'genetic background' of inbred mouse strains - Nature

© 2007 Nature Publishing Group http://www.nature.com/natureimmunology

C O M M E N TA R Y

New tools for defining the ‘genetic background’ of inbred mouse strains William M Ridgway, Barry Healy, Luc J Smink, Dan Rainbow & Linda S Wicker There is general appreciation that ‘genetic background effects’ can profoundly affect the immune phenotypes of congenic, transgenic and knockout mice. We suggest that attributing phenotypes to genetic background effects is outmoded and that new databases containing single-nucleotide polymorphisms obtained with a group of inbred mouse strains can be used to define the flanking DNA of nearly all mouse genes.

I

mmunologists analyze the effect of genes on immunity with two fundamental approaches. One involves studying the effect of a ‘candidate’ gene on a phenotype; for example, using transgenesis or homologous recombination (creating ‘knockout’ or ‘knockin’ mice). The other entails using a genetic mapping approach to identify small genomic regions that contain functional alleles of a ‘to-be-defined’ gene that affects the phenotype of interest (for example, generating congenic mice). It is not always appreciated, however, that these approaches share an underlying variable that is obvious in the construction of congenic strains but often not apparent during the analysis of a candidate gene: the influence of potential polymorphisms in genes flanking the gene of interest. Effect of genetic variation The application of PCR to the detection of strain-specific differences in microsatellite polymorphisms revolutionized the development of congenic strains and allowed the development of congenic fine mapping1,2. There are now many examples of congenic mice in which William M. Ridgway is in the Division of Rheumatology and Immunology, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania 15218, USA. Barry Healy, Luc J. Smink, Dan Rainbow and Linda S. Wicker are in the Juvenile Diabetes Research Foundation/ Wellcome Trust, Department of Medical Genetics, Cambridge Institute for Medical Research, University of Cambridge, Cambridge CB2 0XY, UK. e-mail: [email protected] or [email protected]

the introduction or introgression of 1–5 megabases (Mb) of genetic material from a healthy strain of mouse into the genome of a mouse strain susceptible to autoimmune disease prevents autoimmunity. Nonetheless, detecting to within an interval of 1–5 Mb the specific gene that results in decreased disease penetrance can be an arduous task. As the development of smaller congenic intervals relies on the occurrence of recombination events near the gene of interest in ever smaller genomic regions, large colonies of mice are required, which places a practical limit on how small an interval can be made in the congenic mouse (Fig. 1a). The reduction in the size of the interval is also influenced by the fact that recombination is not random. ‘Hot spots’ of recombination, which often cause nearby ‘cold spots’, can dictate the limit of the congenic interval reduction3. At first glance it seems that the use of transgenic or knockout mice (‘knockout’ is used here to indicate one of many genetic modifications in an embryonic stem cell) removes the limitations of recombination-dependent congenic analysis and produces a mouse with a specific and defined genetic alteration. Such a conclusion, however, is misguided. Although the usual specificity control for a transgene-mediated phenotype is the production of at least two independent founder transgenic lines (with the same transgene but different flanking genes) that have the same phenotype, this control can be irrelevant when the transgene is backcrossed onto another inbred strain. In the new experiment, the effect of the transgene (usually from a single founder) is tested by means of backcrossing with an inbred strain with a complex

