Do you have a question you'd like answered on a future episode of Translating Proteomics? Send it to translatingproteomics@nautilus.bio!

On this episode of Translating Proteomics, hosts Parag Mallick and Andreas Huhmer of Nautilus Biotechnology discuss the challenges and opportunities of plasma proteomics. Their conversation focuses on:

·      Why blood plasma may be a good source of protein biomarkers

·      Current methodologies and pitfalls in plasma proteomics

·      The path forward for plasma proteomics

What is Plasma Proteomics?

For those who are new to this topic, plasma is the liquid portion of the blood distinct from fractions containing red and white blood cells. Given the relatively non-invasive ways physicians can collect patient plasma, and the blood’s intimate association with tissues throughout the body, plasma is potentially an excellent source of protein biomarkers. Yet, it is quite difficult to measure the levels of all plasma proteins because their concentrations span over 12 orders of magnitude. This episode features an in-depth discussion of the ways plasma proteomics efforts have and have not lived up to the promise of biomarker discovery and what we can do to advance plasma biomarker discovery efforts in the future.

Chapters

00:00 – 01:01 – Intro

01:02 – 4:55 – What is the promise of plasma proteomics?

04:55  – 07:23 – Is the plasma proteome really the best source of biomarkers?

07:23 – 10:16 – How do proteins get into the blood and what are the implications for biomarker discovery?

10:16 – 13:59 – Is it clear that proteins are the best candidates for blood biomarkers?

13:59 – 19:57 – Advances in and the future of comprehensive plasma proteomics

19:57 – 22:31 – Pros and cons of fractionating the plasma proteome to discover biomarkers

22:31 – 28:14 – Progress in identifying multiomic plasma biomarkers and the path forward

28:14 – End – Outro

Resources

Nano-omics: nanotechnology-based multidimensional harvesting of the blood-circulating cancerome (Gardner et al. 2022)

o   Review from focused on the development multiomics liquid biopsies

Multicompartment modeling of protein shedding kinetics during vascularized tumor growth (Machiraju et al. 2020)

o   Work from Parag’s Lab investigating tumor protein shedding

Simulation of the Protein-Shedding Kinetics of a Fully Vascularized Tumor (Frieboes et al. 2015)

o   Tumor protein shedding work from Parag’s Lab

Mathematical model identifies blood biomarker-based early cancer detection strategies and limitations (Hori and Gambhir et al. 2011)

o   Study modeling how much protein could be shed and detected from different size tumors

The human plasma proteome: history, character, and diagnostic prospects (Anderson and Anderson 2002)

o   Review discussing the clinical importance of the plasma proteome and the wide range of protein abundance in the plasma proteome

Blood Biomarkers to Detect Alzheimer Disease in Primary Care and Secondary Care (Palmqvist et al. 2024)

o   An example of the potential power of blood plasma as a source of biomarkers

Plasma Proteome Project

On this episode of Translating Proteomics, hosts Parag Malik and Andreas Huma of Nautilus Biotechnology discuss the challenges and opportunities of plasma proteomics.

Their conversation focuses on why blood plasma may be a good source of protein biomarkers, current methodologies and pitfalls in plasma proteomics and the path forward for plasma proteomics. Now, here are your hosts, Parag Malik and Andreas Humer.

Welcome back to another episode of Translating Proteomics.

Today we're going to discuss a topic of keen interest to the many in the proteomics community, and particularly researchers who apply proteomics in the clinical setting, plasma proteomics. We'll cover what plasma proteomics is, why it's useful, the pitfalls of doing plasma proteomics and the ideal path forward for plasma proteomics.

Andreas, what is plasma proteomics and why is it useful?

I by now should know that because it's, I think, my third round of trying to do plasma proteomics using different technologies.

The first one obviously was 2D gels, followed by first generation LCMS systems and then ultimately the most modern approach with very high performance mass spectrometers. I think it's a dream that is often a nightmare. On one hand, plasma is probably the best source of biomarkers.

In fact, I would argue that every single molecule that is in plasma is a biomarker. We just haven't figured out how to decipher the signal of the biomarker.

On the other hand, it makes it so complex that it is really difficult to utilize plasma as a reliable source for measuring biomarkers simplistically. So I think we still struggle in particular in proteomics with making plasma a great tissue for biomarkers.

I'm pretty sure you're going to correct me and say we have many other examples that are positive.

Well, no, actually, I think I initially got into proteomics in part because of a series of papers that turned out to have some methodological challenges.

