Written by Matt DiLeo
I’ve been meaning to tell this story for some time. It’s a good example of how not all biotechnology is genetic engineering.
In trad breeding, the breeder/gardener simply crosses two parents that show great (and complementary) traits, grows up the offspring, selects the best and repeats. It’s effective, slow, labor intensive and limited by the perception of the breeder. Most traits are also very heavily impacted by the environment, so each new genotype must be grown in multiple locations +/or multiple years to make sure the recorded phenotype is due to the genetics (not the environment) of the individual. Most of our crops were domesticated and refined this way (quite a success!). Modern breeding has additionally been refined by the development of various statistical techniques and crossing schemes that make the whole process more efficient.
MAS relies on the development of “markers” that co-segregate with traits of interest. Picture a chromosome: an incredibly long stretch of DNA with genes located occasionally along its length. You can develop molecular markers (e.g. SNPs or microsatellites) that act as signposts along the whole length of the chromosome (where each of the signposts look different in each parent). Since sexual recombination moves DNA in big chunks, lots of the nearby markers will be moved with each gene. Statistical techniques can then be used to see which markers are associated with each trait. Since markers that are physically close to the gene along the chromosome are more likely to move with it during recombination (and since you know where on the chromosome each marker sits) you can narrow down where the gene that causes your trait is and then actually have a shot at identifying it! Or you can just use the marker to help make sure your gene moves where you want it to. This has been an extremely useful tool to complement trad breeding but in practice it’s often impossible to pin a trait on just one or two genes (e.g. human height, last time I heard, was associated with huge number of genes that together only explained a small amount of the total population variation – even though it’s extremely heritable).
Genome Wide Selection
In GWS, the breeder doesn’t even bother to try to identify which traits or genes a marker is associated with. She simply picks a population of her crop or livestock and measures each individual with huge numbers of markers. She uses statistics to see which markers are “good” and which are “bad” and decides how good future offspring are just by their combination of markers. New developments in biotechnology are making marker development and measurement absurdly affordable – which makes phenotyping (growing and measuring offspring over multiple sites/years) the bottleneck in many breeding operations.
This is where it gets really interesting…
Even if the markers aren’t as efficient at recognizing “good” offspring, you can more than make up for this with shorter generation times. A typical maize breeding operation will need to grow each generation in multiple sites in some representative climate (probably the Midwest) to see which individuals/lines are really the best. With GWS, you can ship the whole lot to some tropical location and grow three generations a year (picking the best in each round with markers)!
From what I’ve heard this is most advanced in the dairy industry. Artificial insemination (AI) has been a huge advance in animal breeding because dairymen can simply order semen from the best bulls in the country instead of keeping their own mediocre bulls on site.* Breeding elite bulls is BIG business. Currently 9 million Holstein cows in the U.S. are bred with AI from just 500 bulls!** Bulls need to get “proved” to access this market. Traditionally, the quality of a bull was determined by seeing how much milk its female relatives produced (b/c milk quantity is what matters to dairymen). This process traditionally involved waiting for each individual young bull to grow to reproductive maturity, produce several rounds of daughters, let the daughters mature, mate them and then measure their milk production. This took years and cost about $50,000 per bull. Now a genetic marker test give you just as much information about a male calf the day it’s born for just $250!
Dairymen are really excited about this. There’s been talk of developing a marker certification system for dairy bulls for 20 years but only now is the technology cheap and effective enough to make it work. From what I’ve heard, the U.S. government now runs a certification program (AIPL I think…) that will assign official breeding values to any cattle DNA that a farmer sends in. I bet they’ll be a lot more farmers in the bull semen business now!
* I once worked with a guy who did dairy AI. I’m all for getting my hands dirty, but that doesn’t include anything that comes with gloves that go past your elbow…
** Hopefully the animal breeding community is as on top of preserving unique germplasm as the plant breeding community is.
Bernardo, R., & Yu, J. (2007). Prospects for Genomewide Selection for Quantitative Traits in Maize Crop Science, 47, 1082-1090 DOI: 10.2135/cropsci2006.11.0690
Schaeffer LR (2006). Strategy for applying genome-wide selection in dairy cattle. Journal of animal breeding and genetics = Zeitschrift fur Tierzuchtung und Zuchtungsbiologie, 123 (4), 218-23 PMID: 16882088
Written by Guest Expert
Matt DiLeo has a PhD in Plant Pathology from UC, Davis. During his postdoctoral research at Boyce Thompson Institute, he researched unintentional effects of genetic engineering. Matt builds R&D teams and biotech platforms: genome editing, gene discovery, microbials, and controlled environment agriculture.