Translating Proteomics hosts Parag Mallick and Andreas Huhmer discuss our new Nature Methods paper, “Large-scale single-molecule analysis of tau proteoforms.” This publication describes how the Nautilus Voyager™ Platform can be used to quantify proteoforms derived from the neurodegeneration-associated tau protein. These proteoforms are defined by their unique combinations of post-translational modifications and isoform specific sequences. They may make powerful next-generation biomarkers or drug targets, but few technologies can measure them today.

In the episode, Parag and Andreas discuss:

• What proteoforms are and why it’s critical to study them to understand biology.
• How the Nautilus Voyager™ Platform quantifies proteoforms.
• The key findings from the tau proteoform analyses in the paper.
• Potential applications of these findings.
• The Nautilus Voyager™ Platform as a powerful foundational technology.

Access our video series exploring the paper in more detail.

Foreign.

Welcome back.

On this episode of Translating Proteomics.

Andreas and I are proud to discuss some of the key takeaways and exciting applications of the tau proteforms assay and iterative mapping in general described in our recent Nature Methods paper.

Large scale single molecule analysis of Tau proteforms.

This paper represents the culmination of nearly a decade of work at Nautilus and really could not have been possible without the tremendous efforts of everyone at Nautilus, as well as our fantastic collaborators at Genentech, the Neural Stem Cell Institute and the Ronald Loeb center for Alzheimer's Disease at Mount Sinai.

So I'm excited.

Andreas, are you ready to dive in?

Yes, Very excited to share the news here.

Awesome.

Let's get started.

So maybe I'll start us off with just some background to center the conversation.

First off, a lot of the manuscript focuses on this method, iterative mapping where we immobilize single molecules on a hyper dense array and probe them repeatedly over and over and over again, being able to learn extra detail about each molecule cycle after cycle after cycle.

And the first application of that iterative mapping technology has been the analysis of proteoforms.

Proteoforms are the functional forms of proteins that actually exist in cells and they're defined by the combination of splice variation and post translational modifications and any other modifications that might occur.

Even though they're derived from the same gene, these proteforms can have very different functions.

The proteform variations proteoform state proteform code of proteins really defines how a protein operates, where it transits in the cell, who it interacts with, and all of the key aspects that underlie its function.

The protein we focused on in this manuscript, the protein tau or the, which derives from the gene mapt microtubule associated protein Tau, is a critical player in Alzheimer's disease and other neurodegenerative diseases.

Its typical role is to stabilize microtubules in healthy cells, but it's also involved in neuronal development and maintenance.

And one of the key observations is that in Alzheimer's disease it aggregates and forms these tangles that potentially lead to cells misbehaving and being really a core component of neurodegenerative conditions.

It also comes in potentially millions of different proteoforms.

It's known to have six dominant splice variants.

It's also in some cases truncated.

It also has a large number of different phosphorylations and citrullinations, acetylations, et cetera and it's also known that that tau protein levels themselves do increase as Alzheimer's disease progresses.

But researchers in general don't know which protea forms are associated with healthy tau function and which are associated with disease.

There also what has been fundamental work going back decades observing that tau is its phosphorylation state in is associated with disease.

Now, our hope is that in understanding the molecular functions of proteins as determined by proteoforms, that we can identify the specific molecules to target to enhance or disrupt biological processes or disease processes.

And in general that measuring these proteforms may be the key to developing highly accurate diagnostics and understanding therapeutic mechanisms.

Targeting proteforms therapeutically may be a way to, to develop ultra precise, very specific precision medicines.

And the challenge that we had when we entered the space was that there really are no technologies today that are capable of measuring proteoforms at scale.

Until now.

In this paper we show that the Nautilus Voyager platform can quantify up to 768 different forms of tau.

And in the manuscript we observed 130 distinct tau proteoforms across a range of cellular models and patient samples with high accuracy, sensitivity and reproducibility.