In particular, there were papers from Leota and Petrocoyne that were published in Lancet and New England Journal and others that promised a biomarker for ovarian cancer early detection with 100% sensitivity and 100% specificity, which should have been our first warning.

That there was a red flag right there.

Red flag.

But I discussed these with my sister who's a practicing surgeon and runs a cancer center, and she said, wow, this is amazing, this is transformative and if we could really have a blood test for cancer, this would change everything.

And so I got into the area because I was so excited about plasma as a readout for physiologic state and excited because you had it was this place where proteins shed all throughout the body make it into the bloodstream.

But I think the challenge that we have, and it's not to say we don't most of our circulating protein biomarkers, most of our circulating biomarkers are proteins, things like PSA and insulin and troponin and phospho tau217. However, because it is a sink and proteins are coming from absolutely everywhere, we can't know where they came from.

We also don't understand the processes by which they got there. And so I would love for plasma to be the pinnacle of where we look for biomarkers.

But I actually think there are substantial challenges to studying the plasma proteome.

I totally agree. The question is, how do we make inroads in that area?

If you approach this from a non proteomics perspective, finding viral infection in plasma is trivial these days and contributes to a lot of early detection of infection. It's standard today in clinical practice.

We certainly have a lot of good biomarkers on the protein side as well, but there's so many other proteins that are hidden. We simply don't understand where they come from and when they actually emerge in blood.

And I think that requires really the next generation discovery tools that we still don't have access to because of the complexity. It's just astronomical.

Well, I'd like to just also push on this point.

We hear a lot that the plasma proteome is a great place for clinical biomarkers because it's really accessible and it's easy to get, it's easy to collect versus imaging or a biopsy. Those are the comparators typically.

But on the other hand we have things like breath, exudate, literally someone breathing into a tube and collecting that even going further than that. There are these sponges now that people can swallow and pull back up to get esophageal scrapings. And those are also pretty easy.

So in order to get a blood sample, well, all right, maybe I can get it from a finger prick, maybe. But you're causing cellular damage. You're getting a bunch of reactive proteins as a function of that finger prick.

The amount of volume that you're getting is very low. So if you're looking for something that's particularly low abundance, maybe it's not even contained in that finger prick.

So you're typically talking about a blood draw. This is not something that people do at home.

On the other hand, maybe we should be thinking about things like patches that measure interstitial fluid. That's how the continuous glucose monitors often work.

So I clearly have done a lot of research on the plasma proteome, but convince me that the plasma proteome is where we should be looking.

I mean, ultimately, as you said, this is end stage of every protein coming out of the tissue. And, you know, it is probably the reservoir to look for a biomarker.

I think what we do wrong is we take a single blood sample because we're used to the biomarkers who are either constantly elevated, like inflammation markers, because you have already a symptom of an inflammation. But instead, what we really would have to do is take many, many time points. And so that comparison would ultimately tell us what's going on.

And I think because we keep sticking to the single point measurement, we'll have really hard time measuring significant biomarkers. Obviously, you can study these things on a population basis, and a lot of technology has been developed to enable that in the past few years.

But I think for significant progress in the biomarker field, we'll have to do a lot more time trials rather than, you know, trying to go really deep and finding, you know, the smoking gun.

So for things like neurofilament, light chain, NFL, which is used to look for brain injury, I think we understand that the larger the amount of brain injury, the larger amount of neuronal death, the more of that protein is shed into the brain region and then traverses to the blood. So we have a clear pathway, we have a clear mechanism that allows us to trace back and say, okay, this is a good biomarker.

Here's its path, here's why it's a good biomarker.

And I think that argues for what we may have discussed in a previous episode, that really what you would like to do is start investigating the disease or the biomarker at the tissue of origin and then really think through the process how it actually is going to be a biomarker. Plasma.

I'm surprised there is really no computational models or any other approaches today to say, all right, this is how much protein would have to escape per hour from this tissue to be able to be a measurable biomarker.

And so I'm surprised that we have done this really repetitive process of just trying to look, because we have given the tools to look instead of really approaching this from a much more rational perspective.

Well, what's interesting about that is, so there are publications in that space. My lab has had several of them. But what makes it challenging are the biophysical properties of proteins.

So the fact that you have large proteins and small proteins, you have highly charged proteins and more hydrophobic proteins, you have proteins that degrade biochemically very easily because they have a lot of exposed digestion sites. That prediction is actually reasonably complex to predict what is the protein stability. So you have two issues.

