Andony Melathopoulos: [00:00:00] Astute listeners of this podcast will know that I have a deep and embedded passion for the breeding of honeybees. I find the whole topic fascinating. It's like unlike breeding in any other system, for one thing, most of the traits are expressed at the colony level, not at the individual level making, measuring those traits very difficult in.
It's hard to trace pedigrees. The, as many if you remember the episode with Julia Mahood, the mating takes place in the air. It's hard to control the mating. There's multiple fathers. It's really complicated and that's where using molecular tools for identifying genotypes of importance are really, I.
I find the whole area of molecular biology in contrast, very confusing, frustrating, and difficult to understand, and that's why I am so excited to introduce our guest, Robert Lou. Now Robert is currently going into the second year of his Masters of Science in Biology at the University of Alberta, and I caught out with him in the Entomological [00:01:00] Society of America meeting in Vancouver, British Columbia, where he won an award for a student presentation.
It was very impressive. I was. Totally amazed at how he was able to break down some very complicated concepts into something that someone like myself could understand. He's in the, in LF Ruples lab at the University of Alberta, and this week he's gonna tell us about his research into breeding using these new molecular Technology, molecular biology technology, molecular biology tools, , that's how complicated it's folks, molecular biology tools to be able to select for virus resistance in honeybees.
Maybe one of the most complicated phenotypes that are out there. So this week in pollination, at pollination, we. Deep dig deep
Robert Lu: down
Andony Melathopoulos: into the wonderful world of bee breeding with molecular tools and markers. Assisted selection this week on pollination.[00:02:00]
Okay, Robert, welcome to Pollination.
Robert Lu: Great to be here. Thank you for having me
Andony Melathopoulos: now. You were at Entomological Society of America. It is foggy. We're overlooking broad inlet here. Supposedly. Supposedly you can't see anything. But we're at the tail end of the Entomological Society of America meeting.
You were in the student competition. That is correct. And you, it was really impressive. You gave a great talk and it was won an award here in the student competition. So congratulations. Thank you.
Robert Lu: I actually did not know I won the award. Because, I was thinking, like you said, today is the last day of the conference.
Of course the award ceremony will happen then, but it actually happened yesterday. And so I had no. , any of this was taking place until a buddy of mine who had attended, because he's more responsible than me and actually looked at the program he texted me and was like, congrat, congratulations on your award.
And I was like, what award? And yeah, it just snowballed into me trying to figure out how exactly stuff went down. But,
Andony Melathopoulos: It was an impressive talk and I think you did a really good [00:03:00] job. You if, for those of you who've never. A scientific talk at a conference. You only get 10 minutes to explain something intricate.
And you had wonderful animation and elements to describe what we want to talk about today. Breeding for virus resistance in bees. And maybe just to start with, to get our listeners on, on some footing. It strikes me that, there's some different, there's livestock breeding. We consider bees as a form of livestock.
There's livestock breeding that's far advanced. I always remarked on this, you can pick up a catalog and you can, the expected breeding values of a, a straw of semen that you buy and you know that your the cows that are bred to that are gonna have a few more pounds. And they're gonna be that color.
And you have these kind of detailed information, which was done over, decades of data collection and pedigree data sets to be able to figure out what would happen if you crossed this with that. And I know honeybees don't [00:04:00] have that. Luxury because it's very difficult to control the crosses.
But this whole area that you're talking about of genetic tools seems to allow us to leap decades over, and catch up a little bit. Can you just give us a brief overview of of how you can use genetic. To be able to infer the relationship of a trait to some underlying genetic thing that you can
Robert Lu: select on.
Certainly. So I'm glad you brought up. Other livestock? No I should clarify it technically, bees are livestock because we treat them as such. These are animals that we're ferrying around to different crops such as blueberries, and we release the bees and they pollinate for us and produce our fruit, and in turn, they also.
Produce honey, which is, as a very sweet award for the beekeepers to go through the difficult process of managing bees and getting stung by them. Yeah. Which does happen even with the protectives coding a lot, [00:05:00] a lot more than I would like to admit, but yeah, it happens. But anyway so the interesting thing about bees, maybe I shouldn't say interesting, but the reason why livestock, For bees in particular is so important is you have these large livestock animals like.
