lupus attributable strains together with B10.Yaa, B10 and BXSB as controls, to the specific chromosome 1 congenic interval rather to direct our future studies. than effects associated with Yaa. Our results therefore B10.Yaa.Bxs2/3 mice develop a phenotype that is show those genes with altered expression levels in both closest to the parental BXSB strain, though with later the congenic line and BXSB, but not in B10.Yaa. All onset, 12 months for 50% mortality compared to 6 differentially expressed probe sequences were BLAST months for BXSB.8 This congenic strain displays acceler- searched using NCBI Genome Blast (http://www.ncbi.- ated mortality, a broad spectrum of autoantibodies and nih.gov/entrez/) to identify chromosomal locations for age-dependent glomerulonephritis with focal mesangial each sequence, and checked against the Ensembl hypercellularity accompanied by a matrix increase that database. As many Affymetrix probe sequences origi- was qualitatively similar to the lesions observed in BXSB nated from complementary DNA (cDNA) sequences, mice.8 B10.Yaa.Bxs1/4 mice also developed similar ne- NCBI Unigene was used to identify the genes repre- phritic lesions, along with a broad spectrum of auto- sented by some probes. Differentially expressed se- antibodies, but the onset was later still (12 months quences that were located on chromosome 1 were mortality 25%).8 B10.Yaa.Bxs1 mice developed glomer- plotted with respect to the congenic intervals on ulonephritis in the absence of autoantibodies, whereas chromosome 1 (Figures 2 and 3). B10.Yaa.Bxs1/2 mice developed autoantibodies without concomitant nephritis. X Kell Blood Group 4-related protein is a candidate for Bxs4 B10.Yaa.Bxs1/4 mice developed a wide spectrum of Results autoantibodies at high titres, in addition to nephritis associated with the Bxs1 locus.8 Candidate genes for the Congenic mice as a microarray resource Bxs4 autoantibody–susceptibility locus must map to the The congenic strains examined in this study have interval specific to B10.Yaa.Bxs1/4, excluding any overlap considerable overlaps with each other (Figure 1). Each with B10.Yaa.Bxs1, that is, from the centromere to congenic strain contained between 2.6 and 4.4% of the D1Mit235 (46.8 Mb) (Figure 1). Bxs4 candidates can only BXSB autosome. In addition, each strain was consomic be differentially expressed in the B10.Yaa.Bxs1/4 con- for the BXSB Y-chromosome. In parental BXSB mice, genic. Any sequences mapping to the interval between disease onset, as measured by the severity of glomer- D1Mit212 (40.6 Mb) and D1Mit235 cannot be excluded ulonephritis, was not detectable prior to 2 months of age, from consideration, because we are uncertain how much and thus congenic strains were sampled at this time of this interval is present in B10.Yaa.Bxs1. The interval point to obtain material for microarray analysis. from the centromere to D1Mit235 contains some 198 In autoimmune diseases, it is likely that multiple cell known genes (Ensembl v32), of which 132 (66.7%) were types play a role in disease; however, the effect of any interrogated by the Affymetrix MG-U74v2 chipset. individual locus, as contained in the congenic mice, may In the Bxs4 target interval, we identified three over- be restricted to a particular cell type. We chose to profile expressed and 10 underexpressed sequences (Figures 2 splenocytes by reason that the major immune cell types, and 3). However, only one of these candidates (AI854359) B cells, T cells and macrophages, would be present in our remained after comparison between the multiple con- samples. In any gene expression analysis, it is crucial that genic strains when analysed using Genespring, illustrat- the cellular profile of samples should be as closely ing the power of the overlapping congenic approach. matched as possible. Fluorescence-activated cell sorting AI854359, a clone that is the mouse homologue of the (FACS) analysis was used to eliminate mice that showed Human X Kell Blood Group 4-related protein (Xkr4), was abnormal cellular profiles, such as elevated macrophage 3.8-fold underexpressed (P ¼ 0.033) in B10.Yaa.Bxs1/4 counts, which may occur as a result of infection. No mice mice. We have confirmed this result using real-time were excluded from the B10.Yaa.Bxs1/4 or B10.Yaa polymerase chain reaction (PCR) (data normalised to cohorts, one was excluded from the B10.Yaa.Bxs1 cohort, B10 ¼ 170.30, BXSB ¼ 0.00470.002, B10.Yaa.Bxs1/ two each were excluded from the B10, B10.Yaa.Bxs1/2 4 ¼ 0.02070.010). A further candidate was identified and BXSB cohorts, and three from the B10.Yaa.Bxs2/3 using the significance analysis method (SAM) analysis cohort, all on the basis of elevated macrophage counts approach, valosin containing protein complex interacting (percentage Mac-1-stained cells 45.5%). Our detailed protein (Vcpip1) at 9.8 Mb, which was downregulated phenotype study of these mice revealed that there was an only in B10.Yaa.Bxs1/4 mice. We did not obtain a age-associated increase in B220 þ cells in the consensus between the two analysis methods on any of B10.Yaa.Bxs1/4 and B10.Yaa.Bxs2/3 strains, and monocy- the possible genes in this region.

Genes and Immunity Microarray analysis of BXSB congenic mice MEK Haywood et al 253

Figure 2 Overexpressed sequences mapping to chromosome 1. Each sequence is denoted by the Gene name or Genbank accession number. The number in brackets indicates the level of fold change for each probe (where two values are shown, it indicates the results for two probes representing the same gene). All genes with twofold or greater change are depicted, those for which the t-test P-value o0.1 are shown in bold, with a black bar indicating the location of the gene, other probes are shown with a grey bar. For each congenic interval the marker-delineated interval selected during breeding is shown in dark grey, and the margins of each congenic are shown in pale grey. Genes which are underlined have altered expression levels in more than one congenic strain.

The Bxs3 interval contains two candidate genes, Gas5 and Two candidate genes were identified for this interval Ifi202 (Table 1), each highlighted by two separate probes in the B10.Yaa.Bxs2/3 mice contain the powerful lupus interval, fold-change analysis: interferon inducible 202 (Ifi202) and Bxs3. These mice rapidly develop severe disease.8 In growth arrest specific 5 (Gas5). The two probes for Ifi202 contrast, the overlapping congenic strain, B10.Yaa.Bxs1/2, (177.83 Mb) showed between 6.8- and 11.5- (Po0.05) fold develops only late-onset, moderate titre, anti-double- overexpression in the congenic mouse compared to stranded DNA (dsDNA) antibodies with no concomitant control B10, consistent with the greater than 10-fold nephritis.8 The phenotypic overlap of these strains overexpression of Ifi202 identified in NZB mice.15 Gas5 therefore localises Bxs3 to the interval between (164.6 Mb) was underexpressed at between 3.3- and 13.5- D1Mit33 (161.9 Mb) and the telomere. There are ap- fold (Po0.05) in congenic B10.Yaa.Bxs2/3 mice and proximately 330 known genes in the Bxs3 interval, comparably in BXSB (À3.5 to À12.5 fold, Po0.05). This of which 239 (72.4%) were represented on the gene has not previously been implicated in the patho- microarray. genesis of SLE. All other candidates were eliminated on

Genes and Immunity Microarray analysis of BXSB congenic mice MEK Haywood et al 254

Figure 3 Underexpressed sequences mapping to chromosome 1. Each sequence is denoted by the Gene name or Genbank accession number. The number in brackets indicates the level of fold change for each probe (where two values are shown, it indicates results for two probes representing the same gene). All genes with twofold or greater change are depicted, those for which the t-test P-value o0.1 are shown in bold, with a black bar indicating the location of the gene, other probes are shown with a grey bar. For each congenic interval the marker-delineated interval selected during breeding is shown in dark grey, and the margins of each congenic are shown in pale grey. Genes which are underlined have altered expression levels in more than one congenic strain.