You have what is its transfer rate from its tissue of origin, then what's its clearance rate, whether that's by biochemical degradation or hepatic clearance, et cetera. On the other hand, other analytes, things like DNA, rna, even some metabolites, are much more consistent in their biophysical properties.

Circulating DNA is a very specific size because of the way that it wraps around nucleosomes. So that makes it just much more amenable to study.

And so when we think about circulating biomarkers in blood, is it clear that proteins are the best type of biomolecule we should be looking at?

Well, certainly, if you want to look at proteins as the causative or as a place where the disease starts, I mean, you have to look at proteins. I think where proteins may be, as you mentioned, the early cancer detection tests, as an example there.

I think where proteins may play an important role in the near future is that we combine measurements with DNA and then get actually a better read out.

Some of these early cancer detection tests, about 80 to 85% correct in terms of detecting early versions of cancer, but they're not very specific with respect to tissue.

This is where proteomics, for example, can come in and say, by the way, the DNA change that you observe is actually coming from a particular tissue, because we have supporting evidence on the proteomics level. So maybe that's the first way to sort of create specificity, but also the right biomarker.

Well, I'm going to argue the other side of this for a moment and just argue that proteins are what we should be looking at. Realizing I've spent the last 10 minutes arguing the other Side.

But one of the advantages that you have of proteins having such a wide range of biophysical properties and shedding rates and also abundances is that you potentially can look at processes on different timescales.

For instance, if what I want is, I want a marker for cancer early detection, what I want is a protein that builds up over time in the blood so that when I have a small tumor, I potentially get an outsized amount of this protein in the circulation.

And that can happen if a protein from an even very small tumor is secreted at a very high rate and it's very stable and so its clearance rate's really low. That's a property that CTDNA cannot have. There's no particular gene that is going to be more abundant or less abundant in that way.

Likewise, on the other hand, if what I want is I want a marker of therapeutic response, I want something that changes very quickly so that within 24 hours of giving somebody a treatment, I can see a response again.

That's where maybe what I want in that case is I want a protein that upon treatment spikes and then goes away in responders and does nothing in non responders. So I want to look for that spike at a defined time.

So I think the acuity of temporal dynamics that proteins potentially have the range of concentrations, which on some level is really frustrating because you have this wide range of concentrations, that same thing that is a challenge can also be a benefit.

So double clicking on a topic that we just discussed where we understand how biomarkers go from their tissue into a bloodstream and that they have certain properties, it should be relatively straightforward then to capture those biomarkers based on the properties that we already. No.

So no. Unfortunately, part of the challenge is what in the world should I look at? So I have a discovery challenge.

Once I know what to look at, sure, I can say I only want to look at the positively charged ones. But a priori we have this challenge of how do I look at as much as possible.

And so those properties may help you later, but they don't help you in the front.

So this is where we basically want to look at what we're doing today in terms of technology applied to plasma proteomics and maybe what we're going to be able to do in the future.

One of the breakthroughs that happened in the last few years is that the capability to look at many proteins and to look at the plasma protein very deeply has been massively increasing. That was a combination of better detectors, but also better sample preparation approaches.

What do you think where we are in terms of chasing down these biomarkers?

Well, it's funny that you asked that question today.

We actually just finished up our summer internship program and had a set of summer interns at the Canary Crest program that were specifically benchmarking four different plasma proteomics workflows. These were mass spectrometry based workflows and they were looking at.

All right, if we do a deep fractionation method, or we use a method from Hanosteen's lab where we do a perchloric acid precipitation which preferentially precipitates high abundance protein, or if we do a bead enrichment for extracellular vesicles, or if we use a kit that does a very targeted enrichment and some cleanup.

We looked at four different methods as part of this and asked in particular, if we're looking for tumor derived proteins in those samples, which method was able to give us the greatest enrichment in tumor derived proteins. So the way we did the experiment was we took mouse blood and then we spiked in human cell lysate at varying concentrations into it.

So we could tell which were the human proteins and which were the mouse proteins. And the human proteins were mimicking those that would have come from a tumor.

And what we found is that each of the methods performed just completely differently.

All right, interesting.

Interestingly, the extreme perchloric acid precipitation method, it flattened out everything. So it really did a great job of taking out the high abundance proteins.

And as a consequence of that, you were able to see lower abundance proteins and you saw the relative. It was the only method where we saw a huge, huge increase in human tumor derived proteins relative to mouse.

Interesting.

