Biology

Grape Parentage from DNA: How Microsatellite Profiling Works, and the Gouais Blanc Problem

Cabernet Sauvignon's parents, Chardonnay's despised mother, Merlot's lost one. How microsatellite profiling identifies a vine's parents, and where it stops.

Cabernet Sauvignon is a cross of Cabernet franc and Sauvignon blanc. John Bowers and Carole Meredith established it in 1997 from DNA alone, with no breeder's notebook to check it against, since nobody is on record as having made the cross. The result is now repeated everywhere, usually without the method behind it or any sense of how far that method can be trusted. The answer to the second question differs from one entry in the pedigree to the next.

Why a variety has parents

A grape variety is one seedling, multiplied ever since by cuttings. Every Merlot vine is a piece of the first Merlot vine, which is why ampelographers speak of a variety's birth. At some point a pip germinated, somebody judged the plant worth keeping, and propagated it. That pip had a mother, the vine that bore the berry, and a father, the vine that supplied the pollen.

Cuttings do not recombine DNA. If the two parents were also kept in cultivation, they are still growing somewhere with the genomes they had on the day of the cross. Parentage analysis works because viticulture keeps its own ancestors alive.

Reading a microsatellite profile

The tool is the microsatellite, or simple sequence repeat (SSR): a site in the genome where a short motif is repeated a variable number of times. The count differs between individuals, so the DNA fragment spanning the site differs in length, and that length can be measured to the base pair. A vine carries two copies of each site, one from each parent. Its profile at one marker is therefore a pair of numbers.

Profiles have two separate uses.

Identification. Two vines with the same pairs at every marker are the same variety, whatever they are called. Collections need this constantly. When Sean Myles and colleagues genotyped the USDA grape collection in 2011, 551 of its 950 vinifera accessions turned out to be clones of at least one other accession. The Vitis International Variety Catalogue (VIVC) currently lists 233 synonyms for Gouais blanc alone.

The numbers must also match between laboratories, and at first they did not. In the 2004 study published by Patrice This and colleagues, ten laboratories in seven countries profiled the same 46 cultivars at six markers. They did not always obtain the same absolute fragment sizes, and in some cases not even the same relative ones. The fix was to stop trusting raw sizes and to code each allele against reference cultivars that any laboratory can grow. Those six markers, later extended to nine, form the core of the profile VIVC publishes for each variety. They are an agreed standard, not a finding. Without them, two laboratories could not be sure they were describing the same vine.

Parentage. The rule is Mendel's. At every marker, a child received one of its two alleles from its mother and the other from its father. A proposed pair of parents is excluded as soon as one marker shows an allele neither could have passed on. Surviving a few markers proves little, since common alleles are shared by chance. Surviving thirty is different, and the last step puts a number on it: a likelihood ratio comparing "these two are the parents" with "two random vines, or close relatives of the candidates, are the parents".

Cabernet Sauvignon, 1997

Bowers and Meredith, at the University of California, Davis, were not looking for parents. They were building profiles to identify California's commercial varieties. According to the university's account of the work, the comparison covered 51 varieties at 30 markers, and at every one of them Cabernet Sauvignon's two alleles could be explained by one from Cabernet franc and one from Sauvignon blanc. Their Nature Genetics paper reports that likelihood ratios support the parentage "to a very high degree of probability". It notes that the link to Cabernet franc had long been suspected, whereas Sauvignon blanc's contribution, "despite its similar name, is a surprise".

The paper establishes who the parents are. It does not say when or where the cross happened: the abstract states only that the variety has been grown in Bordeaux since at least the seventeenth century. That the cross was spontaneous, between neighbouring vines, is an inference from the absence of any record of deliberate breeding at that date. Nor do microsatellites alone say which parent was the mother.

Sixteen siblings and a despised mother

In 1999 the same Davis group, now working with Jean-Michel Boursiquot and Patrice This on the French national collection near Montpellier, reported in Science on a screen of more than 300 cultivars. Sixteen varieties of north-eastern France, among them Chardonnay, Gamay noir, Aligoté and Melon, have genotypes consistent with a single pair of parents, Pinot and Gouais blanc, tested at 32 markers. Both parents were widespread in the region in the Middle Ages, so the same cross had many chances to happen again.