the basis of inappropriate expression patterns across the downregulated, Bat2d (104782_at), Cdca1 (135371_at) and congenic strain panel, for example, Cd84, or high t-test P- Sdccag8 (116178_at), and one was upregulated values (P40.1) indicating unreliable or inconsistent data, (112013_at), for which there is no identified gene. for example, Hnrpu. The SAM analysis also identified one of the Gas5 Bxs2-driven anti-double-stranded DNA antibody production probes (98531_g_at) as significantly differently ex- was associated with three underexpressed sequences pressed, confirming this result. We did not confirm the B10.Yaa.Bxs1/2 mice, in common with B10.Yaa.Bxs2/3, Ifi202 probe result by this second method; however, we develop an enhanced level of anti-dsDNA antibodies, did identify probe 115624_at, which lies in the closely but do not develop nephritis or a wider spectrum of related Ifi203. Our investigations into the complex Ifi autoantibodies. The overlapping B10.Yaa.Bxs1 strain does family of genes are continuing. SAM analysis identified not develop anti-dsDNA antibodies. Thus, this pheno- four further probes located in this region that showed typic trait maps to the overlap between B10.Yaa.Bxs1/2 significant differences in B10.Yaa.Bxs2/3 mice. Three were and B10.Yaa.Bxs2/3, namely D1Mit305 (90.7 Mb) to

Genes and Immunity Microarray analysis of BXSB congenic mice MEK Haywood et al 255 Table 1 Bxs3 candidate genes

Gene name Position Affymetrix Gene expression fold changes (42) (Genbank ID) (Mb) probe ID Congenic strain

B10.Yaa.Bxs1/4 B10.Yaa.Bxs1 B10.Yaa.Bxs1/2 B10.Yaa.Bxs2/3 BXSB

Ifi202 (M31418) 177.83 94744_at n/c (P ¼ 0.058) +1.8 (P ¼ 0.273) +1.2 (P ¼ 0.273) +11.5 (P ¼ 0.009) +5.1 (P ¼ 0.168) Ifi202 161173_f_at +1.2 (P ¼ 0.540) À2.0 (P ¼ 0.004) À2.5 (P ¼ 0.110) +6.8 (P ¼ 0.049) +2.1 (P ¼ 0.894) (AV229143)

Gas5 (AI849615) 164.60 98531_g_at +1.4 (P ¼ 0.095) À1.2 (P ¼ 0.025) +1.2 (P ¼ 0.047) À13.5 (P ¼ 0.005) À12.5 (P ¼ 0.016) 98530_at +1.3 (P ¼ 0.027) À1.4 (P ¼ 0.210) À1.2 (P ¼ 0.479) À3.3 (P ¼ 0.009) À3.5 (P ¼ 0.102)

Differentially expressed sequences in B10.Yaa.Bxs2/3 and BXSB congenic mice in comparison to B10.Yaa.Bxs1/4, B10.Yaa.Bxs1 and B10.Yaa.Bxs1/2. The BXSB-derived intervals for each congenic strain are given. Only fold changes greater than twofold in magnitude are shown, and the numerical values represent fold change relative to the normalised B10 value of 1, n/c indicates there was no change between the expression level in the congenic strain and B10. The t-test P-value statistic indicates the significance of the variation from a B10 normalised value of 1 for three replicate samples.

Table 2 Bxs2 candidate genes

Gene name Position/ Affymetrix Gene expression fold changes (42) (Genbank ID) Mb probe ID Congenic strain

B10.Yaa.Bxs1/4 B10.Yaa.Bxs1 B10.Yaa.Bxs1/2 B10.Yaa.Bxs2/3 BXSB

EST (AI427984) 100.47 131703_at +1.7 (P ¼ 0.097) À1.8 (P ¼ 0.377) À9.4 (P ¼ 0.006) À2.2 (P ¼ 0.095) À5.0 (P ¼ 0.049) Marco (U18424) 123.03 102974_at n/c (P ¼ 0.979) À1.6 (P ¼ 0.078) À2.4 (P ¼ 0.006) À8.5 (P ¼ 0.012) À2.9 (P ¼ 0.094) EST (AA756546) 159.89 115058_at +2.4 (P ¼ 0.027) n/c (P ¼ 0.904) À2.1 (P ¼ 0.005) À4.6 (P ¼ 0.127) À2.5 (P ¼ 0.019)

Differentially expressed sequences in B10.Yaa.Bxs1/2, B10.Yaa.Bxs2/3 and BXSB congenic mice in comparison to B10.Yaa.Bxs1/4 and B10.Yaa.Bxs1. Only fold changes greater than twofold in magnitude are shown, and the numerical values represent fold change relative to the normalised B10 value of 1, n/c indicates that there was no change between the expression level in the congenic strain and B10. The t-test P- value statistic indicates the significance of the variation from a B10 normalised value of 1 for three replicate samples.

D1Mit107 (165.9 Mb).8 This is a large region of approxi- MG-U74v2 microarrays. Candidate genes for this locus mately 74 Mb containing 631 known genes, of which 425 should be differentially expressed in B10.Yaa.Bxs1/4 and (67.3%) were assayed here (Figures 2 and 3). B10.Yaa.Bxs1, but not in the partially overlapping Comparison of the B10.Yaa.Bxs1/2 and B10.Yaa.Bxs2/3 B10.Yaa.Bxs1/2 mice, which do not develop nephritis. strains by fold change analysis revealed three under- There were no probes that satisfied these criteria and expressed sequences that satisfied our screening criteria thus we could not identify a candidate gene for Bxs1,by and no overexpressed sequences (Table 2). Two of these either method of analysis. probes represented expressed sequence tags (ESTs), AI427984 and AA756546. The former showed 50% Validation of microarray results homology at a protein level to Mus musculus immuno- The approach taken in this study was to include probe globulin E-binding protein, while the latter was origin- sets in the analysis, irrespective of Affymetrix flag data. ally derived from a thymus library and lies Also, we did not set a lower threshold for the signal approximately 60 kb 50 upstream of sterol O-acyltransfer- intensity, so that all signals were included. However, this ase (Soat1). The remaining underexpressed probe, Marco, necessitated a careful validation of the results. We took is a macrophage scavenger receptor,18 which functions to two approaches to this validation. Firstly, we designed recognise both bacterial cells and apoptotic cells. SAM real-time quantitative PCR assays for the candidate analysis did not identify these sequences, but did genes. The primers were designed to amplify the region highlight an overexpressed probe (163011_at) that repre- of the gene that was represented by the Affymetrix sents a tRNA methyltransferase-like sequence at probe. Secondly, we reasoned that the primary genetic 151.3 Mb. basis of any mRNA expression level difference was likely to be mutation(s) in the promoter of the gene, so we No Bxs1 candidate genes sequenced upstream of the transcription start site of each The Bxs1 locus contains a novel BXSB nephritis-suscept- candidate gene in BXSB and B10 to identify potential ibility gene, which is capable of directing the develop- sequence differences. We focussed on the two novel ment of nephritis in the absence of significant candidate genes with known function for this approach. autoantibody production.8 This gene must lie in the Real-time quantitative PCR analysis of the Bxs3 overlapping interval between D1Mit212 (40.6 Mb) and candidate, Gas5 (Table 3), confirmed the underexpression D1Mit303 (64.5 Mb). This interval contains 109 known detected by the microarray data (Table 1), with a genes, of which 80 were interrogated in this study using downregulation of 11.0–58.1-fold for BXSB and

