0:08
Today I will be talking about our internal research project in which we successfully sequenced antibody proteins de novo using mass spectrometry from human blood for the first time ever.
0:24
But before I get into presenting the paper, I figured I'd share why we're introducing new technology into the antibody discovery landscape.
0:34
And I'm also recognising that there are a broad group of attendees here.
0:40
So for context, what it's really all about is trying to get more information earlier in the discovery and selection process to avoid wasted efforts and get more efficient.
0:53
There's also going to be some goodies in there for people who are looking at vaccines and doing vaccine development.
0:59
So with those things in mind, here's how we're going to explore the topics.
1:04
I'm going to introduce the selection problem in antibody discovery.
1:07
Of course, I'll present our Nature Communications paper, but it's not just about the paper.
1:13
I'm going to try and share some practical applications of this technology that you might be able to use or consider today.
1:20
And if you're still interested by that point, then I'll tell you a little bit about Rapid Novor and us.
1:26
OK, so starting out with the selection problem and I'll start with an analogy.
1:33
If I have an infinite room filled with non-human primates and typewriters, they will eventually come up with the optimal sequence for the antibodies that we need for our therapeutic or diagnostic tool.
1:48
But it's probably not going to be this guy because you haven't even put the paper in this typewriter yet.
1:54
No matter how you come to the diversity that you need; you somehow have to select down.
2:01
So the upper limit of useful diversity, there's many estimates of it, but could be somewhere in the range of 10 to the 16, 10 to the 18 antibodies that are produced naturally in response to immune threat.
2:15
And we have some tools to access them.
2:20
So the problem is from that almost inconceivably large set of diversity, how do we come down to the one antibody that we need for this needle in a haystack problem.
2:31
We do have some wonderful tools available to us today and I've listed some of them here on the slides.
2:38
Cell sorting, of course you can express lots of things and test them.
2:42
You can do SPR work, cell based assays, et cetera.
2:47
And I try to put these on some kind of continuum that describes how expensive versus how high throughput some of these are.
2:57
Please forgive me if you've got slightly different categorization, but in principle it's sort of easier to do affinity based testing, you know, does it bind or not?
3:09
Towards the end of these kind of campaigns, we might be doing the more expensive slower things with cell based assays or detailed characterization of glycans and PTMs.
3:20
And then there's something in the middle with maybe exploring epitopes, maybe the doesn't bind at certain temperature ranges and things like that.
3:29
So what we hope to do is try to move some of the stuff that's listed in the middle towards the left so that it's more information earlier.
3:40
OK, so those are all useful selection criteria, but we also have to acknowledge that there are some sort of considered unavoidable selection just to nature of technology.
3:49
And you know, a lot of B cells die before we can actually sequence them.
3:54
Many B cells are hidden in the bone marrow.
3:57
We can't actually access them from blood and there's only a certain amount that we can process at a time.
4:05
So if we want to propose new technology, new solutions to the selection problem, we're looking for something that gives us more information earlier in the process, is cheaper and faster than what we can already do, and without creating new unintentional limits to diversity.
4:24
So our proposed solution that we decided to explore was sequencing the functional antibodies, the proteins themselves from human serum.
4:36
And if you can do that, it can tell you potentially some things or at least indicate that there's a better chance that the resulting sequences might be ones that are binding their efficacious in vivo.
4:50
They have some kind of neutralisation properties.
4:54
And also if you have those proteins, you can play with them, right?
4:56
Unlike B cells, you can play with them a lot.
4:58
You can run them at certain temperature ranges or pH ranges.
5:01
You can check for binding for different off targets and things because those proteins will last.
5:09
So before we go digging into the human serum proteome, I'll present the case in a minute.
5:16
Keep in mind we're trying to reduce the expense of trial and error and we're looking for can we find anything new that was hidden from the sort of B cell oriented technologies.
5:30
OK, so this is the case study, this is the meat of it.
5:35
And this was published in Nature Communications late last year.
5:40
The full title is up there.
5:41
It's about de novo protein sequencing of antibodies for neutralising against the SARS-CoV-2 vaccination.
5:49
I know SARS-CoV-2 is boring at this point, but we started this work quite a long time ago.
5:54
There's the QR code, so I'll leave that up on the screen for just a second.
5:58
If you want to take out your phone and you can take a picture of that and download the paper.
6:03
This is also excellent for me because it shows everyone's really engaged in the content, holding their phones up, OK, and spoiler alert, OK, we did it right, so I'm not going to wait, make you wait till the end.
6:15
We did actually manage to find some neutralising antibodies for this method.
6:21
Here's the rough numbers, I'll go through them again later.
6:23
We accessed lots of diversity.
6:25
We came up with 12 sequences that we really liked.
6:29
We tested and expressed some of them and six of them were found to have neutralising effect against the ACE 2 receptor.
6:39
OK, so here's how we went about doing it.
6:43
We started with three human blood samples.
6:49
These patients had been immunised with a vaccine, and we did a simple ELISA test to see that on the bottom left there sample #522 had the best results for ELISA.