Mammals such as your pigs and your cows and your birds, like your chickens. Yeah. Advanced vertebrates as the taxonomists will sometimes refer to them. These are animals with very highly evolved immune systems that are able to adapt to infections by certain viruses. . . And in doing so, they're able to fight off these viruses and we can give them medicine that helps them in this process.
But bees don't really have that luxury. You can vaccinate them all. Exactly. Yeah. Yeah, exactly. And bees unfortunately don't really have this luxury. They don't, they have a very minimal adaptive immune response as far as we can tell. And they, which is honestly already better than most insects that don't have an adaptive immune system.[00:06:00]
but primarily they rely on these cells of the innate immune system that just engulf anything that they find that they identify to be a foreign pathogen, for example. But a lot of these viruses are very good at circumventing these, , oh. . Yeah. So a lot of these viruses are very good at circumventing these innate immune responses and make it very difficult for the bees to fight off.
So a lot of the time these viruses are very severe. They'll just quickly kill the bees and. This is also compounded by the fact that in recent years there was a mite called Ver Variety Struc, which jumped in from the eastern honeybee, a API Sana into our Western honeybee, a APIs Mora. And that brought with it a whole.
Slurry of viruses, which are now emergent threats that are rapidly propagating through colonies of the world.
Andony Melathopoulos: Okay so this point, this is almost the imperative for breeding. So not only are, bee breeding is behind, [00:07:00] but there's also these challenges that other livestock systems can deal with much more effectively because there's an immune system in the, in them, whereas honeybees have to rely on, I guess the frustration that beekeepers have is they have these mites and these viruses floating around.
There's almost seemingly nothing they can do about it.
Robert Lu: Exactly. Yeah. And the, we have to rely on. To be able to kill off the viruses themselves because we can't just give them some sort of medication like we can with cattle. We can't just dose 'em with antibiotics, for example. Not that doesn't have problems in its own right, but it's certainly a more immediate solution than having all your bees die off and then trying to find a stock that doesn't just die off.
So it is, you're right, it is much more imperative that we. resistance or rather survivorship. I should choose my words carefully and I'll elaborate. Okay.
Andony Melathopoulos: And so this problem is let, we'll get into survivorship in just one second. Yeah. But this problem is the solution has appeared in terms of molecular tools that analyze the DNA [00:08:00] and associate it with traits.
Tell us how that
Robert Lu: process works. Yeah, so in theory we know that if a trait such as survivorship when infected by a virus . If that trait is genetic invasive, then throughout a population we would see that at certain genes, variation in the individual's genotype of the bees will correlate with variation of a phenotype.
For example, you may have. This one B here with a particular genotype that correlates to high survivorship or the ability to survive viruses better than say another B with a different genotype. That is much worse at surviving infection by some virus. So it dies quicker. Exactly. Or you have a popul.
with different genetic compositions that a lot less of them die when a virus rampages through them. Okay, gotcha. Than another population. It varies depending on [00:09:00] how we look at the trait.
Andony Melathopoulos: Define it. So the word trait and phenotype, they're
Robert Lu: synonyms. Not quite, they're very closely related, but it, whereas, so whereas a trait is a very general term of what characteristic we're looking at.
Yeah. The phenotype is how. Trait ex expressed in an individual. So as a perhaps easy example, we can talk about something that most people should know about Gregor Mendel and his pea plants. Oh, yeah.
Andony Melathopoulos: It's very famous that late 19th century monk who did these crosses. And understood inheritance through looking at patterns of variation in these
Robert Lu: peas.
Exactly. So what, one thing that Mendel noted was that the color of the peas could be dependent on the parents that you use to breed that pea plant Uhhuh. . And so you would have, for example, green and yellow peas. Yeah. Now the trait that Mendel was looking at was pea color. Pea color. Oh. [00:10:00] But
Andony Melathopoulos: they, you had yellow ones and.
Green one. Yeah. Different. Those were the phenotypes.