We also benchmarked the assay really showing those figures of merit.

So I'd love to dive in with that as sort of preamble.

And Andreas, I'd love to hear your key takeaways from the manuscript.

What are the things that you really, really jumped out at you from the manuscript?

Happy to do so.

But before I go into diving in the details of what we saw around tau, I would strongly encourage everybody reads the paper, step back a little bit and think about what actually is underlying all of these tau portiform insights.

It's millions of single molecule measurements.

So to me that's still breathtaking.

Having spent most of my career measuring ensembles and averages.

And so the fact that we can talk about single molecule measurements is just fascinating.

So with that in mind, let me talk a little bit of what my key takeaways are from, particularly from the tau measurements we did.

The important biological information is actually in the intact molecule, not in an isolated phosphorylation site.

For decades, the field of Alzheimer's has really measured either total tau individual isoforms or the average of abundances of certain phosphorylation sites, phosphorylation217 being one of the best knowns.

But those measurements do not reveal modifications, that modifications occur together on the same T molecule.

And so this study shows that combination matters.

And each of those phosphoration sites by itself is only part of the story.

And so you really have to measure the intact molecule and measure all of the sites to really come up with a conclusion that might be biologically relevant.

And so the assay revealed, for example, that tau phosphorylation is not randomly distributed.

Right.

One of the conclusions you could have thinking about it from first principle that phosphorylations are just random.

And so, you know, there could be many, many combinations of that, but our paper clearly shows that they're not random.

And so figure 4G and H, for example, give a very, very nice insight into that.

And maybe you talk a little bit about that in more detail that, you know, certain sites actually CO occur preferentially, for example, 181 and 217.

And the data clearly shows this interesting site aspect of this is both of those sites have been, you know, been very interested or people have been interested in no size as a potential biomarker.

So very clearly another level of understanding of what's going on at a single molecule level.

Any, any other key takeaways you'd like to add?

Yeah, I mean, I can take additional ones.

You know, the field in Alzheimer's has the story of more phosphorylation as the disease progresses.

We did a few brain tissue samples as part of the study.

It looks like there's more of a remodeling going on rather than a progressive increase in phosphorylation.

The conventional model that the disease becomes more aggressive as you add more phosphorylation is probably too simplistic.

And so the post mortem brains that we show that, you know, phosphorylation, not simply cranked up, the actual nuance is actually in the proteform distribution.

So I should caution here, right.

None of this data can be immediately taken and used for, you know, maybe for, you know, for conclusions yet because we haven't done sufficient number of samples there.

I should also say that we only probe, in this particular case, seven sites.

But what we see is that the abundant protein forms with one or two phosphopeptides in our case actually decreased as the disease progressed, while the more rare, highly coordinated multifossosphere species, most of the ones actually containing 231, which is very interesting, get enriched in cognitively impaired people.

Again, this shows that this is not just simply a progressive phosphorylation event, but it's a remodeling of that.

Maybe as a third observation, biological context matters in this case.

We've looked at teleports in different model systems, the different genotypes in brain regions in follow up studies.

That's not part of this paper.

But the takeaway here clearly is we need to stop treating every neural model system, assuming that it's the same tau biology.

So as they showed that the organoids in M brains in mouse brain and adult human brain have each very characteristic peripheral landscape.

And obviously in some of the neural cells we looked at the eye neurons, they have more of the immature part, immature form of tau.

While for example, human brain has less phosphorylation in general, but more mature isoform distribution.

So I think the strength of the paper is that we investigated tau across several model systems, not just the post mortem brain tissue.

And I think that really opens up a new perspective on how do we actually should think about tau as an intact protein with many of these fossil F events and how should we actually approach tau biology research in the future.

So I thought these were some of the key takeaways I had.

What about your thoughts there?

Those are some very interesting takeaways.

I think for me at the highest level as, as part biochemist, being able to look at billions of single molecules, millions to billions of single molecules at, you know, at this scale really is just, just a thrilling thing.

