Chromosomal Regions Underlying Noncoagulation of

Copyright Ó 2008 by the Genetics Society of America
DOI: 10.1534/genetics.107.083964
Chromosomal Regions Underlying Noncoagulation of Milk in Finnish
Ayrshire Cows
Anna-Maria Tyrisevä,1 Kari Elo, Arja Kuusipuro, Veijo Vilva, Isto Jänönen, Heidi Karjalainen,
Tiina Ikonen and Matti Ojala
Department of Animal Science, University of Helsinki, FI-00014 Helsinki, Finland
Manuscript received October 31, 2007
Accepted for publication August 5, 2008
ABSTRACT
About 10% of Finnish Ayrshire cows produce noncoagulating milk, i.e., milk that does not form a curd
in a standard 30-min testing time and is thus a poor raw material for cheese dairies. This phenomenon is
associated with peak and midlactation, but some cows produce noncoagulating milk persistently. A
genomewide scan under a selective DNA pooling method was carried out to locate genomic regions
associated with the noncoagulation of milk. On the basis of the hypothesis of the same historical
mutation, we pooled the data across sires. Before testing pools for homogeneity, allele intensities were
corrected for PCR artifacts, i.e., shadow bands and differential amplification. Results indicating association
were verified using daughter design and selective genotyping within families. Data consisted of 18 sire
families with 477 genotyped daughters in total, i.e., 12% of each tail of the milk coagulation ability. Data
were analyzed using interval mapping under maximum-likelihood and nonparametric methods. BMS1126
on chromosome 2 and BMS1355 on chromosome 18 were associated with noncoagulation of milk across
families on an experimentwise 0.1% significance level. By scanning gene databases, we found two
potential candidate genes: LOC538897, a nonspecific serine/threonine kinase on chromosome 2, and
SIAT4B, a sialyltransferase catalyzing the last step of glycosylation of k-casein on chromosome 18. Further
studies to determine the role of the candidates in the noncoagulation of milk are clearly needed.
M
ILK coagulation ability of dairy cows has been an
extensively studied trait because of its association with cheese yield (e.g., Martin et al. 1997; Ikonen
et al. 1999b; Johnson et al. 2001; Malacarne et al.
2006). It is a heritable, quantitative trait; 40% of the
variation among animals is caused by genetic factors
(Ikonen et al. 2004). Because a major part of the milk
currently produced is used for cheese production
(Eurostat 2007), selection of breeding animals for
milk coagulation ability is a tempting choice to produce
high-quality milk for cheese dairies. The best selection
option is still under research (Ikonen 2000; Tyriseväet al.
2003, 2004; Ikonen et al. 2004; Ojala et al. 2005) since due
to a lack of high throughput, automated measuring device
possibilities for direct selection are limited.
In the course of the above-mentioned studies, we
found that noncoagulation of milk is a common problem in Finnish Ayrshire (Fay) cows. About 10% of Fay
cows produce noncoagulating milk (Ikonen et al. 1999a,
2004; Tyrisevä et al. 2004), and some cows produce it
persistently (Tyrisevä et al. 2003). This milk does not
coagulate at all in a standard 30-min testing time. It is
thus a poor raw material for cheese dairies (Ikonen et al.
1
Corresponding author: Department of Animal Science, Koetilantie 5,
P.O. Box 28, University of Helsinki, FI-00014 Helsinki, Finland.
E-mail: [email protected]
Genetics 180: 1211–1220 (October 2008)
1999b) and has even been reported to be lethal to calves
(Johnston and MacLachlan 1977). However, nowadays, calves are normally fed with bulk milk from several
mothers, and the probability of the mixed bulk milk
being noncoagulating (NC) is lower than that of individual milks. Noncoagulation of milk has been observed
in Holstein and Friesian breeds as well (e.g., Okigbo et al.
1985; Van Hooydonk et al. 1986; Malossini et al. 1996;
Kübarsepp et al. 2005), and it seems to be as common a
problem in Italian Holstein–Friesian (Malossini et al.
1996; Cassandro et al. 2008) and Estonian Holstein and
Red-and-White Holstein cows (Kübarsepp et al. 2005) as
in Fay cows.
The ultimate cause of the phenomenon has not yet
been established. Moreover, an exact casein micelle
structure has not been established, rendering the task
more difficult (for review, see Farrell et al. 2006 and
Horne 2006). Some consensus of the micelle structure
does, however, exist: (a) the stabilizing role of hydrophilic k-casein on the micelle surface and (b) Ca-P
bonds between hydrophobic as1 -, as2 -, and b-caseins in
the core of the micelle (Farrell et al. 2006; Horne
2006). How the caseins are bound to each other to form
a three-dimensional structure and whether there are
submicelles or not are still under debate, mostly because
the structure of the micelle cannot be crystallized
(Kumosinski et al. 1991).
1212
A.-M. Tyrisevä et al.
Regardless of the above-mentioned problems, some
differences between NC and normal milks have been
established. On the basis of several studies, calcium
content (colloidal, active, total) is lower in NC milks
than in normal milks (Tervala and Antila 1985; Van
Hooydonk et al. 1986; Resmini et al. 1995; Tsioulpas
et al. 2007), and the micelles seem to be more hydrated
(Resmini et al. 1995). This is likely a direct consequence
of the lower Ca content because it increases the
hydrophilicity of Ca-binding caseins (Resmini et al.
1995). Another observation is that the k-casein content
seems to be lower in NC milk (Ikonen et al. 1999b;
Wedholm et al. 2006), and some indications of higher
pH in NC milk have also been reported (Okigbo et al.
1985; Van Hooydonk et al. 1986). Evidently, something
is wrong with the micelles. However, the first stage of the
coagulation process, the cleavage of the micelle-stabilizing k-casein into two parts, is more or less normal; on the
basis of several studies, NC and normal milk differ only
marginally in glycomacropeptide content (Tervalaand
Antila 1985; Van Hooydonk et al. 1986; Resmini et al.
1995). The glycomacropeptide is a C-terminal end of the
k-casein that goes to whey after cleavage of k-casein.
The noncoagulation of milk is associated with peak
and midlactation (e.g., Tyrisevä et al. 2003; Ikonen et al.