On the other hand, it also decimated the number of proteins that we were looking at. So other methods that we looked at looked at two or three times as many proteins by that, that method.

But that again, could have been just the way that we ran it in our lab with our interns. So that's not to say that's the limits of that method.

The EV method was in our particular setup where we were looking for tumor derived circulating proteins. It was a very interesting method for finding tumor derived proteins, but it was limited to things that were encapsulated in EVs.

There are lots of interesting proteins that are not encapsulated in EVs. They're secreted or they're cleaved or others.

And so that method is incredibly powerful for finding things that did come from the tumor, but it's limited in that it can't look at everything. The extreme fractionation Method has been around for a very long time.

Initially developed by Vitor Fascia and Sharon Piteri and Sam Hinash's lab continues to be a powerhouse method for really digging deeply. But it is extremely time consuming and instrument consuming in terms of what is required to execute on it.

Yeah. We still essentially have the same problem of either depth or coverage. Right.

But what is fascinating to me is really the ability to enrich for particular, maybe tissue through the EVs.

I think that's a very promising method to find relevant Biomarkers because these EVs don't stick around for a long time and so they actually have a timestamp on and so to speak. And I think they're really reflecting a lot of the things going on in tissue.

And so I think that EVs propose an elegant method to go around the problem we have in just looking straight at plasma where you have so many mixed signals, besides the fact that plasma also reflects just a functional organism by itself, not just disease.

Yeah. I think the question is each one of these methods is segmenting a piece of the plasma proteome.

So whether it's the EVs that are looking at just the things that are exported in a very specific packaging, or whether it's the throwing away the high abundance proteins and whatever's carried along with them, you're still selecting for a subset of the plasma proteome.

But I think there's nothing wrong with it. Right.

I mean, to go full circle where we started, I mean there's, there's the obsession with finding a biomarker by covering every single protein at every concentration level.

And obviously that's the most difficult way to approach this, versus maybe I have a more intelligent approach to finding a biomarker because I have a prior information about it, whether it's, you know, where it comes from or whether it has a particular fraction, it is appearing in plasma or it's bound to another protein. So I think the technologists love the challenge of building a tool that sees everything.

And I think for the, you know, serious biomarker discovery one might have to compromise and say, maybe I'm just going after, you know, the fraction that really contains the information of that I want to really study.

Using biophysical properties to enrich or deplete proteins is a strategy that certainly has been explored quite a bit.

There was some really nice work by Costos, Costarelos and Marillana had in the UK where they used nanoparticles and nanoparticle corona to, to try to enrich for lower abundance proteins. They also used a similar method to look at enriching for circulating tumor DNA. And really interesting approach.

They found that it was really just an adsorption property of the EVs and whether they changed the size or charge or shape, it didn't materially change what proteins were pulled down. But this was a method to begin to look at.

If I'm interested in a particular subset of proteins that have specific biophysical properties, let me put something into the matrix that can grab those, capture those, and allowed looking at a different subset of the circulating proteome.

Is there any downside to that approach for plasma proteomics?

Yeah, I think it's still not comprehensive. It's limited by the capabilities of the detector on the back end.

Also, it potentially biases the quantification on the other end because you're necessarily biasing what you're analyzing. So proteins that stick really well to the nanoparticle have a really strong corona.

There are going to be more of than proteins that don't stick very well.

So if we would like to have an absolute quantification measure where we say there's more of this protein than that protein, it goes through this convolution of the nanoparticle and we've completely lost our ability to look at the sample in an unbiased manner.

So is the conclusion that it's just too hard?

No, I think it's just that we need to continue to advance the technologies.

When we think about dynamic range, we often talk about, oh, the dynamic range of blood is 8 orders of magnitude or 9 orders of magnitude or 10 orders of magnitude.

The reality is that there are proteins that may be present in one copy in the blood, in your whole body, in liters of blood, because it happened to sneak its way out from some peripheral tissue. So what is our dynamic range? Is it 18 orders of magnitude because it's one molecule in five liters?

I think we need to define the problem and then we do need to continue to work on technologies to dig deep.

Yeah. So maybe to end on a positive note, I mean, acknowledging that this is really, really difficult to do, I think there's a lot of progress.

This has happened over the last few years in plasma proteomics, but also by not just using plasma proteomics, but other measurements in plasma, whether it was DNA or measuring, for example, microbiome changes that holistically, when you look at the entire dataset, have given us really good information about the health status of humans. I would think as we continue to find the perfect technique for plasma proteomics.