NATURE IMMUNOLOGY VOLUME 8 NUMBER 7 JULY 2007

disease phenotype caused by multiple genes. Through the use of backcrossing to a different inbred strain, the experiment tests not only the transgene but also the donor strain alleles at genes flanking the transgene. In the development of many transgenic mice, transgenes are injected into (C57BL/6 (B6) × BALB/c)F2 eggs and the genomic source (B6 or BALB/c) of the integration site is not defined (Fig. 1b). With increasing frequency, investigators avoid such ambiguities by injecting transgenes directly into eggs obtained from the inbred strain to be studied, even though mating efficiency as well as egg yields and quality are much lower when the most common inbred strains are used. The same problem arises during the development of knockout mice. Because of the limited number of genetic backgrounds permissive for viable embryonic stem cells, many knockout mice are constructed with embryonic stem cells from one of the many substrains of 129 mice4. Chimeras are produced by the injection of embryonic stem cells into blastocyts, often of the B6 strain, which are then implanted into the uterus of a foster mother. The resulting mouse pups have some tissues derived from the embryonic stem cells and some tissues derived from the blastocyst. Chimeric mice in which germ cells are derived from the embryonic stem cells are able to transmit the targeted allele. If the chimeras are crossed to the 129 strain from which the embryonic stem cells were generated, a heterozygous knockout mouse on a homogeneous background is created and homozygous knockout mice can be developed to evaluate the effect of the targeted gene (one or two doses) on the 129 strain. However, most investigators

669

C O M M E N TA R Y


a

b

c

NOD

NOD

NOD

B6 or BALB/c B6

Transgene

B6 129 Knockout gene

want to test the effect of their targeted mutation on established immune processes defined in other inbred strains of mice, such as BALB/c or B6. In this case, heterozygous knockout mice are repeatedly backcrossed to the inbred strain with the desired phenotype on which the effect of the targeted gene will be tested. However, in fact a ‘B6.129’ congenic strain is developed with the disrupted 129-derived targeted allele and an unpredictable amount of flanking 129derived DNA. Often after the targeted allele is moved to one background strain, the investigator then begins another backcross to a third inbred strain, such as the nonobese diabetic (NOD) strain. Such a strain would have unpredictable amounts of both 129- and B6-derived alleles flanking the targeted gene in the newly developed incipient ‘NOD.129’ congenic strain (Fig. 1c). Therefore, the process of backcrossing a genetically modified genomic region from knockout or transgenic strain onto a different inbred strain faces exactly the same problem that attempting to decrease a congenic interval in the process of fine-mapping faces: linked genetic material from the original strain accompanies the manipulated genomic region. The amount of this genetic material can be dozens of megabases and can be minimized only by monitoring and selection of the random meiotic recombination events that occur in the DNA flanking the targeted gene. Even eight to ten generations of backcrossing is no guarantee that the flanking region will be substantially reduced. For example, in one study Cd38 in 129 mice was ‘knocked out’ and then the knockout mice were backcrossed ten generations onto the B6 strain5. Even after ten generations of backcrossing, at least 20 centimorgans (cM) of linked 129-derived DNA remained associated with the disrupted Cd38 allele. On average, 1 cM corresponds to approximately 2 Mb of DNA, and the average gene density in mice is ten genes per megabase (ref. 6); therefore, an additional 400

670

Figure 1 The flanking-gene problem, demonstrated by hypothetical congenic, transgenic and knockout intervals crossed onto the NOD background. (a) Introgression of B6 DNA containing a gene or genes of interest onto the NOD background produces a congenic mouse with a B6 region defined by genetic markers. An additional example online (http://t1dbase. org/cgi-bin/dispatcher.cgi/DrawStrains/display?taconic_line = 1106) shows a congenic strain used to fine-map the Idd9.3 region12. (b) Transfer of a transgene (red line) originally injected into (B6 × BALB/c)F2 eggs to the NOD background generates an interval of arbitrary size containing the transgene plus either B6 or BALB/c flanking genes (depending on whether the integration site is in B6-derived or BALB/c-derived DNA). The distal or proximal flanking regions can be long or short if left to chance during the backcrossing period. (c) A knockout generated in 129 embryonic stem cells, introduced into a B6 blastocyst, backcrossed to the B6 strain and then transferred to the NOD background results in random lengths of 129 and B6 genetic material extending outward from the knockout region. Once again, the contributions are completely random unless recombination events are detected and selected during the backcrossing process.

