0:00
Thank you for the introduction.
0:02
Good morning and thank you for coming to our session.
0:05
Can you hear me well?
0:07
OK, my name is Rapolas Zilionis.
0:11
I'm a Co founder and Chief Scientific Officer at a Atrandi Biosciences.
0:16
And before I go and share my excitement with you about a high throughput single cell DNA and RNA seq method that we're developing, I'd like to tell a few introductory words.
0:28
So we at Atrandi Biosciences were founded in 2016 in Vilnius, Lithuania.
0:35
We'll be opening an office in Boston, MA.
0:38
We currently are a team of 80 people and growing and you can meet three of us here at the conference, my colleagues Monica, Jan and myself.
0:49
We have close to 200 labs using our instruments and when we look at the composition of the team in terms of technical expertise, we have hardwood engineers, software engineers, molecular biologists, chemists, experts in droplet microfluidics and all these people with diverse background are united by a passion to innovate in high throughput single cell analysis.
1:19
Now I would like to start by acknowledging all the progress that has already been made in the field of single cell multiomics.
1:33
There is this beautiful review by Thierry Voet and colleagues listing methods that were developed from 2015 to 2022.
1:43
And you know, that continues.
1:45
I'm sure we'll be hearing about new multiomic methods at this very conference.
1:50
And there are also commercial solutions out there.
1:54
But developing these methods requires deep technical expertise.
1:59
It requires creativity, it requires careful optimisations.
2:03
What I'm getting to under the hood, developing a single cell multiomic workflow is hard and it does not need to be this way.
2:13
But before I go into that, why is it hard?
2:15
We believe that it's as hard because of the very way we set up reactions.
2:21
We can choose either between the versatility of microwell plates.
2:26
So there you have a full freedom.
2:28
You can perform one reaction, perform nucleic acid clean up and proceed to another reaction and do it virtually any number of times.
2:37
And on the other end of the spectrum, we have throughput.
2:41
We have throughput offered by microfluidics nano wells, but ideally we want to have both and we believe to have a solution for that.
2:51
That solution is called semi permeable capsules and it combines the best of both worlds.
2:56
It combines the versatility of micro well plates with the throughput of droplet microfluidics.
3:04
Now in the next few slides, what I want to do is get technical.
3:07
I want to share with you what these capsules are.
3:11
So what is this new way of setting up reactions?
3:15
As their name suggests, they're capsules.
3:17
They're capsules in the sense that they have a liquid core and a semi permeable hydrogel shell.
3:23
Now this hydrogel shell, it will retain cells and nucleic acids, but it will allow smaller molecules, DNTPS, proteins, including molecular biology, enzymes, oligos, of course, buffer components, nutrients, metabolites to diffuse through freely.
3:42
Now what that means is that once you've put your cells into capsules, into these compartments, you can perform multi step reactions, ensuring that every reaction happens under optimal conditions.
3:56
When I say, what does it look like to do it in the lab?
4:00
What does it feel to handle capsules?
4:03
Let's get into lab in our minds.
4:06
Well, once you put your sample into capsules, all you need is a pipette and a centrifuge.
4:10
You set up a first reaction, you allow it to occur and when you spin down the capsules, remove the supernatant and proceed to the next reaction and then repeat that virtually any number of times.
4:22
I also want to bring your attention to the panel on the left, which is so the showing how much 1 million of these semi permeable capsules or SPCs takes up.
4:31
So that's 300 microliters and that's enough to encapsulate up to 100,000 cells.
4:36
So really the high throughput and the miniaturisation is there.
4:43
Now once you have them, what can you do with them?
4:47
So first of all, I want to emphasise that they're robust compartments.
4:50
I don't know if you worked with droplets, you know the concerns that they may coalesce.
4:55
You have to be quite delicate and careful despite their power.
4:58
We come from a background of droplet microfluidics.
5:00
We love that technology as well.
5:02
But capsules are different.
5:03
You don't have to worry about them coalescing.
5:08
You can centrifuge them, vortex them.
5:10
You can expose them to extreme pH values to harsh lysis conditions.
5:15
You need to break your nucleus open and strip it from proteins by exposing to SDS and prod K No problem, that can be done there.
5:23
You can use guanidinium salts for RNA friendly cell Iysis, that's another option.
5:30
You can perform PCR.
5:31
You can even do something as extreme.
5:33
I don't know why one would do that, but we've tried exposing them for 15 minutes at 100° of the pH 13 and they stayed there.
5:42
But once you want to break them open and release their content, that is a very mild step.
5:47
That is an enzymatic degradation of the shell.
5:50
So that can be in a cell friendly and nucleic acid friendly manner.
5:59
We and our customers have performed a number of molecular biology reactions involved.
6:04
So these are just a few examples and the list is ever growing.
