Content uploaded by Sigridur Bjornsdottir
Author content
All content in this area was uploaded by Sigridur Bjornsdottir on Oct 26, 2017
Content may be subject to copyright.
ORIGINAL ARTICLE
The effect of the ‘Gait keeper’ mutation in the DMRT3 gene on
gaiting ability in Icelandic horses
T. Kristjansson
1
, S. Bjornsdottir
2
, A. Sigurdsson
1
, L.S. Andersson
3
, G. Lindgren
4
, S.J. Helyar
5
,
A.M. Klonowski
5
& T. Arnason
1
1 Agricultural University of Iceland, Hvanneyri Borgarnes, Iceland
2 Icelandic Food and Veterinary Authority, Selfoss, Iceland
3 Capilet Genetics AB, €
Oster Skogsta, V€
aster
as, Sweden
4 Department of Animal Breeding and Genetics, Swedish University of Agricultural Sciences, Uppsala, Sweden
5 Mat
ıs, Reykjavik, Iceland
Keywords
Gaiting ability; genotype effect; genotype
probability.
Correspondence
T. Kristjansson, Agricultural University of
Iceland, IS-311 Borgarnes, Iceland.
Tel: 00345-8662199;
Fax: 00354-4335001;
E-mail: thorvaldurk@lbhi.is
Received: 22 April 2014;
accepted: 30 June 2014
Summary
A nonsense mutation in DMRT3 (‘Gait keeper’ mutation) has a predomi-
nant effect on gaiting ability in horses, being permissive for the ability
to perform lateral gaits and having a favourable effect on speed capacity
in trot. The DMRT3 mutant allele (A) has been found in high frequency
in gaited breeds and breeds bred for harness racing, while other horse
breeds were homozygous for the wild-type allele (C). The aim of this
study was to evaluate further the effect of the DMRT3 nonsense muta-
tion on the gait quality and speed capacity in the multigaited Icelandic
horse and demonstrate how the frequencies of the A- and C- alleles
have changed in the Icelandic horse population in recent decades. It
was confirmed that homozygosity for the DMRT3 nonsense mutation
relates to the ability to pace. It further had a favourable effect on scores
in breeding field tests for the lateral gait t€
olt, demonstrated by better beat
quality, speed capacity and suppleness. Horses with the CA genotype
had on the other hand significantly higher scores for walk,trot,canter
and gallop, and they performed better beat and suspension in trot and
gallop. These results indicate that the AA genotype reinforces the coordi-
nation of ipsilateral legs, with the subsequent negative effect on the syn-
chronized movement of diagonal legs compared with the CA genotype.
The frequency of the A-allele has increased in recent decades with a
corresponding decrease in the frequency of the C-allele. The estimated
frequency of the A-allele in the Icelandic horse population in 2012 was
0.94. Selective breeding for lateral gaits in the Icelandic horse popula-
tion has apparently altered the frequency of DMRT3 genotypes with a
predicted loss of the C-allele in relatively few years. The results have
practical implications for breeding and training of Icelandic horses and
other gaited horse breeds.
Introduction
One of the major characteristics of horse breeds is
their ability to perform specific gaits. A gait is a coordi-
nation pattern of the limbs identified by timing and
sequence of the footfalls. The gait chosen by a horse
depends on speed, genotype and environmental fac-
tors (Alexander 1988; Clayton 2004). The Icelandic
horse is a multigaited horse breed showing the stan-
dard gaits of all domestic horse breeds that are walk,
©2014 Blackwell Verlag GmbH •J. Anim. Breed. Genet. (2014) 1–11 doi:10.1111/jbg.12112
J. Anim. Breed. Genet. ISSN 0931-2668
trot,canter and gallop. In addition, it has t€
olt and pace.
T€
olt is a four-beat running gait with lateral sequence
of footfalls and without suspension. Pace is considered
a two-beat gait with a moment of suspension where
lateral legs move almost synchronously back and
forth and is optimally a very fast gait. Icelandic horses
that possess walk, trot, canter, gallop and t€
olt are
referred to as four-gaited horses, whereas horses that
additionally have the ability to perform pace are called
five-gaited horses.
A nonsense mutation in DMRT3 (DMRT3_Ser301-
STOP), also referred to as the ‘Gait keeper’ mutation,
has been shown to have a great impact on gaiting abil-
ity in horses (Andersson et al. 2012). Previous work
has indicated that the mutation is permissive for the
ability to perform lateral gaits, such as t€
olt and pace,
and homozygosity for the mutation is required
although not sufficient for the ability to pace. More-
over, the mutation was reported to have a favourable
effect on speed capacity in trot and seemed to inhibit
the transition from trot to gallop in a study on Stan-
dardbred horses used in harness racing (Andersson
et al. 2012). The DMRT3 mutant allele (A) was found
in high frequencies in gaited breeds and breeds bred
for harness racing, while tested non-gaited horse
breeds were found homozygous for the wild-type
allele (C) (Andersson et al. 2012). Comparison of
wild-type and Dmrt3-null mice showed that DMRT3 is
crucial for the normal development of a coordinated
locomotor network that controls limb movement. It
was concluded that DMRT3 neurons are essential for
left/right coordination as well as for coordinating the
movement of fore- and hind legs (Andersson et al.
2012).
The Icelandic horse is bred for leisure riding as well
as for sport competitions (Albertsd
ottir et al. 2007;
FIZO 2012), with the international breeding goal for
the Icelandic horses promoting five-gaited horses. The
breeding assessment system is based on breeding field
tests for both riding qualities and conformation,
where assessment of riding qualities includes judging
of the five gaits (FIZO 2012). Scores are also given for
slow t€
olt and canter although they are not weighed into
the total score, but influence the scoring for t€
olt and
gallop, respectively. The horses are judged on a scale
from 5 (not presented) to 10 (best) with intervals of
0.5, the average being 7.5. The judges can also give
standardized comments on the assessed traits that
describe certain attributes of the traits and substanti-
ate the scoring (listed in Table S1 for the five gaits).