We might just take advantage of the many other biomarker signals that are present in our humans. Right, yeah.

So I guess maybe two points there. The first is on the use and integration of multiomic data and multiomic signals from plasma.

One thing that I've seen quite a bit of that I don't really understand is people doing pathway analysis on circulating proteins, or in particular multiomic analysis. Multiomic pathway analysis on circulating proteins.

The DNA and the RNA and the protein that come into the blood are coming from all throughout the body. They unlike in cellular systems where the DNA and the RNA and the protein are coupled together in a.

Functionally, functionally there is no clear functional link in the circulation.

So those kinds of studies, the analysis for pathways that are upregulated or down regulated or influential, they're going through this communication channel of how things make it from site of origin to the blood. So we just have to be really careful when we're thinking about functional pathway analysis of circulating things.

Yeah, I would totally agree that certainly applying a multi omics approach has to be done with carefully and with a lot of insights.

When we think about what is our ideal path forward for plasma proteomics, I think there are two key aspects of it that we need to address. The first is the pre analytical variables.

We need to make sure that we're collecting samples in a way that is reproducible and consistent and that we understand that we're not mixing tubes that have a lot of hemolysis with tubes that don't. We want to make sure that tubes are sitting out for similar amounts of time before being frozen. The methods of freezing need to be consistent.

And so part of our path has to be how do we collect the best samples? And then the second part is recognizing. All right, what is the question I'm trying to ask of these samples? Am I trying to go as wide as possible?

Am I looking at a subset of the proteome? Am I looking for the larger proteins or the smaller proteins?

Those that'll allow us to develop methods that really allow us to ask the most impactful clinical questions.

Yeah, sampling is definitely a big aspect of how do we solve the problem in a biomarker?

Because some of them may just be obfuscated by that sample handling and literally they're there, but we can't see them because we have messed up the sample handling process in the case. Yeah.

I think the other part is that there is obviously new affinity techniques coming that allow to measure several hundred or several thousand Proteins that have given us a lot of new insights in larger population studies.

And while these population studies are always going to be limited, being informative about individual, this certainly will give us a framework, what we should expect as biomarkers in blood.

So I think there will be tremendously insightful to say, yeah, if you are looking at inflammation, maybe these are the set of proteins you might want to specifically focus on. And maybe that way we can design experiments around sample handling, around fractionation, and then specifically around detection.

That gives us the insights we're looking for.

Yeah.

So I think two pieces to that as we think about what do we want for the future, one of which is really on the data analysis side of how do we integrate and analyze all this data together so that we can understand what does variation look like? There have been a number of instances where so many people have done plasma proteomics on this population or that population.

Aside from efforts like the plasma protein atlas that Eric Deutsch's group put together, there have been very few efforts to bring all of that data together. And what we've seen with these emerging AI methods is that they really benefit from having large scale, organized, clean data.

And so as we think about the future of plasma proteomics, we really need to consider that anytime we do a study, it's part of a larger whole. The other place where you've mentioned is time, that we understand that unlike DNA, proteins change over time.

And if we want to look at a tumor growing sampling at multiple time points will help us obscure some of the just general, I'm a human and there's variation from that.

Yeah. So fulfilling.

As we discussed the challenges that come with finding the proper biomarkers, how we know how actually biomarkers make it from the tissue to the plasma, and then the technical challenges come with the actual detection of the biomarker is that I have the feeling we're going to have another episode on plasma proteomics in the future.

I think you're right. I think this is such a deep and rich topic and I think we're making a tremendous amount of progress.

I think in the last five years it's been amazing to watch as we have made so much progress in the field. Thank you again for joining us on today's episode of Translating Proteomics. Today, we dove deep into the plasma proteome.

And to recap, we discussed how plasma may be a great source of protein biomarkers, but we need to understand how those proteins made it from their original tissues into the blood. We also talked about how current technologies each capture very different views of the plasma proteome.

Lastly, we talked about how new technologies and technologies that are on the horizon may ultimately give us a much deeper and better understanding of the plasma proteome, how it changes, and help us find those elusive biomarkers. We'd love to hear your thoughts on the plasma proteome.

We'd love to hear your experiences on work that you've done, or challenges that you faced, or things that you're excited about. Please join the conversation and let us know in the comments.

We hope you enjoyed the Translating Proteomics podcast brought to you by Nautilus Biotechnology. To contact us or for further information, please email translatingproteomicsautilus Bio.