20 Nov 2017
Adam Martin discusses the introduction and evolution of AI throughout history, as well as its uses in herd breeding.
Adam Martin
Job Title
Single nucleotide polymorphism chip genome testing.
Genomic evaluation has transformed the dairy industry worldwide, having been implemented in the UK and Ireland, as well as North America, New Zealand, Australia, Germany, France, the Netherlands and Scandinavia. International collaborative agreements to share genomic information have hastened the progress of the technology, while rates of genetic gain in the countries and breeds with effective genomic breeding programmes have significantly improved.
Genomics has revealed the number of genes linked to economically important traits is enormous and, consequently, the potential for continued improvement without achieving a genetic plateau due to the loss of diversity within a population is low.
The importance of selection to improve domestic livestock by agriculturist Robert Bakewell in the 18th century is generally regarded as being the start of modern day selection practices in the UK.
Bakewell played an important role in the agricultural revolution; he lived around 100 years before Darwin, who cited Bakewell’s work to “create” new livestock breeds – most notably, the now extinct New Leicester sheep – in On the Origin of the Species as an example of artificial selection, as opposed to the more famous natural selection.
Bakewell’s innovation was separating males and females, which had traditionally been kept together, and introduced selected males to the females to control breeding. This early phenotypic selection process is still practised in many parts of the world, including the UK. However, as our knowledge of genetics has increased, so has the sophistication of the practice.
Around the same time Bakewell was developing his New Leicester sheep, Catholic priest, biologist and physiologist Lazzaro Spallanzani, and his colleagues, were working on the concept of AI and, in 1784, the first successful AI was reported to have occurred when puppies were born to a bitch 62 days after being artificially inseminated.
It took another 100 years or so before real advancements were made by Walter Heape and his colleagues at the University of Cambridge in the late 19th and early 20th centuries, who expanded the technique across many species.
However, the first recognisable step in AI began in Denmark in 1936, when a breeding cooperative including more than 1,000 cows using AI was established. The concept of a breeding cooperative was taken up by Cornell University in the US and a cooperative centred around the university drove forward scientific process in the field in the 1940s.
Successful freeze – thawing of semen and development of the AM-PM/PM-AM insemination rule – were discovered in the 1940s and 1950s, which made AI a potential mass market technique.
Although AI is primarily used in the UK to promote genetic advancement, it is worth noting disease control was the reason many farmers began using AI in the early days when a commercial service was available.
Genetic progress per year = (accuracy of selection × selection intensity × genetic variation).
Until about 10 years ago, genetic improvement in cattle populations was based on the concept of progeny testing. This involved a cohort of young bulls being selected from the population on the basis of their pedigree, which was around 35% reliable.
The bulls were taken to a breeding station, then semen was collected from those deemed best suited for semen production and distributed for AI on farms. This technique is still used in many countries and continues to be used in smaller breeds that do not have the resources of the population size to harness the power of genomic selection.
In progeny testing, the phenotypic performance data of the offspring of the test bulls was gathered and bulls ranked accordingly. The best performing bulls were selected to be elite, marketed sires, while the non-selected bulls were slaughtered.
Progeny testing took a long time – most bulls would be 1.5 to 2 years old before their semen would be used to inseminate cows 2.5 to 3 years old when their first daughters were born, and 5.5 to 6 before their daughters had completed their first lactations and their test proof was complete.
This meant, by the time elite bulls entered the market, they had either been kept alive and non-productive for years, or the semen from the bulls was complete and the bulls slaughtered, meaning only a limited semen supply was available from that bull.
Either way, the progeny testing process delay meant the generation interval (a denominator in the rate of genetic progress; Panel 1) was high, as calves born to two-year-old heifers and elite sires would have a generation interval of four years or more.
The idea of genomic testing really flourished in the 1980s, when various simulation models looking at testing around 100 genes related to production was put forward as being a viable future technology.
However, in the late 2000s, technology became available to test 50,000 genes simultaneously on a single nucleotide polymorphism (SNP) chip. This 50,000 SNP chip genome testing originally cost around £400 per animal in 2010; however, costs have fallen to around £30 per animal.
This progression has led to the genotyping of more than 1.6 million Holsteins in the North American genome consortium (which includes Canada, the US, the UK, Italy, Switzerland and Japan). Much of the early work went into assessing the correlation between breeding values derived by genomic analysis and those by traditional progeny testing.
A high degree of correlation was found, so we can confidently use genomic evaluations of bulls to predict breeding values. However, while the general correlation is good, it is worth noting some individual bulls perform considerably better or worse on genomic evaluations when compared to a progeny testing calculated breeding value.