genes may have been ‘carried along’ with the mutated Cd38 allele. The ‘flanking-gene’ problem The problem of flanking genes is often ignored, with the assumption that the ‘small’ amount of genetic material linked to targeted alleles must be irrelevant. However, many examples have been published showing that linked genetic material has influenced the immunological phenotype of interest7; probably dozens more unrecognized cases of this ‘flanking-gene’ problem have occurred and may have obscured the true immune functions of targeted genes. In one well known case, Wang et al. attempted to definitively assess the function of interferon-γ (IFN-γ) γγ) in autoimmune diabetes by knocking out the gene encoding the IFN-γγ receptor (Ifngr). They generated the knockout in the 129 mouse strain and then backcrossed it to NOD mice for eight generations8. They also constructed the proper control, backcrossing the Ifngr+/+ locus from the 129 strain to NOD mice. The NOD.129 Ifngr–/– mice developed neither insulitis nor diabetes, whereas the ‘NOD.129 Ifngr+/+’ mice (NOD mice with the 129-derived Ifngr region) succumbed to insulitis and diabetes. Thus, these data seem to provide definitive evidence of a central function for IFN-γγ in the diabetogenic process. However, Kanagawa et al. later reported that the diabetes prevention in the NOD.129 Ifngr–/– mice was not due to the lack of the IFN-γ receptor but was instead due to a linked 129 gene9. Why were the original control mice not also protected from diabetes? The key point is that there is insufficient information to answer this question. Kanagawa et al. showed that a critical recombination event that occurred during backcross 13 separated the protective 129-derived allele of an unknown gene from the disrupted 129 Ifngr allele. PCR-based genotyping indicated that this recombination event, which presumably replaced a 129 ‘resistance allele’ with a NOD allele, occurred

an estimated 1 cM distal to the targeted mutation9. Therefore, after ten generations of backcrossing, the Ifngr–/– mice generated by Kanagawa et al.9 were protected from diabetes just as the mice were generated by Wang et al. were8, but after 13 generations of backcrossing, the protection was lost. It is possible that when Wang et al. made their control strain, a similar recombination event occurred in the first generation. A reasonable question is how this scenario could have been prevented. Fortunately, many more tools are now available for the analysis of genetic crosses to avoid exactly such potential pitfalls. A quick check of the Ensembl website shows that some immunologically relevant receptors are located adjacent to the Ifngr locus (http://www.ensembl.org/Mus_ musculus); genes encoding components of the interleukin 22 and interleukin 20 receptors are located within 200 kilobases (kb) of Ifngr10. Moreover, the DNA marker used by Kanagawa et al.9 to identify the recombination location was further upstream of the Ifngr locus than the 1-cM mapping distance suggested; it was actually 6 Mb upstream. Thus, although gaps remain in the sequence of the mouse genome, results of sequencing of the B6 strain based on bacterial artificial chromosomes now allow much more accurate determination of the genes near any particular sequence of interest. In retrospect, we conclude that during the backcrossing of the disrupted Ifngr allele, a set of informative genetic markers should have been used to select for recombination events occurring close to the gene. However, this information was less readily obtainable when that study was done than it is today. Mouse genealogy It is important for immunologists using mouse models to understand the underlying genetic architecture of mouse strains used in their research. Two types of mice are available from facilities that breed mice for research purposes:

