Accelerating Breeding Cycles

• Realise the challenges associated with increasing the rate of genetic gain in wheat.

• Explore technologies that can reduce the length of a breeding cycle.

• Understand the challenges associated with adapting new technologies to breeding programs.

• Explore opportunities for integrating technologies to improve breeding efficiency.

Wheat breeding programs are designed to prioritise the release of cultivars better adapted to drought and heat, and resistant or tolerant to pests and diseases, with the ultimate goal of increasing yield potential [1]. The estimated rate of genetic gain in wheat is 0.9% per year and an overall agronomic and genetic yield improvement of over 2% per year is needed to meet the increased demand. A typical wheat breeding cycle can take over 12 years for crossing, inbreeding, testing, and selection. Broadly, a breeding cycle involves three distinctive phases, starting and finishing with parental lines used for crossing. These phases include: (i) The crossing and inbreeding phase, which requires the progenies to go through six generations of self-pollination to reach homozygosity and minimise segregation of the traits of interest; (ii) The testing phase, which involves screening for biotic and abiotic traits followed by multi-environment trial evaluation. Lines that are stable across a wide range of environments and carry desirable traits can be recycled and used as parents for making new crosses; (iii) Bulking up seed of the most successful line and releasing a new cultivar available to growers. Due to the lengthy process of cultivar development, breeding programs must adopt new technologies to reduce the time required for completing breeding cycles. Therefore, efforts must focus on speeding up the rate of genetic gain by optimising the components that fit in the breeder's equation. This includes increasing the intensity and precision of selecting individuals within a population that has a high level of useful genetic variation for traits of interest, while also accelerating breeding cycles by reducing the time required to achieve a single cycle [2].

Modern wheat breeding programs have adopted several technologies and breeding strategies that are instrumental in increasing the rate of genetic gain. For example, high-throughput phenotyping platforms can be used to screen large numbers of selected candidates for target traits more precisely, which leads to enhanced selection accuracy and selection intensity [3]. The evolution of next-generation sequencing platforms has resulted in cost-effective genotyping services that have increased the efficiency and accuracy of selection in breeding programs [4]. GS in combination with high-throughput phenotyping platforms shows promise for enhancing predictive breeding approaches, particularly for complex traits such as grain quality and yield [5]. Furthermore, adopting GS in wheat breeding programs can lead to an increase in selection accuracy and intensity, while reducing the breeding cycle concurrently [6]. Shuttle breeding, and most recently ‘speed breeding’ (SB), has transformed breeding programs by shortening the crossing and inbreeding phase [7, 8]. The faster generation turnover enables evaluation of selection candidates sooner, ultimately reducing the length of the breeding cycle (Fig. 30.1). Likewise, DH technology enables the development of homozygous lines in only two plant generations and has dramatically reduced the length of breeding cycles, particularly for winter wheat programs [9].

Illustration of different wheat breeding technologies (DH, SB, and SpeedGS) and their impact on the length of the breeding cycle. Combining speed breeding and genomic selection (i.e. SpeedGS) further reduces the length of the breeding cycle by reducing the need for extensive phenotyping. Black arrows indicate a single plant generation. Green indicates steps performed under conventional growing conditions, whereas pink indicates steps performed under speed breeding conditions. (Modified with permission from Ref. [8])

Full size image

This chapter provides an overview of the technologies and strategies available to wheat breeders for reducing the length of the breeding cycle. Furthermore, the opportunity to combine technologies to accelerate genetic gain is discussed.