Robert Lu: Exactly. The phenotype were the different ways that All right. Gotcha. They express an individual. And likewise, genotype is different from genes. In a similar sense. The gene is a functional region of the genome. This physical bit of DNA that does something, for example, in most cases, encoding for a protein, or I shouldn't say most cases, but in a lot of cases it produces a.
Now this gene, just like the trait can vary across a population and how that gene varies, is known as the variations of the gene is known as alleles, uhhuh and most organisms not most again, but many organisms that we're interested in, such as bees have two copies of the genome each with one allele of a part.
. And so the genotype refers to the particular alleles, the particular variants of this gene that are in an individual. . So when I'm talking about genotype and phenotype, [00:11:00] it's specifically referring to how the genetic and trait components of what we're testing for present within a individual organism.
Andony Melathopoulos: Okay, so we got that out of the way. So how do you, I imagine you can't just go with the microscope and look at this variation in these in the genetic sequences of these individuals that are ex exhibiting different survivorships. How do you find them? Yeah,
Robert Lu: In theory we could do that, but I would say it would take you forever to, but real reality is that you'd run out of money before you were able to screen every single.
gene in the genome to try to find one that correlates to whatever trait you're looking for. Uhhuh, . So we can do something called quantitative trait, Loki analysis, loci analysis or quantitative trait locus analysis. Locus is just the singular word for loci and what a locus is. This is allele? Not quite.
Okay. Alright. To clarify, [00:12:00] locus. Is not the grasshopper that swarms and eat all the different plants. It, it doesn't have the tea at the end, like the location. A locus is a particular location. It's somewhat arbitrarily defined based on our purposes, Uhhuh . But it's essentially just. A region of a gene.
Okay. So an allele or a gene could certainly be a locus, but generally we take larger chunks of the genome as the low size. Okay.
Andony Melathopoulos: Okay. We just break this down though. Quant, why is it quantitative? And what's the trait have to do is three words in this thing that you told, you've told us
Robert Lu: about.
So a quantitative trait is a trait that. Expresses quantitatively throughout a population survivorship. Survivorship.
Andony Melathopoulos: Live 14 days, live 24 days. It's a continuous, quantitative, measurable trait.
Robert Lu: Exactly. Gotcha. Okay. And so the loci that correspond to that quantitative trait are these regions of the genome.
That will vary and that variation will correlate to the variation of the trait. [00:13:00] Now, how do we know whether Locus varies or not? While we can use these things called genetic markers, which are essentially regions of non-coding dna, so DNA that. Essentially his junk DNA doesn't do anything but it There's junk dna.
Yeah. A lot of, in fact, most of our genome is junk dna. Really, I believe something like 50% is, or more is just one type of junk dna called the transposon that is a, essentially a gene that replicates itself and then inserts itself into another region of the genome. It's very interesting stuff.
But if it, if
Andony Melathopoulos: It's junk, then selection isn't, it's stable then. Is that the idea?
Robert Lu: The idea is, Junk. It actually is instable because it doesn't really matter whether or not it mutates, right? It's not for example, the gene that controls oxidative phosphorylation, which is the process that happens in the mitochondria at the powerhouse of the cell, everyone.
That if that gets mutated, The organism is finished because it can't produce
Andony Melathopoulos: energy. Oh. So there's no consequence. So if a mutation happens, it just gets, it just, it's just drifting along
Robert Lu: in history. Exactly. . And what's great about that [00:14:00] is that there's a lot of variation within a population at these regions, which make it very easy to tell individuals apart by their markers.
But it's also very easy to tell different markers within the genome. And so we used a type of marker called a single nucleotide polymorphism, an s and p, which is just one base pair of DNA gets mutated. And that's a very common variation because it's such a very simple mutation that can happen.
Okay. So when we track the variation of the quantitative trait loci, and we see that one particular locus, or I should say when we track the variation of loci and one particular locus correlates to a trait that we're looking at, we call that locus a qtl, a quantitative trait locus, because it.
statistically related to the quantitative trait that we're measuring and the relevance of that locus now is that we can infer that somewhere on that region of dna. , there's probably a gene that [00:15:00] encodes for the trait we're looking for now, to be fair, a single locust can still contain hundreds of gene.