Horses can only receive scores above average if they
present one or more of the listed advantages and
horses below average have one or more of the listed
disadvantages. Horses can attend the breeding field
tests from the age of four, the majority being five and
6 years old. Approximately 12.5% of the Icelandic
horse population is presented based on a preselection
by the breeders (Albertsd
ottir et al. 2011).
The aim of this study was to evaluate the effect of
the DMRT3 nonsense mutation on the gait quality and
speed capacity in the multigaited Icelandic horse and
demonstrate how the frequency of the A- and C-
alleles has changed in the Icelandic horse population
in recent decades.
Material and methods
Estimation of DMRT3 genotype effect on gait traits
Selection of horses
Horses were selected for genotyping on the basis of
their scores for the different gaits in breeding field
tests. For practical reasons, the selection was limited
to horses judged in Iceland and Sweden, in the years
2000–2012, with a stored DNA sample according to
the global database WorldFengur (http://www.world-
fengur.com). The first criterion was the score for pace
including both horses with scores below average
(5.5–7.0) and higher performing (7.5–10). The number
of horses in each score for pace was in accordance with
the proportion of horses getting each score annually
for the last 5 years in Iceland. This provided 390 five-
gaited horses with a wide distribution of scores for the
other gaits. In the next step, four-gaited horses (with
the score 5.0 for pace and various scores for the other
gaits) were added until at least 20 horses showed each
score for each gait in the range of 7.0–9.0 and as many
as possible in the range of 9.5–10. This added 243
horses and enabled comparison of four- and five-gai-
ted horses of different gait quality with respect to the
DMRT3 genotype. Horses were selected at random
when possible, but in the case of few available candi-
dates, all were selected. Finally, 34 horses were
selected on the basis of the judges’ comments describ-
ing quality and speed capacity of the gaits (FIZO
2012) (Table 1). As comments are not necessarily
recorded for each trait for all horses, comparison with
the genotype data was limited to the following
variables: good beat in walk; good speed capacity in
trot versus lack of speed; clear beat in trot and good
suspension in trot versus four-beated trot; good speed
capacity in gallop versus lack of speed; clear beat in
gallop; good suspension in gallop versus lack of suspen-
sion in gallop; good speed capacity in t€
olt versus lack of
speed; good beat in t€
olt versus trotty t€
olt; and supple
t€
olt versus stiff t€
olt.
©2014 Blackwell Verlag GmbH •J. Anim. Breed. Genet. (2014) 1–112
The effect of a mutation in DMRT3 on gaiting ability T. Kristjansson et al.
Description of data
The data set consisted of 667 horses, of which 360
were stallions and 307 mares. Where a horse had
been scored more than once at breeding field tests,
only the record where the horse obtained its highest
total score for riding ability was used. Scores for t€
olt
and age at first evaluation at a breeding field test were
also investigated. The age ranged from 4 to 14 years,
with a mean of 6.4 1.7 years. The horses were born
in 1986–2008 with a mean birth year 2001 4.3.
The total number of sires was 271, with an average of
2.5 3.7 offspring per sire (range: 1–43). The data set
included 404 five-gaited horses (pace score ≥5.5) and
263 four-gaited horses (pace score =5.0).
SNP genotyping
Samples from the 667 horses were obtained from two
repositories in Iceland and one in Sweden. DNA was
extracted from nose swabs and blood using mag
TM
kit
(AGOWA GmbH, Berlin, Germany) and Gentra Pure-
gene Blood Core Kit (QIAGEN Inc., Venlo, Limburg,
the Netherlands), respectively.
Custom TaqMan SNP Genotyping assays (Applied
Biosystems, Inc., Foster City, CA, USA) were used to
genotype the DMRT3_Ser301STOP SNP with the follow-
ing primers and probes: Forward primer: 50-CCTC
TCCAGCCGCTCCT-30; reverse primer: TCAAAGATG
TGCCCGTTGGA-30;wild-typeprobe:5
0-CTGCCGAA
GTTCG; mutant probe: 50-CTCTGCCTAAGTTCG-30.rt-
PCRs were carried out on a 384-well ABI PRISM 7900
HT sequence detection system (Applied Biosystems)
and a 96-well Stratagene Mx3005P.
Statistical analyses
Statistical analyses were performed using SAS (SAS
Institute Inc. 2009). The mean, standard deviation
(SD), skewness and kurtosis were calculated to
describe the variation of the gaits within the sample of
667 horses. To ascertain whether the distribution of
the gaits deviated significantly from zero measure of
skewness and kurtosis, the following calculations
were made: Estimated skewness >1.96√6/n for
p<0.05 and estimated skewness >2.33√6/n for
p<0.01; and estimated kurtosis >1.96√24/n for
p<0.05 and estimated kurtosis >2.33√24/n for
p<0.01.
Effects of the age*sex interaction (four age classes:
4, 5, 6 and ≥7 years old horses, two sex classes: mares
and stallions) and the genotype of the horse (two clas-
ses: AA and CA genotypes) on the gait traits were esti-
mated with analysis of variance using PROC GLM
(SAS Institute Inc. 2009). The following model was
assumed for each gait trait:
y
ijn
=l+age sex
i
+genotype
j
+age –sex *
genotype
k
+e
ijn
where y
ijn
is a gait trait (six traits: walk,trot,gallop,
canter,t€
olt and slow t€
olt) for the nth horse, lis the pop-
ulation mean, age-sex
i
is the combined effect of the i
th
age–sex group (i=1,...,8), genotype
j
is the effect of
the j
th
genotype (j=1, 2; 1 =AA, 2 =CA), age-
sex*genotype
k
is the effect of the interaction of k
th
age–sex by genotype (k=1,...,16) and e
ijn
is a ran-
dom ~NID (0, r2
e) residual effect. Because of their low
number, horses with the CC genotype (n =8) were
not included in this analysis. A Student’s t-test was
used to ascertain whether both scores for t€
olt and age
at first evaluation differed significantly between
horses with AA and CA genotypes. Then a chi-square
test with 1 df was performed to study whether propor-
tions of genotypes within a subgroup of 28 horses that
had received scores of 9.0–9.5 for t€
olt as 4 years old
deviated significantly from the proportion of the
genotypes within the whole data set.