7:01
So we decided, OK, let's go ahead and use that one.
7:06
We this, we then did a protein a purification to isolate the IgG population from there.
7:14
And then we used RBD on a column to isolate those IgGs that are specific to our target.
7:24
OK, so here's a high level view of what the sequencing process looks like.
7:30
And I know it's a big diagram and it's only going to get more complicated, so bear with me.
7:36
The first thing we would do is going to separate that sample out into various fractions.
7:41
And the top stream there you're going to look at the intact antibody molecules.
7:45
The middle stream you're going to do a simple digestion and you're going to look at the subunits, the Fab unit, the FC region for instance.
7:53
And in the bottom stream you're going to digest with cocktail of proteases to produce peptides.
8:01
All of those things separately are going to be separated using chromatography, gels, et cetera.
8:08
And then you're going to run each of those things on mass spectrometry.
8:11
This is an extraordinary amount of work, hundreds of mass spectrometry runs and that's where you need some pretty intense algorithms.
8:19
And this is sort of like the secret sauce behind it to pull all that data together.
8:24
You can see at the bottom we're layering in B cell sequencing data.
8:27
So there's a B cell repertoire sequencing information and compiling that into the sequencing results and it comes out with full length mAbs at the end.
8:37
And I'll just zoom in a little bit on the de novo sequencing part of that workflow.
8:45
So you can see along the top we've separated out the sample and we're digesting with a whole number of episode of enzymes at trypsin, lysine, pepsin, chymotrypsin, et cetera.
8:58
And you can see that in some of those digest conditions, we're also adding a custom modification which is called cysmod, which helps us to create the kind of ions that we need for leucine, isoleucine identification.
9:13
If you know anything about sequencing from mass spec, those two amino acid residues are isobaric.
9:18
So you need to create a special ions called w-ions to fragment them.
9:24
So all that's done and then there's the middle down workflow and then the top down workflow as well, which is used for the chain pairing efforts.
9:36
OK, so you're getting the sense that this is quite a complicated thing.
9:40
I can't talk about everything, but I can describe a little bit about why de novo sequencing is hard and some of the problems that we had to overcome to get there.
9:50
So the simplest part is sequencing peptides, right?
9:53
So you run the peptide through mass spec, you've got a nice series of ions that describe the amino acids.
9:58
And then you can, because you've used different enzymes, you can layer that information over top and you can get like, you know, contigs like that form the sequence.
10:11
So these are the CDR 1, 2, and 3 regions of one of the particular antibodies.
10:18
And each of those coloured bars represents a peptide that we found from mass spec.
10:24
OK, so that's the start.
10:25
And this is sort of fairly routine for us at this point.
10:31
Newer technology that we had to invent was bridging between CDR 2 and CDR 3, for instance.
10:37
So for this, you're going to need an extraordinarily long peptide, which you can't really de novo sequence on its own.
10:44
You have to use that in combination with those shorter peptide reads in order to come up with the bridge between those two regions.
10:52
So here's an example, here's what a mass spectrum looks like, and the lines up top represent the cleavage points between those or fragmentation between those residues.
11:07
And then there's the final challenge, which is, or part 2 of that challenge, which is bridging between the CDR 3 and CDR 1.
11:17
This is typically involving non linear peptides.
11:21
Now you're sequencing across disulfide bonds and that's really complicated.
11:28
And again, you need to have done some small peptide work first before you can put all that together.
11:37
And then the final thing that you need to do is to have your light chain pairing.
11:40
And so for this, you need the measurements of the intact molecules themselves, chromatography information, as much separation as you can and then marry that up with all the sequence information that you've already generated to come up with the hits so that we're not just proposing every single possible combination of heavy and light chain pairing.
12:01
So we'll come back to the results.
12:04
And I shared that we got 12 and we got six of those were binding.
12:09
So our question is, was this a good result?
12:12
So we did our best to compare against another published work that relied on B cell sequencing and their results indicated that they had sequenced 5.8 million B cells.
12:27
They got down to 187 that they felt good about expressing and they got at the end nine of those were neutralising.
12:38
OK, so much credit to them, extraordinary amount of work and those went on to be very successful antibodies.
12:44
But the point that I want to make is that 50% of the ones that came through our process were neutralising, whereas only 5% of the ones that came through the other process were neutralising.
12:55
So I don't want to say throw out B cell sequencing.
12:59
What I would like to propose is you layer on top both of the sets of information.
13:04
OK, this is a huge data table.
13:06
It's just showing some of the data that we have.
13:09
If you look at the chain source column, you can see that some of the data was found within the RNA sequencing information and confirmed by mass spectrometry.
13:18
And those are marked RNA and others are coming from pure de novo, right There weren't found in the RNA.
13:28
If you look at the heavy and light chains, you can see that there are some alternate chain pairing combinations that were proposed as well.
13:37
And here's some of the test results we've got.
13:40
On the top right, there's a line with the black dots marked PD 124.
13:44
That was the original pAb and we can see how it was binding and all of the recombinant ones except one followed kind of a similar profile in terms of binding.
