Académique Documents
Professionnel Documents
Culture Documents
o r i g i n a l r es e a r c h
S
the context of predicting breeding values (BVs). Using the data
everal important agronomic and horticultural
from this blueberry population, we found for all the traits that tet-
crops such as alfalfa (Medicago sativa L.), potato
rasomic inheritance creates a better fit than disomic inheritance.
(Solanum tuberosum L.), and highbush blueberry (Vac-
In one of the five traits studied, the level of double reduction was
different from zero, decreasing the estimated heritability, but it did
cinium corymbosum L.) are autotetraploids, with four
not affect the prediction of BVs. We also discovered that different
homologous chromosomes present in each linkage group.
depths of pedigree would have significant implications on the As more than two homologous chromosomes exist in
estimation of double reduction using this approach. This freely autotetraplopids, during meiosis chromosome pairing can
available R package is available for autopolyploid breeders to occur between a pair of randomly chosen chromosomes
estimate the level of double reduction present in their populations (bivalents) or between more than two homologous chro-
and the impact in the estimation of genetic parameters as well as mosomes (multivalents) (Fisher, 1947). This polysomic
to use advanced methods of prediction and selection.
a m adeu et al .: aghm atrix : r pack age to construct rel ationshi p m atri ces 3 of 10
considering the proportion of parental gametes that are resources in most autotetraploid species. Therefore, the
identity by descent as a result of double reduction w. Thus AGHmatrix package presented in this study may be the
for every individual ki with parents s and d, solution for many autotetraploid breeders to obtain an
If s and d of ai are unknown, they are assumed unre- estimation of double reduction in their populations.
lated, and Using blueberry as an example, we demonstrated the
method to achieve such goals. We first built seven rela-
kij = kji = 0; for j = 1:(i 1) tionship matrices (A-matrices) base on the pedigree: one
assuming a diploid system and the other six assuming
kii = (1 + w)/4 an autotetraploid system with different levels of double
reduction ranging from 0 to 0.25. Each of these matrices
If only s is known, d is assumed nonrelated, and was fit in the same mixed model equation using the same
phenotypic data for each of the traits considered. We then
kij = kji = 0.5(kjs); for j = 1:(i 1) obtained the AIC value from the REML. The AIC dif-
ference between the disomic and tetrasomic model (Fig.
kii = [5 + 7w + 4k ss(1 w)]/24 1) was used to indicate the model that best fit the data.
Yield, weight, and diameter had similar negative slopes,
If s and j are known, then each decreasing more dramatically with increasing dou-
ble reduction. The pattern of scar was similar but with a
kij = kji = 0.5(kjs + kjd); for j = 1:(i 1) marginal decrease from double reduction varying from
0 to 0.25. Firmness was the only trait that had the best fit
kii = [1 + 2w + (1 w)k ss + (1 w)kdd + 3k sd]/6 when double reduction was larger than zero; the fitness of
the model was maximized when double reduction ranged
Finally, the matrix Aw is given by Aw = 4K. Six Aw matri- from 0.15 to 0.20. The tetraploid model assuming w = 0
ces were constructed from the pedigree assuming vary- was the best fit for four of the five traits. The absence of
ing levels of double reduction (w): 0, 0.05, 0.10, 0.15, 0.20, double reduction for the majority of traits is similar to
and 0.25. Additionally, we also built the traditional dip- the findings of Krebs and Hancock (1989) when study-
loid pedigree-based relationship matrix (A-matrix) using ing blueberry based on segregation of a limited number
the AGHmatrix R package, which is calculated according of isoenzyme loci in V. corymbosum. The most plausible
to the recursive method presented in Mrode (2014) and explanation for the null double reduction comes from the
described by Henderson (1976). The AGHmatrix R pack- physical barrier imposed by the size of the chromosomes.