Discriminant analysis was performed using stepwise
selection to obtain a subset of the gaits to be able to dis-
criminate between the genotype classes (AA and C-).
Only gaits that were significant in the stepwise discri-
minant function procedure and that had partial R
2
values ≥0.01 were retained in the final model. These
gaits were then included in a canonical discriminant
analysis to find a linear combination of the gaits that
best summarized the difference between the genotype
classes (AA and C-). Mahalanobis distance between
the class means was estimated. The analyses were
Table 1 Number of horses in the data set within scores for each gait assessed at breeding field tests for Icelandic breeding horses
Trait 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 Total
Walk 0 9 24 47 114 140 176 105 29 2 1 647
Trot 1 0 2 19 71 95 194 172 87 25 1 667
Gallop 0 0 1 2 28 109 235 197 84 11 0 667
Canter 26 0 0 10 62 162 197 93 35 11 1 597
T€
olt 0 0 2 4 21 71 182 217 112 55 3 667
Slow t€
olt 3 0 5 8 43 130 220 150 63 10 1 630
Pace 263 44 47 59 80 30 35 46 50 12 1 667
©2014 Blackwell Verlag GmbH •J. Anim. Breed. Genet. (2014) 1–11 3
T. Kristjansson et al. The effect of a mutation in DMRT3 on gaiting ability
performed using the PROC STEPDISC and PROC CAN-
DISC (SAS Institute Inc. 2009), respectively.
Chi-square tests with 1 df were performed to study
whether proportions of genotypes within groups of
horses receiving certain judges’ comments describing
the gaits deviated significantly from the proportion of
the genotypes within the whole data set.
Change in allele frequency over time
When the mode of inheritance is known, genotype
probabilities at individual loci in large animal popula-
tions can be estimated from genotyped data or pheno-
typic data on a part of the population. For this
purpose, efficient computing algorithms have been
created (van Arendonk et al. 1989; Fernando et al.
1993; Janss et al. 1995; Kerr & Kinghorn 1996).
WorldFengur (http://www.worldfengur.com) pro-
vided a pedigree file containing 410 285 horses for
this study (birth years 1860–2012) where of 83%
were born after 1989. The average pedigree depth for
horses born 2009–2012 was 5.1 generations (max.
value 15), and the corresponding five generations
pedigree completeness index (PEC) was 85% accord-
ing to the method of MacCluer et al. (1983).
So far, only a very small proportion of the popula-
tion of Icelandic horses has been genotyped for the
DMRT3 mutation (706 genotyped horses were avail-
able for this study). However, recording of pace scores
in breeding field tests may provide approximate infor-
mation on the conditional probabilities of the geno-
types for a larger number of horses. As a starting
point, pace scores were extracted from the breeding
field test records kept in WorldFengur based on the
following conditions: a) pace score ≥6.0 (indication of
a carrier of the A-allele), b) pace score =5.0 (no pace
shown) and trot score ≥7.0 (initial indication of a
C-allele carrier). Data were excluded for horses
receiving 5.5 for pace and for horses with no pace
shown in combination with limited or bad trot.In
total, 55 073 records on 33 036 horses fulfilled the
required conditions. The highest pace score for horses
with repeated observations was selected, and one
record per horse was used.
Six phenotypic classes were formed based on the
information content of the genotype data and the
phenotypic data (first two columns in Table 2). The
genotype data consisted of 521 AA, 177 CA and
eight CC horses. The phenotypic pace scores ≥7.0
were taken to indicate the AA genotype (score 1),
and pace scores 5.0 were used as a preliminary indi-
cation of the CA or CC (C-) genotype (score 4),
while pace scores 6.0–6.5 were assumed to exclude
the CC genotype and leave scope for CA or AA geno-
types (score 5).
The present genotype data and earlier results (An-
dersson et al. 2012) have shown that horses with pace
scores ≥7.0 are almost certainly of genotype AA, while
a large part (>30%) of the horses receiving pace score
of 5.0 (shown as four-gaited horse) are also AA
although pace was not presented, for various reasons.
By use of pedigree data and the laws of Mendelian
inheritance, the preliminary phenotypic scores can be
updated and improved. It is feasible to use the Geno-
type Elimination (G–E) algorithm of Lange (1997) to
improve the phenotypic score data by creating a legal
data set compatible with the pedigree and Mendelian
models (Table 2). The G-E algorithm is an iterative
procedure for eliminating genotypes where incompat-
ibility is observed between any offspring–parent pairs
in the pedigree list. The G-E algorithm was run
repeatedly (seven times), and inconsistency was
listed. Unlikely scores of offspring were adjusted to
score 1 whenever both parents had genotype AA con-
firmed on the basis of genotype or phenotypic data.
As many horses shown as four-gaiters are truly AA,
this procedure of updating seems important (Anders-
son et al. 2012). The data with G-E updated scores
were used as an input in the segregation analysis by
the Geneprob Fortran program of Kerr & Kinghorn
(1996). The Geneprob program is based on the con-
cept of ‘iterative peeling’ and, as all such algorithms,
which are based on probability equations, the method
is sensitive to inconsistency in the data. The prior use
of the G-E procedure is highly recommendable before
segregation analysis in large data sets where errors in
pedigree and/or data recording are inevitable.
The resulting changes in the phenotypic scores are
shown in Table 2. The increase in fully and partly
informative scores was from 8.03% in the initial data
to 58.55% in the G–E updated scores.