2004), but none of the environmental factors can
unambiguously explain this. Thus, the cause of the
noncoagulation of milk cannot be purely environmental. In the course of our studies, evidence of a genetic
cause emerged. The first indication was the observation
that many of the NC milk-producing cows in one data set
were daughters of two closely related sires (Ikonen et al.
1999a). Stronger evidence turned up as we sampled
5000 cows sired by 91 bulls (Ikonen et al. 2004). Many
bulls had large daughter groups, and we detected clear
differences in the proportion of the NC milk-producing
cows among sires (Figure 1). Further, a moderate
heritability estimate of milk coagulation ability as a
binary coagulate–noncoagulate trait (0.26) hinted at a
genetic cause of the noncoagulation of milk. Due to the
commonness of noncoagulability in Fay cows, the main
dairy breed in Finland, and to NC milk’s poor suitability
for cheese production, we decided to carry out a genomewide scan to locate and ultimately identify the genes
affecting this phenomenon. It would then be possible to
use this information to eliminate the carriers from the
breeding population, thus effectively improving the overall
milk coagulation ability in the Finnish dairy cattle population. The objective of this study was to locate the genomic
regions associated with the noncoagulation of milk.
MATERIALS AND METHODS
Data sets and stages of research: The sire families were a
subsample of the data set by Ikonen et al. (2004). Only bulls
most likely to be heterozygous for the hypothesized NC genes
and with large daughter groups were selected. The classifica-
tion of bulls as heterozygous for NC genes was based on their
daughters’ distribution for milk coagulation ability (Figure 1,
sire B-type bulls; the proportion of NC daughters among bulls
selected for gene mapping ranged from 4 to 19%, mean 11%).
We conducted a genome scan with selective DNA pooling
(Darvasi and Soller 1994) and association analysis methods,
verifying the results by genotyping a larger group of sires and
their daughters, and analyzed the collected genotypes with
linkage analyses (Figure 2).
Genome scan: On the basis of a hypothesis that only a few
genes cause the noncoagulation of milk and that they are same
in all families, we pooled the data across sires. Our hypothesis
relies on three observations from the data. First, the distribution of milk coagulation ability in the whole data set, as
described by Ikonen et al. (2004), is clearly divided into a spike
of noncoagulated samples and a normally distributed part
consisting of coagulated samples. Second, a large variation
existed in the daughters’ distribution for milk coagulation
ability among sires (Figure 1), indicating that the sires could
be classified into homozygous, heterozygous, and noncarriers
for the NC genes. Third, bulls were relatives in many ways since
some widely used, popular bulls occurred repeatedly from
pedigree to pedigree and even within pedigrees.
The pool of NC milk-producing cows consisted of 33
animals, and the pool of excellently (E) coagulating milkproducing cows consisted of 49 animals. Only cows with
negative indexes for milk coagulation ability were accepted
into the NC pool to ensure that the cows were real carriers for
the NC genes. The indexes for the cows were predicted from
the data set for 4664 cows following Ikonen et al. (2004). No
limitation for lactation stage was set for NC cows. Further, only
cows with positive indexes for milk coagulating ability and in
peak or midlactation were accepted into the E pool to
minimize the risk of pool members being carriers for NC
genes. The cows were sired by 17 bulls. The overall mean
number of daughters per sire was 4.8, with means 1.9 and 2.9
daughters per sire in the NC and E pools, respectively.
Verification of genome-scan results: The genomic regions
selected for further analyses were analyzed using daughter
design and selective genotyping (Darvasi and Soller 1992).
The individually genotyped data consisted of 18 sire families
and 477 cows. Data included 188 NC cows with a mean of 10.4
daughters per sire and 289 E cows with a mean of 16.1
daughters per sire. Further, 1561 cows with only phenotypic
information and with a mean of 86.7 daughters per sire were
included in statistical analyses. The genotyped cows represented 12% of each tail of the distribution. Detailed information of the number of cows in the sire families is
presented in Table 1.
DNA samples: Sire DNA was extracted from semen samples
following the protocol by Zadworny and Kuhnlein (1990).
Milk did not prove to be as good a DNA source as we had
expected; moreover, the chloroform-phenol extraction protocol (Lipkin et al. 1993) worked unsatisfactorily. We changed
to a Chelex protocol (Amills et al. 1997), which worked
reasonably well. Due to problems with the DNA extracted from
milk samples, we also obtained blood and hair samples from
the cows still alive. DNA from blood was extracted following a
slightly modified protocol by Miller et al. (1988), and DNA
from hair samples following a Chelex protocol by Walsh et al.
(1991). DNA pools were made only of good-quality blood
DNA. Individual samples for DNA pools were diluted to equal
concentrations and tested in control PCR and electrophoresis
before pooling. Samples running out of DNA during the study
were amplified using a GenomiPhi DNA amplification kit (GE
Healthcare, Chalfont St. Giles, UK).
Markers and genotyping: Microsatellite markers were chosen from the Marc (http://www.marc.usda.gov/genome/
Mapping Noncoagulation of Milk in Cows
1213
Figure 1.—Three sires
differing in milk coagulation
ability on the basis of their
daughters’ distributions. To
some degree, the larger the
value for curd firmness
(millimeters) is, the better
the milk coagulation ability.
A zero refers to noncoagulating samples. The number of
daughters is provided in
parentheses.
cattle/cattle.html) and NCBI databases (http://www.ncbi.nlm
nih.gov/genome/guide/cow/index.html), both for the genome scan of the 29 autosomes and for the next analysis stage
of the selected regions. The density of the map in the genome
scan was on average 15 cM and included 194 markers. At the
next stage, the marker density ranged from 2 to 18 cM, with a
mean of 9.8 cM, and included 47 markers. Three (two in one
case) markers were chosen for each of the selected regions,
except for chromosome 24, which was completely covered.
For further information, see supplemental Tables 1 and 2 at
http://www.helsinki.fi/animalscience/english/supplement.
html.
The pooled samples of the genome scan and the individual
samples of chromosome 24 were amplified with PTC100 PCR
machines (MJ Research, Waltham, MA) and run with a Li-Cor
Gene Reader 4200 DNA analyzer (LI-COR, Lincoln, NE).
Markers were individually amplified in a reaction volume of
10 ml and then multiplexed. A basic PCR protocol included
20 ng of template DNA, 200 mm of each dNTP, 0.16 unit of
Dynazyme II polymerase (Finnzymes, Espoo, Finland), the
buffer provided with the enzyme, and 0.25 mm of each primer.