Genes and Immunity Microarray analysis of BXSB congenic mice MEK Haywood et al 256 B10.Yaa.Bxs2/3 compared with B10 and the other con- B10.Yaa.Bxs1/2 mice do not. The histological changes genic strains (excluding Gapdh results). Sequence analy- are qualitatively the same for all three nephritis strains, sis of the promoter of Gas5 spanned 1621 nucleotides indicating potential convergence of the disease path- upstream of the transcription start site.19 Six single- ways. Commonly differentially regulated genes in the nucleotide polymorphisms were observed in BXSB, nephritis-prone congenic strains may indicate compo- compared to both B10 and B6, which were identical: nents in this pathway. À1285 (G in B10, A in BXSB), À1122 (T to G), À314 (G to We have identified by fold change analysis four genes C), À277 (A to C), À267 (A to T) and À258 (G to A). The potentially involved in nephritis (Table 4). A single polymorphism at À267 was predicted to lead to a loss of overexpressed sequence corresponded to the 50UTR of an Sp1 consensus-binding site. the ribosomal protein S6 kinase a1(Rps6ka1 gene) on The Bxs2 candidate, Marco, was excluded by real-time chromosome 4. Rps6ka1 may mediate the activation of quantitative PCR (data not shown). Splenic cDNA levels CREB transcription factor and lies in an interval were found to match those observed in the microarray identified as a lupus nephritis susceptibility locus in 20 analysis, with B10.Yaa.Bxs2/3 and BXSB levels being the (NZB Â NZW)F1 model, Nba1. Three underex- consistently lower than B10. However, fluctuations were pressed sequences were identified. One of these probes also observed in the Marco levels in the other chromo- (derived from AW050000) does not correlate with any some 1 congenic strains and in B10.Yaa, such that lower known or predicted gene nor has significant homology to levels were observed in all the congenic strains. While any other gene(s). However, this sequence maps to the this does not fully exclude Marco from a role in disease, Bxs3 interval on chromosome 1. A second underex- its candidacy as the Bxs2 allele has been undermined. pressed sequence maps to the 50UTR of a component of the STAG transcription coactivator–hypoxanthine–ami- Disease phenotype-associated candidate genes nopterin–thymine complex, ST6G (STAGA complex 65 g We have previously phenotyped each of the congenic subunit praline-rich extension) on chromosome 5. The strains presented here.8 The primary genetic defects third underexpressed probe represented exon 12 of the leading to the development of the autoimmune pheno- St13 tumour suppressor gene on chromosome 15.21 type of each congenic strain are attributable to the By contrast, SAM analysis identified a more limited set particular BXSB intervals that they contain. However, of nephritis-associated genes; only the underexpression there is a convergence of phenotype observed for some of Als2cl (11038_at, chromosome 9, 113.7 Mb) was strains and it is likely that common downstream path- detected. ways are involved in disease expression. We have used our data to look for expression level changes associated Bxs1-driven nephritis candidate genes. B10.Yaa.Bxs1/4 and with specific aspects of the phenotype (Table 4). These B10.Yaa.Bxs1 partially overlap, but are distinct from genes may indicate disease-associated pathways, and B10.Yaa.Bxs2/3. Hence, the primary genetic basis of the may be candidate genes in other lupus mouse models or nephritis associated with Bxs1 may drive an additional indeed humans. We selected genes that exceeded the set of genes to those associated with a common nephritic twofold level of expression change in BXSB and in pathway. Indeed, we found three overexpressed and various combinations of congenics according to common six underexpressed probes, that were common to shared phenotypic characteristics, for example, nephritis the Bxs1-driven nephritis strains (Table 4). One of the was associated with Bxs2/3, Bxs1/4 and Bxs1, but not overexpressed genes, Mast cell protease 6 (Mcp6, with Bxs1/2 intervals, so we examined the data to look chromosome 1722), maps to a region close to the Asbb3 for commonly over- or underexpressed genes in the three proteinuria locus on chromosome 17.23 One of the nephritis-prone strains. underexpressed genes plays a role in G-protein-coupled receptor protein signalling pathways; Fzd4 mapped to Nephritis-associated candidate genes. B10.Yaa.Bxs2/3, the Nba3 interval on chromosome 7.24 Jip-1b (chromo- B10.Yaa.Bxs1/4 and B10.Yaa.Bxs1 mice all develop ne- some 2) is known to be important for DNA-dependent phritis to varying degrees of severity,8 whereas transcription regulation. Lysosomal trafficking regulator

Table 3 Validation of Gas5 as a candidate gene by QRT-PCR

Mouse strain Fold change relative to B10

Microarray RT-QPCR

98530_at 98531_g_at 18SRNA Gapdh b-actin Ubc

BXSB À3.5 À12.5 À13.6378.07 À1.8271.00 À11.0076.96 À38.76722.82 B10.Yaa.Bxs1/4 +1.3 +1.4 À1.2470.71 4.6271.12 À1.0370.62 1.3070.41 B10.Yaa.Bxs1 À1.4 À1.2 À1.2070.69 2.4070.42 1.3470.52 À1.0270.59 B10.Yaa.Bxs1/2 À1.2 +1.2 À1.2070.71 1.9170.30 1.2370.55 À1.0470.59 B10.Yaa.Bxs2/3 À3.3 À13.5 À14.6578.26 À2.2871.23 À9.3375.66 À58.09733.73 B10.Yaa À1.1 +1.1 À1.4170.91 4.8471.4 1.7270.90 À1.4370.78

QRT-PCR, quantitative reverse transcriptase-polymerase chain reaction. Gas5 mRNA was measured in BXSB congenic strains, relative to the control B10 strain. The data for the microarray probes for Gas5 are shown for comparison. The fold change relative to B10 is indicated for each strain, with a negative sign indicating down-regulation.

Genes and Immunity Table 4 Pathway candidate genes

Trait B10.Yaa. B10.Yaa. B10.Yaa. B10.Yaa. Chromosomal Mb Gene Genbank Affymetrix Description Function Over/Under B10.Yaa. B10.Yaa. B10.Yaa. B10.Yaa. BXSB Bxs1/4 Bxs1 Bxs1/2 Bxs2/3 location ID Bxs1/4 Bxs1 Bxs1/2 Bxs2/3

Nephritis ++ + À +++ 1 183.20 — AW050000 106903_at — Under 3.68 2.10 1.37 2.25 2.53

Nephritis ++ + À +++ 4 132.30 Rps6ka1 AV321321 161419_r_at Ribosomal protein S6 Serine/threonine Over 2.03 2.61 1.56 2.51 6.24 kinase a1 kinase, CREB activation

Nephritis ++ + À +++ 5 29.93 ST6G AV122294 169004_at STAGA complex 65 Component of STAG Under 2.55 2.51 1.54 2.80 2.26 gamma subunit transcription praline-rich extension coactivator–HAT complex

Nephritis ++ + À +++ 15 81.70 St13 AW124318 97310_at St13 tumour suppressor Under 2.37 2.38 1.70 2.19 2.72 gene

Nephritis ++ + ÀÀ 2 121.30 Hisppd2a AW125881 110587_at Histidine acid phosphatase Histidine acid Over 2.46 2.33 0.86 0.71 2.20 due to Bxs1 domain containing 2A phosphatase

Nephritis ++ + ÀÀ 2 92.28 Jip-1b AV354894 161619_f_at JNK interacting Antiapoptosis, DNA- Under 3.28 2.37 1.72 1.69 3.85 due to Bxs1 protein-1b dependent transcription (Mapk8ip1) regulation, signal transduction mice Haywood congenic MEK BXSB of analysis Microarray

Nephritis ++ + ÀÀ 3 69.20 AW049065 106876_at Spermine synthase Under 2.85 2.46 1.95 1.92 4.07 due to Bxs1

Nephritis ++ + ÀÀ 4 98.09 AV254209 164349_r_at Under 2.99 5.42 1.73 1.84 6.84 al et due to Bxs1

Nephritis ++ + ÀÀ 7 76.58 Fzd4 U43317 95771_i_at Frizzled homolog 4 G-protein-coupled Under 2.16 3.65 1.02 1.05 4.81 due to Bxs1 (Drosophila) receptor signalling pathway, wnt signalling pathway

