optimised microsatellite marker set
Introduction
Since the first report of osteochondrosis (OC) 56 years ago (NILSSON 1947) its importance to the horse industry has been well demonstrated. The condition occurs in growing animals of all species (GLADE 1986) and appears to be the most common primary cause of degenerative joint diseases in domestic animals (DIK et al. 1999). Alterations that are associated with OC are subchondral fractures and cyst-like lesions, chondromalacia, wear lines, cartilage flaps, synovitis and osteochondrosis dissecans (OCD).
Available epidemiological data indicate that osteochondrosis is present in Warmblood and Trotter horse populations at unacceptably high levels (i.e. between 10 and 25%) across a range of different breeds (JEFFCOTT 1997). The importance of OC is well recognised, as its incidence appears to be steadily increasing, and is therefore associated on a global scale with corresponding enormous economic losses (JEFFCOTT 1996).
OC is defined as a disturbance in the process of endochondral ossification of growing cartilage of the growth plates and/or the articular/epiphyseal complex (VAN DE LEST et al.
1999). The specific causes are still unknown but OC appears to be multifactorial in origin, including the factors growth rate, body size, nutrition, mineral imbalance, endocrinological dysfunction and biomechanical trauma. Furthermore, a hereditary disposition to osteochondrosis has been demonstrated in Trotters, Coldblood and Warmblood horse breeds (SCHOUGAARD et al. 1990, GRØNDAHL and DOLVIK 1993, PHILIPSSON et al. 1993, WINTER et al. 1996, WILLMS et al. 1999, PIERAMATI et al. 2003). OC may be supposed to be a quantitatively inherited trait as it is likely that its development is influenced by different genes and environmental factors. Heritability estimates vary in a wide range from h² = 0.01 to h² = 0.64. The reason for these large differences in heritability estimates may be seen in the type of method applied for estimation, the sampling scheme of the data and, in some studies, in the low numbers of animals included. Despite a considerable number of studies on heritability estimates, the mode of inheritance has not been determined by complex
Genome scan for QTL of OC in Hanoverian Warmblood horses using an optimised microsatellite marker set 38
segregation analysis. Therefore, the number of genes influencing OC or OCD remains unknown.
The aim of this study was to carry out a whole genome scan using an optimised microsatellite marker set in order to identify quantitative trait loci (QTL) with significant influence on the development of OC or OCD.
Material and methods Family material
A first radiographic survey was performed from February to July, 2001, at the Clinic for Horses, Surgery and Radiology, Department of Veterinary Medicine, Free University of Berlin (Director: Professor Dr. B. Hertsch), There, 629 foals (the progeny of 168 stallions) and their mares were examined for radiographic signs of osteochondrosis (OC) and osteochondrosis dissecans (OCD). In a second radiographic survey carried out from March to June, 2003, 405 of the 629 foals from the first survey were x-rayed again in order to make it possible to study the development of the osteochondrotic lesions.
Digital radiographs were obtained of the metacarpo- and metatarsophalangeal (fetlock) joints, tibiotarsal (hock) joints, and femoropatellar joints (stifles). The radiological examination comprised eight views chosen to reveal fetlock OC/OCD, hock OC/OCD and stifle OC/OCD.
Lateromedial views were used for the fetlock and the stifle joints. The tarsocrural joints were examined by using plantolateral-dorsomedial views. Signs of OC (irregular bone trabeculation with variable radiopacity and changes of the regular bone contour such as exostoses, cartilage flaps, irregular bone margins and flattened or concave outlining) and OCD (bony fragments in the joint space) were examined at the predilection sites in each articulation. The predilection sites in the fetlock joints are the sites of attachment of the short sesamoidean ligaments to the proximal phalanx, the palmar/plantar bone margin of the proximal phalanx, and the dorsoproximal rim of the proximal phalanx. The predilection sites in the hock joints are the intermediate ridge of the distal tibia, the lateral/medial trochlear ridge of the talus, and the lateral/medial malleolus of the tibia. In the stifle joints, the predilection sites are the lateral/medial trochlear ridge of the femur, the trochlear groove of the femur and the patella.