Table 2 Phenotypic scores and corresponding possible genotypes for
Icelandic horses included in the pedigree list. The distribution of the
scores is shown before and after G-E updating, for the 410 285 Icelandic
horses included in the pedigree list
Phenotypic
scores Genotypes
Initial scores
G–E updated
scores
N%N%
1 AA 17 284 4.21 67 019 16.33
2 CA 176 0.04 7910 1.93
3 CC 8 0.002 8 0.002
4 C- 8857 2.16 4661 1.14
5 A- 6620 1.61 160 633 39.15
9 (no score) –377 348 91.97 170 054 41.45
©2014 Blackwell Verlag GmbH •J. Anim. Breed. Genet. (2014) 1–114
The effect of a mutation in DMRT3 on gaiting ability T. Kristjansson et al.
The frequency of the C-allele in the founder popu-
lation, p(C), was assumed to be either 0.13 as in the
sample of 706 genotyped horses or 0.30, which may
be a more probable value in the founder population
(ca. 5 generations back) according to preliminary
results. These values were used as priors in the geno-
type probability computations. Many horses in the
population were slightly inbred (average inbreeding
coefficient was 2.5%). The prior genotype probabili-
ties for inbred animals are not exact as the algorithm
in Geneprob does not account for the increased proba-
bility of inbred animals being homozygous. For ani-
mals with sufficient phenotypic or genotypic
information, the effects of prior allelic frequency or
inbreeding level on the posterior genotypic probabili-
ties are negligible.
The accuracy of the genotype probability estimates
was evaluated by the genotype probability index
(GPI) developed by Kinghorn (1997) to indicate the
information content from the segregation analysis.
Mean genotype probabilities within each year were
the estimate for the genotype frequencies within each
cohort, and from these, the annual development in
the frequencies of the A- and C-alleles was plotted for
the birth years 1980–2012. A chi-square test with 1 df
was performed to evaluate whether the genotypes
would conform to the Hardy–Weinberg proportions.
The chi-square value for each year (cohort) was
regressed on year for the period 1980–2012. These
calculations included 146 763 horses with a GPI of
≥30% (Kinghorn 1997).
Results
Effect of DMRT3 genotype on gait traits
The majority of the 667 horses genotyped for the
DMRT3_Ser301STOP mutation, or 509 (76.3%), were
homozygous for the A-allele (AA) and 150 (22.5%)
were heterozygous (CA) while only 8 (1.2%) were
found homozygous for the wild type (CC). Accord-
ingly, the frequency of the A-allele was 0.88 and of
the C-allele 0.12 in this data set and the genotypes
conform to the Hardy–Weinberg proportions. Among
the four-gaited horses, 118 of 263 (45.0%) were
homozygous AA, 137 (52.0%) heterozygous CA and
8 (3.0%) homozygous CC, while 391 of 404
(96.8%) five-gaited horses were homozygous AA
and 13 were heterozygous CA (3.2%). The 13 five-
gaited horses with the CA genotype had scores from
5.5–7.0 for pace with a mean score of 5.92, compared
with a mean score of 7.30 for horses of the AA
genotype.
The mean, range and variation of six gait traits are
presented in Table 3. The distribution of the traits
walk, trot and canter deviated significantly from zero
measure of skewness and kurtosis.
The DMRT3 genotype had a significant effect on all
gaits except slow t€
olt (Table 4). Scores for walk, trot,
gallop and canter were significantly higher among
horses with the CA genotype compared with AA
horses which had significantly higher scores for t€
olt.
The interaction term between the age–sex classes
and genotype (two classes: AA genotype and CA
genotype) proved to be non-significant for all gaits
except for t€
olt. Stallions aged four and 5 years with
the AA genotype had significantly higher scores for
t€
olt than their contemporaries with the CA genotype.
Mean scores of 4-year-old stallions with the AA and
CA genotype were 8.55 and 7.90, respectively
(p <0.01), and mean scores of 5-year-old AA and CA
stallions were 8.48 and 8.18, respectively (p <0.05).
Moreover, 6-year-old mares with the AA genotype
had significantly higher scores for t€
olt (8.25) than 6-
year-old mares with the CA genotype (7.93)
(p <0.05). Mean scores for t€
olt at first evaluation in
breeding field tests for CA and AA horses were 8.11
(mean age: 5.5 years) and 8.15 (mean age: 5.1 years),
Table 3 The mean, range and variation of six gait traits of the 667
horses included in the data set
Trait Mean SD Min Max Skewness Kurtosis
Walk 7.66 0.73 6.00 9.50 0.25* 0.37
Trot 8.09 0.69 6.00 10.00 0.18 0.39*
Gallop 8.16 0.55 6.50 9.50 0.14 0.07
Canter 7.89 0.60 6.50 10.00 0.29* 0.11
T€
olt 8.36 0.64 6.00 10.00 0.19 0.29
Slow t€
olt 8.05 0.59 6.00 10.00 0.04 0.13
Levels of significance: *p <0.05.
Table 4 Results of analysis of variance for the effect of DMRT3 geno-
type on gait traits (667 horses). Least square means of six gait traits of
homozygous mutant (AA) and heterozygous (CA) horses. The p-values
indicate where there is significant difference between least square
means
Trait Number of AA Number of CA AA CA p-value
Walk 502 143 7.52 7.71 *
Trot 509 150 7.99 8.24 ***
Gallop 509 150 8.08 8.36 ***
Canter 474 119 7.61 8.32 ***
T€
olt 509 150 8.39 8.26 *
Slow t€
olt 488 136 8.01 8.04 NS
Levels of significance: *p <0.05; ***p <0.001.
©2014 Blackwell Verlag GmbH •J. Anim. Breed. Genet. (2014) 1–11 5
T. Kristjansson et al. The effect of a mutation in DMRT3 on gaiting ability
respectively. CA horses were significantly older at first
evaluation than AA horses, while the difference in
mean score for t€
olt was not significantly different. Fur-
ther, it was shown that significantly more horses had
the AA genotype (93%) compared with the CA geno-
type (7%) within a subgroup of 28 horses that had
received 9.0 or higher for t€
olt at the age of 4 years.