age can also build the genomic-base relationship matrix, Blueberries have small chromosomes (1.52.5 m), which
G-matrix, for diploids using either the method proposed could limit multivalent formation, an essential step for
by VanRaden (2008) or from Powell et al. (2010). To cal- double reduction to occur (Krebs and Hancock, 1989;
culate the G-matrix, molecular data needs to be in table Hall and Galletta, 1971). Also, cytogenetic studies of
format where n is the number of individuals (in rows) meiotic pairing in V. corymbosum found the majority
and m is the number of loci (in columns). The loci infor- of pairing to be bivalent with few quadrivalent forma-
mation is set to 1, 0, 1, for the genotypes aa, Aa, and tions (Jelenkovic and Harrington, 1971; Jelenkovic and
AA, respectively. Although the G-matrix capability is Hough, 1970; Vorsa and Novy, 1995; Qu et al., 1998). Qu
present in the package, the genomic matrix was not built et al. (1998) found that >90% of the chromosomes formed
or used in this study. bivalents in a highbush cultivar and a wild tetraploid
The autotetraploid matrices with different levels of selection. Jelenkovic and Hough (1970) found similar
double reduction were used in the linear mixed model results when studying three V. corymbosum cultivars,
described above. These autotetraploid models were com- where up to four quadrivalents were found in a pollen
pared against the diploid counterpart using the Akaike cell, but most cells had only one. Additionally, preferen-
information criteria (AIC). As the level of double reduc- tial pairing lowers the expected frequency of multivalent
tion w is unknown, we inferred its level based on the formation between homologous chromosomes (Sybenga,
model that maximized the likelihood of the data given 1996). Preferential pairing has been reported in V. corym-
the parameters of the linear mixed model. bosum and V. corymbosum V. darrowii hybrids (Draper
and Scott, 1971; Vorsa and Novy, 1995).
Results and Discussion While the above explanations are feasible for the
zero double reduction found for four of the five traits,
Using AGHmatrix to Determine Trait-Specific they do not explain the high double reduction estimated
Levels of Double Reduction for firmness. However, a similar study in autotetra-
A unique challenge to plant breeders dealing with auto- ploid potato found that the model best estimating the
tetraploid species has been the determination of the proportion of double reduction varied between 0 and
degree of double reduction and its impact on heritabil- 0.25 depending on the trait considered (Slater et al.,
ity and BV estimates. This challenge has been exacer- 2013). Double reduction is expected to differ from trait
bated with the lack of molecular markers and genetics to trait because it is a position-dependent phenomenon
fluctuating between chromosomes based on the fre- Table 1. The heritability of five breeding traits calcu-
quency of multivalent formation and within chromo- lated assuming disomic and tetrasomic inheritance.
somes depending on where the measured loci reside Tetrasomic inheritance was estimated assuming no
(Wu et al., 2001). Bourke et al. (2015) found a parallel double reduction for all traits except firmness where
increase between the rate of double reduction and the 0.15 was used.
distance from the centromere using a high-density set
Disomic inheritance Tetrasomic inheritance
of markers in autotetraploid potato. The rate of double
reduction increases toward telomeres because there is a Trait Heritability Standard error Heritability Standard error
greater probability of a crossover to occur between the Yield 0.46 0.02 0.45 0.02
centromere and the loci (Bourke et al., 2015; Luo et al., Weight 0.58 0.02 0.58 0.02
2006; Welch, 1962; Wu et al., 2001). The high propor- Firmness 0.40 0.03 0.35 0.02
tion of double reduction for firmness suggests that genes Fruit diameter 0.26 0.03 0.26 0.03
controlling this trait may reside toward the distal end Scar diameter 0.54 0.02 0.54 0.02
of a chromosome. Thus, in the same line, the genes con- Tetrasomic inheritance was calculated assuming double reduction of 0.15.
trolling yield, weight, diameter, and scar may be located
closer to the centromere, be quantitative in nature and
dispersed over the whole genome, or located on chromo- showed a better fit with the tetrasomic model consider-
somes that do not exhibit much multivalent formation. ing higher levels of double reduction, thus the heritability
for the tetrasomic model was estimated using w = 0.15,
Impact of Increasing Double Reduction Levels which was the level of double reduction that fit the data
on Heritability and Relationship Estimates best (Table 1).