Nephritis ++ + ÀÀ 12 3.81 AI835404 116363_at Transmembrane Over 2.27 2.82 1.84 1.24 3.10 due to Bxs1 protein

Nephritis ++ + ÀÀ 13 12.95 Lyst AA987135 130142_at Lysosomal trafficking Under 4.81 2.74 1.37 1.98 2.36 due to Bxs1 regulator (beige protein)

Nephritis ++ + ÀÀ 16 89.39 Krtap16–10 AI642522 111298_at Keratin associated Under 3.03 2.20 1.93 0.86 2.45 due to Bxs1 protein 16–10

Nephritis ++ + ÀÀ 17 24.00 Mcp6 M57625 92711_at Mast cell protease 6 Inflammatory response Over 2.10 2.07 1.31 0.53 2.03 due to Bxs1

Anti-ssDNA ++ ÀÀ+++ 1 — AI852681 105644_at — Predicted pseudogene Over 3.54 1.00 1.00 5.09 2.23 Abs ee n Immunity and Genes

Anti-ssDNA ++ ÀÀ+++ 10 20.70 Ahi1 AA985771 103734_at Abelson helper Provirus integration Over 2.43 1.97 1.92 3.46 2.53 Abs integration site (Humans). 257 ee n Immunity and Genes 258

Table 4 Continued

Anti-ssDNA ++ ÀÀ+++ 10 100.26 AI256761 112247_at Known transcript, Over 3.16 1.78 1.35 3.20 7.83 Abs centrosomal protein Cep290

Anti-ssDNA ++ ÀÀ+++ 12 19.02 AU023384 133781_at Novel transcript cd11c+ Over 2.71 1.06 0.86 2.26 2.17 Abs dendritic cells

Anti-ssDNA ++ ÀÀ+++ 15 99.36 Lmbr1l AV303675 164818_at Lipocalin-interacting Under 2.47 1.50 1.75 2.60 2.21 Abs membrane receptor

Anti-ssDNA ++ ÀÀ+++ 16 20.42 Ece2 AW050013 138204_at Endothelin-converting S-adenosyl methionine- Over 2.45 0.62 1.72 2.98 2.76 mice congenic BXSB of analysis Microarray Abs enzyme 2 dependent methyl transferase activity

Anti-ssDNA ++ ÀÀ+++ 19 30.40 Cstf1 AV340271 170579_at Cleavage stimulation Under 10.45 1.12 1.90 7.43 2.91 Abs factor 30 RNA subunit 2

Anti-dsDNA ++ À ++ +++ 3 40.82 Pgrmc2 AI842482 116987_at Progesterone membrane Steroid receptor protein Over 2.37 1.58 2.32 2.24 2.49 Haywood MEK Abs binding protein

Anti-dsDNA ++ À ++ +++ 4 152.83 Hes5 AV166814 171606_r_at Transcription factor Under 3.40 1.65 2.19 2.44 2.19 Abs Hes5 tal et Anti-dsDNA ++ À ++ +++ 5 140.45 Fscn1 AV160393 169468_i_at Fascin (singed-like 55kDa actin binding Under 3.40 1.91 2.36 3.31 2.50 Abs protein) protein

Anti-dsDNA ++ À ++ +++ 7 29.76 Gprc2a-rs2 AF022251 101351_at G protein-coupled Transmembrane Under 17.21 1.78 11.05 3.36 14.90 Abs receptor, family C, receptor group, member A related sequence

Anti-dsDNA ++ À ++ +++ 9 21.32 Ilf3 AV045900 165969_at Interleukin enhancer Nuclear factor of Under 3.08 1.83 4.06 4.28 4.10 Abs binding factor 3 activated T cells

Anti-dsDNA ++ À ++ +++ 9 76.23 Bmp5 AI426445 163125_at Bone morphogenic Under 6.59 0.79 2.45 2.31 2.48 Abs protein 5

Anti-dsDNA ++ À ++ +++ 9 110.68 Dhx30 AV237136 171315_at DEAH box polypeptide ATP-dependent RNA Under 4.09 1.87 2.43 2.33 2.60 Abs 30 isoform 1 helicase

Anti-dsDNA ++ À ++ +++ 9 136.19 Mtmr2 AA756466 130472_at Myotubularin-related Possible tyrosine Under 6.03 1.18 2.77 4.18 5.59 Abs protein 2 phosphatase

Anti-dsDNA ++ À ++ +++ 19 16.50 — AI839720 106171_at — Similar to Bcl-2 Over 3.98 0.63 2.92 5.74 2.62 Abs interacting protein, Nip2