In genomic testing, young bulls can be sampled at birth and an estimate of their genetic merit can be made at a 65% reliability level. To put this into perspective, a 65% reliability is analogous to an 80% accuracy of selection (reliability = accuracy squared) – a number not uncommon in elite beef and sheep breeding programmes.
For a moderately heritable trait, such as milk protein yield (heritability = 0.2), genomic testing provides a similar amount of information as would be gathered after the data of 32 daughters has been gathered. For traits with lower heritability – for example, reproductive performance traits – genomic testing can provide the similar level of accuracy as would otherwise be available, only gathering data from around 130 daughters.
This means genomic testing on a day-old calf can predict its genetic merit, as well as years of progeny testing – more so than what occurs today for many elite beef bulls and rams.
Using practical examples of how much information can be gathered from one test performed on an animal immediately after birth can be very useful in helping farmers understand the power of genomic selection.
Another clear benefit of genomic testing is, at the herd level, farmers can – for a relatively low level of investment (approximately £30) – have genomic testing performed on their replacement heifers.
This means heifer selection can become a highly focused, quantitative exercise, meaning either the number of heifers reared as replacements can be reduced considerably, or all but the best heifers can be reared and sold. At a time when margins are being squeezed, knowing, with a high degree of accuracy, which heifers will succeed in a herd and which will not can greatly affect efficiency of production.
The equation for genetic progress per year in Panel 1 shows generation interval is a key driver of genetic progress. Where progeny testing is used, it is unlikely the offspring of determined elite sires will have a generation interval below four years.
The generation interval is half the combined ages of the progeny’s parents at its birth. In the case of a heifer calving at two years old, to a sire proved to be elite at six years old by progeny testing, the generation interval will be around 4.5 years (the dam is aged 2 years and the sire 7 at the time of the progeny’s birth).
Genomic testing allows the generation interval to be reduced to around 2.5 years (2 years for the dam and 3 for the sire), which, if all other factors were equal, would allow for a doubling in the rate of genetic gain on a farm.
Unfortunately, things are not that simple, as the reliability of the predictions of genetic merit determined by genomic testing is not quite as high when the bull is unproven as a progeny-tested elite bull.
Therefore, the advice from the majority of experts inthe field is herds should choose a larger number of bulls than they would have traditionally chosen with progeny-tested bulls.
A typical UK dairy herd should be using around eight bulls to maximise the potential genetic gains available through genomic selection, while reducing the risk of an inaccurate bull selection damaging the entire herd’s genetic merit.
Ideally, bulls chosen will vary in age – around half will be in their first year, a quarter in their second and the others older (more proven bulls).
One noticeable effect of genomic selection is fewer bulls becoming AI sires. In Germany, for example, the number of Holstein AI sires recruited has fallen from more than 1,000 bulls to fewer than 400, annually. This poses the very real danger of a reduction in genetic diversity, as fewer bulls are selected using ever-more demanding criteria. This means – in theory, at least – the genetic variation could be substantially reduced, eventually resulting in diminishing genetic gains.
However, so far, evidence seems to suggest the number of genes involved in each economically viable production trait is so large, this is unlikely to be a real risk if breeding indices are used, rather than specific breeding for a single production trait (such as milk yield).
Indeed, it is likely as information on the genome expands and knowledge increases, breeding indices will become more complicated, as traits with a low heritability – such as fertility or lameness – will increasingly be added to selection indices and bred for.
Other problems exist with genomic evaluations. For example, fewer bulls have two generations of daughters, which means genetic ties between generations will prove harder to find. This will, therefore, make intergenerational assessments of genetic change harder to accomplish. Other drawbacks include a reduction in the number of niche sires that had enough – 50 to 100 – daughters to be proven, but not the characteristics required to become mainstream elite sires.
Perhaps, worryingly, a real risk exists that larger breeds with the resources and population size to use genomic testing may accelerate away from smaller breeds without the population size to compete with the very rapid genetic progression seen in the genomic selected breeds. However, little evidence exists of this happening at the moment, and a number of the smaller, but well-known, breeds – such as the Brown Swiss, Norwegian Red and Swedish Red – have invested in genomic breeding programmes heavily and look to be in a position to continue this investment.
To summarise, the impact of genomics in cattle breeding has been enormous, given the technology is still in its infancy. The potential to make on-farm decisions about individual animals’ genotypes, combined with the population level impact of increased selection accuracy and reduced generation intervals, has increased the potential of cattle populations.
The question as to how much we limit a cow’s genetic potential through suboptimal management and exposure to endemic disease remains; however, without the genetic potential to perform, the phenotype will almost certainly not.