This is the Gouais blanc problem. Pinot is the archetype of a noble variety. Gouais, in the account the geneticists take from the ampelographic literature, was a high-yielding peasant grape, poor enough that there were attempts to ban it. The genetics papers cite that history; they do not document it. Half of Chardonnay's genome is Gouais all the same.

The 1999 authors proposed that the pairing succeeded because the parents are genetically distant, so that the offspring gained hybrid vigour. It is a reasonable hypothesis, and no study has tested it. Frequency alone could explain the result: two varieties planted side by side across a whole region for centuries will leave many seedlings.

The mother question was settled in 2010. In Vitis, chloroplasts pass through the seed parent, so chloroplast DNA marks the maternal line. Harriet Hunt and colleagues at Cambridge typed chloroplast markers in Pinot, Gouais and twelve of their offspring. Nine, including Chardonnay, Gamay noir and Aligoté, carry the Gouais chloroplast. Three carry Pinot's: Aubin vert, Knipperlé and Roublot. The sample is small. But the study's two kinds of chloroplast marker agree on every variety, and the inference rests on a known mechanism of inheritance, not on a statistical fit. The crosses went both ways. In most of them, the despised variety bore the fruit.

The list of offspring has kept growing. Erika Maul's group at the Julius Kühn-Institut, which maintains VIVC, counts Riesling among the offspring of Gouais (Heunisch Weiss in German), and in 2015 added five more, confirmed at 35 markers.

Merlot's missing mother

Parentage analysis can only find a parent that has been sampled. For years Merlot had a known father, Cabernet franc, and no mother. In 2009 Boursiquot and colleagues found her in a vine that no ampelographer had described, sampled first in northern Brittany and then at four sites in the Charentes. Inheritance at 55 nuclear and three chloroplast markers makes it Merlot's mother. From the name its growers used and from the literature, they called it Magdeleine Noire des Charentes. They also showed it to be a parent of Cot (Malbec), with Prunelard as the other. A variety absent from every catalogue turned out to be central to the history of Bordeaux's grape varieties.

Missing parents are common. In 2013 Thierry Lacombe and colleagues ran all 2,344 distinct genotypes in the Montpellier collection against one another, at 20 markers, accepting only logarithm-of-odds (LOD) scores above 18. They resolved both parents for 828 cultivars. For 1,087 they found one parent only, the other being absent from the collection. For 276 they found no direct relative at all. Much of the old varietal range was abandoned after oidium, phylloxera and mildew, when growers replanting on grafted vines kept only what paid. Some of the missing parents are probably extinct.

How far to trust it

The 2013 study doubles as the best audit of the method. It confirmed 100 parentages previously published from markers and invalidated 32. Against breeders' own records, it confirmed 255 pedigrees and invalidated 126. The breeders' written pedigrees were wrong about a third of the time, and roughly a quarter of the earlier marker-based claims failed on the larger dataset.

In 2018 Valérie Laucou and colleagues checked the microsatellite results against 10,207 single-nucleotide polymorphisms (SNPs) in 783 cultivars. Of the full two-parent trios proposed from 20 microsatellites, about 5 percent failed. Of the one-parent "duos", 19 percent failed.

That gap is the practical point. In a trio, every allele of the child is accounted for by two named parents, and the well-known trios rest on thirty markers or more. A duo is weaker. Two vines that share an allele at every marker may be parent and child, but microsatellites cannot tell which is which, and with few markers they cannot always rule out siblings. Myles's team, working from 5,387 SNPs, judged that barely half of the first-degree links in their sample were parent–offspring rather than "sibling or equivalent". They called full genealogical reconstruction intractable for a crop with selfing and crosses between generations.

Cabernet Sauvignon, the Pinot × Gouais family and Merlot are trios, and the Pinot × Gouais family has also been cross-checked with chloroplast DNA. The statements of the form "X is a parent or offspring of Y" that fill the rest of the published charts are hypotheses of a lower grade. VIVC records, on each variety's page, whether its pedigree has been confirmed by markers.