dsDNA, double-stranded DNA; HAT, hypoxanthine–aminopterin–thymine; ssDNA, single-stranded DNA. The data for phenotype associated candidate gene microarray results. The fold change relative to B10 is indicated for each strain, with BXSB included for comparison, changes greater than twofold are shown in bold. Microarray analysis of BXSB congenic mice MEK Haywood et al 259 (Lyst, also known as the beige protein;25 chromosome 13) overlap between the pathway genes for anti-ssDNA and was also downregulated in Bxs1 nephritis. Mutations in anti-dsDNA. SAM analysis highlighted that Colec12 the human homologue of Lyst result in Chediak–Higashi (168379_i_at, chromosome 18, 9.9 Mb) was significantly Syndrome (OMIM 215000), although the function of the underexpressed; the potential role of Colec12 as a Lyst protein is still unclear.26 membrane protein with scavenger receptor activity SAM analysis identified just one downregulated makes it an attractive candidate for further study. sequence mapping to 113.7 Mb on chromosome 9 (111038_at), representing the 30 end of the Clasp2 gene, though any potential role for a CLIP-associating protein in nephritis is not immediately apparent. Discussion The dissection of the mechanisms and pathways in- Specific autoantibody-associated genes. Only two of four of volved in complex genetic diseases such as SLE will the congenic lines, B10.Yaa.Bxs1/4 and B10.Yaa.Bxs2/3, necessitate several different experimental strategies. developed high levels of autoantibodies directed at Classical genetic studies have been invaluable for single-stranded DNA (ssDNA). There were five com- defining disease-susceptibility intervals. Subsequently, monly overexpressed sequences in these strains (Table 4). congenic strains have been successfully used to delineate AU023384 is a novel transcript derived from Cd11c þ smaller genetic intervals with the aim of identifying dendritic cells, which maps to a lupus-susceptibility positional candidate genes.32–34 Typical congenic inter- locus identified as Nbwa123,27 and Asbb2,27 both asso- vals contain at least 400–600 known or predicted genes ciated with antinuclear antibody production. Another, and hence some strategy for prioritisation of candidate Ahi1, is the mouse orthologue of the human Abelson genes is essential. The approach that we present here, helper proviral integration site, and a third is Ece2, with expression analysis of overlapping congenic strains, methyl transferase activity, which mapped close to the provides an excellent method for narrowing the focus anti-dsDNA locus, Lprm5.28 The remaining two probes even further and identifying differentially expressed represent ESTs. One of the two underexpressed se- positional candidate genes. However, this approach is quences mapped to a known transcript for lipocalin- not without its limitations. Any expression analysis is interacting membrane receptor implicated in sterol and only as comprehensive as the array on which it is hormone transport, while the other is Prkg1 protein performed and there is a possibility that a disease gene is kinase, the orthologue of which is known to play a role in not included on the array. Furthermore, expression platelet activation in humans. Interestingly, Prkg1 maps profiling does not readily identify functional changes in to the anti-ssDNA specific locus, Nwa2.6 proteins. However, differential expression undoubtedly Our second method of analysis based on RMA underlies some susceptibility loci14,15,35 and the determi- identified a set of eight significantly underexpressed nation of differentially regulated pathways may aid the sequences (data not shown), of which the only candidate identification of any primary functional variant. gene of any interest was Btla, a B- and T-lymphocyte- The primary aim of this study was the identification of associated sequence represented by probe 133663_at and chromosome 1 SLE disease genes. We have successfully mapping to 44.1 Mb on chromosome 16. This gene is identified a number of novel genes as candidates for the known to play a role in the regulation of B-cell BXSB disease-susceptibility loci on chromosome 1, using proliferation, and will thus be further studied as a two very different analysis techniques. A single Bxs4 candidate gene. candidate was identified, Xkr4. The sequence of Xkr4 is Antibodies specific for dsDNA were only detected in highly evolutionarily conserved, and it is homologous to B10.Yaa.Bxs1/4, B10.Yaa.Bxs1/2 and B10.Yaa.Bxs2/3.Two Ced8 from Caenorhabditis elegans. It may function to commonly overexpressed candidate genes were identi- degrade DNA during apoptosis, in a pathway parallel to fied along with seven underexpressed sequences (Ta- the Ced9 pathway.36 Defects in apoptosis have been ble 4). Pgrmc2, a progesterone membrane-binding demonstrated to affect the availability of autoantigens protein, was overexpressed, but the other upregulated which drive autoantibody production,37 and genes in the sequence (AI839720) did not map to any known or apoptotic pathway, such as Fas, have been implicated in predicted gene, although it showed homology to the Bcl- the development of lupus in the MRL model.38 The Bxs3 11 2 interacting protein, Nip2. Four underexpressed se- locus is homologous to Nba2 in (NZB Â NZW)F1 and quences mapped to chromosome 9, although they were Sle1 in NZM2410, two models based on the New Zealand not positionally clustered on the chromosome. These mouse.39 The Sle1 interval has subsequently been were RNA helicase, Dhx30, myotubularin-related subdivided into Sle1a, b, c, d and e,40 indicating the protein 1 (Mtm1), bone morphogenic protein (Bmp5) presence of several susceptibility loci within this interval. and interleukin enhancer binding factor 3, Ilf3, a nuclear Ifi202 was identified by microarray analysis as a major factor of activated T cells. We have previously identified candidate susceptibility gene in this interval in the New a chromosome 9 lupus-susceptibility interval, Gp1, Zealand model.15 The replication of Ifi202 overexpression associated with the production of anti-gp70 immune in a second lupus-prone strain, BXSB, indicates that this complexes and anti-dsDNA antibodies centred around is indeed a strong candidate for one of the susceptibility 104.7 Mb.29 Dhx30 (110.68 Mb) is a candidate for Gp1. The loci in this interval, and lends support to the validity of remaining underexpressed genes commonly found in both data sets. It has been suggested that increased strains with anti-dsDNA antibodies included transcrip- expression of Ifi202 transcription factor may contribute to tion factor, Hes5, actin-binding protein, Fascin and a G lupus through inhibition of lymphocyte apoptosis.41 A protein receptor, Gprc2a-rs2, which mapped to the Lbw5/ further novel Bxs3 candidate, Gas5, was also identified. Sle3 lupus-susceptibility region30,31 implicated in mortal- While the function of Gas5 remains unclear, it has been ity and anti-dsDNA antibody production. There was no suggested that it plays a role in the cellular decision to

Genes and Immunity Microarray analysis of BXSB congenic mice MEK Haywood et al 260 enter the apoptotic pathway.19 The potential loss of an pathways. Three chromosome 1 genes, which have been Sp1 binding site from the promoter of the BXSB allele of previously implicated in the development of SLE, Cr2,45 Gas5 may be the cause of the down regulation. Further Il1046 and CRP,47 showed no change in their expression work is underway to characterise which splenocytic cell pattern and we can make no comment on their role in population Gas5 is normally expressed in and the disease. Conversely, differential expression of a gene in functional relevance of the promoter polymorphisms. all the congenic strains implies that such a gene plays a We have also identified three potential candidates for key role in the disease process per se. We would argue Bxs2. One of these candidates, Marco, was not validated that such a gene is not a primary cause of disease, but by the strict criteria that we set for expression because it reflects an unbalanced physiological pathway. This is showed differential expression, albeit limited, in other exemplified by overexpression of the immunoglobulin strains not carrying the BXSB allele. However, a role for genes in all strains, overproduction of antibodies being a Marco is attractive because of its function in clearance of characteristic of the disease in all the congenic strains apoptotic cells. We are looking in greater detail at the including B10.Yaa. On chromosome 1, this effect was also tissue specificity of Marco expression, which may illustrated by Pdcd1, another gene recently implicated in explain these results. We did not identify any candidates the development of SLE.48 Our data therefore support an for Bxs1, and it may be that a functional polymorphism important role for Pdcd1 in the development of SLE in underlies this disease-susceptibility locus. our strains, but it is not an initial causative factor in The use of multiple congenic strains has also allowed BXSB. us to identify genes that may play a role in specific This study has sought to extend the use of microarray disease pathways. Thus, 29 genes were identified that expression data by placing it into the wider context of were consistently differentially expressed in strains with murine lupus-susceptibility studies. The overlap of common phenotypic characteristics by fold change several analytical approaches, classical genetics, pheno- analysis and 12 by SAM analysis. There is limited typic studies and gene expression analysis, has enabled overlap between these two methods; however, genes us to prioritise candidate genes that merit further study. such as Ifi202 have been directly confirmed in other strains.15 Consequently, these methods cannot be looked on as mutually exclusive, but rather complimentary to other methods of analysis. The use of two methods of Materials and methods analysis has allowed us to extract the maximum information from our data set. Mice Some of these phenotype-related expression differ- Mice were bred and maintained under identical condi- ences were found for genes that map to known SLE tions at Imperial College from original stocks: BXSB from susceptibility intervals in BXSB, such as Dhx30 in the Gp1 The Jackson Laboratory (Bar Harbor, ME) and B10 from 7 interval. Other genes were identified that map to Harlan Olac Ltd (Bicester, UK) as described previously. susceptibility intervals identified in other lupus-prone Microsatellite marker-directed breeding was used strains, for example, Rps6ka1, which lies in the NZB/ to generate four overlapping congenic strains, BXSB BXSB W-derived Nba1 disease-susceptibility interval,20 Mcp6 B10.Y BXSB-Bxs1/4 (B10.Yaa.Bxs1/4), B10.Y BXSB- BXSB mapping to the Asbb3 locus23 and Prkg1 in the Nwa2 Bxs1 (B10.Yaa.Bxs1), B10.Y BXSB-Bxs1/2 (B10.Yaa.Bxs1/ BXSB 7–9 interval.6 Thus, we have used data generated in one 2) and B10.Y BXSB-Bxs2/3 (B10.Yaa.Bxs2/3). The lupus-prone strain, BXSB, to identify potential candidate nature of this approach relies upon the use of micro- genes that may represent strong candidates for aetiolo- satellite markers to position recombination events. gical lupus-susceptibility alleles in other lupus models. However, it was not possible to position the recombina- In addition, we have found a number of novel genes that tion between the selection markers for each congenic do not localise with known lupus loci, but are indicative line. Hence, a region of indeterminate origin was created of important physiological pathways. at the end of all congenic intervals, and putative There is now increasing interest in interferon-regu- candidate genes in such intervals were also considered. lated target genes that may be involved in the SLE All congenic mice analysed were male, consomic for the þ disease process42 due to the identification of an ‘inter- BXSB Y-chromosome (Yaa ) and free from disease BXSB feron signature’ in SLE patients. Several of the candidate pathogens. Age-matched B10, BXSB and B10-Y genes that we have identified in this study may be (B10.Yaa) mice were included as controls. implicated in this pathway, and it is likely that general deregulation of the interferon response genes may be as important in murine models of lupus as has been Flow-cytometric analysis suggested for human lupus.43 Hence, Ifi202 has been Flow cytometry was performed using two-colour stain- shown to play a negative regulatory role in cell growth ing of splenocytes, and analysed with a FACSCalibur and can be induced by interferon. The human orthologue (BD Biosciences, San Jose, CA, USA). The following of Rps6ka1 has been shown to be induced by interferon g antibodies were used: anti-cluster of differentiation in peripheral blood mononuclear cells.42 Both mouse and (CD)4-fluorescein isothiocyanate (FITC), anti-CD8-phycoer- human mutations in Lyst are associated with reduced NK ythrin (PE), anti-B220-FITC (RA3-6B2) and anti-CD11b-PE cell activity and a reduced response to interferon.26,44 (BD Biosciences Pharmingen, San Diego, CA, USA), The failure to detect differential expression of a gene along with isotype-matched controls, rat IgG2a-FITC and does not preclude it from playing an aetiological role in rat IgG2a-PE (Caltag, Burlingame, CA, USA). Staining disease. Functional mutations cannot be detected directly was performed in the presence of a saturating concen- by expression profiling, but may be inferred from tration of 2.4G2 anti-FcgRII/III mAb. Data were analysed downstream perturbations in the expression of specific using WinMDI (http://facs.scripps.edu/software.html).