The selection procedure of STEPDISC was used to
select the subset of the gaits that best reveals the dif-
ference between the genotype classes. All gaits were
selected in the final model, and multivariate tests
(Wilks’ lambda and Pillai’s trace) indicated highly sig-
nificant (p <0.001) differences between horses with
the AA and C- genotypes. However, pace followed by
canter,t€
olt,gallop and trot had, according to their R
2
and F-values, more discriminant power than walk and
slow t€
olt (results not shown),so the latter were
removed from the final model. In the canonical discri-
minant analysis, the canonical coefficients generated
were significant (p <0.001). The adjusted canonical
correlation between the resulting discriminant func-
tion and the classification variable of the AA or C-
genotype was 0.58. The structure of the discriminant
function is shown in Table 5. The traits with the high-
est absolute canonical coefficients or loadings contrib-
ute the most to the divergence between genotype
classes. Canter had the highest positive coefficient, fol-
lowed by gallop and trot; high scores for these traits
indicated a C- genotype. Pace and t€
olt had negative
coefficients, with pace having a higher loading value;
high scores for these traits indicated an AA genotype.
The mid-point between the group centroid scores,
which may be used as a cutting point to assign a pre-
viously unclassified horse to a genotype group, was
0.52; a horse with a value >0.52 would be classified as
a C- horse and a horse with a value <0.52 would be
classified as an AA horse. The Mahalanobis distance
(D
2
) was 3.23 and showed a significant difference
between the genotype classes, indicating that horses
would be correctly classified in 82% of all cases.
The proportions of the AA and CA genotypes dif-
fered significantly within groups of horses receiving
judges’ comments describing beat in trot, gallop and t€
olt;
suspension in trot and gallop; and speed capacity and
suppleness in t€
olt. Horses with the CA genotype more
often had good suspension in trot and gallop, better beat
in gallop and were less likely to be four-beated in trot
while AA horses were more likely to be supple in t€
olt
and to possess good speed capacity in t€
olt (Table 6).
Change in allele frequency over time
The results of the segregation analyses with the two
different prior allele frequencies (p(C) =0.13 and p
(C) =0.30) were compared in terms of information
content (Table 7). The true allele frequency for the C-
allele is probably closer to 0.3 in the founder popula-
tion (see Figure 1). The frequency of accurately esti-
mated genotypes was slightly higher, and therefore,
only the results from the analysis with p(C) =0.3 will
be presented and discussed further.
The increase in exactly evaluated genotypes in the
segregation analysis compared with the G-E proce-
dure is shown in Table 8. The data included eight
Table 5 Total canonical structure of the discriminating function sepa-
rating horses of the AA and CA genotypes
Trait Coefficient
Pace 0.79
Canter 0.59
Gallop 0.40
Trot 0.27
T€
olt 0.11
F-value 62.72
p-value ***
Levels of significance: ***p <0.001.
Table 6 The proportion of genotypes within groups of horses receiving
certain judges’ comments for the gaits. The p-values indicate where the
proportions deviate significantly from the expected proportions of 0.76
for the homozygous mutant genotype (AA) and 0.24 for the heterozy-
gous genotype (CA) according to a chi-square test with 1 degree of free-
dom
Trait Number AA C/- p-value
Walk
Clear beat 132 0.73 0.27 NS
Trot
Good speed capacity 123 0.79 0.21 NS
Lack of speed capacity 88 0.83 0.17 NS
Clear beat 131 0.74 0.26 NS
Four-beated 64 0.91 0.09 *
Good suspension 92 0.45 0.56 ***
Gallop
Good speed capacity 166 0.79 0.21 NS
Lack of speed capacity 35 0.83 0.17 NS
Clear beat 65 0.6 0.4 *
Good suspension 77 0.42 0.58 ***
Lack of suspension 71 0.96 0.04 ***
T€
olt
Good speed capacity 222 0.83 0.17 *
Lack of speed capacity 38 0.53 0.47 ***
Clear beat 257 0.78 0.22 NS
Trotty beat 27 0.19 0.81 ***
Supple 99 0.87 0.13 *
Stiff 33 0.82 0.18 NS
Levels of significance: *p <0.05; ***p <0.001.
©2014 Blackwell Verlag GmbH •J. Anim. Breed. Genet. (2014) 1–116
The effect of a mutation in DMRT3 on gaiting ability T. Kristjansson et al.
genotyped individuals with the CC genotype, and
only one horse with exactly confirmed CC genotype
was additionally revealed in the segregation analysis.
Then, 24 horses were found to have estimated geno-
type probability >0.85 for the CC genotype.
The trend from 1980 to 2012 in the frequency of
the alleles A and C was estimated in the whole popu-
lation through calculation of genotype probabilities.
The frequency of the A-allele was estimated to be
0.72 in 1980 and 0.94 in 2012 (Figure 1).
The chi-square test was used to evaluate whether
the genotypes reflected Hardy–Weinberg proportions.
This was performed for each birth year from 1980 to
2012, and the chi-square value regressed on birth
year (Figure 2). For the years 1980–1993, the geno-
types were not in Hardy–Weinberg equilibrium as the
values were above 3.84, which is the 0.05 significance
level for one df. In this period, the proportion of the
CA genotype was higher than expected and the pro-
portion of the homozygotes was subsequently lower.
In the years 1994–2012, the genotypes were esti-
mated to be in Hardy–Weinberg equilibrium in the
population. The results showed that the proportion of
the genotypes in the selected material of 667 horses
conformed to the Hardy–Weinberg proportions. These
findings therefore agree well with the fact that the
majority of horses in the selected material are born in
1997–2005.
Discussion
The population of the multigaited Icelandic horse
allows for detailed estimation of the effects of the
DMRT3 nonsense mutation (Ser301STOP) on gaiting
ability. The assessment of the different gaits is system-
atic and standardized (FIZO 2012), and the population
is not fixed for the mutation (Andersson et al. 2012).