But that's a much smaller amount of genes that we'd have to screen through than say the millions in a full genome. Is it
Andony Melathopoulos: significant that it's so small the snmp, the single nucleotide polymorphism, that it's so small, it's like at the finest scale. Does that, is that afford some kind of functionality when you're looking for traits?
Robert Lu: So when you say functionality, do
Andony Melathopoulos: you mean, when you're looking, you're trying to find the needle in the haystack? Is it useful to have, what's the, to help me wrap my head around the concept? Yeah. To have something that's so fine. Some a very fine, as opposed to a very kind of bigger region.
Robert Lu: modern genetic techniques are very good at picking up fine. To scale details like this, and especially something like SNPs where yes, you're right, they're very small and it's very hard to hone in on one, but because they're so small, they're also all over the place. So it gives you a lot
Andony Melathopoulos: of
Robert Lu: resolution then.
Yeah it's not quite [00:16:00] the needle in the haystack sort of situation. It's more like you have a whole bunch of needles in a haystack and where they. You don't necessarily know where they are, but you can infer that if you put your hand in, you're gonna get poke a fair amount. Okay. So to say.
Alright. And so obviously we're not looking at every single s and p that exists in the genome. We can't pick them up to that level, but we can still. Quite easily pick up tens of thousands. And in fact said, which I'll talk about in a second, we had exactly that, tens of thousands of snmp. But well before you
Andony Melathopoulos: go on, yeah, so let me just get the, let me just see if I got the concept.
, you got these two bees two populations of bees or two, colony. Yeah, we've got similar genetic background and you give them some virus , and you notice that, these ones die three days earlier than these other ones. And you know that the virus, these ones have a shorter survivorship or a longer survivorship.
And then you take their, you go in and you extract out their DNA and you're looking for these, A [00:17:00] lot of the DNA will be, there'll be a lot of DNA that. Small little vari variation. So you know, this one is, you look at the two bees and this one's slightly different than this one. And which allows you to have all these needles in the haystack.
And then what? And did I get the concept right there?
Robert Lu: Yeah. Okay. Exactly. Yeah. And then we. Take a look at each of these SNPs. These, we call them SNPs, which is probably easier to say than s and p anyway. Snip, yeah. Okay. Snip. And we track their variation across all the bees that we sampled, and then we.
statistically compare them to the variation of the trait that we're looking at. Ah-huh . Okay. Now one thing to note is that despite the method we use being called quantitative trait locus analysis, QTL analysis, it actually can be used on binary traits as well. Okay. So if we're looking at something, say, just able to survive infection, or not able to survive an infection, that's a very binary trait.
Quantitative the way we would [00:18:00] traditionally think it would be, but that's actually not a problem. We can still, the concept of, okay, the genotype correlating with the trait that still exists, that correlation will still exist no matter whether we consider the trait to be binary or not. We just keep the terminology the same because that's how it was originally developed.
now the advantage of QTL analysis is that you actually don't need a very large population. You can just do a self-contained cross where you take BS of different parentage, you read them together and you create a hybrid B that you then cross with one of the parents, and then you have a whole bunch of vari.
In the lineage of your resulting bees, and you can use that variation to directly compare to the phenotype that these bees are expressing. So let me
Andony Melathopoulos: get this straight. So you've got variation in the phenotype. Yeah. But then you have variation in the, which SNPs are associated [00:19:00] with that trait? With that phen.
and that allows you to infer a relationship,
Robert Lu: right? Yeah. So the variation is going to be present in all the different SNPs, but we're mostly interested in SNPs that vary in correlation with the trait, because that is the association that we're looking for. Okay. So in our experiment, I think I think it's a more concrete example.
For example's, do it, the let's do experiment that I, we. Now this breeding part was actually not done by me. It was done by our collaborators at the U S D A, but what they did was they had a whole bunch of colonies in the B yard, and they took a bunch of worker bees from each one. And after that they dosed each B with a certain amount of virus.
In our study, it was a virus called Israeli Acute Paralysis Virus, which is a virus. That was a great tool for us because it's a very, Severe virus, which expresses the survivorship very [00:20:00] quickly. Bees will really quickly die within the span of days if
Andony Melathopoulos: they're, oh, with some viruses it can be latent or it doesn't express exactly under certain conditions, but this one's reliable.