A marker-specific MgCl2 concentration ranged from 1.5 to
2.25 mm. The amount of IRD700/IRD800 labeled forward
primer, ranging from 5 to 25 nm, was subtracted from the total
amount of forward primer. The basic PCR program started
with 4 min of denaturation at 94°, followed by a cycle of 1 min
at 94°, 1 min at a marker-specific annealing temperature,
1 min at 72° repeated 30 times, 10 min at 72°, and cooling to
4°. The annealing temperature ranged from 50° to 64°. For the
lysated samples, the amount of DNA varied, and the number of
cycles ranged from 30 to 45, depending on the marker.
Marker-specific PCR protocols and programs are available at
http://www.helsinki.fi/animalscience/english/supplement.
html, supplemental Table 1.
The individual samples of the rest of the markers were
amplified with the PTC100 and PTC200 PCR machines (MJ
Research) and analyzed with an ABI Prism 3130 Genetic
Analyzer (Applied Biosystems, Foster City, CA). Markers were
individually amplified in a reaction volume of 5 ml and then
multiplexed. A basic PCR protocol included 10 ng of template
DNA, 200 mm of each dNTP, 0.16 unit of Dynazyme II
polymerase (Finnzymes), the buffer provided with the enzyme, and 0.05–0.2 mm of each primer. A marker-specific
MgCl2 concentration ranged from 1.5 to 2.25 mm. Forward
primer was fluorescently labeled with 6-FAM, NED, VIC, or
PET. As earlier, the amount of DNA varied in the lysated
samples and the number of cycles ranged from 30 to 45,
depending on the marker. Detailed information on markerspecific conditions is available at http://www.helsinki.fi/
animalscience/english/supplement.html, supplemental Table 2.
Two persons analyzed allele sizes and intensities of the pools
and allele sizes of the individual samples run with Li-Cor 4200,
Figure 2.—Study design and structure of the data set.
1214
A.-M. Tyrisevä et al.
TABLE 1
Number of samples and their usability in the study by family
NC samplesa, genotypic and
phenotypic information
E samplesb, genotypic and
phenotypic information
Usablee samples
M samplesc, phenotypic
information only:
Usable samples
Family
nd
Minf
Maxf
Mean
n
Min
Max
Mean
n
37811
38031
38069
38393
38432
38451
38459
38533
38544
38564
38571
38585
38588
38634
38651
38673
38711
38734
Mean
Total
7
15
4
12
9
12
9
5
7
3
7
12
5
15
11
18
2
35
10.4
188
1
2
1
1
1
2
1
1
2
1
1
3
1
3
2
4
1
9
2.1
7
13
4
11
8
10
9
5
7
3
7
11
5
12
11
15
2
35
9.7
3.0
4.3
1.7
4.6
2.8
5.5
3.2
1.7
3.2
1.0
2.8
4.2
2.2
5.7
3.7
6.9
0.9
13.6
4.0
9
20
12
15
13
11
13
7
11
19
10
16
16
19
16
17
27
38
16.1
289
2
5
2
6
2
2
1
1
3
3
1
5
5
6
4
1
6
10
3.6
8
17
12
12
13
10
13
5
10
16
8
14
13
13
15
12
25
27
13.5
3.7
7.2
4.5
6.5
4.6
4.3
4
2.2
5
6.7
3.8
6.5
6.8
6.8
6.1
5.6
12.1
12.6
6.1
64
99
33
104
80
48
63
85
50
44
51
96
85
104
135
122
100
198
86.7
1561
a
Cows producing noncoagulating milk.
Cows producing excellently (E) coagulating milk.
c
Cows producing moderately (M) coagulating milk with only phenotypic records.
d
Number of daughters by family.
e
An informative genotype result on both a sire and a daughter.
f
A minimum (maximum) number of usable samples if a sire was informative.
b
using Gene ImagIR software Version 3.5.6 (Scananalytics,
Fairfax, VA). Allele sizes of the samples run with ABI Prism
3130 were automatically analyzed using GeneMapper Version
4.0 (Applied Biosystems). Two persons checked the correctness of the genotypes.
Phenotypes and background information: Collection and
laboratory analyses of the phenotypic data [milk coagulation
ability, somatic cell score (SCS), and pH of milk] were performed
according to Ikonen et al. (2004). The trait used to describe milk
coagulation ability was curd firmness after a 31-min testing time.
Modeling of the environmental factors associated with the traits
was not possible with the programs used for statistical analyses.
We therefore precorrected curd firmness values for fixed parity,
lactation stage, age of sample, and measuring unit of renneting
device effects as well as random herd effects.
Milk coagulation ability and SCS and pH of milk are
moderately genetically correlated (Ikonen et al. 2004). Further, the milk of cows suffering from mastitis is associated with
a higher pH than that of healthy cows (Bergère and Lenoir
2000). It is thus possible that some genes are common to these
traits, and we decided to analyze SCS and pH of milk as well.
SCS of milk was corrected for fixed parity and lactation stage
effects as well as random herd effects, and pH of milk for fixed
parity, lactation stage, and age of sample effects as well as
random herd effects.
The fixed-effect classifications were according to Ikonen
et al. (2004). Solutions for the fixed and random effects were
obtained from the total 4664-observations data set using a
univariate mixed model with a PEST program (Groeneveld
1990). The variance components for herd effects were
estimated with a VCE4 program, using a restricted maximum-likelihood methodology (Groeneveld 1997).
The distribution of curd firmness was clearly bimodal—even
after correction for environmental effects. To avoid spurious
quantitative trait locus (QTL) effects with the maximumlikelihood method, we tried to normalize it. The transformation was done by balancing between normalization and
minimal information loss. The best transformation proved
to be a square-root transformation of the precorrected curd
firmness values (transformed curd firmness, TCF) (Figure 3).
Before transformation, we turned the distribution around by
subtracting the trait values from the maximum trait value to
enhance the transformation of the more deviating tail.
Statistical analyses of DNA pooling data: Correction for
shadow bands and differential amplification: Two kinds of artifacts
are typical during PCR: shadow bands and differential
amplification of differently sized bands. Due to misfunctioning of the polymerase, bands differing between 3 and 11
dinucleotides from the actual allele size are typical. Differential amplification of the bands is purely a physicochemical
phenomenon. Polymerization of smaller alleles is faster than
that of larger ones, and their quantity is thus larger at the end
of the polymerase reaction. These artifacts cause errors in the
intensities, i.e., allele frequencies of the marker alleles, and are
especially harmful when pooled data are analyzed.
