How Many Strains in a Farmhouse Yeast?

<< 2026-04-12 11:48 >>

The new Verstrepen lab paper on farmhouse yeast (blog post introducing it) gives us a level of insight into these cultures that we've never had before. And one of the things it gives us a much stronger handle on is exactly how many strains there are in each farmhouse yeast culture. (If you don't know what I mean by farmhouse yeast, see this example.)

Several labs have done work on this before, and their results were used in earlier versions of the farmhouse yeast registry. Generally they find 2-4 strains, at most up to 10. However, different labs have found different results. For Terje Raftevold's kveik, #5, the NCYC found 8 strains. However, the de Proef brewery found 17 strains. And Carlsberg told me they found 42. So, how many strains are there really?

Since this paper took 40 random samples from every one of the 44 cultures studied, and then did PCR fingerprinting to tell which samples were the same strains we get far better data on this than we've ever had. (Explanation of the method.)

Figure 1B gives an in-depth overview of the composition of the cultures:

Figure 1B from the paper, showing the contents of each culture.

Each bar is one culture, broken into one box per strain. You can think of this as being a row of 40 boxes, one per sample, but with adjacent boxes merged where they are the same strain. So you'll see on the left a lot of the boxes are long, while on the right most of the boxes are really small. This shows you how much of the culture is made up of different strains.

The red boxes with strain codes are the strains that were sequenced (more on that later). The strain codes are: "[number of culture]R[sample number]". The culture codes come from the farmhouse yeast registry. The boxes with hatching are the strains of outside yeast, that is, non-farmhouse yeast that's made it into some of these cultures.

As you can see, the compositions vary wildly. At the bottom are some cultures that contain just a few strains. #43 has 3 strains, and a single strain makes up almost the entire culture, while AE has two strains, but they make up about 60% and 40%, respectively. At the top, #5 and #68, have 33 and 32 strains, and none of them make up even 10% of the culture.

(If you're now wondering how Carlsberg could find 42 strains in #5 if this analysis says it has 33, hold that thought. We'll come back to it.)

So how many strains are there really in a farmhouse yeast culture? The histogram below summarizes what we know.

Histogram of the number of strains per culture. The sideways axis is the number of strains, and the height of each bar shows how many cultures have that number of strains.

What you see is that the number of strains is really quite evenly distributed. Having 5 strains is as common as having 25, but having 30 or more is a little unusual.

But How Many Strains Are There Really?

What's always been bothering me about these analyses is that there's no way to tell if we've dug deep enough to hit the bottom. Think of this as having a swimming pool full of ping-pong balls in different colours. Blindfolded, you pick out a certain number of balls, then go into a different room and count how many different colours you found. How do you know if you got all the colours? Well, unless you empty the swimming pool you can never be sure. It's always possible that the last ball is a different colour.

Obviously, you can't check every cell in a yeast culture, since there are trillions of them. But how many do you need to check to get a good estimate? We don't know — it depends how many strains there are, which is exactly what we're trying to find out. This bothered me sufficiently that in the supplemental note to the paper I dug deeper into this.

Let's say a brewer harvests 100g of dried yeast from their beer. Richard Preiss estimated the number of cells per gram, which allows me to estimate that that might be 1.4 trillion cells. So roughly 170 cells for every human being on the planet. Taking 40 cells from that means sampling 0.000000003% of the culture.

Clearly there is a possibility that we didn't get everything. They took 40 samples from each culture. That means a strain making up 1% of the culture has a two-thirds chance of being missed. At 0.1% it will 96% certainly be missed. At 0.01% 99.6% certainly.

To put this another way, let's say you come back from the swimming pool with 40 balls, and find that 33 of them are in different colours, as we did with #5. You obviously can't conclude that there's definitely no more than 33 different colours. Odds that you missed some are very high, but how to get a handle on how many were missed?

Since we deduplicated the strains across all the cultures (more on that later) I figured I could take a look at how many times each strain was picked up.

Figure 10 from the supplemental note. Sideways axis shows how many times each strain was found, and the height of the bars is how many strains were found that often.

As you can see from the diagram above, nearly all of the strains were found 5 times or less. In fact, 64%, nearly two thirds, of all the strains were only picked up once. That makes it pretty obvious that some strains were not picked up at all. But how many? Well, we don't know. But let's say the lab had taken 120 samples from each culture. What would the shape of this graph be then? It's completely possible that the shape would be the same, in which case we'd have at least twice as many strains.

It's completely possible that the true number of strains is ten times or even a hundred times more than what this analysis has found. Or maybe just twice as many. Until someone takes enough samples that the peak moves up to at least 3 we will not know.

Is There a Dominant Strain?

There's been a lot of speculation over whether cultures of brewing yeast have a dominant strain or not, and many microbiologists have taken it as a given that obviously yeast cultures must have a single dominant strain. We had that discussion in the group while we were working on this paper, and I thought "well, maybe we could answer that with the data we have."

