Freezing Tolerance in the Triticeae - Translational Genomics for Crop Breeding: Abiotic Stress, Yield and Quality - page 116

specific environments by selecting for appropri-
ate alleles at the major loci determining long- or
short-day sensitivity (that is PPD-H1 on chro-
mosome 2H and PPD-H2 on chromosome 1H,
respectively). Long-day sensitivity could max-
imize yield by delaying maturity in those envi-
ronments not affected by terminal drought stress,
while short-day sensitivity is appropriate for all
environments, as it will delay the vegetative-to-
reproductive transition. By contrast, LD-MAS of
functional polymorphisms in CBF genes has still
not yet been validated. In barley, Akar and col-
leagues (2009) found that only one out of the
three markers designed on CBF genes was mod-
erately associated with frost tolerance in bar-
ley, and Rapacz and colleagues (2010) did not
find significant associations between a HvCBF4
polymorphism and frost tolerance. Due to the
observed CNV of specific CBF family elements
differentiating tolerant and susceptible wheat
and barley (Knox et al. 2010), fast and reli-
able methods for CNV detection could also be
sought for application in LD-MAS. Other LD-
MAS studies used PCR-based markers associ-
ated with FR-1 and FR-2 (Toth et al. 2004), or
the dehydrins Wcs120 and Dhn13 (Holkov´aetal.
2009).

should be selected, containing cumulated posi-
tive effects from all contributing genes and minor
effect QTLs. Heffner and colleagues (2011)
demonstrated that for 13 agronomic traits in a
population of 374 winter wheats, the average pre-
diction accuracies for GS would be 28% higher
than MAS. Since a high-density SNP panel could
be excessively costly per single analysis, the
adoption of GS by breeders strongly depends
on the increasing availability of cheap high-
throughput marker systems. Moreover, GS can
be better proposed for species where genomic
constitution is known, in terms of sequences
and their physical position, and for which cul-
tivar resequencing projects are in progress, as in
apple (Kumar et al. 2012). In this view, the suc-
cess in sequencing all gene-containing regions
of barley and wheat is a necessary requirement
to allow GS-based schemes. Recently, Paux and
colleagues (2011) reported that GS methods are
under evaluation for crops such as maize and
wheat and, in some cases, are being applied in
commercial breeding programs, although details
have yet to be published. Genomic selection
could be particularly useful for accumulating
durable (quantitative) disease resistance QTLs
in wheat, as proposed by Rutkoski and col-
leagues (2011) for stem rust, where the multi-
genic nature of adult plant resistance hampers the
efficiency of MAS-based pyramiding. Because
of the lack of mapped minor QTLs affecting the
final level of freezing tolerance, GS for freez-
ing tolerance could be an option for Triticeae,
albeit together with genome-wide selection for
other abiotic stress tolerances and agronomically
relevant traits. Once high-density SNP panels
can be made available and at reasonable assay
costs, it should not be necessary to know all
the (minor) QTL positions to select associated
markers. While GS should substantially accel-
erate the breeding cycle, it would also dramati-
cally change the role of phenotyping (including
automated phenomics facilities), which could
be used more to update the prediction models
driving GS than to select lines (Heffner et al.
2009).

Genomic Selection (GS)

Genomic selection was proposed as a means
of overcoming the limits of LD-based MAS
to polygenic trait selection. It was defined by
Meuwissen and colleagues (2001) as a method
of predicting the breeding value of genotypes by
analyzing phenotype together with high-density
marker scores. Basically, GS simultaneously
estimates all locus, haplotype, or marker effects
across the entire genome to calculate genomic
estimated breeding values (GEBVs). The index
incorporates all marker information in a pre-
diction model, thereby avoiding biased marker-
effect estimates and capturing more of the varia-
tion resulting from small-effect QTLs. By using
high-density SNP panels, the genotype that
would best fit with the genomic prediction model