We corrected the pooled allele intensities of the markers for
the shadow bands and differential amplification before the
homogeneity tests. The correction for the shadow bands was
Mapping Noncoagulation of Milk in Cows
Figure 3.—Distribution of the square-root transformed precorrected curd firmness values. Before transformation, the
original distribution was turned around by subtracting the trait
values from the maximum trait value to enhance the transformation of the more deviating tail.
made according to Lipkin et al. (1998). We estimated from
individual data the relative intensities of 3, 2, 1, and 11
shadow bands for each marker using a simple regression
model. We then corrected the allele intensities of each marker
for the shadow bands of the neighboring alleles. Instead of
adding the intensities of the shadow bands to the allele
intensities themselves, we corrected the effect of differential
amplification according to Kirov et al. (2000). Using a
regression model, we estimated the relative allele intensities
for each marker and allele from the data of heterozygous
individuals. The relative intensity (RI) for each allele size n is
RI ¼ a 1 bn 1 e;
where a is the intercept and b is the regression coefficient. The
allele intensities corrected for the differential amplification
(DASCI) were calculated from the allele intensities corrected
for the shadow bands (SCI), using the regression equation
DASCIn01i ¼
a 1 bn0
SCIn01i ;
a 1 bn01i
i $ 0;
where n0 is the allele size of the shortest allele.
The number of individuals used to make the corrections
ranged from 17 to 64, with a mean of 12 usable individuals for
shadow corrections and 7 usable individuals for differential
amplification corrections.
Homogeneity tests: The corrected allele intensities were tested
for homogeneity between the two pools with a CLUMP
program (Sham and Curtis 1995). The program is especially
suitable for large and sparse 2 3 N tables that no longer follow
x2-distribution, as the distribution of the test statistic is
simulated using a Monte Carlo method. Following the
authors’ recommendation, we used two test statistics of the
four presented in the article: T1 and T4. T1 is Pearson’s x2statistic of the original 2 3 N contingency table, whereas all
combinations of alleles are compared against the others in T4,
and the largest test statistic value is selected for the final T4
value (Sham and Curtis 1995). Our selection criterion for the
further analyses of a marker was a 1% risk level for the T1 or
the T4 test statistic.
Statistical analyses of individually genotyped data: A
marker map was constructed with ANIMAP programs (Georges
1215
et al. 1995), slightly modified by V. Vilva to take into account the
situation where a sire is noninformative for a marker (available
at http://www.helsinki.fi/animalscience/english/supplement.
html). According to Broman (2003), large distances among
markers can produce spurious LOD score peaks under maximum-likelihood methodology when spike data are analyzed.
This was actually observed as we conducted preliminary data
analyses. Thus, if more than one region in the chromosome was
selected for further analyses and the regions were .20 cM apart
from each other, they were analyzed as two separate linkage
groups. We used two different interval-mapping methods developed for daughter/granddaughter designs: a standard maximum-likelihood method of the ANIMAP software (Georges
et al. 1995) and a nonparametric method of the HSQM software (Coppieters et al. 1998). The nonparametric method
utilizes ranks of the phenotypes instead of the real phenotypes
(Kruglyak and Lander 1995; Coppieters et al. 1998).
The LOD significance thresholds in maximum-likelihood
analyses were based on a simulation study by Van Ooijen
(1999). Due to the selective genotyping method used, they are
probably too conservative (Manichaikul et al. 2007). The
statistical significance of the test statistic in nonparametric
analyses was based on 10,000 permutations and on a Bonferroni correction for 11 studied chromosomes. We considered
an at least 10% experimentwise significance level to be a
statistically significant result in this study. SCS and pH of milk
were analyzed using the same methods and software as those
for milk coagulation ability.
RESULTS
We discovered several loci affecting either noncoagulation of milk or milk coagulation ability. Some of the
loci found were also associated with pH of milk.
Genome scan: Coverage of the markers throughout
the genome was fairly good. Of 194 pooled markers,
only 10 were discarded due to poor quality. This resulted
in seven gaps that were wider than the intended 15 cM,
ranging from 31.8 to 41.5 cM. Some differences in allele
intensities between pools were detected for 130 of 184
markers. Of these, 120 could be corrected for PCR artifacts, and the remaining 10 were only visually inspected.
In total, 16 markers in 11 chromosomes reached the
#1% risk level (Table 2) and were chosen for further
analyses.
Verification of genome-scan results: The statistically
significant genome-scan results were also significant in
the linkage analyses, in at least one family in each of the
studied chromosomes, but not in each linkage group
(Tables 3 and 4). Thus, the DNA pooling method seemed
to work well. The most significant findings were discovered on chromosomes 2, 18, and 24. On chromosome 2,
the result across families using the nonparametric
method reached the experimentwise 0.1% significance
level near marker BMS1126 (Table 3), but none of the
results in individual families were significant, even
though the maximum was in the same position in many
families and with both methods. Further, the result across
families using the nonparametric method on chromosome 18 reached the experimentwise 0.1% significance
level at marker BMS1355, and the results of the four
1216
A.-M. Tyrisevä et al.
TABLE 2
TABLE 3
Homogeneity test results reaching the maximum 1% risk level
for T1 or T4 test statistics
Interval-mapping results of selected regions using the
nonparametric method
Level of significance in
a
Chr.
2
2
4
5
5
13
13
15
16
18
19
21
21
24
24
27
b
c
Marker
T1 (%)
T4 (%)
BMS1126
BMS1866
BMS648
BMC1009
BMS1658
BM720
BMS2319
JAB1
INRA48
BMS1355
BM17132
BP33
BMS743
BM7151
BMS466
INRA027
0.03
0.00
0.85
0.35
0.92
0.23
0.50
0.14
0.07
0.01
1.06
1.50
0.03
0.11
0.05
0.76
0.11
0.00
0.73
0.44
0.45
1.12
0.82
0.06
0.08
0.03
2.96
0.44
0.01
0.27
1.16
0.46
a
Chromosome.