VOLUME 8 NUMBER 7 JULY 2007 NATURE IMMUNOLOGY

C O M M E N TA R Y Chromosome 4 149,760 kb

149,770 kb

149,780 kb

149,790 kb

Ensembl transcripts Tnfrsf9 Tnfrsf9

B6 TAAC C T C C

ACG A G

G T G G CTA T CG CA TC AG

TCTG ATC CC AAT C GT TCAC

AT G T AG CTAG TTG TC TCC

CT G

G

T

T

A CTT AG TA CAG TG T T G G C

ATTG A

ACG A G




CT G

G

T

T

A CTT ANT A CAG TG T T G G C

ATTG A

ACG A G




CT G

G

T

T


ATTG A

ACG N G

G T G G CTA T CG NA TC AG



CT G

G

T

T

A CTT CG TA CAG TG T T G G C

ATTG A

ACG A G



AT G T CG CTAG TTA TC TCC

CC G

G

C

T


ATTG A

ACG A G




CC G

G

C

T


ATTG A

ACG A G




CC G

G

C

T


ATN G A

ACG A G



AT G N CG CTAG TTN TC TCC

CC G

G

C

T

A CTT ANT A CAG TG T T G G C

ATTG A

ACC G

G T G ACTA T CG CA TC AG

TCTG AAC CAAT C GT TCAC

AT G T AG TTAG CTA TC TCC

CC G

G

C

T

A CTT AG TA CAG TG T T G G A

ATTG A

ACG G

NA ACTA T T CG CA TC AG

TCTA

AAC CC AAT T

AC TCAC

AT G C AG CTAG CTA TCCT C

CC T

A

T

N

G TCT AG CN TAN C C T A N AC

TCN G

ACG G

TA ACTA T CG CA TC AG

TCTA

AAC CC AAT T

AC TCAC

AT G C AG CTAG CTA TCGT C

CC T

A

C

G

G TCCG CG TGCC C A A AC

TCAG

ACG G


TCTA

AAC CC AAT T

AC TCAC

AT G C AG CTAG CTN TCNT C

CC T

A

N

N

G TCN AAC N TNN C C N A A AC

TCN G

ACG G


TCTG AAC CC AAT T

AC TCAC

AT G C AG CTAG CTA TCGT C

CC T

N

N

N

N NNN CNN NNNN N N N NN

NCAG

G TC G A

NG G C CN C TG A CCC AA

C T CG CTC TCGT C AC TG G C

G C A C AACCG G CCA CT CNC T

CC G

N

N

N

N NNNNN NNNN N N N NN

NCAA G

ACC G A

NG G C CN C TG A TCC AA

C T CG CTC TCGT C AC TG G C

G N A C AACCG A CCA CT CNC T

TC G

N

N

N

N NNN CAN NNNN N N N NN

NCAA G

G TC G

TA G TC T CACCT G G G

C T TG ATACCGC C AC C G G T

G C G C AG CTG G CTA C TCNC T

CC G

N

N

N

N NNN CNN NNNN N N N NN

NCN G A

BALB_cByJ TAAC C T C C

KK_HlJ


TAAC C T C C

A_J TAAC C T C C

DBA_2J TAAC C T C C

FVB_NJ TAAC C T C C

BTBR_T+_tf_J TAAC C T C C

NZW_LacJ TAAC C T C N

129S1_SvImJ TAAT C A C A

C3H_HeJ A T AAC C T C C

AKR_J A T AAC C T C C

NOD_LtJ A T AAC C T C C

WSB_EiJ A T AAC C T C C

MOLF_EiJ TAG C C T C C

PWD_PhJ TAAC C A C C

CAST_EiJ T CAC T G N C C

Tnfr sf9 Mb

Mb

Mb

Mb

Mb

Mb

Mb

Mb

Mb

Mb

Figure 2 SNP maps generated from the T1Dbase website (http://dil.t1dbase.org/page/PerlegenSNPs). These maps cover 30 kb (a) and 1 Mb (b) surrounding Tnfrsf9 in 16 strains of mice. B6 is the reference strain here. Vertical blue lines indicate identity with B6 sequence at a SNP, red lines indicate that the SNP is different from B6, and yellow lines indicate no information at that SNP. Narrow horizontal lines separate different haplotypes in a; these same horizontal lines are present in b, but in several cases the haplotypes present in Tnfrsf9 shift to a different haplotoype. Horizontal red lines emphasize the haplotype shift in three examples.

inbred laboratory strains and ‘wild-derived’ inbred strains. Many reviews have detailed the origins of the inbred laboratory strains11 and thus we will not cover such information here other than to emphasize the point that inbred

laboratory strains such as B6, BALB/c, DBA, NOD and NZW are derived from essentially the same founder stock originally selected by ‘mouse fanciers’ for attributes such as coat color, ear size and behavior. This genetic ‘bot-