3.1 Shuttle Breeding

Shuttle breeding is an off-season field-testing technique whereby genetic material is grown in contrasting environments to turn over two plant generations per year. By implementing this simple, yet effective technique, breeders have successfully reduced the time required to complete a breeding cycle by 50% [10]. In this method, segregating populations are subject to screening, selection and simultaneous generation advancement [11]. The strategy was first developed by Norman Borlaug at the International Maize and Wheat Improvement Centre (CIMMYT) Mexico in 1946. The goal was to speed up breeding cycles and develop varieties faster for the Mexican wheat farmers [12]. To achieve the two generations each year, the material is grown at two contrasting locations in terms of precipitation, altitude and latitude (Fig. 30.2). Today, the technique is still adopted by CIMMYT. In the winter season (November to April), breeding populations are grown in the Sonora Desert (Yaqui Valley, North-Western Mexico) at 39 m.a.s.l. under short days, and selection is applied for yield, agronomic type, leaf and stem rust disease, grain quality and photoperiod insensitivity. In summer (May – October), populations are grown at the Toluca station at 2649 m.a.s.l.), to ensure the crop experiences cooler temperatures during grain filling, and selection is applied for resistance to yellow rust (Puccinia striiformis f. sp. tritici) and speckled leaf blotch (Septoria tritici) disease [7]. In addition to the key advantage of shortening the breeding cycle, the strategy enables selection of breeding materials exposed to different soil types, temperatures, disease pressure and most importantly, photoperiod. The semi-dwarf, rust-resistant and photoperiod insensitive varieties developed by Borlaug and his colleagues resulted in widespread adoption and adaptation to the wheat mega environments of the world. The global success of these varieties is the foundation of what is known as the Green Revolution of the 1960s and 70s, transitioning CIMMYT to internationalisation [12]. Several winter wheat breeding programs have adopted shuttle breeding, for example, the material is shuttled from breeding programs in North America and Northern Europe to New Zealand [13]. The Japanese breeding program initiated shuttle breeding in the 1970s, taking advantage of the wide variation in latitude [14]. Shortly after the emergence of Ug99, a modified shuttle breeding program was implemented by CIMMYT and partners to incorporate screening of the highly virulent race of stem rust in Noro, Kenya [15]. A similar application of shuttle breeding that incorporates selection for biotic and abiotic stresses has been implemented at the International Centre for Agricultural Research in the Dry Areas (ICARDA). For instance, winter × spring crosses are generated at Terbol Station, Lebanon (34° N; 36° E, 900 m.a.s.l.) during winter and summer and the segregating F₂ material is shuttled to Sids station (29° N; 31° E, 32.2 m.a.s.l.) in Egypt during winter for early yield potential trials, the F₃ material is then shuttled to Kulumsa (08° N; 39° E, 2220 m.a.s.l.) in Ethiopia during summer for rust screening, and finally, the F₄ segregating populations are shuttled to Merchouch station (33.6° N; 6.7° W, 430 m.a.s.l) in Morocco during the winter season and screened for insect tolerance and disease resistance [11]. This process can generate broadly adapted inbred lines enriched with target traits prior to yield testing, thus assisting the development of robust cultivars for farmers.

Overview of the shuttle breeding strategy developed by Dr. Norman Borlaug at CIMMYT, Mexico. The arrows indicate the shuttling of genetic material between contrasting North-West and South-East Mexican environments over the winter and summer seasons

Full size image

3.2 Doubled Haploid Technology

A doubled haploid genotype is created when the chromosomes in haploid cells (n) are doubled. DH wheat was first developed by culturing immature haploid pollen and generating diploid homozygous plants using Colchicine (Fig. 30.3). However, this technique is limited due to several factors related to growth conditions of the donor plants, their genetic stability and anther-culture ability, resulting in a low success rate to regenerate healthy and green DH plants [9]. This has led to the search for alternative and high-throughput routes for developing DH plants more suited to applications into breeding programs. Chromosome elimination of wide crosses, such as the cross between wheat and maize (Zea mays L.), was first trialled by Zenkteler and Nitzsche [16]. Following pollination, maize chromosomes are eliminated during cell division and the haploid wheat is induced and the chromosomes are doubled using Colchicine. This technique requires less effort and tripled the success rate, providing a DH production system that is more suitable for breeding purposes. The success of this technique is underpinned by high efficiency and low genotype specificity in comparison to the anther-culture technique. The technology of DH has been integrated into many global largely winter wheat breeding programs which required the upscaling of DH production and the development of a large number of individuals necessary to increase the rate of genetic gain. The advantage of this technology is that the breeding cycle is significantly shortened due to the efficiency in developing inbred lines. The DH homozygous lines could be available for evaluation and breeding within two generations in comparison to six generations or more when using traditional self-pollination methods in the field [11] (Fig. 30.1). Despite the breakthrough in accelerating breeding cycles by rapidly developing homozygous lines, DH technology has drawbacks when implemented in breeding programs. For example, it does not allow evaluation and selection for important traits during the early segregating generations of a classical inbred line development approach. The technique requires specialised labs, is quite labour intensive and requires a high level of experience. Furthermore, there is often significant variability in successfully producing DH lines, due to the genotype dependency. However, the main limitation for DH technology to become a mainstream tool in wheat breeding programs is cost. For example, if a large-scale wheat breeding program were to generate 40,000 DH lines, at a very conservative rate of US $15 per line, it would cost US $600,000. Notably, the cost of DH production for wheat can be even higher, up to US $50 per line (see Chap. 5). Despite this, DH technology is common in breeding programs in Europe and the UK, due to the strong vernalisation requirement of winter wheat that extends generation turnover time. On the other hand, for spring wheat DH technology has been largely adopted in research and pre-breeding programs for QTL mapping studies.