Pearson’s x2-statistic of the original 2 3 N contingency
table (Sham and Curtis 1995).
c
Comparison of all combinations of alleles against the
others and selection of the largest test statistic value for the
final T4 value (Sham and Curtis 1995).
b
(one) families using the nonparametric (the maximumlikelihood) method were also significant (Tables 3 and
4). On chromosome 24, one family, family 38651,
reached the experimentwise 0.1% significance level
near marker BM7151 with both methods (Tables 3 and
4). The number of usable samples, i.e., informative
genotype results for both sires and daughters, was rather
low (Table 1 and supplemental Figures 1 and 2 at http://
www.helsinki.fi/animalscience/english/supplement.
html), affecting the extent of the findings. Further, because the number of usable daughters was very small in
family 38533 (Table 1, supplemental Figures 1 and 2),
we excluded it from the final results.
On the basis of maximum-likelihood analyses, the loci
on chromosomes 18 and 24 were associated with noncoagulation of milk (Table 4). However, on the basis of
the extent of the results, chromosomes 2 and 18 were
the most promising candidate chromosomes since the
results across families were highly significant (Table 3).
Further, on the basis of inspection of the distribution of
sire alleles to NC and E daughters, marker BMS1355 on
chromosome 18 was the most promising region (Table
5). We decided to scan chromosomes 2, 18, and 24 for
candidate genes using the NCBI database (http://www.
ncbi.nlm.nih.gov/genome/guide/cow/index.html).
Searching was done in the regions encompassing the
nearest markers (or the beginning of chromosome 18)
around markers BMS1126, BMS1355, and BM7151. We
Family
Chr.a
Trait
log(P)
Max position
Acrossb
38571
38585
38651
38031
Across
38451
38585
38673
38711
38585
38031
38432
38634
38651
Across
38585
38634
38031
38651
38585
38588
38651
2
4
4
4
5a
5a
5b
13b
15
15
16
18
18
18
18
18
19
19
21a
24
24
27
27
TCFc
TCF
TCF
TCF
TCF
TCF
TCF
TCF
TCF
TCF
TCF
TCF
TCF
TCF
TCF
TCF
TCF
TCF
TCF
TCF
pHd
TCF
pH
4.3
3
2.4
2.7
2.2
2.8
2.0
3.2
2.1
2.6
2.4
2.1
3.7
3.4
4.3
4.3
2.2
3.2
3.0
4.3
2.4
2.9
3.7
BMS1126
BMS2809-BMS648
BMS648
BMS648-BR6303
DIK2718
DIK2718
DIK4070-DIK2281
BMS2319-DIK5243
JAB1
JAB1-BB1539
IDVGA49
MNB_74
BMS1355
BMS1355-MNB_74
BMS1355-MNB_74
BMS1355
BM17132
BM17132-STAT5B
BP33
BM7151
AGLA269
DIK5134
DIK5134
log(P) 2.0 corresponds to the experimentwise P ¼ 0.10
value. log(P) 2.3 corresponds to the experimentwise P ¼
0.05 value. log(P) 3.0 corresponds to the experimentwise
P ¼ 0.01 value. log(P) 4.0 corresponds to the experimentwise P ¼ 0.001 value.
a
Linkage group.
b
Across families.
c
Transformed curd firmness.
d
pH of milk.
discovered two potential candidate genes: LOC538897
on chromosome 2 and SIAT4B on chromosome 18.
LOC538897, located 1.4 Mbp downstream from
BMS1126, is a protein kinase, catalyzing the phosphorylation of amino acids, and SIAT4B, located 1.2 Mbp
downstream from BMS1355, is a sialyltransferase, catalyzing the glycosylation of amino acids.
As we studied SCS and pH of milk, we found
indications on chromosomes 4 and 27 that the same
regions associated with milk coagulation ability were
also associated with pH of milk (Table 3), although the
results on chromosome 4 did not quite reach the
experimentwise 10% risk level. Further, a QTL associated with pH of milk was segregating in one family on
chromosome 24 (Table 3). The maximum was located
near marker AGLA269, 20 cM from marker BM7151,
which was associated with milk coagulation ability. We
also found evidence of a locus affecting SCS of milk in
one family in linkage group 13a (Table 4). The region
was not, however, associated with milk coagulation
ability in any family.
Mapping Noncoagulation of Milk in Cows
DISCUSSION
Milk coagulation ability as a study trait: Milk coagulation ability is a quantitative trait, and thus, many
genes affect it, some with larger effects than others. On
the basis of our understanding, some alleles of these
genes can even have dramatic effects, causing noncoagulation of milk. This inevitably affects the structure
of the collected data set. NC cows may be a very
heterogeneous group for the genes affecting milk
coagulation ability (MCA) in general, whereas E cows
likely carry predominantly good genes for MCA. Thus,
even if the main aim was to look for NC genes, some of
the genomic regions found are likely those affecting
milk coagulation ability, not noncoagulability.
Structure of the data set and statistical methods
used: Due to the small number of useful daughters in
the sire families, the results were statistically significant
in only a few individual families. However, when the
results of individual families were summed, they were
highly significant on chromosomes 2 and 18. Inclusion
of animals with only phenotypic information increased
the statistical power as well, resulting in a strengthening
of the result for family 38651 on chromosome 24
relative to the earlier result based on only genotyped
animals (Elo et al. 2007).
The nonparametric and the maximum-likelihood
methods gave partially different results. Considerably
more loci reached statistical significance with the nonparametric method than with the maximum-likelihood
method. In several cases, the maximum was located in
the same position with both methods but was not
statistically significant using the maximum-likelihood
method (or vice versa). The structure of the data gives a
possible explanation for these differences. Despite data
transformation, the distribution of the transformed
curd firmness was still nonnormal (Figure 3). The
extreme phenotypes strongly affect the residual variance, decreasing the ratio of the QTL variance to the
residual variance, and thus, decreasing the power
1217
(Kruglyak and Lander 1995; Coppieters et al. 1998).
The replacement of the phenotypes with the ranks of
the phenotypes in the nonparametric method decreases
the effect of the extreme phenotypes and increases the
power. The advantage of the maximum-likelihood
method is that an estimate of the QTL effect is obtained
as well. All of the loci found with the maximum-likelihood method had a major effect, almost two phenotypic
standard deviations (Table 4). Due to the limited size
of the data, finding loci with small effects it is not even
possible. It is, however, likely that the effects are overestimated because of the low power (Georges et al. 1995).