Robert Lu: Exactly. It's very gotcha. It's generally very quick if it takes hold of a bee, and if it doesn't, then after a few days you'll know whether you'll be survived or not. And so what, and it's also found in lab grade samples. So we had a bunch of pure virus that we could work with. Now what Mike Simone Linstrom at the U S D A did was he dosed each B with I A P V, Israeli acute paralysis virus.
And he waited and he waited for three days to see how many of the bes from each hive dyes from the. Hive with the lowest amount of death. He took a queen, which he labeled as the resistant queen, and from the hive with the highest amount of death, he took a drone, which is a male pee. And then he extracted the semen from the drone and artificially inseminated the queen with it.
Interestingly enough, we can't seem to figure out how [00:21:00] to naturally mate queens and drones. . So artificial insemination is currently the method of, but that
Andony Melathopoulos: must give you a lot of control cause you can. , I in the wild, I imagine multiple males, mate with a female, but here you can get one single male inseminated.
So you've got a perfect cross between that one line where they had low survivorship Yeah. And high and make a progeny that are some kind of mixture of the two. Oh,
Robert Lu: exactly. Yeah. And the the upside of that is certainly, yeah, the control. But the downside is if you. If you're trying to cross a whole bunch of bees, it's a lot more work.
Technically hard. Yeah. Yeah. Okay. But getting back to the study so this would've resulted in a hybrid queen in theory, which would have half of her DNA from the susceptible colony and another half of her DNA from the resistant colony. So this bee was then bred back with a susceptible. Because we think that the resistant genes are, the resistance genes are going to be dominant over the [00:22:00] susceptible genes.
. And so what we did was we did, this is called a back cross because her father was a susceptible phenotype and now her mate is also a susceptible phenotype. And so what the result is a bunch of worker bees who their fathers and the chromosomes they inherit from their fathers are all susceptible, but from their.
There's some myosis going on, some recombination going on when she's making her eggs. And this scrambles the genetics of the resistant susceptible lineages. Oh, I see. Which then gives us a very heterogeneous. Makeup in our worker bees in terms of their maternal genetics. So let me
Andony Melathopoulos: get this straight.
So the goal of doing all of this cross crossing extremes and then back crossing is to really create as much, to maximize the variation available in that pool.
Robert Lu: Exactly. We want to create a very diverse set of genetics, which in. Hypothetically if our [00:23:00] hypothesis is correct we'll create a very diverse set of phenotypes.
Okay. So with these bees another collaborator was Uhhi Batia at, who was at the University of North Carolina Greensboro at the time. She took a bunch of these war piece and she. Dosed them with I A P V and after that she waited three days to see which ones would die and which ones would survive.
The bees that died were considered susceptible and they were collected. And the bees that survived were dosed again with I A P V just to increase the amount a little bit. And then we waited another two days and those that died were considered intermediate while those that survived were considered to be high survivorship.
We called it resistant at the time. But resistance has a very strict meaning, and I'll get into that a little bit later, but they were high survivorship. They. Survivors now. We took the survivors and the susceptibles and we collected them. We actually though, discarded the intermediates. And the reason why is because [00:24:00] we want to compare the two extremes and say, okay, what's going on with deeds?
Really hardy bees that deeds are very susceptible. Bees don't have in our genetics, and so we sequence them throughout their entire genome just to try to find as many SNP markers. . But we also collected a tissue sample from each beat. And that tissue sample was used to check for the presence of any other viruses.
We're measuring our survivorship based on death. . But we're also assuming that the death was caused by the virus that we're dosing them with. And the problem is if there were any other viruses within that population of these that were causing. That would confound our experiment.
That would screw up our data. Okay. Okay,
Andony Melathopoulos: so you take this sample out and you can analyze it for viruses. You've got these two extremes based on, the test, you've got the colony and you just take a few some sub sample of workers out. You feed them some virus and you confirm that these ones die very quickly and these ones die very late.
So you've got these pools, they get two sets of samples.