Genes and Immunity Microarray analysis of BXSB congenic mice MEK Haywood et al 261 Microarray analysis using only genes flagged present or marginal). All At least nine mice for each of the congenic lines and congenic strains were normalised to the median of the control strains were killed at 2 months of age, before the B10 control samples (calculated using only those onset of disease.8 The spleen was removed, divided and measurements that had at least a normalised value of half was snap frozen in liquid nitrogen prior to RNA 0.01). Values below 0 were set to 0. extraction using Trizol (Invitrogen, Paisley, UK) as per Fold change analysis was used to identify probe the manufacturer’s recommendations. The remaining signals which were more than twofold over- or under- half was analysed by FACS to determine the CD4/CD8 expressed relative to B10 control signals. Absent, margin- ratio and macrophage count. Mice that showed any al or present Affymetrix flag calls were not used in the abnormality in either respect were excluded from further selection of candidate genes. The likelihood of any gene’s analysis. At the age of 2 months, none of the congenic expression value being different from B10 was evaluated strains show measurable signs of disease; however, any using the t-test P-value, which gives an indication of the imbalance in the cell populations used for the prepara- variation between the control data and the test data for tion of RNA for microarray analysis may have given all three replicates for each congenic strain. Genes were false results. We prepared the individual RNA from nine excluded if they also showed differential expression in screened mice for each strain and then randomly B10.Yaa mice to ensure that only congenic effects were assigned three of the RNAs to each of three pools to measured. This measure was used to exclude potential generate the biological replicates which were applied to candidate genes when it indicated that the data were three arrays. highly variable between the replicate samples (P40.1).49 All experiments were performed using Affymetrix This level of cutoff was chosen as a compromise between MG-U74v2 oligonucleotide arrays, A, B and C chips the false-positive and false-negative rates; however, we (totalling 36767 probe sets) (http://www.affymetrix. felt that, given the importance of positional information com/products/arrays/specific/mgu74.affx). Total RNA with respect to the interpretation of our data, we could from each sample was used to prepare biotinylated deal with a predicted 12% drop in our true positive rate49 target RNA, with minor modifications from the manu- by subsequent analysis. facturer’s recommendations. Briefly, 10 mg of RNA was We further analysed our data using the significance used to generate first-strand cDNA using a T7-linked analysis method.50 Microarray data were extracted from oligo(dT) primer. After second-strand synthesis, in vitro .cel files and normalised using the expresso function of transcription was performed with biotinylated UTP and the Affy package51 in R;52 in particular, pm/mm CTP (Enzo Diagnostics, Farmingdale, NY, USA), result- correction was performed using the rma option and ing in approximately 100-fold amplification of RNA. A quantile normalisation was used. After normalisation, complete description of procedures is available on CSC significant gene expression between groups was assessed MRC Web site (http://microarray.csc.mrc.ac.uk/proto using SAM,50 a technique which takes into account the cols/Other%20GeneChip%20Protocols/front_page_- variance and absolute magnitude of a gene’s expression whole.htm). The target cDNA generated from each pattern as opposed to its relative fold change. Further- sample was processed as per the manufacturer’s more, rather than a (multiple testing corrected) P-value, recommendations using an Affymetrix GeneChip Instru- SAM produces a false detection rate (FDR) value, the ment System. Spike controls (detailed below) were crucial difference being that while with a P-value of 0.05 added to 10 mg fragmented cDNA before overnight one would expect five false positives for every 100 tests hybridisation. Arrays were then washed and stained (regardless of the outcome of the tests), an FDR of 0.05 with streptavidin-PE, before being scanned on an requires 100 positive tests before one expects five of these Affymetrix GeneChip scanner (http://www.affymetrix.- to be falsely positive and thus avoids the multiple testing com/support/technical/manual/expression_manua- correction issue. Significantly differently expressed genes l.affx). Additionally, the quality and amount of starting were identified at the minimal FDR at which they were RNA was confirmed using agarose gel electrophoresis. deemed significant. After scanning, array images were assessed by eye to confirm scanner alignment and the absence of significant Real-time quantitative polymerase chain reaction bubbles or scratches on the chip surface. 30/50 ratios for Strain-specific splenocyte cDNA was synthesised from glyceraldehyde-3-phosphate dehydrogenase (Gapdh) and 1 mg RNA using 2 ml oligo dT12–18 primer (Invitrogen) and b-actin were confirmed to be within acceptable limits 400 U Superscript II reverse transcriptase (Invitrogen) and Escherichia coli biotin synthetase gene B (bioB) spike according to the manufacturer’s recommendations and controls were found to be present on all chips, with E. coli in the presence of 52.8 U RNAGuard (Amersham biotin synthetase gene C (bioC), E. coli biotin synthetase Biosciences, Amersham, UK). Quantitative real-time gene D (bioD) and Bacteriophage P1 gene for recombi- PCR was performed using 12.5 ml SYBR Green master nase protein (cre) also present in increasing intensity. For mix (Applied Biosystems, Warrington, UK), 2.6 pmol of quality control purposes, data were scaled to a target each primer and 5 ml of cDNA in a total reaction volume intensity of 100 using Affymetrix MAS 5.0 array analysis of 20 ml. Samples were tested in triplicate using a 96-well software; scaling factors for all arrays were within reaction optical plate (Applied Biosystems) and run on acceptable limits, as were background, Q-values and an Applied Biosystems 7700 Sequence Detection System. mean intensities. Data were analysed using the manufacturer’s Sequence All subsequent analysis, including normalisation of Detection software (v1.7a) and Microsoft Excel (Micro- the data, was carried out using the global scaling method soft Corporation, CA, USA). Primers for control genes within Genespring (Agilent Technologies UK Limited, were: 18SRNA: 843F, CCTGGATACCGCAGCTAGGAA Cheshire, UK). Each measurement was divided by the TAA; 936R, CCCCGGCCGTCCCTCTTAATCAT; b-actin: median of all measurements in that sample (calculated 915F, TGTGACGTTGACATCCGTAAA; 1143R, CCTGCT