Casein genes and results of the genome scan:
Suggestions have been made that the A- or E-alleles of
k-casein are associated with the noncoagulation of milk
(e.g., Van Hooydonk et al. 1986; Kübarsepp et al. 2005;
Wedholm et al. 2006). However, the results of Van
Hooydonk et al. (1986) were based on 23 cows only, and
those of Kübarsepp et al. (2005) (n ¼ 87 cows, four
breeds) and Wedholm et al. (2006) (n ¼ 134 cows, three
breeds) were confounded with breed effects. According
to our earlier studies (Ikonen et al. 1999a; Tyriseväet al.
2003) and a more recent long-term sampling in the
University of Helsinki experimental herd, no major
differences in k-casein allele frequencies were detected
among NC and E cows. For example, the allele frequencies of the extreme cows for milk coagulation ability in
the latest sampling were A-allele 46.9%, B-allele 15.6%,
and E-allele 37.5% for 16 NC cows and A-allele 45%, Ballele 22.5%, and E-allele 32.5% for the 20 E cows.
Further, we found no association with the casein superloci in this study. It is thus unlikely that the k-casein A- or
E-alleles would be the ultimate cause of the noncoagulation of milk.
Since the casein genes themselves seem not to be the
cause of the noncoagulability, the next obvious hypothesis is the occurrence of an unfavorable mutation in the
gene/genes responsible for the post-translational modification (phosphorylation and glycosylation) of the
TABLE 4
Interval mapping results of selected regions using the maximum-likelihood method
Family
Chr.a
Trait
LOD
Max position
QTL effect
SE of the
QTL effect
QTL effect as a
phenotypic SD
Type of allele
effect
38069
38451
38588
38651
38651
13a
16
16
18
24
SCSb
TCFc
TCF
TCF
TCF
2.7
4.5
3.3
2.7
4.2
ACC24
INRA48-MB103
INRA48-MB103
BMS1355
BM7151
1.37
2.70
2.52
2.62
2.65
0.6
0.4
0.7
0.7
0.3
1.8
1.9
1.7
1.9
1.8
—
MCAd
MCA
NCe
NC
LOD 2.3 corresponds to the experimentwise P ¼ 0.10 value. LOD 2.7 corresponds to the experimentwise P ¼ 0.05 value. LOD
3.4 corresponds to the experimentwise P ¼ 0.01 value. LOD 4.4 corresponds to the experimentwise P ¼ 0.001 value.
a
Linkage group.
b
Somatic cell score of milk.
c
Transformed curd firmness.
d
Milk coagulation ability.
e
Noncoagulation of milk.
1218
A.-M. Tyrisevä et al.
TABLE 5
Distributions of the BMS1355 sire alleles to the NC and E daughters within sires
NC
First allele
E
Second allele
Total:
First allele
Second allele
Total:
Family
n
%
n
%
n
n
%
n
%
n
37811
38031
38069
38393
38432
38451
38544
38564
38571
38585
38588
38634
38651
38673
38711
38734
3
2
4
3
0
6
3
1
0
4
3
5
6
4
0
9
100
28.6
100
60
0
85.7
50
100
0
100
75
71.4
100
36.4
0
56.3
0
5
0
2
3
1
3
0
3
0
1
2
0
7
1
7
0
71.4
0
40
100
14.3
50
0
100
0
25
28.6
0
63.6
100
43.8
3
7
4
5
3
7
6
1
3
4
4
7
6
11
1
16
2
10
4
3
5
4
5
5
1
3
1
1
3
2
10
8
50
71.4
50
42.9
83.3
100
55.6
38.5
16.7
33.3
11.1
12.5
33.3
25
52.6
57.1
2
4
4
4
1
0
4
8
5
6
8
7
6
6
9
6
50
28.6
50
57.1
16.7
0
44.4
61.5
83.3
66.7
88.9
87.5
66.7
75
47.4
42.9
4
14
8
7
6
4
9
13
6
9
9
8
9
8
19
14
Only informative sires and daughters were included.
casein genes or in the genes controlling their activity.
The post-translational modification of the caseins to a
large extent influences their ability to bond calcium and
their degree of hydrophilicity, both of which further
affect the micelle structure (e.g., Horne 1998; Farrell
et al. 2006). We found two candidate genes as we studied
regions around the maximum log(P) values on chromosomes 2, 18, and 24: LOC538897 on chromosome 2
and SIAT4B on chromosome 18. Both are enzymes
catalyzing the post-translational modification of the
amino acids. LOC538897 is a predicted gene that functions as a nonspecific serine/threonine kinase (Ensembl
database at http://apr2007.archive.ensembl.org/Bos_
taurus/index.html). Golgi casein kinase, responsible
for the phosphorylation of caseins, has not yet been
molecularly characterized, nor has it been located in the
bovine genome. According to Tibaldi et al. (2006), the
Golgi casein kinase recognizes the Ser-x-Glu/pSer sequence and accounts for all casein kinase activity of the
Golgi apparatus with nonspecific kinase activity. It is thus
possible that the novel gene is the Golgi casein kinase.
The second candidate gene, SIAT4B, is a sialyltransferase that catalyzes the addition of N-acetylneuraminic
acid (NeuAc) to galactose (Gal) (Ensembl database and
KEGG database at http://www.genome.jp/kegg/kegg2.
html). k-Casein is the only glycosylated casein and its
major glycoform is a branched tetrasaccharide: NeuAca(2-3)Galb(1-3)[NeuAca(2-6)]GalNAc, attached to
the threonine (Holland et al. 2006). Thus, sialyltransferase is the enzyme catalyzing the last step in the glycosylation of k-caseins. Both of these genes are strong
candidates, but their roles in the noncoagulation of milk
must still be verified.
The power in the analyses of milk SCS and pH was
lower than that of milk coagulation ability because the
animals were selected on the basis of the latter. However,
some loci associated with these traits were detected. We
found indications that the same regions on chromosomes 4 and 27 were associated with both milk coagulation ability and pH of milk. The maxima of milk
coagulation ability and pH on chromosome 24 were 20
cM apart. However, because of the long confidence
interval in the study, the same gene plausibly could be
associated with both of them. Due to the limited size of
the data, further analysis of whether the same or two
closely linked loci were associated with milk coagulation
ability and pH of milk on chromosomes 4, 24, and 27 was
not possible.