Part of Figure 1C from the paper. It's pretty tricky to read, but the text will explain.

How much of a culture does a strain need to make up to be considered dominant? Well, I didn't want to argue over what proportion to pick, so I did a graph across all proportions. The result is a figure that's not trival to read, so let's start with the edges. On the far left, at 0.0 propotion the curve shows what percentage of the cultures have a strain making up 0.0 (nothing) or more of the culture. Of course, they all do: 100%. On the far right, we see how many cultures have a strain making up all (1.0) of the culture. Of course, none of them do, 0%.

Let's say you think a dominant strain needs to make up at least half the culture. That will be at 0.5, and we find that something like 35% of the cultures have a strain that dominant. But maybe to be really dominant a strain needs to be two thirds of the culture? Well, at 0.67 we find that about 17% of the cultures have a strain that dominant. About 5% of the cultures have a strain making up 90%, being totally, utterly dominant. Is one third of the culture enough to be dominant? Well, then about half of the cultures have that.

If you want a really short summary of this I'd say most cultures don't have a dominant strain, but quite a few do.

Wet or Dried?

Throughout the project one thing we wanted to check was whether it made any difference to the number of strains whether the brewer preserved the culture wet or dried. I already had information about what most of the brewers did, and managed to collect it for a few more. But every time we tried to check the result was the same: no difference.

I tried the analysis above, separating the cultures by dried/wet, and again found no difference. But then Andrea Del Cortona suggested we separate the cultures not by how the brewer preserved the yeast, but by how the Verstrepen lab received the samples. I thought this couldn't matter, but tried anyway.

You see the result below. Obviously the preservation of the sample made a big difference.

Part of Figure 1C from the paper, this time separated by whether the Verstrepen Lab received the culture liquid or dried.

What you're seeing here is that, no matter how you define what makes a dominant strain, the liquid cultures have a lot more of them. In other words, diversity is greater in the dried cultures than in the liquid cultures. You see this in the number of strains per culture, too. The dried cultures had a median of 22 strains, while the liquid ones had a median of 14 (see Figure S3). We tested it statistically, and it's definitely significant (p=0.0013, meaning odds of getting this result at random is one to 783).

Plot of culture diversity. Each dot is a culture, with the colour indicating which group it's from (white = outside yeast), and the height giving the number of strains. Position on the X axis is random.

This means that if you want to preserve a farmhouse culture, the best way to do it is to dry the yeast. Keeping it in liquid form means you lose more of the the yeast, and more of the diversity in the culture. Drying also has the benefit of killing off a lot of potential contaminants, like bacteria and modern brewers' yeast.

Note carefully: this advice is for farmhouse yeast. If you try drying a Beer 1 yeast you'll almost certainly kill it.

What this also means is that probably this analysis has not seen the full complexity of many of these cultures, because they were stored and sent wet, and so parts of the cultures died before the lab was able to look at them. So the strain counts here are in many cases probably too low, and the estimate of how many strains in total is probably also too low.

How could we be so unprofessional as to collect and mail around liquid samples? Well, for one thing we didn't know about this result, but the samples were also collected by unpaid volunteers. There's limits to what we could do.

How Are the Strains Related?

We have a lot of data (coming in a later post) about how the different yeasts relate to each other overall, but I think it's also interesting to look at what the structure is like inside each culture. For some of the cultures it's clear that all of the strains are pretty similar to each other. For other cultures, however, things are very different.

Figure from the Verstrepen Lab, not used in the paper, showing the family structure inside #17 Midtbust, based on the PCR fingerprinting.

The diagram shows the result for the culture called #17 Midtbust, based on the PCR fingerprinting. The dark bands are the actual, optical fingerprinting results for each strain (shown as a column). Above that is the family tree built from the fingerprints (which, to be clear, is not as reliable as a tree based on whole genome sequencing).

What the diagram shows is that this culture consists of several different groups of strains. On the far right is one group of 6 strains, then on the far left is a few more groups. In the middle it kind of gets hard to say, but there's at least one more group with a whole bunch o different strains.

In short, the relationships between the strains in this culture are very complex, and the flavour they make during fermentation is likely a mix of what all the strains produce.

The first batch of Bygland Framgarden was one of my favourite beers. Apart from the juniper and the raw ale flavour, it had a beautiful fruitiness, and I particularly loved the subtle aroma of pineapple that kept peeking through, then disappearing, then coming back. It made me think of the sun coming and going behind clouds, but inside my mouth.

In the second batch that subtle pineapple aroma was gone, and the complexity was not the same. I asked the brewer, and he said the difference was that the first batch was made with the original culture from Petter Øvrebust, but the second was fermented with a commercial single strain version.

So, yes, it matters. At least sometimes.

(And, no, we're still not done going through the results from this paper. More is coming.)

Similar posts

Comments

No comments.

Add a comment