All radiographic findings considered abnormal were registered. Two categories of radiographic changes were classified, (1) those showing radiographic signs of
Genome scan for QTL of OC in Hanoverian Warmblood horses using an optimised microsatellite marker set 39
osteochondrosis, and (2) those showing radiographic signs of osteochondrosis dissecans (joint mice). In 403 foals (64.1%) there were no signs of OC/OCD; 118 foals (18.8%) were affected with OCD, and 108 (17.2%) with OC.
We used an affected half-sib design. Thus, only foals exhibiting OC or OCD were used for the present study.
The progeny of one breeding season for 32 stallions was chosen for genotyping due to the high number of affected foals among them. The size of the affected half-sib groups ranged from 2 to 15; the average value was 3.8. All selected foals (n = 123) were affected either by OC or OCD, or both. Of the 123 foals 66 (55.7%) were female and 57 (46.3%) were male.
The average age at examination was 6.7 months. Two foals were four months old, while the other 121 foals were from five to ten months of age. All mares of the foals were also genotyped. In 13 of the 32 families, blood samples of the stallions were available for analysis, but the phenotypes of the stallions were unknown.
Genome scan panel
All animals were genotyped using a genome scan panel consisting of 157 microsatellite markers. The markers spanned the 31 autosomes and the X chromosome with an average distance of 19.1 cM. The average observed heterozygosity (HO) of all 157 markers was 65.7%, the average expected heterozygosity (HE) was 64.6%, and the average polymorphism information content (PIC) was 60.3%. Appendix 2, Table 10 contains a list of the markers including chromosome location, primer sequences, primer concentration, HO, HE, PIC and references.
Genotyping
For genotyping, 1.35 µg genomic DNA was isolated from 75 µl EDTA blood using the QIAamp® 96 DNA Blood Kit (QIAGEN, Hilden, Germany).
PTC 100™ or PTC 200™ (MJ Research, Watertown, MA, USA) thermocyclers and a general PCR programme with variable annealing temperature (Ta) were used for the PCR amplification of the 157 markers (Table 10). The reaction started at 94 °C for 4 min, followed by 35 cycles at 94 °C for 30 s, Ta for 1 min, 72 °C for 30 s, and at 4 °C for 10 min.
Genome scan for QTL of OC in Hanoverian Warmblood horses using an optimised microsatellite marker set 40
The PCRs were performed in 15-µl volumes using 10 ng DNA, 10x incubation buffer containing 15 mM MgCl2, 5% DMSO, 100µM each dNTP, 1.0 U Taq Polymerase (Qbiogene, Heidelberg, Germany).
The amount of primers in the multiplex groups ranged from 1.0 to 18.0 pmol. All forward primers were flourescently labelled at the 5’ end with IRD700. To increase efficiency, 94 primer pairs were pooled into PCR multiplex groups of two pairs, and the 63 remaining primer pairs were PCR amplified separately.
For the analysis of the marker genotypes, the PCR products were size-fractionated by gel electrophoresis on an automated sequencer (LI-COR 4200/S-2, Lincoln, NE, USA) using 6%
polyacrylamide denaturing gels (rotiphorese®Gel 40, Carl Roth GmbH & Co, Karlsruhe, Germany). Prior to loading, dilution of PCR products with formamide loading buffer was determined empirically and was carried out correspondingly. To ensure adequate size fractionation of alleles, the diluted PCR products were denatured at 94 °C for one minute. The multiplex groups and the separately amplified PCR products were pooled according to their size and labelling prior to electrophoresis. Allele sizes were detected using an IRDye700-labelled DNA ladder, and the genotyping data were analysed manually.
Statistical analysis
The marker genotypes of all animals in combination with pedigree data and phenotype data of the foals were analysed using Merlin software (multipoint engine for rapid likelihood inference, version 0.9.10) (ABECASIS et al. 2002). In this way it was possible to detect significant, nonparametric test statistics (Zmeans and LOD scores) for the co-segregation of marker alleles and OC/OCD in fetlock and/or hock joints. OC/OCD in the stifle joints could not be considered, as the number of affected foals among the half-sib groups was too small.