We also scanned the QTL databases of Texas A&M
University and University of Sydney (http://bovineqtlv2.
tamu.edu/index.html and http://www.vetsci.usyd.edu.
au/reprogen/QTL_Map/) to collect information on
the SCS findings by other studies in chromosomes 4, 13,
24, and 27.
The most interesting findings were observed on
chromosomes 24 and 27. On chromosome 24, the
maximum of the pH of milk in our study was located
at the same place as that of SCS of milk in Fay cow data
reported by Schulman et al. (2004). It is thus possible
that the gene underlying both of these traits is the same.
Several QTL affecting SCS have been discovered on
chromosome 27. Many of them were positioned at the
beginning of the chromosome, but one of the findings
in the Holstein sample was located in the vicinity of our
finding around marker DIK5134. The two genes can
thus be hypothesized to be associated with milk co-
Mapping Noncoagulation of Milk in Cows
agulation ability, milk pH, and SCS of milk on chromosomes 24 and 27.
To summarize, this was the first study aimed to locate
genomic regions associated with noncoagulation of
milk in dairy cattle. Two highly significant gene mapping results associated with the noncoagulation of milk
emerged over families on chromosomes 2 and 18, near
loci BMS1126 and BMS1355. We also discovered some
other loci associated with coagulation ability, pH, and
SCS of milk.
We thank Nanna Anttila, Marika Lehtinen, Satu Leppänen, and Heli
Kivimäki-Manner for their contribution to laboratory analyses and
Tapani Ala-Tossava and Johanna Vilkki for useful discussions. This
work was financially supported, in part, by the Ministry of Agriculture
and Forestry (Dnro 4636/501/2003), Faba Breeding, the Artificial
Insemination Cooperatives in Finland, and Valio Ltd.
LITERATURE CITED
Amills, M., O. Francino, M. Jansa and A. Sanchez, 1997 Isolation
of genomic DNA from milk samples by using Chelex resin. J.
Dairy Res. 64: 231–238.
Bergère, J. L., and J. Lenoir, 2000 Cheese manufacturing accidents
and cheese defects, pp. 447–508 in Cheesemaking: From Science to
Quality Assurance, Ed. 2, edited by A. Eck and J.-C. Gillis. Lavoisier Publishing, Paris.
Broman, K. W., 2003 Mapping quantitative trait loci in the case of a
spike in the phenotypic distribution. Genetics 163: 1169–1175.
Cassandro, M., A. Comin, M. Ojala, R. Dal Zotto, M. De Marchi
et al., 2008 Genetic parameters of milk coagulation properties
and their relationships with milk yield and quality traits in Italian
Holstein cows. J. Dairy Sci. 91: 371–376.
Coppieters, W., A. Kvasz, F. Farnir, J.-J. Arranz, B. Grisart et al.,
1998 A rank-based nonparametric method for mapping quantitative trait loci in outbred half-sib pedigrees: application to milk
production in a granddaughter design. Genetics 149: 1547–1555.
Darvasi, A., and M. Soller, 1992 Selective genotyping for determination of linkage between a marker locus and a quantitative trait
locus. Theor. Appl. Genet. 85: 353–359.
Darvasi, A., and M. Soller, 1994 Selective DNA pooling for determination of linkage between a molecular marker and a quantitative trait locus. Genetics 138: 1365–1373.
Elo, K., A.-M. Tyrisevä, P. Anttila, V. Vilva and M. Ojala,
2007 Genomic mapping of non-coagulation of milk in the
Finnish Ayrshire. J. Anim. Feed Sci. 16(Suppl. 1): 195–199.
Eurostat, 2007 Statistics under theme: agriculture and fisheries,
6.8.2007. http://epp.eurostat.ec.europa.eu/portal/page?_pageid¼
1090,30070682,1090_33076576&_dad¼portal&_schema¼PORTAL.
Farrell, Jr., H. M., E. L. Malin, E. M. Brown and P. X. Qi,
2006 Casein micelle structure: What can be learned from milk
synthesis and structural biology? Curr. Opin. Colloid Interface
Sci. 11: 135–147.
Georges, M., D. Nielsen, M. Mackinnon, A. Mishra, R. Okimoto
et al., 1995 Mapping quantitative trait loci controlling milk production in dairy cattle by exploiting progeny testing. Genetics
139: 907–920.
Groeneveld, E., 1990 PEST User’s Manual. Institute of Animal Husbandry and Animal Behaviour, Federal Agricultural Research
Centre, Neustadt-Mariensee, Germany.
Groeneveld, E., 1997 VCE4 User’s Guide and Reference Manual. Institute of Animal Husbandry and Animal Behaviour, Federal Agricultural Research Centre, Germany.
Holland, J. W., H. C. Deeth and P. F. Alewood, 2006 Resolution
and characterisation of multiple isoforms of bovine k-casein by 2DE following a reversible cysteine-tagging enrichment strategy.
Proteomics 6: 3087–3095.
Horne, D. S., 1998 Casein interactions: casting light on the Black
Boxes, the structure in dairy products. Int. Dairy J. 8: 171–177.
Horne, D. S., 2006 Casein micelle structure: models and muddles.
Curr. Opin. Colloid Interface Sci. 11: 148–153.
1219
Ikonen, T., 2000 Possibilities of genetic improvement of milk coagulation properties of dairy cows. Ph.D. Dissertation, University of
Helsinki, Helsinki. http://ethesis.helsinki.fi/julkaisut/maa/
kotie/vk/ikonen/.
Ikonen, T., K. Ahlfors, R. Kempe, M. Ojala and O. Ruottinen,
1999a Genetic parameters for the milk coagulation properties
and prevalence of noncoagulating milk in Finnish dairy cows.
J. Dairy Sci. 82: 205–214.
Ikonen, T., O. Ruottinen, E.-L. Syväoja, K. Saarinen, E. Pahkala
et al., 1999b Effect of milk coagulation properties of herd bulk
milks on yield and composition of Emmental cheese. Agric. Food
Sci. Fin. 8: 411–422.
Ikonen, T., S. Morri, A.-M. Tyrisevä, O. Ruottinen and M. Ojala,
2004 Genetic and phenotypic correlations between milk
coagulation properties, milk production traits, somatic cell count,
casein content, and pH of milk. J. Dairy Sci. 87: 458–467.