In this study, the effect of the DMRT3 nonsense muta-
tion on the gaiting ability of the Icelandic horse was
estimated using more detailed information and a lar-
ger sample of assessed breeding horses than previ-
ously published (Andersson et al. 2012). The horses in
the data set were selected with regard to scores and
judges’ comments referring to the individual gaits to
include as detailed information about both gait qual-
ity and speed capacity as possible. The mean scores for
Table 7 Illustration of information content in the studied data for esti-
mation of genotype probabilities. The genotype probabilities are
derived from segregation analyses with two different prior allele fre-
quencies (p(C) =0.13 and p(C) =0.30)
GPI
P(C) =0.13 P(C) =0.30
N%N%
100 75 059 18.29 75 232 18.34
90 84 214 20.53 85 106 20.74
80 94 623 23.06 98 536 24.02
70 109 773 26.76 118 732 28.94
60 119 944 29.23 135 489 33.38
50 134 662 32.82 161 567 39.38
40 183 269 44.67 199 211 48.55
30 254 340 61.99 226 410 55.18
GPI, Genotype Probability Index.
Figure 1 Frequencies of the A- and C-alleles in DMRT3 in the Icelandic
horse population from 1980–2012; the red line refers to the A-allele,
and the green line refers to the C-allele.
Table 8 Number of exactly estimated genotypes in the segregation
analysis when prior p(C) =0.3
Genotypes N %
Increment compared
with G–E updated scores
AA 67 273 16.40 254
CA 7950 1.94 40
CC 9 –1
Figure 2 Development of chi-square values over time, testing Hardy–
Weinberg equilibrium of DMRT3 genotypes in the Icelandic horse popu-
lation, with indicated 0.05 significance level for one degree of freedom
(green line).
©2014 Blackwell Verlag GmbH •J. Anim. Breed. Genet. (2014) 1–11 7
T. Kristjansson et al. The effect of a mutation in DMRT3 on gaiting ability
the gaits (Table 3) were, however, higher in the
selected material than in all Icelandic breeding horses
presented for breeding assessment in a similar period
(Albertsd
ottir et al. 2008), except for walk and pace.
No single sire is believed to have a great impact on the
results as the average number of offspring per sire is
low and it is assumed that the data set reflects the esti-
mated proportion of genotypes in the population. The
rider has been shown to have a significant effect on
gait quality (Albertsd
ottir et al. 2007). The rider effect
was, however, not included in the model where the
genotype effect on the gaits was estimated because of
obvious risk of confounding effects of the rider and
genotype, as 74% of riders rode only one or two
horses. Kerr and Kinghorn’s method (1996) of calcu-
lating genotype probabilities facilitated the estimation
of the development in the frequency of the A- and C-
alleles, which shows how breeding decisions have
shaped the distribution of DMRT3 genotypes over
time.
Effect of the DMRT3 nonsense mutation on gait traits
This study confirmed favourable effects of the DMRT3
nonsense mutation on the lateral gaits t€
olt and pace.
Almost all horses with a pace score of 5.5 or higher
were homozygous for the DMRT3 nonsense mutation,
confirming that the AA genotype is a prerequisite for
the ability to pace. The AA genotype is, however, not
sufficient for the ability to perform pace as 45% of
horses classified as four-gaited were homozygous
mutants. This high proportion of AA horses presented
as four-gaited (without pace) could thus be influenced
by other genetic and environmental factors. Present-
ing all gaits at breeding field tests will give the possi-
bility of highest total score and is therefore the main
goal. However, the score for t€
olt is the most valuable
trait for the marketing price of the horse, so t€
olt
has the highest weight in the total score. Presenting
four gaits is an alternative, preferable for horses that
do not have outstanding performance in pace. Train-
ing of t€
olt receives the highest priority and it is well
known that pace training can in some instances
impair the t€
olt quality. Therefore, many horses are rid-
den as four-gaiters even if they could perform pace up
to a certain level (
Arnason & Sigurdsson 2004). The
presence of few CA horses receiving a score of 5.5 or
higher for pace (3.2% of the five-gaited horses), which
all received scores below average for pace, is most
likely a phenotypic misclassification or in some
instances presumably resulting from training, confor-
mation or other factors that can facilitate CA horses to
perform low quality pace. T€
olt and pace are very
similar gaits, both being in fact four-beat, lateral gaits
(Wilson et al. 1998), and phenotypic misclassification
is therefore not unexpected. The main features that
separate them is the shorter time between ground
contact of lateral legs in pace than t€
olt and a moment
of suspension in pace, which should be non-existing in
t€
olt (Zips et al. 2001).
The results clearly showed a positive effect of the
AA genotype on the t€
olt ability. This seems to depend
both on superior speed capacity and suppleness of the
AA horses compared with the CA horses. Speed
capacity and suppleness greatly impact the scoring for
t€
olt (FIZO 2012). The significant interaction between
genotype and the age–sex classes in the analysis of
variance indicates that AA horses have more natural
ability to t€
olt. CA horses were also significantly older
when presented at breeding field tests for the first
time. This could indicate that they need longer train-
ing than AA horses to develop an acceptable t€
olt
capacity, as the quality of t€
olt is one of the main crite-
ria for the preselection of horses to the breeding field
tests (Albertsd
ottir et al. 2011). Moreover, AA horses
are overrepresented in the group of 28 horses in the
data set that had received a score of 9.0 or 9.5 for t€
olt
at the age of four. Heterozygous horses had signifi-
cantly higher scores for the basic gaits walk, trot, gallop
and canter. A previous study (Andersson et al. 2012)
has shown that Icelandic horses with the CA genotype
had significantly higher scores for trot compared with
homozygous mutant horses. This was confirmed in
the current study and further related to correct beat
and suspension. It was also revealed that CA horses
had significantly higher scores for gallop and canter
compared with AA horses, possessing more often cor-
rect beat and suspension in canter/gallop. Correct beat
in canter (a pure three-beat) depends on synchronized
movement of diagonal legs in much the same way as
in trot (Clayton 2004). These results indicate that the
AA genotype reinforces the coordination of ipsilateral
legs, with the subsequent negative effect on the syn-
chronized movement of diagonal legs. This agrees well
with previous suggestions that DMRT3 neurons play a
critical role in left/right coordination, as well as in
coordinating the movement of fore- and hind limbs
(Andersson et al. 2012). The negative effect of the AA
genotype on beat and suspension in trot as well as in
canter/gallop has probably the same cause, suggesting a
negative effect of the AA genotype on the synchro-
nized movements of diagonal legs. The genotype
effect on scores for canter is strong (Table 4), but high
scores for canter demand correct beat and suspension
(FIZO 2012). This is further supported by the higher
proportion of horses with the CA genotype among
©2014 Blackwell Verlag GmbH •J. Anim. Breed. Genet. (2014) 1–118
The effect of a mutation in DMRT3 on gaiting ability T. Kristjansson et al.