And the, the that we can see things on each of those molecules that are, are completely hidden in bulk.

That was just.

I spent so much of my training thinking about populations moving.

When we learn reactions, we're often thinking about concentrations of bulk molecules shifting from one thing to another.

But in practice, in a cell, yes, those forces are at work, but there are also forces that are really driven at a single molecule talking to another molecule and interacting and an enzymatic reaction happening at that scale.

And that is truly the scale that biology actually happens.

And so to be able to look at things at the scale that biology actually happens en masse, I think that's an important takeaway from the paper, is that this is now not just a theoretical construct, but this is a very practical, accessible technology.

So for me, that just foundationally is a really important takeaway from the manuscript that there was so much innovation that had to happen in order to figure out how to deposit individual molecules en masse, how to pattern the arrays, how to cycle and probe them over and over and over again and to make that a robust, stable, reliable process.

I think that for me is one of the really critical takeaways, is that this method, this approach, iterative mapping, is ready for prime time.

And accordingly that there's a specific power that comes from doing single molecule analysis at scale.

One of my other key takeaways, I was quite honestly terrified that we were going to do this analysis and we were going to find that there actually were no Co occurring PTMs and that every molecule we saw would just have one PTM on it and that was it.

And so that was not the case.

There were a large percentage of molecules that had 2, 3, 4, 5, 6 CO occurring phosphorylations on a single molecule.

And that really blew my mind.

I really had not considered just that how much multiplicity of co occurrence of PTMs occurred in nature.

And this really was my first look at the extent to which that phenomena occurs and is an important driver of biology.

I think my third takeaway really, I think another takeaway that comes from thinking about multiple modification aligns with one of yours, which is it really does not appear to be this completely random process.

Instead it suggests that there is some order and timing where a modification at one site potentially opens up the protein to make it amenable to being modified at a different site.

And that there is this cascade on a per molecule basis that, you know, ultimately makes sense.

It comes back to the structure, it makes sense.

It comes back to the binding of enzymatic partners that these modifications fundamentally alter the shapes, interactions, locations of molecules.

And so it, it's important to look at those cascades, but it's also important to do it quantitatively because it's not just about having these molecules present in the cell.

One of the other things that we saw is the quantities changed.

And in some systems you saw more of the triple phosphorylated states than you saw the double phosphorylated states.

And that's really interesting because again, if you were to follow sort of a random process model, you wouldn't really ever expect to see that.

You would expect to see this just decreasing number of molecules that had 1, 2, 3, 4, 5 phosphorylations.

And we didn't strictly see that, which is really interesting.

So again, the other aspect that matters so much is the quantification that we're able to look at these processes to make delineations about what is truly random versus non random.

So those were my key exciting takeaways from the manuscript.

Yeah, let me hop back to your original argument or observation that in fact, looking at millions or billions of individual molecules, I'm still blown away how relatively easy it is to do that.

I mean, as soon as you have the sample prepared, the library process done, put it into the Voyager system and you get the readout.

It is extremely complex to get any type of information around, let's even say isoform distribution by any other methodology.

And there's papers out that describe all of that.

The fact that we can do this so well and then actually get very robust relative quantitation across these different samples very reproducibly, that is just such a big progress for analytical science.

To the.

Level where we can now really study, as you say, individual molecules and how they actually may interact with automolecules.

So to me that's mind blowing progress.

And I need to keep reminding myself that we can now operate at a level when we create hypothesis that was completely impossible before.

Right.

And so I think this is one of the things I keep digesting as I talk to more customers or collaborators.

What are the types of studies we can really initiate now, now that we have these capabilities?

And, and so, you know, the takeaways keep evolving in my head now that I, you know, seen some of data, but you know, now that I also see the power of it.

Yeah, I think that that point of.