Overview of the steps involved in generating wheat doubled haploid lines using the anther culture method, including (a) harvesting anthers from the selected wheat plants; (b) culturing anthers; (c) callus initiation; (d) transferring embryos to rooting medium; and (e) transplanting rooted plants into the soil

Full size image

3.3 Speed Breeding

SB involves growing plant populations under controlled environmental conditions that are conducive to early flowering and generation advance. The concept was inspired by research funded by the National Aeronautics and Space Administration (NASA) at Utah State University in the 1980s, which explored the possibility of growing a fast crop of wheat on space stations [8]. This research resulted in the development of ‘USU-Apogee’ wheat which can flower within 25 days after sowing when grown under 23°C and continuous light [17]. In 2003, researchers at the University of Queensland coined the name SB, and the technique was first applied to wheat breeding for fast-tracking the introgression of yellow spot resistance. With advances in science and technology over the last few decades, techniques for SB crops have evolved [8]. In particular, advances in light-emitting diodes (LED) technology have led to the widespread availability of affordable grow-lights that provide healthy and efficient plant growth in controlled environment facilities. SB regimes that use LED lighting systems have been developed for a range of long-day or short-day crops [18]. Furthermore, LED lights with ‘tunable’ wavelengths are now available, opening the door for optimising wavelengths to further manipulate plant growth, such as plant height and flowering [19]. This technology has reduced energy consumption and costs, but also facilitated the delivery of optimal light quality in fully controlled SB facilities, resulting in improved plant growth, increased quality and enhanced seed production. SB facilities can take many shapes and forms, including growth cabinets for small-scale research, and modified glasshouses or warehouses with multi-tiered growing spaces for large-scale programs. In glasshouses fitted with supplemental lighting systems, sensor technologies coupled with automated systems can adjust the light intensity depending on the cloud cover, sky luminance and radiation. This provides an integrated light control system to promote flowering, maximise growth rate, and importantly, minimise costs associated with generation turnover.

Speed breeding enables wheat researchers and breeders to grow up to 4–6 generations per year instead of 1–3 generations in the field or under regular glasshouse conditions. Protocols have been made available [20], including a step-by-step guide for establishing large-scale facilities [18]. Exposing plants to extended photoperiods (22 h light, 2 h day) and controlled temperatures (22/17 °C, day/night) respectively, promotes early flowering and seed formation. A common feature in many rapid generation advance methods is embryo rescue, which is a laborious process. However, it is possible to avoid the need for embryo rescue using a pretreatment to break seed dormancy (e.g. 4 °C for 3 days) and achieve high germination rates [18]. For example, to short-cut generation time, spikes can be harvested green (just 2 weeks post flowering) and placed into an air-forced dehydrator at 35 °C for 3 days to artificially mature and dry the seed prior to re-sowing. The premature harvest technique is very effective when applied on a small scale, making it suitable for research and pre-breeding activities. However, for large-scale wheat breeding programs, harvest at maturity is preferred because it avoids the need for multiple harvests for populations that are typically segregating for flowering time, and as such minimises labour (Fig. 30.4). To hasten maturity, water supply can be reduced after flowering to enable harvest approximately 4 weeks later [18].

Illustration of a wheat breeding population growing at high-density under speed breeding conditions. Seven to ten days after sowing, wheat seedlings reach 2–3 leaf growth stage. Time to flowering typically ranges between 4 and 6 weeks for most spring wheat genotypes. To complete the plant generation within eight weeks, wheat heads can be harvested prematurely two weeks after flowering and dried down in an air-forced dehydrator at 35 °C for three days. The slightly shrivelled seeds germinate well provided a cold treatment is applied, and can be used to fast-track generation turnover for pre-breeding and research applications. Alternatively, for breeding applications that involve larger and more diverse populations, after flowering water supply can be reduced to accelerate maturity and enable harvest of mature/well-filled seed four weeks later

Full size image

In addition to accelerating breeding cycles, SB can be used to fast-track research and pre-breeding outcomes. The tool is particularly useful to accelerate the development of populations suitable for trait dissection and mapping QTL for important traits. For example, nested-association mapping populations (NAM) suitable for dissecting drought adaptive traits were rapidly generated using SB at The University of Queensland, requiring only 18 months from crossing to development of F4-derived lines [21]. To support trait screening and selection for pre-breeding applications, a number of protocols have been developed for disease resistance traits, such as wheat rusts [22], yellow spot [23] and crown rot [24]. These techniques incorporate rapid generation advance and enable screening of large segregating populations all year round. Selected plants can be advanced to develop inbred lines or backcrossed to elite parents for trait introgression. When selection is applied in early generations, the resulting inbred lines are enriched with desirable allelic combinations, which enables more targeted and efficient field testing. To support breeding, SB can reduce the number of years required for trait dissection and introgression of new traits into elite genetic backgrounds.