horses that received the judges’ comment trotty t€
olt,
which involves too much association of diagonal legs
in t€
olt.
Standardbred trotters with the AA genotype have
been reported to have significantly higher breeding
values for racing performance compared with the CA
genotype (Andersson et al. 2012). It was suggested
that the AA genotype promotes speed capacity at trot.
This was not supported in the current study probably
because riders at breeding field tests for Icelandic
horses are not always riding them to their limit in
speed in trot to maintain correct beat, as correct beat
counts more than high speed in the scoring for trot
(FIZO 2012). It has been suggested that the transition
from trot to gallop is triggered when musculoskeletal
forces reach a critical level and that peak forces are
reduced at a certain speed by the transition from trot
to gallop, as gallop is a more compliant gait with the
sequential ground contact of limbs (Farley & Taylor
1991). This critical level could be avoided at high
speed in trot by dissociating diagonal legs (become
four-beated) and therefore placing the legs more
sequentially on the ground. It has, indeed, been
shown that the magnitude of diagonal dissociation
increases with speed (Drevemo et al. 1980). Therefore,
the superiority of AA horses in trot racing could be
explained by their ‘ability’ to be four-beated in trot as
shown in the current study. That could be an advan-
tage when high speed in trot is required but a disad-
vantage when qualities such as correct beat and
suspension are required.
The results of the canonical discriminant analysis
supported the findings of the analysis of variance for
the effect of the DMRT3 genotypes on the gaits. Based
on scores for trot,canter,gallop,t€
olt and pace, it was
possible to discriminate between AA and CA horses
with high confidence. Pace had the highest negative
loading and, along with a high score for t€
olt, suggested
an AA genotype. High scores for trot canter and gallop
suggested a CA genotype, with canter having the
greatest discriminating power of the basic gaits.
Change in allele frequency over time
The change in the frequency of the A- and C-alleles
indicated a selection in favour of the A-allele in the
Icelandic horse population over the last decades. This
result must be interpreted in the light of the effect of
the AA genotype on t€
olt and its crucial role for the
ability to pace. Since the definition of the official
breeding objective for the Icelandic horse in 1950 and
until now, an excellent five-gaited horse has been the
main aim (Hugason 1994). This entailed a heavy
emphasis for many years on the selection of Icelandic
breeding horses with good capacity for both t€
olt and
pace. Breeding value estimations for Icelandic horses
are based on breeding field test scores, where t€
olt has
the highest weight and pace a relatively high weight.
In addition, both traits have high genetic variation
compared with other assessed traits, especially pace
(Albertsd
ottir et al. 2008). The quality of horses with
respect to these traits therefore greatly impacts their
ranking, and breeders have to a greater extent based
their selection on breeding values since 1984 (
Arnason
& Van Vleck 2001). The observed trend in the geno-
type frequencies implies that the C-allele may be lost
in the Icelandic horse population around year 2030.
In the light of the favourable effect the C-allele seems
to have on the basic gaits, the breeding goal for
the Icelandic horse should perhaps be redefined.
Competitions for four-gaited horses, where the basic
gaits and t€
olt have equal weights, have become more
popular resulting in high market value of high-class
four-gaited horses (Albertsd
ottir et al. 2007). The cur-
rent study indicates how probability calculations can
be used to estimate the genotype of an individual using
genotype and phenotypic information combined with
prior knowledge about genotype effect. The results
demonstrate the strength of genomic methods in
monitoring the effect of breeding decisions on genetic
variability.
In the years 1980–1993, the DMRT3 genotypes were
not in Hardy–Weinberg equilibrium. The observation
of a higher proportion of CA horses than expected
from the gene frequencies suggests compensatory
mating, where the mating of two four-gaited horses
has probably been avoided over these years. The rea-
son for possible reduction in compensatory mating
according to DMRT3 genotypes and a consequent
Hardy–Weinberg equilibrium in the population in the
years 1994–2012 can probably be explained by
decreased selection intensity on the mares’ side due to
expansion in population size (Sigurdard
ottir 2011),
combined with growing popularity of four-gaited
horses among which the AA genotype is more com-
mon.
Conclusions
Homozygosity for the DMRT3 nonsense mutation is
permissive for pace and has a major effect on the qual-
ity of t€
olt,trot and canter/gallop, and speed capacity in
t€
olt. Selective breeding for lateral gaits in the Icelandic
horse population has altered the frequency of DMRT3
genotypes with a predicted loss of the C-allele in
relatively few years. The results have practical
©2014 Blackwell Verlag GmbH •J. Anim. Breed. Genet. (2014) 1–11 9
T. Kristjansson et al. The effect of a mutation in DMRT3 on gaiting ability
implications for breeding and training of Icelandic
horses and other gaited horse breeds.