Genome scan for QTL of OC in Hanoverian Warmblood horses using an optimised microsatellite marker set 41
Results
Linkage analysis
The evaluation of the data using the Merlin software was based on a multipoint analysis of the affected foals and their parents, including IBD (identical by descent) calculations, kinship calculations and nonparametric linkage (NPL) analysis. Furthermore, we estimated haplotypes by finding the most likely path of gene flow (ABECASIS et al. 2002). While the all modus tested the association of all genotyped markers with the phenotype, the pairs modus tested in pairs the alleles of the markers with the phenotypic expression of the disease. A LOD score was used for both methods according to the linear model of KONG and COX (1997). For p (probability) < 0.05, there was a significant chromosome-wide co-segregation of a marker allele with the phenotypic expression of osteochondrosis in the examined population. For the genome-wide type-I error probability (pg), a Bonferroni correlation was applied with pg = 1 - (1 - p)1/r, where r is the ratio of the length of the chromosome harbouring the QTL to the total length of the equine genome in cM.
In the present study, separate analyses were performed for the traits OC (fetlock and/or hock) OCD (fetlock and/or hock), fetlock OC, fetlock OCD, hock OC and hock OCD. These parameters were analysed for both sexes together (Table 6), as well as for each sex separately (Tables 7 and 8). In total, QTL could be identified on 13 chromosomes. The error probabilities for the Zmeans (pZmeans) and LOD scores (pLOD) below 0.05 of these QTL are indicated in boldface in the tables. The remaining error probabilities below 0.10 were noted to permit comparisons between the estimates for both sexes and the estimates for each sex separately, and to indicate genomic regions with suspected QTL. These regions should be further considered in a second step of the marker study using dense microsatellite marker sets.
Genome scan for QTL of OC in Hanoverian Warmblood horses using an optimised microsatellite marker set 42
Table 6. Error probabilities (pZmeans, pLOD) below 0.10 and map positions for the microsatellite markers used in the whole genome scan for OC/OCD in all animals (both sexes)
OC OCD OC f OCD f OC h OCD h
Genome scan for QTL of OC in Hanoverian Warmblood horses using an optimised microsatellite marker set 43
Table 7. Error probabilities (pZmeans, pLOD) below 0.10 and map positions for the microsatellite markers used in the whole genome scan for OC/OCD in male animals
OC OCD OC f OCD f OC h OCD h
Genome scan for QTL of OC in Hanoverian Warmblood horses using an optimised microsatellite marker set 44
Table 8. Error probabilities (pZmeans, pLOD) below 0.10 and map positions for the microsatellite markers used in the whole genome scan for OC/OCD in female animals
OC OCD OC f OCD f OC h OCD h ECA: equine chromosome POS: position on the linkage map Where no modus is indicated in the table, the all modus is noted.
Genome scan for QTL of OC in Hanoverian Warmblood horses using an optimised microsatellite marker set 45
In total, 27 chromosome-wide QTL for the different traits could be mapped on 13 different equine chromosomes. Of these QTL, six were present in both sexes, nine in male animals and twelve in female animals.
A QTL for fetlock OC was found for both sexes on ECA02. There is some indication of this QTL in male and female animals. A QTL for OCD in female animals was also identified on ECA02, which suggests that it applies to male animals and to both sexes.
A QTL for fetlock OC in male animals was found on ECA03, and can be anticipated for both sexes. Here, the QTL was more obvious in the analysis of male animals.
A QTL for hock OCD was identified in male animals only on ECA04.
A QTL for OCD in male animals was found on ECA05, which is also suggested by the analysis of female animals only and both sexes together. Furthermore, a QTL for hock OCD was found in male animals, and for which there is some indication for both sexes and female animals.
A QTL for OCD in female animals was identified on ECA06 which can be anticipated in both sexes but was not present in male animals.