There's one particular figure which I think we call a pseudo western figure, which is what you might have seen if you looked at each of these sites just one site at a time on a western.

And you really don't see a lot of variation across the model systems.

You can't tell as they're aging.

You can tell grossly the difference between human and the cell lines.

But when you go to full proteoform view and you're looking at those co occurring modifications, all of a sudden there's this just incredible clarity that allows you to say, wow, these, there actually is a lot of biology that's different and that there was so much biology hidden from us that we simply couldn't access.

I think that also was a really big epiphany for me of just how much is going on and how important it will be to measure it.

And we only import, cost, look at, you know, use and look at this case, seven phosphorylation sites.

So imagine what might evolve as these capabilities are expanded.

Which brings me to my next thought process.

What are the applications you're thinking about now that we have this capability?

As you said, it's very routine.

We can do many, many samples at the same time and compare them across time points or regions or different model systems.

What are some of your, you know, applications that you see?

Yeah, well, I think level zero is that this really is the introduction of a Foundational technology, we applied it to Tao, but it can be applied to many other things.

It can be applied to oncology proteins, things like AKT1, EGFR, P53, that of course are very interesting proteins and likely have a lot of very interesting proteaform biology.

And so as a, as a first layer, I'm just excited to be able to expand the scope of the technology to many other proteins, many other disease indications.

Coming back to Tao, I think in my mind there are, there are two really key areas that I just scientifically that I'm very curious about.

Number one is differentiating amongst taupathies.

We have these family of tauopathies, ftd, psp, Alzheimer's disease and others that are patients when they present, often have a mix of one or the other.

And I think that's one of the challenges that the field has had in developing their effective therapeutics is if you believe that someone has Alzheimer's disease, but really what they have is FTD or they have a mix of Alzheimer's disease and FTD or a mix of Alzheimer's disease and psp, you're treating them with an Alzheimer's disease targeting drug and they don't respond well.

Maybe it's because they have a blended population of Alzheimer's and something else.

So that's a fundamental barrier when developing therapeutics is if you don't know exactly what disease somebody actually has.

So I firmly believe that in the proteform landscape is the information to differentiate amongst these tauopathies both from a mechanistic side, but also from a downstream how should I treat it side.

And so I think that's a very important thing for us to look into.

Another aspect is in Alzheimer's disease in particular, there's this notion of what's called selective vulnerability.

And so this is a chain where Alzheimer's disease begins in one part of the brain and then moves throughout regions of the brain, typically starting in what's called the entorhinal cortex.

And why Alzheimer's disease starts in the entorhinal cortex, why it moves next to a particular region and then from there to another region.

It is clearly not solely about location and proximity.

There is something more complicated going on.

And it wouldn't surprise me if what's going on is that the different regions contain different baseline distributions of tau proteforms and that under environmental stresses, inflammatory processes, metabolic consequences, that those distributions shift a little bit.

And so maybe a Key part of selective vulnerability is underlying tau proteoform distribution.

And maybe that gives us a lever to actually intercept these neurodegenerative diseases early, as well as even potentially even come up with preventative therapeutics.

So I think really digging deeply into the mechanisms of early pathogenesis and progression of pathogenesis using the proteoform lens is an application that I'm really excited about.

Absolutely would agree with that.

I would probably add, you know, besides you looking at tiopathies, there's also other applications like in dramatic brain injury.

Turns out there's also tau phosphorylation events and so, you know, yet to be investigated whether there is particular phosphor protiforms that of interest or different.

But maybe that particular capability is also helping to address some of these very common points of initiation of neurodegeneration that can be investigated by this technology.

The other application I was thinking about, and that comes really out of the fact that we looked at all the model systems or many of the model systems, is there's an application in this space called nams, it's abbreviation standing for new alternative methods, where basically researchers use human or human derived model systems.

For example, like brain organoids.

They become really very popular because they are much more similar to a human tissue than a mouse might be or another animal model system might be.