Johnson, M. E., C. M. Chen and J. J. Jaeggi, 2001 Effect of rennet
coagulation time on composition, yield, and quality of reducedfat cheddar cheese. J. Dairy Sci. 84: 1027–1033.
Johnston, W. S., and G. K. MacLachlan, 1977 Digestive illness in
the calf associated with non-coagulation of cows’ milk. Vet. Rec.
101: 325–326.
Kirov, G., N. Williams, P. Sham, N. Craddock and M. J. Owen,
2000 Pooled genotyping of microsatellite markers in parent-offspring trios. Genome Res. 10: 105–115.
Kruglyak, L., and E. S. Lander, 1995 A nonparametric approach
for mapping quantitative trait loci. Genetics 139: 1421–1428.
Kübarsepp, I., M. Henno, H. Viinalass and D. Sabre, 2005 Effect
of k-casein and b-lactoglobulin genotypes on the milk rennet coagulation properties. Agron. Res. 3: 55–64.
Kumosinski, T. F., E. M. Brown and H. M. Farrell, Jr., 1991 Threedimensional molecular modeling of Bovine caseins: k-casein.
J. Dairy Sci. 74: 2879–2887.
Lipkin, E., A. Shalom, H. Khatib, M. Soller and A. Friedmann,
1993 Milk as a source of deoxyribonucleic acid and as a substrate
for the polymerase chain reaction. J. Dairy Sci. 76: 2025–2032.
Lipkin, E., M. O. Mosig, A. Darvasi, E. Ezra, A. Shalom et al.,
1998 Quantitative trait locus mapping in dairy cattle by means
of selective milk DNA pooling using dinucleotide microsatellite
markers: analysis of milk protein percentage. Genetics 149:
1557–1567.
Malacarne, M., A. Summer, E. Fossa, P. Formaggioni, P. Franceschi
et al., 2006 Composition, coagulation properties and ParmigianoReggiano cheese yield of Italian Brown and Italian Friesian herd
milks. J. Dairy Res. 73: 171–177.
Malossini, F., S. Bovolenta, C. Poras, M. Dalla Rosa and W. Ventura,
1996 Effect of diet and breed on milk composition and rennet coagulation properties. Ann. Zootech. 45: 29–40.
Manichaikul, A., A. A. Palmer, S. Sen and K. W. Broman,
2007 Significance thresholds for quantitative trait locus mapping under selective genotyping. Genetics 177: 1963–1966.
Martin, B., J.-F. Chamba, J.-B. Coulon and E. Perreard, 1997 Effect of milk chemical composition and clotting characteristics on
chemical and sensory properties of Reblochon de Savoie cheese.
J. Dairy Res. 64: 157–162.
Miller, S. A., D. D. Dykes and H. F. Polesky, 1988 A simple salting
out procedure for extracting DNA from human nucleated cells.
Nucleic Acids Res. 16: 1215.
Ojala, M., A.-M. Tyrisevä and T. Ikonen, 2005 Genetic improvement of milk quality traits for cheese production, pp. 307–311
in Indicators of Milk and Beef Quality, edited by J. F. Hocquette
and S. Gigli. Wageningen Academic Publishers, Wageningen,
The Netherlands.
Okigbo, L. M., G. H. Richardson, R. J. Brown and C. A. Ernstrom,
1985 Variation in coagulation properties of milk from individual cows. J. Dairy Sci. 68: 822–828.
Resmini, P., I. De Noni and G. Sala, 1995 Chemical-analytical characteristics of milk compounds associated with abnormal coagulation aspects. Latte 20: 1348–1353 (in Italian).
Schulman, N. F., S. M. Viitala, D. J. de Koning, J. Virta, A. MäkiTanila et al., 2004 Quantitative trait loci for health traits in
Finnish Ayrshire cattle. J. Dairy Sci. 87: 443–449.
Sham, P. C., and D. Curtis, 1995 Monte Carlo tests for associations
between disease and alleles at highly polymorphic loci. Ann.
Hum. Genet. 59: 97–105.
1220
A.-M. Tyrisevä et al.
Tervala, H.-L., and V. Antila, 1985 Milk with anomalous renneting
properties. Meijeritiet. Aikak. XLIII: 26–32.
Tibaldi, E., G. Arrigoni, A. M. Brunati, P. James and L. A. Pinna,
2006 Analysis of a sub-proteome which co-purifies with and is
phosphorylated by the Golgi casein kinase. Cell. Mol. Life Sci.
63: 378–389.
Tsioulpas, A., M. J. Lewis and A. S. Grandison, 2007 Effect of minerals
on casein micelle stability of cows’ milk. J. Dairy Res. 74: 167–173.
Tyrisevä, A.-M., T. Ikonen and M. Ojala, 2003 Repeatability estimates for milk coagulation traits and non-coagulation of milk
in Finnish Ayrshire cows. J. Dairy Res. 70: 91–98.
Tyrisevä, A.-M., T. Vahlsten, O. Ruottinen and M. Ojala,
2004 Noncoagulation of milk in Finnish Ayrshire and Holstein-Friesian cows and effect of herds on milk coagulation ability. J. Dairy Sci. 87: 3958–3966.
Van Hooydonk, A. C. M., H. G. Hagedoorn and I. J. Boerrigter,
1986 The effect of various cations on the renneting of milk.
Neth. Milk Dairy J. 40: 369–390.
Van Ooijen, J. W., 1999 LOD significance thresholds for QTL analysis in experimental populations of diploid species. Heredity 83:
613–624.
Walsh, P. S., D. A. Metzger and R. Higuchi, 1991 Chelex 100 as a
medium for simple extraction of DNA for PCR-based typing from
forensic material. BioTechniques 10: 506–513.
Wedholm, A., L. B. Larsen, H. Lindmark-Månsson, A. H. Karlsson
and A. Andrén, 2006 Effect of protein composition on the
cheese-making properties of milk from individual dairy cows.
J. Dairy Sci. 89: 3296–3305.
Zadworny, D., and U. Kuhnlein, 1990 The identification of the
kappa-casein genotype in Holstein dairy cattle using the polymerase chain reaction. Theor. Appl. Genet. 80: 631–634.
Communicating editor: C. Haley