13:56
And then on the bottom right is a pseudo neutralisation assay binding against that H2 receptor.
14:04
And there's a negative control in there.
14:07
And again, we can see that they're also binding in there.
14:10
And we did a cell based assay, which is the final column in the table and you can see yes, no, whether it was neutralising or not.
14:18
Finally, we did some SPR work, which was a complex stability analysis to see what were the on rate, off rate, et cetera of these recombinant forms.
14:29
And so the ones towards the top right are the most desirable ones.
14:33
And so we would reject them.
14:35
R1 and R4 are recombinants for this point.
14:39
OK, so those are some of the results.
14:42
And as I promised, I'll go back into, you know, how you can actually practically think about using these sorts of things.
14:49
So the first one that I want to suggest is that you can use this in any species.
14:54
So you're not limited by availability of primers or libraries.
14:58
You can go into different species.
15:02
This case was human, but of course, we've done work in cows, rabbits, sheep, goats, llamas, chickens.
15:10
And our one of our main scientists really wants a shark.
15:13
If you have a shark, please send it to us.
15:15
We'd love to try that.
15:18
The other one is preselecting antibodies.
15:21
So this is that idea.
15:22
Again, bringing this information earlier in the workflow, this is a very rough diagram of what a typical discovery workflow might look like.
15:32
You're generating hits, you're doing some optimization work.
15:35
Eventually you might have dozens or hundreds of sequences to express and then test and that may be costly.
15:43
What we're suggesting is if you layer on top this polyclonal sequencing using mass spectrometry in parallel with that existing workflow, that can surface some of those hits as being important from the proteomics layer.
15:56
And that might help you to select which ones to proceed with which may lower your costs and efforts involved in expressing and testing different variants.
16:08
The third thing is starting at the end, so you may be able to skip re immunisation if you already have a polyclonal antibody that you've obtained or derived that's useful in your work.
16:20
So the options are do polyclonal sequencing, get some recombinants out of it, or start again from scratch and try to come up with a monoclonal or recombinant that recapitulates the activity of the polyclonal.
16:37
I think I'm running a little short on time, so I'll skip through the anti-drug antibody bit.
16:42
But essentially you can use the technology to preselect for antibodies that are binding the idiotype of your drug with a positive selection purification step and then a negative selection purification step.
16:57
So you're not binding other IgG forms.
17:01
And here's some hot off the press data that we've just come out with binding to adalimumab versus not binding on the right to the general IgG population.
17:15
Vaccine development is an interesting one.
17:17
I've had lots of people coming up to me that are interested in the immune response to a vaccine, decoding what that response looks like.
17:24
It is possible to use this technology to do that.
17:27
And as we've seen, you can come up with potentially useful antibodies in response to a vaccine from a human donor.
17:39
And the last piece of tech that I want to share, which was not in covered in the paper, but a subsequent work was about epitope mapping.
17:48
And again, we're using mass spectrometry and some AI approaches to screen for epitope.
17:56
So the point is again to interrogate epitope, you can use SPR binning based approaches, but epitope mapping using mass spec and cryo-EM and such things has typically been reserved for further down the funnel.
18:10
What we're trying to do is move that forward so you can access epitope information earlier in the process.
18:18
So this is representing the cost and effort that we're putting forward for you know, one to two mAbs.
18:24
You're going to put in a lot of effort.
18:26
If you have 20 to 50 antibodies, it's about the 10th of that effort.
18:33
And if you have 100 or 200 or more than it's even significantly less than that.
18:38
And we're layering in some AI to help predict the binding and then working in the wet lab information to further refine those predictions so we can help select from a large set of antibodies using epitope information.
18:55
Here's some of the results of that work.
18:57
This is again the RBD protein and a section of antibodies that bind against it.
19:03
The colour codes indicating the binding regions for various antibodies.
19:09
On the top is our HDX-MS results and then on the bottom is the predicted docking from the AI experiments and there broadly in agreement with each other.
19:25
And I don't have time to go into the whole thing, but some of the information feeds into each other and it's useful confirming information, OK.
19:35
And again there's a QR code up there.
19:37
So you can find more details about this work, this study.
19:42
And if you've by the way, if you've missed any QR codes, you can just e-mail me at the end, and I will get that information to you.
19:50
So briefly about us.
19:53
Company was founded in 2015.
19:55
The core technology is about sequencing antibodies and that's based on work from Doctor Bin Ma, which started almost 25 years ago when I met him.
20:06
And since then, the company has successfully sequenced 10,000 monoclonal antibodies and completed a hundred of these polyclonal sequencing and discovery type jobs that I've been expressing here.
20:19
We're a service provider and happy to provide a number of services in the antibody discovery sort of realm and we've got locations around the world for easy shipping.
20:37
And here's our contact information.
20:40
We are available at booth 45 over there if you want to come and talk to us.
20:44
And if you've missed any of the links or QR codes, then just send me an e-mail which just says simply send me the links and I'll know what to do and I'll send you all those links.
20:54
And that's all I have to say for today.
20:56
If you have any questions I would be happy to answer them.