tleneck’ limits the diversity of DNA sequences in the inbred laboratory strains. In contrast, ‘wild-derived’ strains of mice are ‘snapshots’ of the diversity present in the main taxonomic groups of Mus musculus that gave rise to the common laboratory inbred strains. With the support of the National Institute of Environmental Health Sciences, a major initiative was completed that has supplied the research community with a vast amount of the genetic variation in the common research strains (http://mouse.perlegen.com/mouse/ index.html). Sequences of 11 inbred laboratory strains and 4 ‘wild-derived’ strains, which are estimated to include approximately 85% of the strains used for biomedical research, were compared with the previously sequenced reference strain, B6 allowing the haplotype of 16 strains to be assessed (Fig. 2). The result is 6.9 million to 7.4 million genotypes per strain and a total of 8.3 million single-nucleotide polymorphisms (SNPs). Although not all SNPs are present, because of technical and biological issues, these ‘Perlegen’ SNPs represent a tremendous leap forward from the genetic polymorphisms available before. Haplotypes Detailed knowledge of SNPs in the mouse genome enables researchers to compare the pattern of SNPs, or haplotypes, present in the 16 strains mentioned above. A browser available at the T1Dbase website (http://dil.t1dbase. org/page/PerlegenSNPs)12 can show SNPs for all of the strains simultaneously. Here we will use the Idd9.3 region (containing the gene encoding CD137), Tnfrsf9 and several other genes to demonstrate the complexity of haplotypes with reference to a genetic region studied in our labs. The Idd9.3 region from the B10 strain, when introgressed onto the NOD background, decreases diabetes incidence and increases the development of autoantibodies to the Smith antigen13–15. As with SNPs in Tnfrsf9, multiple inbred laboratory strains are commonly identical or nearly identical across a region encompassing one gene or tens of genes (Fig. 2a). In regions in which two mouse strains have identical SNP haplotypes, the strains are called ‘identical by descent’ (IBD) at this portion of the genome16,17. However, inbred strains often have multiple haplotypes across a given genetic region; additional haplotypes may be present in ‘wild-derived’ strains. This can be evidence of balancing selection, which commonly occurs in immunologically relevant genes and is thought to provide a diverse immune response in the species18. The genes encoding major histocompatibility complex class I and class II are examples of such selective pressures on the immune system.

671


C O M M E N TA R Y What is the practical implication of mouse haplotype structure for understanding genetic background effects in immunology experiments? Here we discuss two main applications. First, if a large genetic region has been identified as producing phenotypic variability between two strains (for example, if a mouse congenic for the region has a immune or disease phenotype distinct from that of the control strain), then the part of the introgressed region most likely to contain the gene causing the phenotypic difference is the region that is not IBD. Given that shared haplotypes between strains have such a low rate of genetic variability, simple probability arguments suggest that the alternative phenotypes arise from a gene in the highly variable regions of an Idd interval. Thus, if two strains are 60% IBD in a congenic region (not necessarily all in one continuous portion), the number of genes that are likely candidates for the phenotypic variation is much lower. It is important to remember that because not all strain-specific SNPs are defined, it remains a possibility that a mutation in a region that is IBD accounts for the phenotypic difference. Knowledge of shared haplotypes is also important for the construction of knockout mice. Of course, the best approach is to ‘knock out’ the gene of interest in embryonic stem cells derived from the strain to be studied. But as this option is often not available because of the difficulty in developing embryonic stem cells from most inbred strains, a viable alternative is to choose embryonic stem cells originating from a strain that is IBD with the region surrounding the targeted gene. Thus, in some cases, the application of haplotype structure to the construction of knockout mice can solve the flanking-gene problem. Yamanouchi et al. have used such an approach to demonstrate that variation in Il2 expression can account for the ‘resistance’ and ‘susceptible’ alleles at the Idd3 region in NOD mice19. Il2 was originally knocked out on the 129 background, and analysis of Idd3 region sequences from NOD and 129 mice showed the regions were IBD, with only 0.68 SNPs per 10 kb. In contrast, the B6 strain showed approximately 100 SNPs per 10 kb in the area and therefore was IBD with neither NOD nor 129. Thus, when the interleukin 2–knockout developed on the 129 genome was backcrossed to the NOD strain, in effect a ‘true’ NOD Il2 knockout was created, because NOD and 129 are IBD at the Idd3 region. There is a critical difference in the knockout breeding of Wang et al.8 and that of Yamanouchi et al.19: in the Ifngr knockout, 129 and NOD were not IBD in the region surrounding Ifngr, whereas in the Idd3 region, they were.