There's a lot of effort in creating these brain organoids.

What we clearly can show is that in some of these brain organoids, when we look at the different genotypes, which may include something like APOE 4 risk factor, for example, we can actually pick up those changes relatively easily.

There's an example in the paper where we looked at these Apoe4me brains and Apoe3mi brains.

Clearly we can actually now use genotypical features that might be existing in some of the patient population that we're talking about and study particular interactions or drug interactions.

But then even beyond that, we might actually be able to now guide these developments of those organized model systems to first of all represent maybe human brain in a much more directed way, but also really help in evolving those models around particular genotypes.

Because many of the neurodegenerative diseases that have been investigated today have particular mutations in tau, and those might be really addressed by specific therapeutic approaches.

And we can help develop those brain organoid systems.

Then typed to that obviously is using the technology would be my third application, maybe I have a fourth one.

But a third application maybe is really guiding therapeutic development.

And so the combination of having the organoid model systems that you can easily interrogate, you can do lots of mechanism interaction studies really will hopefully accelerate the field in creating the right therapeutic target and then testing drug modalities that may be suitable to treat those.

Closing the loop what you said is we need to intervene in this disease relatively early and being able to study the disease in a model system across a long progression might really be one of the breakthroughs that this particular technology can enable because we can really observe the tau biology that's happening otherwise in the human brain where we have no access to.

Yeah, those are really great, great points, Andreas.

Well, thank you so much for sharing your perspectives.

What I might do now is just try and try and coalesce our conversation into a couple key topics that came up.

So number one on the takeaways, I think there's we both agree that just the ability to probe at the intact protein single molecule level at scale is a really critical foundational new piece of measurement technology.

And I think we've seen throughout history that new readers and writers that allow us to look in detail and at scale have been transformative is that with any foundational technology.

The authors, creators of that technology of course had something in mind.

We have thoughts on what one can do with iterative mapping, but it's clear that in genomics, for example, that foundational Illumina sequencing by synthesis method has been adapted for chip seq and hiseq and 3seq and all of these other methods.

It's even been used for protein interaction screening and understanding transcription factor binding.

So one of the this ability to just probe individual protein molecules.

I am certain that when people get their hands on this method, they're going to have a thousand other ideas for how it could be used for asking a wide variety of questions.

And so I'm excited also about the future of where people are just playing with this and able to ask a wide range of questions, not just questions about tau proteoforms.

And then I think a second key takeaway really is about the extent to which multiple co occurring modifications, proteforms in general are drivers of biology, are systematic rather than random processes in terms of applications.

Of course talking about this as a foundational technology and it sounds like really what I'm hearing is really two very distinct directions, both of which are powerful.

One, on the more mechanistic basic science side, understanding early disease, how proteoforms affect cellular behavior, how they affect drugability, how they affect cellular function, all of those core over time is a critical application.

And then there are further applications on the therapeutic development side, the clinical implementation side as well.

And so that we really this really captures that entire life cycle from early discovery all the way to potentially patient facing interaction.

Did I capture that pretty well?

Andreas?

Yes.

I would just add we didn't mention biomarker.

Maybe there is a separate session we should really do on that because now that we have the ability to do that in tissue, we certainly want to do this in biofluids.

And that's probably worthwhile a separate episode to talk about that.

Yeah, absolutely.

Absolutely.

I think that that's clearly an important next step is that once you uncover the core biology at the primary site, being able to measure it remotely in lower cost, higher throughput manners, minimally invasive manners is critical.

So awesome.

Well, thanks so much for sitting down with me today.

This was really fun.

Thanks all to our listeners.

And if you have any thoughts or take a look at the paper, send us your comments.

And I hope to see you next time on Translating Proteomics.

We hope you enjoyed the Translating Proteomics podcast brought to you by Nautilus Biotechnology.

To contact us or for further information, please email Translating Proteomics at Nautilus Bio.