The key to integrating SB into a large-scale wheat breeding program is establishing cost-effective facilities and streamlining operations. With this in mind, SB facilities have been established by private wheat breeding companies, as well as public breeding and research centres, such as CIMMYT and ICARDA. Reducing the cost per plant is important, which ultimately reduces the cost per line development through the SB facility. Ghosh et al. [18] provide detailed protocols and advice for scaling-up wheat SB protocols, including growing plants at high density (e.g. 1000 plants/m²). Such techniques can enable wheat breeding programs to generate large populations in a cost-effective manner. Other important considerations include designing low-cost and energy efficient infrastructure, adopting automation where possible (e.g. automated irrigation systems), and streamlining operations.

3.4 Genomic Selection

The implementation of GS is outlined in Chap. 5. This method was initially developed and applied to accelerate genetic gain in animal breeding. The method is based on statistical models that can predict the genetic merit of individuals based on high-resolution genome profiles, before they are tested in the field. The genetic merit of a genotype, mostly referred to as the genomic estimated breeding value (GEBV), is calculated by simultaneously estimating genome-wide DNA marker effects and then summing them for each individual. Individuals with the highest GEBV can then be selected either for extensive field evaluation or to be used as parents in the next breeding cycle [25]. The method requires a training population where the individuals within the population are phenotyped for the traits of interest and genotyped using high-density DNA markers. By combining the genetic makeup of individuals in the training population and their phenotypes, a statistical model that can estimate the association between markers and quantitative traits is initiated. It is then used to predict GEBV of genotypes in the ‘selection candidates’ that have only been genotyped using the same genome-wide DNA markers [26].

Several growing seasons are usually required for testing breeding material under different environments in conventional breeding methods, with a risk of inaccurate selection due to genotype-by-environment interaction [27]. One strategy by which GS models can help overcome this issue is to incorporate proxy traits that explain an additional amount of the phenotypic variation observed for the target trait. Incorporating such traits in GS models has been shown to improve GEBV prediction [5]. This enables a more accurate selection of superior genotypes with the desired combination of alleles without the need for phenotyping. Besides prioritising genotypes for further field testing based on their GEBV, a recurrent GS method could be applied during early generations of the line development phase. For example, recurrent GS could be used at F₂ or even F₁ generation to select candidates for intercrossing. This would enable more rapid enrichment of populations with desired alleles (e.g. [26, 28]). Thereafter, further shortcutting the testing phase and accelerating the breeding cycle is possible, unlike in conventional breeding methods. The shift from phenotypic selection to GS has resulted in saving resources and time. For example, GEBVs can be obtained in 2 years when GS is applied, instead of 5 years in conventional breeding [4]. In addition to shortening the breeding cycle and improving selection accuracy, GS has increased the rate of genetic gain in breeding programs by increasing selection intensity. This is possible when the number of tested lines is boosted due to reduction in replication [8]. This approach has been implemented in wheat breeding programs around the globe [27]. However, application of GS in plant breeding programs has experienced a significant shift from single-trait/single-site prediction models to multi-trait/multi-site models which incorporate genotype-by-environment effects. The shift to more complex models has enabled more accurate predictions of phenotypes which have been sparsely collected in multiple environments [29]. GS is usually applied to quantitative traits governed by a large number of genes, such as drought adaptive traits [5], grain quality [30] and grain yield [31]. Furthermore, GS has been applied to proxy traits such as flowering time, which has high heritability, is easy to measure, and has a significant impact on yield [32]. GS has been shown to be powerful for improving resistance to fusarium head blight and stem rust in wheat using large historical datasets [33]. In the early stages of GS application, research largely focused on improving selection accuracy based on optimised prediction models by accounting for variation caused by genotype-by-environment interaction in the models [34]. More recently, high-throughput phenotyping platforms that generate very large datasets, such as unmanned aerial vehicle (UAV) platforms, have been used for improving GS models with enhanced accuracies [35]. GS has shifted from being the focus of research to being widely adopted into wheat breeding programs in both private and public sectors. This is due to several factors including advances in molecular biotechnology, a dramatic decrease in genotyping cost and the development of bioinformatic tools that can efficiently calculate GEBVs for very large datasets [4].