672

It is important to understand that relative to that of other inbred strains of mice, the haplotype of a strain is constantly changing due to many ancient recombination events (Fig. 2b). The definition and purity of haplotypes among strains is complex and could break down at the level of individual base pairs20. Reconsidering the Idd9.3 region, but taking a broader view beyond the Tnfrsf9 locus over 1 Mb, it is obvious that the haplotypes identified in the Tnfrsf9 region are often disrupted even in this small 1-Mb interval. Solutions The discussion above indicates many ways that the flanking-gene problem can affect attempts to unambiguously specify the effect of a genetic manipulation on an immunological phenotype. Here we present some practical approaches to address these issues. First, for the creation of a knockout, the mutation should be introduced directly into an embryonic stem cell derived from the strain of interest. This obvious suggestion is constrained by the inability to generate robust embryonic stem cells from most inbred strains of mice. The wide availability of embryonic stem cells derived from the 129 strain has resulted in the development of many genetically manipulated strains on the 129 background with subsequent backcrossing to B6 or to another strain of interest. If the phenotype of interest will be studied in the B6 strain, the best solution to the flanking-gene problem is to use B6 embryonic stem cells. If no embryonic stem cells are available for the strain of interest, the haplotype structure of the two strains should be compared on available databases. For example, Yamanouchi et al. wanted to introduce a mutation already existing on a 129 background onto the NOD background19. Comparison of the DNA flanking the mutated region showed that 129 and NOD were IBD in the region, thus eliminating the flanking-gene problem. If two strains are not IBD in the flanking regions, the mutation should be introduced into an appropriate F1 embryonic stem line. For example, NOD embryonic stem lines have not been produced, but (129 × NOD)F1 cell lines exist21. In this approach, usually the vector is designed with the NOD genomic sequence but targets the NOD and 129 alleles equivalently, thus making it necessary to define which clones have the NOD allele disrupted. With SNPs that define the 129 and NOD alleles in the targeted region, if the homologous recombination of the targeted gene occurs on the NOD allele, the region will be heterozygous, with both 129 and NOD genes present. If the recombination involves the 129 allele, the embryonic stem cell clone will be homozygous for NOD SNPs