6:08
We have grown cells, so what you see here are examples of mammalian cell lines encapsulated into capsules and then forming colonies, spheroids, if you wish, from individual cells.
6:23
And you know, I'm showing all of these examples to trigger your imagination, to trigger the innovator inside of you.
6:30
And if I've achieved my goal, then we have everything to get you started.
6:36
Part of our customers are innovators, and all they want is just to be able to put a suspension of cells into capsules and then develop the workflow they have in mind.
6:48
If we're one of them, please stop by our booth number 35 we're very happy to discuss and make sure to enable you.
6:54
But now coming back to what my talk is about, it's about a single cell multiomic workflow.
7:04
So both our customers and ourselves have a number of multiomic workflows in mind.
7:09
So where do we start?
7:10
We decided to start with single cell RNA and DNA code sequencing.
7:16
When I say DNA here, I want to emphasise that I'm talking about Multiplex PCR panels, which are genotypic readout and that can be combined with a phenotypic readout, which is the transcriptome.
7:31
Now where would you use it?
7:32
Well, maybe you'd perform in CRISPR edits and you want to make sure that the headed happened not by sequencing the guide already and inferring that maybe it have happened, it has happened, but actually going to the edit sites, targeting with your custom PCR panel and validating it for sure.
7:52
But then maybe you also care about, why would you need RNA seq layer on top?
7:57
Maybe you care about the different efficiencies of editing between different cell types.
8:03
Another example is maybe you're working in Cancer Research and you have a cancer panel in mind.
8:10
You need some flexibility to be able to add some extra targets depending on your experiment.
8:17
And you also want an RNA seq readout to really see how these mutations are linked to different cell states and transcriptional responses.
8:26
Another example.
8:27
This one is rather niche, but I really like it.
8:30
We've talked to researchers who are interested in linking DNA viral infections to host responses at the transcription level.
8:40
So that's another example where being able to design a custom panel amplifying the viral genome and then also getting an RNA seq readout of the cell would be useful.
8:53
So what does the workflow look like?
8:58
First we start with cell encapsulation into capsules.
9:04
So again please visit our booth we have their instrument which is needed for that.
9:08
There is just a very simple procedure, non-intimidating and as I was saying from there on your deal with a pipe and a centrifuge.
9:18
Once we have the cells in capsules, we perform cell lysis and then we perform RT with template switching.
9:25
Pretty conventional choice and we proceed to two PCRs.
9:29
First we amplify the transcriptome and then we amplify the multiplex panel.
9:37
So why we're amplifying the Multiplex panel is quite clear, right because we want to sequence it but why are we amplifying the transcriptome?
9:43
Well, that's because we can.
9:45
And why are we doing it is because it will mitigate any downstream inefficiencies in the workflow, allowing conceptually to achieve the highest possible captures of transcripts.
10:01
OK.
10:02
Then next we proceed to split and pull barcoding.
10:05
We just ligate barcodes and assemble them on the transcriptomes and amplicon panels from single cells in the individual capsules.
10:16
And from that point, we don't need the capsules anymore.
10:20
By the way, they're discrete.
10:21
So you can sample any number of them.
10:23
Say you've barcode at 100,000, but you first want to perform a pilot run and just sequence 1000 very discrete compartments.
10:32
You pipe at the equivalent of 1000 cells and prepare a library and maybe perform a small sequencing run.
10:39
And we've also designed the work for us to enable to prepare the transcriptome and amplicon library separately, so you could balance the number of reads that you dedicated to each of these modalities.
10:53
So features we're talking about free primary sequencing of the workflow version we're currently developing is for 100,000 cells per experiment.
11:01
But I'm very curious to talk to you and learn more about what scale would be important for your experiments.
11:08
Capsules are very amenable both to scaling down and scaling up.
11:12
So that's a possibility.
11:15
The amplicon panels are user defined and they need to be at least 300 base pairs in length.
11:22
On the other hand, if you already know a panel that you will use for most of your experiments, also very curious to hear about it because in addition to the user defined panels, we could develop specific panels that would be, you know, off the shelf eventually.
11:37
And I mentioned about the separation already.
11:41
OK, so this is the workflow that's very much work in progress.
11:45
But I already have some data to share with you.
11:48
This is data that we collected a few weeks ago.
11:51
And we challenged ourselves with the following experiment.
11:55
So we took frozen PBMCS from 2 donors and we said, OK, we don't know the genotype and we mixed the cells.
12:04
So we gave ourselves a challenge.
12:06
How can we then separate them downstream from our data?
12:11
To do so, we targeted 6 amplicons.
12:17
Well we designed a panel of 6 amplicons that are targeting polymorphic sites.
12:22
These sites are very likely.
12:23
They have a high minor allele frequency from population studies.
12:26
So they should be different between two individuals.