Heritability estimations were not significantly different The heritabilities estimates for firmness had the larg-
when assuming tetrasomic inheritance with zero double est difference between the two models, decreasing from
reduction (w = 0) instead of disomic inheritance for all 0.40 to 0.35 as a result of the larger proportion of double
traits (Table 1). However, firmness was the only trait that reduction used in the estimation. As seen in Fig. 2, there
a m adeu et al .: aghm atrix : r pack age to construct rel ationshi p m atri ces 5 of 10
Fig. 2. The estimated narrow-sense heritability for firmness assuming disomic inheritance and tetrasomic inheritance with increasing
coefficients of double reduction.
is a small difference between the disomic-estimated heri- Table 2. Average of the relationship coefficient be-
tability for firmness and the tetrasomic-estimated herita- tween two individuals assuming various levels of
bility with no double reduction. The difference becomes double reduction.
more pronounced, and the narrow-sense heritability for
Double First Third
the trait decreases as the proportion of double reduction Reduction Min. quartile Median Mean quartile Max.
increases. This trend was seen in all the traits measured
0 0 0.067 0.106 0.115 0.148 1.000
(data not shown). Comparing disomic and tetrasomic
0.05 0 0.071 0.113 0.123 0.158 1.050
inheritance with w = 0.10 in autotetraploid potatoes,
0.10 0 0.076 0.120 0.130 0.168 1.100
Slater et al. (2013) observed no consistent trend and
0.15 0 0.080 0.126 0.138 0.177 1.150
found the difference in heritability ranged from 0.01 to
0.20 0 0.084 0.133 0.145 0.186 1.200
0.13 depending on the trait being analyzed.
0.25 0 0.088 0.139 0.151 0.194 1.250
The downward trend in heritability observed
for blueberry fruit firmness can be explained by the
increased inbreeding caused by increasing the levels of by comparing the kinship coefficients obtained for w =
double reduction. As double reduction increased, the 0 and those obtained for w = 0.25 (Table 2). The effects
kinship coefficient also increased (Table 2). The maxi- of double reduction are cumulative with larger kinship
mum coefficient was determined by the amount of dou- coefficients more susceptible to its effects. In the first
ble reduction, whereas the minimum kinship coefficient quartile, the difference between the kinship coefficients
was unaltered by the level of double reduction. A large estimated assuming w = 0 and the coefficient estimated
number of the individuals in the pedigree had unknown assuming w = 0.25 was 0.021, while this difference almost
ancestry, which negatively skewed the distribution. doubles (0.046) in the third quartile.
Unknown ancestry was a result of missing information, The BLUP analysis uses these kinship coefficients
unspecified clones, and wild selections. These individu- and estimations of heritability to predict BVs. Thus, inac-
als were assumed to be unrelated and thus unaffected curate estimations of double reduction could lead to bias
by double reduction. As double reduction increased the predictions. Higher relationships should suffer a larger
kinship coefficient increased as well. This can be seen impact as these are more affected by double reduction.