across the targeted interval. In silico mapping across the interval of interest should also be done with online databases. The Ensembl, Perlegen and T1DBase websites can be used as described here to examine the genetic ‘terrain’ of the region of interest. From such in silico mapping, SNPs can be obtained that are useful for monitoring the backcrossing of specific gene regions to a different background. If it is necessary to introduce a gene alteration into 129 embryonic stem cells with subsequent backcrossing to the strain of interest, the progeny of each backcross should be monitored to identify mice with recombination events that reduce the number of flanking 129-derived genes. These mice should be used as breeders for the subsequent backcross generation. Concurrently, a control strain containing the same portion of 129-derived DNA but lacking the targeted region should be developed to confirm that any altered phenotype can be attributed to the targeted mutation rather than to the 129 alleles at neighboring genes. Following these guidelines and using SNP databases to analyze flanking DNA regions at each backcross will provide more precise information about ‘genetic background effects’ and may prevent mischaracterization of the genetic basis of immune phenotypes. ACKNOWLEDGMENTS Supported by the Juvenile Diabetes Research Foundation (L.S.W.), the Wellcome Trust (L.S.W.), the National Institute of Diabetes and Digestive and Kidney Diseases of the National Institutes of Health (W.M.R.) and the Autoimmunity Center of Excellence program (W.M.R.). COMPETING INTERESTS STATEMENT The authors declare no competing financial interests. 1. Wakeland, E., Morel, L., Achey, K., Yui, M. & Longmate, J. Speed congenics: a classic technique in the fast lane (relatively speaking). Immunol. Today 18, 472–477 (1997). 2. Wicker, L.S., Todd, J.A. & Peterson, L.B. Genetic control of autoimmune diabetes in the NOD mouse. Annu. Rev. Immunol. 13, 179–200 (1995). 3. Petes, T.D. Meiotic recombination hot spots and cold spots. Nat. Rev. Genet. 2, 360–369 (2001). 4. Simpson, E.M. et al. Genetic variation among 129 substrains and its importance for targeted mutagenesis in mice. Nat. Genet. 16, 19–27 (1997). 5. Leiter, E.H. Mice with targeted gene disruptions or gene insertions for diabetes research: problems, pitfalls, and potential solutions. Diabetologia 45, 296–308 (2002). 6. Flint, J., Valdar, W., Shifman, S. & Mott, R. Strategies for mapping and cloning quantitative trait genes in rodents. Nat. Rev. Genet. 6, 271–286 (2005). 7. Wolfer, D.P., Crusio, W.E. & Lipp, H.P. Knockout mice: simple solutions to the problems of genetic background and flanking genes. Trends Neurosci. 25, 336–340 (2002). 8. Wang, B. et al. Interferon-γγ impacts at multiple points during the progression of autoimmune diabetes. Proc. Natl. Acad. Sci. USA 94, 13844–13849 (1997). 9. Kanagawa, O., Xu, G., Tevaarwerk, A. & Vaupel, B.A. Protection of nonobese diabetic mice from diabetes by gene(s) closely linked to IFN-γγ receptor loci. J. Immunol. 164, 3919–3923 (2000). 10. Hubbard, T.J.P. et al. Ensembl 2007. Nucleic Acids

VOLUME 8 NUMBER 7 JULY 2007 NATURE IMMUNOLOGY


C O M M E N TA R Y Res. D610–D617 (2007). 11. Witmer, P.D. et al. The development of a highly informative mouse simple sequence length polymorphism (SSLP) marker set and construction of a mouse family tree using parsimony analysis. Genome Res. 13, 485–491 (2003). 12. Hulbert, E.M. et al. T1DBase: integration and presentation of complex data for type 1 diabetes research. Nucleic Acids Res. 35, D742–D746 (2007). 13. Lyons, P.A. et al. The NOD Idd9 genetic interval influences the pathogenicity of insulitis and contains molecular variants of Cd30, Tnfr2, and Cd137. Immunity 13, 107–115 (2000). 14. Cannons, J.L. et al. Genetic and functional associa-

tion of the immune signaling molecule 4–1BB (CD137/ TNFRSF9) with type 1 diabetes. J. Autoimmun. 25, 13–20 (2005). 15. Irie, J., Wu, Y., Sass, D.A. & Ridgway, W.M. Genetic control of anti-Sm autoantibody production in NOD congenic mice narrowed to the Idd9.3 region. Immunogenetics 58, 9–14 (2006). 16. Cuppen, E. Haplotype-based genetics in mice and rats. Trends Genet. 21, 318–322 (2005). 17. Wiltshire, T. et al. Genome-wide single-nucleotide polymorphism analysis defines haplotype patterns in mouse. Proc. Natl. Acad. Sci. USA 100, 3380–3385 (2003). 18. Edwards, S.V., Chesnut, K., Satta, Y. & Wakeland,


E.K. Ancestral polymorphism of Mhc class II genes in mice: implications for balancing selection and the mammalian molecular clock. Genetics 146, 655–668 (1997). 19. Yamanouchi, J. et al. Interleukin-2 gene variation impairs regulatory T cell function and causes autoimmunity. Nat. Genet. 39, 329–337 (2007). 20. Yalcin, B. et al. Unexpected complexity in the haplotypes of commonly used inbred strains of laboratory mice. Proc. Natl. Acad. Sci. USA 101, 9734–9739 (2004). 21. Brook, F.A. et al. The derivation of highly germline-competent embryonic stem cells containing NOD-derived genome. Diabetes 52, 205–208 (2003).

673