Robert Lu: What then well, then [00:25:00] we sequence 'em at their SNP markers. We take the tissue sample to check for other viruses, and then we run our QTL analysis. Now there are technically many different mathematical models you can use for your QTL analysis in order to find these quantitative trait loi.
The simplest of 'em is something called a single marker analysis, where you take each snp each genetic marker, and you compare them directly. You compare the genotype at that snip directly. The trait. And the idea is that each SNP represents its locust that is found on , which is bordered by other s snips.
Now this can be done with a very simple statistical test. I just give you a p value. But the the problem with this test is that these SNPs are not as evenly distributed as we may want them to be, and they're not. Tightly packed together as we may like them to be. Sometimes you have large gaps between your SNPs and if you're only using the SNPs as markers, then anything [00:26:00] between those two SNPs where there's a large gap, we can't infer anything about what genes might be going on there.
And also genotyping is very good. Like sequencing is very good, but it's not perfect and we have a fair amount of our SNPs as missing data because there was just something wrong with the sequencing that went on okay. During that time. And that's also lost data, right? Because we can't use a snap if we don't know what the genetics at that snap are.
Okay. . And so we can use a different model called interval mapping. And what interval mapping is you use the genetic markers to create a genetic map, which you can use many different software for. Most of the time QTL analysis software has that function built in. Now with this genetic map, it simulates what's going on in these large gaps using what information is already available.
Oh, I see. And in doing we can then have essentially a perfectly covered genome, uhhuh, that we can then subdivide into intervals. I see. [00:27:00] So
Andony Melathopoulos: where the single nucleotide comparison is just looking, it's agnostic to position in all of these things. This interval mapping thinks about how they are related on the genome, and that allows you to say I know.
Calgary, sorry. This is, I know that Eugene, I know that Corvallis lays between Eugene and Portland so that if the trade is somewhere, north of Corvallis, but south of Portland, you'll know that because you can look at
Robert Lu: the Exactly. We're essentially inferring what we cannot directly observe using what evidence we have.
Okay. Which to be fair is a lot of science anyway. , Yeah, this method is a very, it allows us to get a very fine resolution of what's going on in the genome, but it's, as you can imagine, very computationally intensive for the computer to take in all of our markers and spit them out. And most of the time with QTL analysis, You don't have more than say, a couple [00:28:00] hundred markers.
Yep. We had 54,000. Okay. So I tried putting that thing into, I tried putting that data set into my computer. and it basically said, no, I'm not doing that. Uhhuh . What we did then was we came up with a bit of a solution which was something called consensus marker creation. Now, what this essentially means is that we took groups of 20 markers from our raw data set of 54,000, and we used the majority rules.
Algorithm to say, okay, most of these markers say that they're coming from the resistant lineage. Some of them say they're coming from the susceptible lineage, but majority rules. We're gonna say that this marker is this genotype that relates to back to the resistant grandmother. This allows us to concatenate our data set into a data set of 2,700 markers, 2,700, really big.
Keep going, but a lot more manageable than our but [00:29:00] without getting too into the specifics of measurement we were able to then create this map of the full genome that represents a full genome that is made from these majority rules markers, and then we could control this full map. For the presence of any quantitative trait loci.
Okay. Any loci that are statistically correlated to our trait. And once we did find a particular interval or a particular locust that did correlate to our trait, we could then go back to our raw data set, take that chromosome that we found that QTL on, and use the single marker analysis on that ch. to find out exactly where this peak would be, okay.
In the real genome. And that would allow us to do things like say we could go into a gene browser to figure out if there are any genes within that region that might be, say, an immune system gene or something else that would allow us to specifically breed for this survivorship [00:30:00] against this virus.
Okay. Let's take a quick
Andony Melathopoulos: break and then we'll come back and want to hear what you found. Sure. It is. Okay. We're poor back. We're just talking about risky science. Which kind of leads us into your findings. Okay, so you've got these, you got these traits mapped using these very ingenious methods and making it compromise between computationally.
Complex but accurate, gives you more coverage in, using a rubric to adjudicate this. So you end up with, so tell us what you found.
Robert Lu: So just to re
Andony Melathopoulos: Did they map perfectly? Did you have a bunch of, I imagine. Loci that were just so tightly linked to the trait that they just shown up, like a marquee at a downtown
Robert Lu: Broadway plate.