12:29
There are good chances these sites will be different between individuals, and we perform our workflow and obtain free data layers of our RNA counts, the DNA counts and from the DNA counts, we obtain the single nucleotide variant counts as well.
12:48
But we are not limited to single nucleotide variants in terms of mutations that can be targeted by a PCR amplicon would work as well.
13:00
And we now ask the following questions.
13:02
Can we identify the cell types that we care about?
13:06
So that's probably the lowest bar for RNA seq.
13:09
Does it tell you the cell types then once and looking at the RNA data only now out of these cells that we identified in RNA data, how many of them, what fraction of them also have the amplicons detected?
13:23
And finally, are these amplicons useful?
13:26
Can we find variants that allow us to, well, separate the two donors and essentially solve the challenge that we gave ourselves?
13:35
So this is what the RNEC data looked like.
13:37
I'm showing you a U map.
13:39
We can see the populations that we expected from a PBMC data set, actually with quite some structure, for example, in the T cell compartment, including T Regs, which I believe deserve being mentioned given this year's Nobel Prize on their study.
13:57
So, but the RNA seq data is there and we can see the different populations.
14:02
So that was one check.
14:03
OK, that's an important part to pass.
14:06
Next, out of these transcriptomes, how many of them also have the amplicon detected?
14:12
And the answer is 87%.
14:15
But here presence I'm defining as having at least 10 reads and I'm asking how many cells have all 6 detected?
14:26
Now if you look more carefully, you can notice and there's a little arrow there showing that one of the amplicons is actually bottlenecking with detection.
14:34
And if we look at the plot at the bottom, it's just showing you know how uniformly the panel is performing in this experiment.
14:41
And honestly, it's not very uniform.
14:43
We have an order of magnitude difference between the best and worst performing amplicons, which is a reminder amplicon panel design, the effort that it requires is not to be neglected.
14:55
And the way we would solve this and improve this is either by brute force deeper sequencing or by optimising a prime appears to achieve a more uniform amplification.
15:07
OK, so again we have the amplicon data.
15:10
So let's now look at the variance.
15:11
And the question is, and you know, what I'm about to show you is it's going to be, you know, a heat map.
15:17
And the question is, can we see C2 genotypes?
15:20
And a few weeks ago when we were analysing this data, I was waiting for the heat map and I was so pleased to see it because it immediately showed two groups of cells.
15:30
What you're seeing on the left is a heat map where every row is a cell.
15:34
And then you it shows you essentially what fraction of reads are supporting 1 variant versus another.
15:40
So for example, at the bottom right, we can look at one of the genes which is metric K1 at position 416 in the on the amplicon we have other an A or a G And you can see that essentially we have two groups of cells by utilisation of this particular variant.
16:01
And then another amplicon is also supporting the same information.
16:05
Now the one in around the middle CFH.
16:09
It's not helping us to separate the two genotypes, but I chose to include it in this heat map because it nicely shows what a head site looks like.
16:18
You can see that around 50% of the reads support one variant versus the other.
16:24
So head sites can also be detected using this method.
16:28
OK, so now we have CVs 2 genotypes.
16:30
We can go back to our RNA seq data and now assign an extra label.
16:36
Which genotypes do these cells belong to?
16:40
And we could assign a genotype to 90, close to 97% of the cells.
16:47
And you know, I just told you that 87% of the cells had all amplicons detected and that I can assign a genotypes to 97% of the cells.
16:56
This is because there's redundancy in the amplicons.
16:59
I don't need to detect all six to be able to split the heat map in two.
17:06
OK, so we can see on the top right that there is some donor specific batchiness in there.
17:11
By the colouring we can also see that, you know, there are donor specific population abundances.
17:18
But it would still be good to have a sanity check.
17:22
OK, like the amplicon data is seems to be, you know, very convincing, two sharp groups.
17:29
But can we somehow use the RNA seq data to check if we really performed the split correctly?
17:35
Turned out our donors were of different sex.
17:38
So we can look at sex specific gene expression after splitting this U map by donor and check whether we can see two populations.
17:50
Sorry we can see this not only two populations but like a donor specific gene expression.
17:55
And while the answer was yes, donor A clearly is a female and donor B clearly is a male.
18:04
Now with that, I want to I'm getting to the takeaways.
18:09
So semi permeable capsules combine the versatility of place with a high throughput of droplet microfluidics.
18:17
They are available today as a platform of innovation if you're an innovator and want to create your own workflow, but we're also working on workflows ourselves.
18:26
And by the way, when I say workflows is for going from sample two sequencing to insights from the data and the first one we're tackling is Multiplex amplicon panels plus RNA seq, please stop by our booth number 35 to chat further.
18:46
And I was believe we still have some time for questions.
18:50
Thank you.
18:50
Thank you for this talk.