Genes and Immunity Microarray analysis of BXSB congenic mice MEK Haywood et al 262 TGCTGATCCACATCTGC; Ubiquitin c-terminal hydroxy- References lase (Ubc): F, GTTACCACCAAGAAGGTCAAACAG; R, ATCACACCCAAGAACAAGCACAAG; glyceraldehyde- 1 Murphy ED, Roths JB. New inbred strains. Mouse News Lett 3-phosphate dehydrogenase (Gapdh): 879F, AGGTTGTCT 1978; 58: 51–52. CCTGCGACTTCAACA; 956R, GAAATGAGCTTGACA 2 Andrews BS, Eisenberg RA, Theofilopoulos AN, Izui S, Wilson AAGTT. Primers for the candidate genes Gas5 and Marco CB, McConahey PJ et al. Spontaneous murine lupus-like syndromes. Clinical and immunopathological manifestations Gas5 were as follows: : 988F, GTATTAATGGGTCACCT in several strains. J Exp Med 1978; 148: 1198–1215. CAAGTGAA; 1115R, CAGTCACTGCATGTCCACTTGT 3 Vyse TJ, Kotzin BL. Genetic basis of systemic lupus erythe- CA; Marco: 1561F, ATCTTTCCAACGCGTCC GGATCAT; matosus. Curr Opin Immunol 1996; 8: 843–851. 1645R, CAAATTGTCCCCCACTCATTGTTA. A standard 4 Kono DH, Theofilopoulos AN. Genetics of systemic auto- curve using 10-fold serial dilutions was included for the immunity in mouse models of lupus. Int Rev Immunol 2000; 19: control and candidate gene on each plate. Candidate 367–387. gene results were normalised to the control gene result 5 Mohan C, Alas E, Morel L, Yang P, Wakeland EK. Genetic for each strain, and then normalised to a B10 value of 1. dissection of SLE pathogenesis. Sle1 on murine chromosome 1 The data shown represent three separate measurements leads to a selective loss of tolerance to H2A/H2B/DNA of a pool of three individuals for each strain. These subnucleosomes. J Clin Invest 1998; 101: 1362–1372. individuals were not the same as used for the microarray 6 Vyse TJ, Morel L, Tanner FJ, Wakeland EK, Kotzin BL. Backcross analysis of genes linked to autoantibody production analysis, but were age-matched to those samples and in New Zealand White mice. J Immunol 1996; 157: 2719–2727. screened in a similar way. Multiple control genes were 7 Hogarth MB, Slingsby JH, Allen PJ, Thompson EM, Chandler used to overcome any variation in control genes which P, Davies KA et al. Multiple lupus susceptibility loci map to may confound the result.53 chromosome 1 in BXSB mice. J Immunol 1998; 161: 2753–2761. 8 Haywood ME, Rogers NJ, Rose SJ, Boyle J, McDermott A, Rankin JM et al. Dissection of BXSB lupus phenotype using mice congenic for chromosome 1 demonstrates that separate Candidate gene promoter sequencing intervals direct different aspects of disease. J Immunol 2004; Promoter regions of candidate genes were sequenced 173: 4277–4285. from male spleen genomic DNA isolated as described 9 Haywood ME, Hogarth MB, Slingsby JH, Rose SJ, Allen PJ, previously.35 Polymerase chain reaction amplification Thompson EM et al. Identification of intervals on chromosomes 1, 3, and 13 linked to the development of lupus in BXSB was performed using NH4 reaction buffer (Bioline, mice. Arthritis Rheum 2000; 43: 349–355. London, UK), 30 mM MgCl2 (Bioline, London, UK), 4 mM deoxynucleoside triphosphates (Bioline), 0.2 U Biotaq 10 Vidal S, Kono DH, Theofilopoulos AN. Loci predisposing to autoimmunity in MRL-Fas lpr and C57BL/6-Faslpr mice. polymerase (Bioline), 6.6 nM each primer and 80 ng J Clin Invest 1998; 101: 696–702. DNA. Primers for Gas5 were as follows: F164600667, 11 Vyse TJ, Rozzo SJ, Drake CG, Izui S, Kotzin BL. Control of GAATGCTCGTGTTAGAGGTG; F164601872, AACGA multiple autoantibodies linked with a lupus nephritis suscept- GCCCAGTCAAGGAG; F164602174, GCTTGAGGAGGA ibility locus in New Zealand black mice. J Immunol 1997; 158: GTCTGA; F164602592, AGAGCATGCTGAGTCACCT; 5566–5574. R164601393, GTGCCGACGCTAGAGAAAG; R164602380, 12 Murphy ED, Roths JB. A Y chromosome associated factor in ACCAGCTTCCAGTGCTTCAG; R164602790, GAGTTG strain BXSB producing accelerated autoimmunity and lym- CACATATCCAGCAG; R164603374, TGCTTACATTGCG phoproliferation. Arthritis Rheum 1979; 22: 1188–1194. CTCGCT. All reactions were performed in polypropy- 13 Hudgins CC, Steinberg RT, Klinman DM, Reeves MJ, Stein- lene tubes using an MJ Research PTC225 thermocycler berg AD. Studies of consomic mice bearing the Y chromosome of the BXSB mouse. J Immunol 1985; 134: 3849–3854. (Reno, NV, USA). Samples were denatured at 941C for 1 1 14 Eaves IA, Wicker LS, Ghandour G, Lyons PA, Peterson LB, 3 min and then 30–35 cycles of 94 C for 1 min, 55–65 C Todd JA et al. Combining mouse congenic strains and for 1 min and 721C for 1 min. In all, 30–90 ng of PCR microarray gene expression analyses to study a complex amplicon was purified using a PEG precipitation proto- trait: the NOD model of type 1 diabetes. Genome Res 2002; 12: col,54 prior to sequencing using 3.2 pM primer, and 232–243. BigDye V2.0 terminator mix (Applied Biosystems) for 25 15 Rozzo SJ, Allard JD, Choubey D, Vyse TJ, Izui S, Peltz G et al. cycles of 951C for 10 s, 501C for 5 s and 601C for 4 min. Evidence for an interferon-inducible gene, Ifi202, in the susceptibility to systemic lupus. Immunity 2001; 15: 435–443. 16 Tang Z, McGowan BS, Huber SA, McTiernan CF, Addya S, Surrey S et al. Gene expression profiling during the transition Transcription factor binding site prediction to failure in TNF-alpha over-expressing mice demonstrates Transcription factor binding sites were predicted using the development of autoimmune myocarditis. J Mol Cell TESS (http://www.cbil.upenn.edu/tess) and Matinspec- Cardiol 2004; 36: 515–530. tor55 with default parameters. 17 Kagami Y, Furuichi T. Investigation of differentially expressed genes during the development of mouse cerebellum. Gene Expr Patterns 2001; 1: 39–59. 18 Kangas M, Brannstrom A, Elomaa O, Matsuda Y, Eddy R, Shows TB et al. Structure and chromosomal localization of the Acknowledgements human and murine genes for the macrophage MARCO receptor. Genomics 1999; 58: 82–89. This work was supported by an Arthritis Research 19 Coccia EM, Cicala C, Charlesworth A, Ciccarelli C, Rossi GB, Campaign Program Grant. We thank all the staff at the Philipson L et al. Regulation and expression of a growth arrest- CSC/IC Microarray Centre, Hammersmith Campus, specific gene (gas5) during growth, differentiation, and Imperial College, for their technical help with the development. Mol Cell Biol 1992; 12: 3514–3521. microarray experiments. We also thank Drs Tim Vyse 20 Vyse TJ, Drake CG, Rozzo SJ, Roper E, Izui S, Kotzin BL. and Nicola Rogers for critical reading of the manuscript. Genetic linkage of IgG autoantibody production in relation to