Just to quickly recap, what we did was we sequenced bees at a bunch of different Markers in their genome. And then we use those markers to create a simulated genetic map, which we then subdivided into individual sections. And each of these sections, we compared variation across RB population [00:31:00] in that section to a variation in survivorship when infected with I A P V.
Okay. And the results were not, yeah, just not . We did not find a single. Qtl that was significantly correlated to virus survivorship of I A P.
Andony Melathopoulos: Does this I guess at one point, like we were joking at the interval like that. It's but I, does this mean that you can rule out that there's one strong.
Trait that governs this is that one of the things you can infer if it was hard to find using such a rigorous methodology, that it's unlikely that there's like a Uber trade out there
Robert Lu: now if taken by, when taken at face value, the answer would technically be yes. We could rule it out. However, before we conclusively say that we have to go back and look at our experimental design, I think it's important.
It's always important to scrutinize. How this experiment was created, because we have to keep in mind that an experiment is not necessarily reality. It's a [00:32:00] model of reality, and When an experiment tells us one thing, we can't say, this is what the reality of it tells us. We instead have to say, this is what our model of the reality tells us, and we proceed with how we apply that model to reality.
So in our situation, our model had the underlying assumption that, okay, while this colony had a lot more bees survived. When dosed with I A P V than this other colony, that means this colony. is much more susceptible than this other than the susceptible colony. But if you actually look at the data that we had, the nons susceptible colony had 70% surviving.
While the susceptible colony only had 50% surviving resistant or yeah the susceptible colony had 50% survivors while the re resist, the resist 70 quote unquote resisted. And so it.
Andony Melathopoulos: The thing is that perhaps the experiment didn't really capture the full breadth of variability that exists
Robert Lu: in the B population.
Exactly. And I think with studies like this, ideally you want to have a [00:33:00] lineage which you've bred and bred some more in order to really homogenize it. Homogenize exactly what traits, yeah, what genotypes it has, which in turn, At all your traits that are related to those genotypes should also be very homogenous.
And while 30% dying to this very severe disease is actually reasonably good survivorship, it's not that, it's not really that great.
Andony Melathopoulos: It's not. So you can't rule. That there is, because you may have not had those alleles for very high survivorship in the population to start with. So you wouldn't be able to detect them because they weren't there.
Robert Lu: Exactly. Or it's entirely possible that they were there, but they were quite distributed between the, they were also distributed within the susceptible population. And so we wouldn't be able to dis ent. The associations,
Andony Melathopoulos: right? Because you hadn't really concentrated it and so it may it, those traits may be confounded because they haven't there's still, if you, with increasing breeding, they would've left the susceptible [00:34:00] population
Robert Lu: theoretically.
Yeah. Or we would've ended up with another resistant population. Okay. That would be, and that is also how a lot. The breeding has been done historically is you don't actually know what the genes that you're selecting for are you just see that, okay, these this bowl is very resistant to some.
Disease that they get. And this cow is also very resistant. Let's breathe them together and let's take the healthiest of their young and let's breathe them together again. And that's how a lot of the selective breed occurs. And of course we're trying to streamline it, so we're not doing that. But this is one of the problems is we don't have these very nice homogenized populations to work with these pure red lines.
Andony Melathopoulos: Just maybe we should probably wrap up here, but just to return, there was silver lining so you. Fortunately somehow you thought to take those samples to see what viruses were in the bees and how did you use that to yeah,
Robert Lu: so I'm glad you brought that up. For those who've forgotten, remember that we [00:35:00] took a tissue sample from each bee that we tested to screen for any other viruses that could be killing, and we didn't find any other viruses that were high enough levels to.
Bees. But we did find that for some of these viruses in our sample we had eight viruses in total. And for some of these viruses, there was a lot of variation in the amount of virus that each be was carrying. And this is despite the fact that they all came from the same hive. So we thought let's do a QTL analysis on these viruses as well.
And for three of them Specifically deformed wing virus B cashmere B virus and Lake Sinai virus. We found that there was actually quite a significant correlation between what genetics the bee had and. how much of this particular virus that they were carrying, and this actually was a quantitative trade, so it truly was quantitative trade analysis in this case.