The test statistics indicated the presence of a QTL on ECA10 for OC in female animals and in both sexes. The analysis over both sexes together gave lower error probabilities than the separate analyses by sex.
A QTL for hock OC was found in both sexes and in female animals on ECA15. This QTL, for which there was also some indication in male animals, had a lower error probability in the analysis for both sexes than in female animals alone. Furthermore, a QTL for hock OCD was present in female animals that was not indicated elsewhere.
ECA16 was the site of a QTL for OC and fetlock OC in female animals only.
ECA19 included a QTL for OC in female animals, for which there is some indication in male animals; it was not be found in the analysis of both sexes together. Additionally, a QTL for fetlock OC was identified in female animals and in both sexes.
On ECA21, a QTL was detected for OC in male animals. The same QTL was also present in the analysis for both sexes. A QTL for OCD was identified in male animals only.
Furthermore, a QTL for hock OC was found in male animals, for which there was some indication in female animals. A QTL for hock OCD was present in male animals and in both sexes, but the error probability was lower in male animals.
The QTL identified on ECA23 for OC and fetlock OCD were present in female animals only.
A QTL for fetlock OC was located on ECA24 in male animals only.
Genome scan for QTL of OC in Hanoverian Warmblood horses using an optimised microsatellite marker set 46
On ECA25, a QTL was detected for OCD in female animals only.
A genome-wide error probability (pg) was calculated for the QTL with the smallest error probability on each chromosome (Table 9).
Table 9. Genome-wide error probability of 13 chromosome-wide significant QTL
ECA Zmean pvalue LOD pvalue pg
2 1.47 0.07 0.74 0.03 0.46 3 1.38 0.08 0.80 0.03 0.44 4 2.21 0.014 0.84 0.02 0.24 5 2.33 0.010 0.68 0.04 0.41 6 1.82 0.03 0.98 0.02 0.39 10 1.75 0.04 1.11 0.012 0.32 15 2.49 0.006 1.41 0.005 0.12 16 2.29 0.011 1.49 0.004 0.14 19 2.06 0.02 1.05 0.014 0.33
21 4.20 0.00001 1.07 0.013 0.0002
23 1.44 0.07 0.90 0.02 0.39 24 1.34 0.09 0.92 0.02 0.46
25 1.42 0.08 0.89 0.02 0.55
A single QTL, located on ECA21, showed a genome-wide significant error probability (p < 0.001).
Discussion
The linkage analysis based on IBD mapping indicated the existence of 13 chromosome-wide significant QTL for OC and OCD. The QTL were heterogeneously distributed between hock and fetlock joints.
QTL for hock OC/OCD were identified on ECA04 and ECA05 but mainly on ECA15 and ECA21. QTL for fetlock OC/OCD were present on ECA02, ECA03, ECA16, ECA19 and ECA24. In no case was a QTL for fetlock OC/OCD found on the same chromosome as for hock OC/OCD. This may indicate that the genetic influences on the development of fetlock OC/OCD and hock OC/OCD are due to different loci. This seems likely, as the genetic
Genome scan for QTL of OC in Hanoverian Warmblood horses using an optimised microsatellite marker set 47
correlations between fetlock OC/OCD and hock OC/OCD were close to zero or even negative; these traits may thus be considered genetically independent. PHILIPSSON et al.
(1993) found negative genetic correlations between height at the withers and the incidences of both hock OC and fetlock OC in trotters, but their estimations included large standard errors.