Nevertheless, there was a strong correlation (r = 0.999) (Fig. 4). By using only 5 yr of data (back to 2009) the level
between the diploid and tetraploid (w = 0.15) estimated of double reduction is estimated to be the maximum
BVs for firmness (Fig. 3). This strong correlation was coefficient tested in this study, 0.25, for four out of five
seen for all traits when analyzed with tetraploid models traits, and 0.20 for the last trait. In this case, all the traits
assuming w = 0 (data not shown). Similarly, Slater et al. had the worst fit assuming no double reduction. When
(2013) found a strong correlation (>0.95) for eight of the 10 yr of pedigree information was used, three traits fol-
nine traits when comparing the diploid estimated BVs lowed the same pattern as when using 5 yr of pedigree
with the autotetraploid estimated BVs. The depth of the information, no double reduction was estimated for one
pedigree could help balance and compensate errone- trait, while the fit for the final trait was maximized at a
ous estimations, resulting in small changes to the BVs double reduction level of 0.05. As more pedigree infor-
over various levels of double reduction. In this case, the mation was added, the slopes continued to change, but
pedigree covered the whole breeding program beginning even when 30 yr of pedigree information were used,
with its origin in 1908. it was not enough to observe the estimation of double
reduction seen when all the pedigree information was
Effect of Pedigree Depth on Double used. Thus, shallow pedigrees underestimate the actual
Reduction Estimate kinship coefficient between two individuals. For example
In the example data set we have used, there is an extensive Boyce (1983) found that shallow pedigrees miscalculate
pedigree available beginning with known founder clones the level of inbreeding when analyzing 10 Standardbred
dating back to 1908 (Coville, 1937). However, the detail of stallions with 30 generations worth of pedigree informa-
pedigree information available for blueberry may not be tion. In blueberries, there has been a steady increase in
available for other autotetraploid plant breeders. To study inbreeding with each consecutive generation as a result
the consequences of using truncated or incomplete pedi- of recurrent selection (Brevis et al., 2008; Hancock and
gree information while studying double reduction using Siefker, 1982; Ehlenfeldt, 1994). Relationships are further
these methods, we truncated the pedigree to 5, 10, 20, and intensified by the narrow gene pool created in the found-
30 yr (back to 2009, 2004, 1994, and 1984, respectively). ing event. Based on pedigree information, Ehlenfeldt
Different levels of double reduction maximize the (1994) concluded that three cultivars, Brooks, Sooy, and
fitting of the data when using different pedigree depths Rubel, have contributed the largest portion of genes in
a m adeu et al .: aghm atrix : r pack age to construct rel ationshi p m atri ces 7 of 10
Fig. 4. In order from left to right, pedigree depths of 5, 10, 20, and 30 yr and their influence on the estimated amount of double
reduction for each trait.
79 highbush cultivars (23.5, 12.5, and 28.6%, respec- coefficients and thus incorrect determination of the level
tively) and V. darrowi contributions can be traced back of double reduction for each trait. This will also impact
to germplasm sources Florida 4B, Florida 4A, Florida the estimated heritability, inbreeding, and BVs.
9A, and clone D (Brevis et al., 2008). The limited pedi-
gree information cannot accurately account for these Conclusion
intertwined relationships. Thus, the higher proportions
of double reduction compensated for the lack of informa- We have developed an R package, AGHmatrix, that can
tion by increasing the assumed relationships between be used to quickly develop relationship matrices for
the individuals and thus better fitted the model. Atkin autotetraploid crop species. To illustrate the utility of
et al. (2009) and Mehrabani-Yeganeh et al. (1999) found this package, we used the pedigree and phenotypic data
the accuracy of estimating additive variance components from a blueberry breeding program. The majority of
and BVs improved when using more pedigree informa- traits (yield, weight, diameter, and scar) had the best fit
tion. Atkin et al. (2009) discovered that the variances assuming tetrasomic inheritance with no double reduc-
were estimated with the same precision once five genera- tion. The only exception was firmness, which had the
tions were used, suggesting older generations do not con- best fit when double reduction was 0.15. Changes in the
tribute much information to the model. This is mainly estimation of heritability occurred with larger propor-
because the coefficients of relationships with the cur- tions of double reduction. However, there was no sig-
rent population become smaller as generation distance nificant improvement in BV estimates for the blueberry
increases. However, using shallow pedigrees can lead the traits assessed when comparing tetrasomic inheritance
erroneous conclusions in the estimation of relationship accounting for the correct levels of double reduction with
a m adeu et al .: aghm atrix : r pack age to construct rel ationshi p m atri ces 9 of 10
National Center for Genome ResourcesCorvallis. 2012. NCGR-Corvallis Shukitt-Hale, B. 2012. Blueberries and neuronal aging. Gerontology 58:518
Vaccinium catalog: Blueberry cultivars and selections. USDANRCS. 523. doi:10.1159/000341101
http://www.ars-grin.gov/cor/catalogs/vactc.html (accessed day Mo. year). Slater, A.T., G.M. Wilson, N.O. Cogan, J.W. Forster, and B.J. Hayes. 2013.