Now, it doesn't necessarily mean we can [00:36:00] infer anything from these results because we didn't design our experiment specifically to determine the relationship between some genetic element and these viruses. It was just a flash in the pan, oh, this is interesting. We should look at it. But to that end, I think if I were to just throw out what I think is going on or what I don't think is going, rather, what we noticed with these viruses is that the genes that were relating to the amount of virus that be had were on different chromosomes.
And like I said, most of the immune system of a honeybee is just very general. It just sees something that it recognizes a pathogen and. It because these genes were on different chromosomes for each virus. We seem to, we would have to infer that these are probably not the one size fits all immune genes that are reacting to these viruses.
Rather, it's something specific to the virus. What that is, we don't quite know. We would have to. Carry [00:37:00] experiments similar to the I A P V experiment using well bred lines. And that may be possible sometime in the near future with better characterization of some of these viruses, but for now it's our abilities are to test such a thing are limited.
Well, f that is
Andony Melathopoulos: fantastic. I love one thing I really appreciate about your work is, just the kind of doggedness and also, in the face of results that were, maybe not the what one had hoped for being able to process through and think about the next set of questions is coming out of such of a, such a set
Robert Lu: of results.
Now, I was doing one thing that you may have noticed throughout this talk. I was stopping you when you. Resistance to jump in and say, oh yeah. Perhaps not resistance. And I was actually talking to one of my colleagues. She's a fantastic scientist in her own right. Her name is Nina Sk from the University of California Berkeley, Uhhuh.
And, sorry, did I say slo? I meant to say sok. I'm sorry. Nina . Nina Sok. It's [00:38:00] been a long meeting. . I'm just not, I'm honestly not good with names. Faces, yes, names , but Nina's, Nina Sokolov. Sok. . Nina Sok. Yeah. So she is looking at virus transmission from honeybees into wild bees, which is actually a very big problem because it's the wild bees that are really suffering right now.
So one thing that actually wanted to bring up with me that we followed up on, we had a very good conversation over coffee yesterday, was that when we look at bees surviving a dose of virus, we can't infer whether or not this bee is strictly speaking, resisting the virus in that it is suppressing the viruses that are getting into.
and preventing them from establishing, or if they're perhaps just more tolerant of higher virus loads within
Andony Melathopoulos: Oh, so they don't stop the virus from multiplying, but they just, they go about their business even though they've got lots of virus.
Robert Lu: Exactly. And the viruses just aren't affecting them.[00:39:00]
Now, functionally, if you're a beekeeper, no difference. You have less bees dying. You can have more virus, you can have better output, but I imagine
Andony Melathopoulos: you, but you may have bees that are packed full of viruses and transmitting them to something that's susceptible.
Robert Lu: Exactly. If we have higher. Tolerance for this virus.
Then these bees are going out and they're essentially virus bombers that are dropping viral. Is all over the environment and that's just more danger for the native bees that also need our help. . And so I just want to conclude this talk by saying that these results are very preliminary. They are exciting however and there are very many cool implications.
However, I think we need to approach these results very carefully because, When we breed for livestock, I think it's too easy sometimes to get fixated on this system that we're working with to forget that all around us is this big interconnected ecosystem and what we do with the one thing could have huge impacts on everything else in the ecosystem.
. So [00:40:00] science is cool, but it's also a very tricky thing to. I guess
Andony Melathopoulos: well, thank you for helping us navigate through the trickiness of this whole field, which I always I think you've done a wonderful job explaining for for myself and also for our listeners. So good luck with the rest of
Robert Lu: your research.
Thank you. Thank you for having me. I'm very happy to just be able to educate the public on I what I think are very important issues. They're very appreciative, . Thank you.
Breeding honey bees is notoriously difficult. New molecular techniques may help.
Robert Lu is currently going into his second year of his MSc Biology at the University of Alberta. His research topic is the parasitic honey bee mite Varroa destructor - specifically, he is interested in developing novel tools to control varroa infestations in managed honey bee colonies. In his free time he enjoys drawing and horticulture. He is also a fan of seafood.