Genes and Immunity Microarray analysis of BXSB congenic mice MEK Haywood et al 263 lupus nephritis in New Zealand hybrid mice. J Clin Invest explained by defects in Fas antigen that mediates apoptosis. 1996; 98: 1762–1772. Nature 1992; 356: 314–317. 21 Gebauer M, Zeiner M, Gehring U. Proteins interacting with 39 Mohan C, Morel L, Yang P, Wakeland EK. Genetic dissection the molecular chaperone hsp70/hsc70: physical associations of systemic lupus erythematosus pathogenesis: Sle2 on and effects on refolding activity. FEBS Lett 1997; 417: 109–113. murine chromosome 4 leads to B cell hyperactivity. J Immunol 22 Hallgren J, Spillmann D, Pejler G. Structural requirements and 1997; 159: 454–465. mechanism for heparin-induced activation of a recombinant 40 Morel L, Blenman KR, Croker BP, Wakeland EK. The major mouse mast cell tryptase, mouse mast cell protease-6: formation murine systemic lupus erythematosus susceptibility locus, of active tryptase monomers in the presence of low molecular Sle1, is a cluster of functionally related genes. Proc Natl Acad weight heparin. J Biol Chem 2001; 276: 42774–42781. Sci USA 2001; 98: 1787–1792. 23 Bolland S, Yim YS, Tus K, Wakeland EK, Ravetch JV. Genetic 41 Xin H, Geng Y, Pramanik R, Choubey D. Induction of p202, a modifiers of systemic lupus erythematosus in modulator of apoptosis, during oncogenic transformation of FcgammaRIIB(À/À) mice. J Exp Med 2002; 195: 1167–1174. NIH 3T3 cells by activated H-Ras (Q61L) contributes to cell 24 Xie S, Chang SH, Sedrak P, Kaliyaperumal A, Datta SK, survival. J Cell Biochem 2003; 88: 191–204. Mohan C. Dominant NZB contributions to lupus in the 42 Baechler EC, Batliwalla FM, Karypis G, Gaffney PM, Ortmann (SWR Â NZB)F1 model. Genes Immun 2002; 3 (Suppl 1): WA, Espe KJ et al. Interferon-inducible gene expression S13–S20. signature in peripheral blood cells of patients with severe 25 Barbosa MD, Nguyen QA, Tchernev VT, Ashley JA, Detter JC, lupus. Proc Natl Acad Sci USA 2003; 100: 2610–2615. Blaydes SM et al. Identification of the homologous beige and 43 Baechler EC, Gregersen PK, Behrens TW. The emerging role of Chediak–Higashi syndrome genes. Nature 1996; 382: 262–265. interferon in human systemic lupus erythematosus. Curr Opin 26 Jackson IJ. Homologous pigmentation mutations in human, Immunol 2004; 16: 801–807. mouse and other model organisms. Hum Mol Genet 1997; 6: 44 Virelizier JL, Lipinski M, Tursz T, Griscelli C. Defects of 1613–1624. immune interferon secretion and natural killer activity in 27 Rigby RJ, Rozzo SJ, Boyle JJ, Lewis M, Kotzin BL, Vyse TJ. patients with immunological disorders. Lancet 1979; 2: New loci from New Zealand Black and New Zealand White 696–697. mice on chromosomes 4 and 12 contribute to lupus-like 45 Boackle SA, Holers VM, Chen X, Szakonyi G, Karp DR, disease in the context of BALB/c. J Immunol 2004; 172: 4609– Wakeland EK et al. Cr2, a candidate gene in the murine Sle1c 4617. lupus susceptibility locus, encodes a dysfunctional protein. 28 Wang Y, Nose M, Kamoto T, Nishimura M, Hiai H. Host Immunity 2001; 15: 775–785. modifier genes affect mouse autoimmunity induced by the lpr 46 Chong WP, Ip WK, Wong WH, Lau CS, Chan TM, Lau YL. gene. Am J Pathol 1997; 151: 1791–1798. Association of interleukin-10 promoter polymorphisms with 29 Haywood ME, Vyse TJ, McDermott A, Thompson EM, Ida A, systemic lupus erythematosus. Genes Immun 2004; 5: 484–492. Walport MJ et al. Autoantigen glycoprotein 70 expression is 47 Russell AI, Cunninghame Graham DS, Shepherd C, Roberton regulated by a single locus, which acts as a checkpoint for CA, Whittaker J, Meeks J et al. Polymorphism at the C-reactive pathogenic anti-glycoprotein 70 autoantibody production and protein locus influences gene expression and predisposes to hence for the corresponding development of severe nephritis, systemic lupus erythematosus. Hum Mol Genet 2004; 13: 137– in lupus-prone BXSB mice. J Immunol 2001; 167: 1728–1733. 147. 30 Morel L, Rudofsky UH, Longmate JA, Schiffenbauer J, 48 Prokunina L, Castillejo-Lopez C, Oberg F, Gunnarsson I, Berg Wakeland EK. Polygenic control of susceptibility to murine L, Magnusson V et al. A regulatory polymorphism in PDCD1 systemic lupus erythematosus. Immunity 1994; 1: 219–229. is associated with susceptibility to systemic lupus erythema- 31 Kono DH, Burlingame RW, Owens DG, Kuramochi A, tosus in humans. Nat Genet 2002; 32: 666–669. Balderas RS, Balomenos D et al. Lupus susceptibility loci in 49 Troyanskaya OG, Garber ME, Brown PO, Botstein D, Altman New Zealand mice. Proc Natl Acad Sci USA 1994; 91: 10168– RB. Nonparametric methods for identifying differentially 10172. expressed genes in microarray data. Bioinformatics 2002; 18: 32 Wakeland EK, Wandstrat AE, Liu K, Morel L. Genetic 1454–1461. dissection of systemic lupus erythematosus. Curr Opin 50 Tusher VG, Tibshirani R, Chu G. Significance analysis of Immunol 11 1999; : 701–707. microarrays applied to the ionizing radiation response. Proc 33 Podolin PL, Denny P, Armitage N, Lord CJ, Hill NJ, Levy ER Natl Acad Sci USA 2001; 98: 5116–5121. et al. Localization of two insulin-dependent diabetes (Idd) 51 Irizarry RA, Gautier L, Bolstad BM, Miller C with contribu- genes to the Idd10 region on mouse chromosome 3. Mamm tions from Astr M, Cope LM, Gentleman R, Workman C, Genome 1998; 9: 283–286. Zhang J. Affy:Methods for Affymetrix Oligonucleotide Arrays, 34 Denny P, Lord CJ, Hill NJ, Goy JV, Levy ER, Podolin PL et al. 2004 R package version 1.5.8. Mapping of the IDDM locus Idd3 to a 0.35-cM interval 52 R Development Core Team. R: A Language and Environment for containing the interleukin-2 gene. Diabetes 1997; 46: 695–700. Statistical Computing. R Foundation for Statistical Computing: 35 Pritchard NR, Cutler AJ, Uribe S, Chadban SJ, Morley BJ, Vienna, Austria, 2005. http://www.R-project.org. Smith KG. Autoimmune-prone mice share a promoter 53 Suzuki T, Higgins PJ, Crawford DR. Control selection for RNA haplotype associated with reduced expression and function quantitation. Biotechniques 2000; 29: 332–337. of the Fc receptor FcgammaRII. Curr Biol 2000; 10: 227–230. 54 Du Z, Hood L, Wilson R. Automated fluorescent DNA 36 Stanfield GM, Horvitz HR. The ced-8 gene controls the timing sequencing of polymerase chain reaction products. Methods of programmed cell deaths in C. elegans. Mol Cell 2000; 5: Enzymol 1993; 218: 104–121. 423–433. 55 Quandt K, Frech K, Karas H, Wingender E, Werner T. MatInd 37 Botto M, Walport MJ. C1q, autoimmunity and apoptosis. and MatInspector: new fast and versatile tools for detection of Immunobiology 2002; 205: 395–406. Nucleic Acids 38 Watanabe-Fukunaga R, Brannan CI, Copeland NG, Jenkins consensus matches in nucleotide sequence data. 23 NA, Nagata S. Lymphoproliferation disorder in mice Res 1995; : 4878–4884.

Genes and Immunity