Olmstead, J.W., H.P.R. Armenta, and P.M. Lyrene. 2013. Using sparkleberry Improving the analysis of low heritability complex traits for enhanced
as a genetic source for machine harvest traits for southern highbush genetic gain in potato. Theor. Appl. Genet. 127:809820. doi:10.1007/
blueberry. Horttechnology 23:419424. s00122-013-2258-7
Piepho, H.P., J. Mhring, A.E. Melchinger, and A. Bchse. 2008. BLUP for Sybenga, J. 1996. Recombination and chiasmata: Few but intriguing dis-
phenotypic selection in plant breeding and variety testing. Euphytica crepancies. Genome 39:473484. doi:10.1139/g96-061
161:209228. doi:10.1007/s10681-007-9449-8 Tipton, D.A., J.P. Babu, and M.K. Dabbous. 2013. Effects of cranberry
Powell, J.E., P.M. Visscher, and M.E. Goddard. 2010. Reconciling the analy- components on human aggressive periodontitis gingival fibroblasts. J.
sis of IBD and IBS in complex trait studies. Nat. Rev. Genet. 11:800805. Periodontal Res. 48:433442. doi:10.1111/jre.12023
doi:10.1038/nrg2865 USDA. 2015. Noncitrus fruits and nuts: 2014 summary. http://usda.
Qu, L., and J.F. Hancock. 1995. Nature of 2n gamete formation and mode mannlib.cornell.edu/usda/nass/NoncFruiNu//2010s/2015/Nonc-
of inheritance in interspecific hybrids of diploid Vaccinium darrowi FruiNu-07-17-2015.pdf (accessed 23 Oct. 2015).
and tetraploid V. corymbosum. Theor. Appl. Genet. 91:13091315. VanRaden, P.M. 2008. Efficient methods to compute genomic predictions. J.
doi:10.1007/BF00220946 Dairy Sci. 91:44144423. doi:10.3168/jds.2007-0980
Qu, L., J.F. Hancock, and J. Whallon. 1998. Evolution in an autopolyploid Van Vleck, L.D. 1998. Charles Roy Henderson, 19111989: A brief biogra-
group displaying predominantly bivalent pairing at meiosis: Genomic phy. J. Anim. Sci. 76:29592961.
similarity of diploid Vaccinium darrowi and autotetraploid V. corymbo- Vorsa, N., and R. Novy. 1995. Meiotic behavior and RAPD segregation in
sum (Ericaceae). Am. J. Bot. 85:698698. doi:10.2307/2446540 a tetraploid hybrid (Vaccinium darrowi V. corymbosum) suggests
Resende, M.F., P. Munoz, M.D. Resende, D.J. Garrick, R.L. Fernando, J.M. genomic divergence in blueberry. HortScience 30:808808.
Davis, E.J. Jokela, T.A. Martin, G.F. Peter, and M. Kirst. 2012. Accu- Welch, J.E. 1962. Linkage in autotetraploid maize. Genetics 47:367396.
racy of genomic selection methods in a standard data set of loblolly Wu, R., M. Gallo-Meagher, R.C. Littell, and Z.B. Zeng. 2001. A general
pine (Pinus taeda L.). Genetics 190:15031510. doi:10.1534/genet- polyploid model for analyzing gene segregation in outcrossing tetra-
ics.111.137026 ploid species. Genetics 159:869882.
Schindelin, J., I. Arganda-Carreras, E. Frise, V. Kaynig, M. Longair, T. Yousef, G.G., A.F. Brown, Y. Funakoshi, F. Mbeunkui, M.H. Grace, J.R.
Pietzsch, S. Preibisch, C. Rueden, S. Saalfeld, B. Schmid, J. Tinevez, D.J. Ballington, and M.A. Lila. 2013. Efficient quantification of the health-
White, V. Hartenstein, K. Eliceiri, P. Tomancak, and A. Cardona. 2012. relevant anthocyanin and phenolic acid profiles in commercial cultivars
Fiji: An open-source platform for biological-image analysis. Nat. Meth- and breeding selections of blueberries (Vaccinium spp.). J. Agric. Food
ods 9:676682. doi:10.1038/nmeth.2019 Chem. 61:48064815. doi:10.1021/jf400823s