A radical change and redesign of breeding programs incorporating advanced technologies holds the key to improving yield potential for our future wheat varieties. Breeding tools such as, DH, SB and GS have great potential for accelerating breeding cycles, however, their implementation on an industrial scale in mainstream wheat breeding programs is yet to see the light of day. As detailed in Sects. 30.2 and 30.3, these breeding tools can play an instrumental role in accelerating breeding cycles. Implementation of these technologies in early stages of breeding cycles enables rapid production of homozygous lines enriched with desired allele combination for field testing.

The process of DH line development usually requires two years and an additional year of self-pollination for increasing sufficient seed necessary for field trials. The rapid development of DH lines could be possible when SB is integrated at different stages of DH production systems. For example, parental lines used for haploid induction could be grown under SB. Furthermore, SB technology could be used during the self-pollination stage following chromosome doubling and the additional generation for seed bulking. The integration of SB into the DH system facilitates shortcutting time to further accelerate breeding cycles.

The adoption of GS enables the identification of superior individuals based on their GEBV, either for advanced yield testing or as parents used for crossing in the next breeding cycle [8]. The integration of GS in breeding programs has thus resulted in the emergence of highly productive wheat cultivars in a shorter time. Notably, combining these technologies targeting different stages of the breeding cycle may have an additional impact on the rate of genetic gain. For example, integrating high throughput phenotyping with GS methods has been shown to improve efficiency and outcomes for both methods [5]. GS is also employed for detecting and stacking the best haplotypes using parents with optimal genetic variation used for DH line development [36]. Implementation of SB and GS has demonstrated significant increases in the rate of genetic gain when applied separately in the breeding programs. However, combining these two technologies may result in a larger effect on shortening time required for completing a breeding cycle (Fig. 30.1). Despite the potential advantages of combining these technologies, it is yet to be employed in existing breeding programs.

To explore the potential, simulation studies have been performed. The study by Voss-Fels et al. [37] compared the rate of genetic gain for four breeding strategies, including traditional phenotypic selection, GS, the combination of SB and GS (SpeedGS), and SpeedGS with introduced diversity. Overall, genetic merit for grain yield was used to determine the amount of gain that could be achieved by implementing the different breeding tools in isolation or in combination. The results concluded that significant gains were possible using all strategies, however, breeding schemes that implemented SpeedGS displayed a 34% increase in the rate of genetic gain per unit of time when compared to conventional breeding [37]. In these simulations GS was only used to identify improved lines in the breeding cycle for advanced field testing, however, the use of GS in a rapid recurrent selection framework has been shown to hold huge potential for substantially increasing genetic gain [19, 28]. Despite these promising results from combining SB and GS, adoption of these tools by breeding programs needs further investigation. Further empirical and simulation studies are needed to determine a suitable pathway in which these technologies could be applied in the most efficient and cost-effective way to maximise investment return.

This chapter has outlined (1) key strategies and advanced technologies that can reduce the length of a breeding cycle, (2) opportunities to integrate these technologies to further accelerate genetic gain. Widespread adoption of these enabling tools in public and private wheat breeding programs would enhance breeding efficiency and support global food security.

Crop production must increase by 50% by 2050 to meet the future demand for food. Adoption of precision farming systems that use cutting-edge technology to optimise management practices, such as fertiliser, pest and disease management will assist in meeting this goal. Most importantly, plant breeding aimed at improving the genetics of crop varieties with high performance in the face of abiotic and biotic challenges will help secure increments in crop production at a global scale. However, the development of resilient crop varieties requires a huge investment and is notoriously slow. In order to speed up the process, redesign and transformation of crop breeding programs is required. Fortunately, plant breeders now have at their disposal technologies that could become a game-changer and transform traditional plant breeding to being significantly more cost-effective and efficient in releasing high yielding and stable varieties. Employing SB, DH, high-throughput phenotyping platforms and GS in breeding programs has advantages, but could be further exploited by integrating these tools at different stages of research and breeding programs; from trait discovery to rapid introgression and population improvement, field testing and release of new cultivars. Fusing these technologies into our existing breeding programs will play a key role in rapidly improving future wheat crops.