Those investigators found that a large carpal circumference was genetically associated with a high incidence of hock OC (positive genetic correlation) but with low incidence of fetlock OC (negative genetic correlation). They also found a tendency towards a positive genetic correlation between hock joint OC and fetlock joint OC. They indicated however that this tendency towards a positive genetic correlation might be ambiguous, since the progeny groups with the phenotypically higher incidence of fetlock OC also were the ones with the lower incidence of hock OC and vice versa. Those authors found that the rather contradictory results obtained for genetic correlations between the traits in their investigation indicate the need for more research in this area. However, they concluded that, in their study, hock OC and fetlock OC did not seem to be genetically related defects, and should be dealt with as independent traits in possible selection schemes. STOCK et al. (2003) estimated moderately negative genetic correlations between hock and fetlock OCD in Hanoverian Warmblood horses. The positions of QTL for hock and fetlock OC/OCD on different chromosomes conform to this finding. Different QTL on the chromosomes could be identified for both sexes, and the number of QTL found in each sex was nearly the same. In no case was a QTL identified in female animals present on the same chromosome in male animals. Some QTL were present in the analysis over both sexes. The results indicate sex-specific QTL for OC and OCD. The QTL affecting either male or female animals may be due to their different loci.
The number of QTL found in total indicate an oligogenic mode of inheritance, which in turn corresponds to the assumption that osteochondrosis is multifactorial in origin (JEFFCOTT 1996).
This genome scan was a first step towards the identification of genes responsible for osteochondrosis in horses. The positions of the QTL identified in this study have to be refined by increasing the density of the markers used in these specific genomic regions. In this second step, for genotyping approximately three to four additional markers in regions closely flanking a QTL will have to be added from the existing equine maps or newly developed and then integrated into the linkage analysis. In this way, the QTL can be confirmed or excluded.
Furthermore, in genomic regions in which co-segregation of a marker allele and the phenotypic expression of osteochondrosis showed chromosome-wide significance,
non-Genome scan for QTL of OC in Hanoverian Warmblood horses using an optimised microsatellite marker set 48
affected foals of the half-sib groups should be added to improve power of the linkage analysis through discordant half-sib pairs and to estimate the explained variance and the size of the additive and dominance effects of the QTL.
Once a QTL for equine osteochondrosis has been successfully mapped, candidate genes in the areas directly flanking a QTL may be chosen by means of comparative genetics. The equine QTL positions can then be compared with the conserved chromosomal region(s) between horses and humans in order to make it possible to look for candidate genes that show appropriate expression and/or appropriate function. Potential candidate genes have to code for hormones, enzymes, metabolic factors and/or their receptors involved in the complex of cartilage proliferation and maturation during endochondral ossification. Candidate genes may be also those involved in epiphyseal dysplasia such as collagen genes, cartilage oligomeric matrix protein gene, or diastrophic dysplasia sulfate transporter gene.
In a recent study, ANDERSSON-EKLUND et al. (2000) identified a genome-wide and a chromosome-wide QTL for osteochondrosis in pigs using factor analysis for extracting two principal components of OC. QTL for these principal components were found on chromosomes 5 and 13. Human chromosome (HSA) 12q14-q24 was found to be homologous with the region containing the QTL on chromosome 5 and contained a potential candidate gene, cartilage homeoprotein1 (CART1). The homologous regions to HSA 12q14-q24 cannot be unambiguously identified on the equine genome, since 6q22-23, 8p12-13, 8p15-16 or 28q15 might be syntenic. Equine 28q15 contains a gene coding for insulin-like growth factor I (IGF-I). IGFs play an important role in cartilage metabolism and growth (SEMEVOLOS et al. 2001; HENSON et al. 1997b), so that this gene can be considered as a potential candidate
In a recent study, ANDERSSON-EKLUND et al. (2000) identified a genome-wide and a chromosome-wide QTL for osteochondrosis in pigs using factor analysis for extracting two principal components of OC. QTL for these principal components were found on chromosomes 5 and 13. Human chromosome (HSA) 12q14-q24 was found to be homologous with the region containing the QTL on chromosome 5 and contained a potential candidate gene, cartilage homeoprotein1 (CART1). The homologous regions to HSA 12q14-q24 cannot be unambiguously identified on the equine genome, since 6q22-23, 8p12-13, 8p15-16 or 28q15 might be syntenic. Equine 28q15 contains a gene coding for insulin-like growth factor I (IGF-I). IGFs play an important role in cartilage metabolism and growth (SEMEVOLOS et al. 2001; HENSON et al. 1997b), so that this gene can be considered as a potential candidate