Acknowledgements
This work is supported by a funding grant from the
Foundation for the Preservation of the Icelandic
Horse. It was also partially supported by funds from
The Swedish Research Council Formas and the Swed-
ish Research Council. The authors would like to
acknowledge J
on H. Hallsson and
Aslaug Helgad
ottir
for helpful comments on the text and Vilhj
almur
Svansson for contribution of DNA samples.
Competing interests
Lisa S. Andersson and Gabriella Lindgren are co-
inventors on a patent application concerning com-
mercial testing of the DMRT3 mutation. Other authors
do not have any actual or potential competing inter-
ests.
References
Albertsd
ottir E., Eriksson S., N€
asholm A., Strandberg E.,
Arnason T. (2007) Genetic analysis of competition data
on Icelandic horses. Livest. Sci.,110, 242–250.
Albertsd
ottir E., Eriksson S., N€
asholm A., Strandberg E.,
Arnason T. (2008) Genetic correlations between compe-
tition traits and traits scored at breeding field-tests in Ice-
landic horses. Livest. Sci.,114, 181–187.
Albertsd
ottir E., Eriksson S., Sigurdsson
A.,
Arnason T.
(2011) Genetic analysis of ‘breeding field test status’ in
Icelandic horses. J. Anim. Breed. Genet.,128, 124–132.
Alexander R.M. (1988) Why mammals gallop. Am. Zool.,
28, 237–245.
Andersson S.L., Larhammar M., Memic F., Wootz H.,
Schwochow D., Rubin C.J., Patra K.,
Arnason T., Well-
bring L., Hj€
alm G., Imsland F., Petersen J.L., McCue
M.E., Mickelson J.R., Cothran G., Ahituv N., Roepstorff
L., Mikko S., Vallstedt A., Lindgren G., Andersson L.,
Kullander K. (2012) Mutations in DMRT3 affect locomo-
tion in horses and spinal circuit function in mice. Nature,
488, 642–646.
van Arendonk J.A.M., Smith C., Kennedy B.W. (1989)
Method to estimate genotype probabilities at individual
loci in farm livestock. Theor. Appl. Genet.,78, 735–740.
Arnason T., Sigurdsson
A. (2004) International genetic
evaluations of the Icelandic horse. In: I. der van Honing
(ed), Book of Abstracts (No.10) of the 55th Annual
Meeting of the European Association for Animal Produc-
tion. Wageningen Academic Publishers, Bled, Slovenia,
pp. 327.
Arnason T., Van Vleck L.D. (2001) Genetic improvement
of the horse. In: A.T. Bowling, A. Ruvinsky (eds), The
Genetics of the Horse. CABI publishing, Massachusetts,
pp. 473–497.
Clayton H.M. (2004) The dynamic Horse. Sport Horse Pub-
lications, Michigan, USA.
Drevemo S., Fredricson I., Dalin G., Bjorne K. (1980)
Equine locomotion: 2. The analysis of coordination
between limbs of trotting standardbreds. Equine Vet. J.,
12,66–70.
Farley C.T., Taylor C.R. (1991) A mechanical trigger for
the trot–gallop transition in horses. Science,253, 306–
308.
Fernando R.L., Stricker C., Elston R.C. (1993) An efficient
algorithm to compute the posterior genotypic distribu-
tion for every member of a pedigree without loops.
Theor. Appl. Genet.,88, 573–580.
FIZO. (2012) English FEIF Rules for Icelandic Horse
Breeding. (available at: http://www.feiffengur.com/doc-
uments/fizo12e.pdf. last accessed 01 February 2014).
Hugason K. (1994) Breeding of Icelandic toelter horses: an
overview. Livest. Prod. Sci.,40,21–29.
Janss L.L.G., van der Werf J.H.J., van Arendonck J.A.M.
(1995) Computing approximate monogenetic model
likelihoods in large pedigree with loops. Genet. Sel. Evol.,
27, 567–579.
Kerr R.J., Kinghorn B.P. (1996) An efficient algorithm for
segregation analysis in large populations. J. Anim. Breed.
Genet.,113, 457–469.
Kinghorn B.P. (1997) An index of information content for
genotype probabilities derived from segregation analysis.
Genetics,145, 479–483.
Lange K. (1997) Mathematical and Statistical Meth-
ods for Genetic Analysis. Springer-Verlag, New York,
USA.
MacCluer J.W., Boyce A.J., Dyke B., Weitkamp L.R.,
Pfennig D.W., Parson C.J. (1983) Inbreeding and pedi-
gree structure in Standardbred horses. J. Hered.,74,
394–399.
SAS Institute Inc. (2009) SAS/STAT 9.2 User’s Guide, 2nd
edn. SAS Institute Inc, Cary, NC.
Sigurdard
ottir H. (2011) Athugun
a
urvalsstyrkleika og
erfðaframf€
orum
ı einst€
okum eiginleikum
ı
ıslenskri hros-
sarækt [Investigation of selection intensity and genetic
progress within breeding field test traits in Icelandic
horse breeding]. BSc dissertation, Agricultural Univer-
sity of Iceland, Department of Land and Animal
resources.
Wilson B.D., Neal R.J., Howard A., Groenendyk S. (1998)
The gait of pacers 2: factors influencing pacing speed.
Equine Vet. J.,20, 347–351.
Zips S., Peham C., Scheidl M., Licka T., Girtler D. (2001)
Motion pattern of the toelt of Icelandic horses at differ-
ent speeds. Equine Vet. J. Suppl.,33, 109–111.
©2014 Blackwell Verlag GmbH •J. Anim. Breed. Genet. (2014) 1–1110
The effect of a mutation in DMRT3 on gaiting ability T. Kristjansson et al.
Supporting Information
Additional Supporting Information may be found in
the online version of this article:
Table S1. Attritbutes assessed in the scoring of gait
traits in breeding field tests for Icelandic horses and
their weight in the total score.
©2014 Blackwell Verlag GmbH •J. Anim. Breed. Genet. (2014) 1–11 11
T. Kristjansson et al. The effect of a mutation in DMRT3 on gaiting ability
