HLA peptidomics by diaPASEF with a model for measuring quantitative accuracy!

 

Weird. I got back from offgrid camping and my first discussion was about HLA peptides this morning and - boom - cool new preprint! 

I'll admit that HLA's by DIA freak me out a little. You've already got all this chaos from your database explosion, I like to see a good solid MS1 and some isotopes for my own mental health. This group goes through a couple of iterations - running some DDA to generate libraries + trying some predictions  -AND -

most importantly - they break out SILAC (younger people, it's an old technology where you grow cells in the presence of expensive heavy labeled amino acids. For some reason we used to call it "the gold standard for proteomics quantification" or something. While that might be an exaggeration we all liked, SILAC is pretty good for relative quan).

Through a SILAC model they actually test if they can quantify their HLA peptides that they discovered! Because that is absolutely the only thing that people really care about, despite what services are generally offered. Super cool study all around! 

Why write your own proteomics paper? Just copy a well-cited and important one figure by figure!

 

I'm standing in a field with my phone in a tree so it can get enough signal for me to tether my laptop so I can post this. That's what you do when you are in central West Virginia and something this weird happens. 

When someone asks anyone in the field of single cell proteomics (including me) this important question "how many single human cells do you actually have to analyze" we all have our own ideas, but we'll almost always bring up Hannah's paper -

And -- today - another paper just about step by step, figure by figure, seems to be exactly this paper. 

Don't take my word for it - check it out for yourself here.

BUT DO NOT - DO NOT - DO NOT - CITE THE NEW PAPER. 

It's paywalled, but you can go through the figures from the 2 papers and sit them side by side and - they're the same.

Original is left panel. New paper is right panel. 

Crazy, right? Okay, I'm going to go back to my vacation whether or not this thing posts. 

Extract nuclei from fixed tissues - optimized protocols!

 

More nuclei extraction proteomics! This is for bulk analysis and this team went all out in testing different procedures for the best yield. Super cool stuff! 

"SomaSignal tests: The next step in the evolution of multiplexed proteomics..."

 

..... (come on Ben, type something nice you can do it....got one!) The illustrations in this new advertisement review for aptamer based proteomics are very nice!

...and....some peer reviewers were fine with letting this thing through, so there must be merit in it somewhere. 

Unfortunately, I was skimming it and saw section headings like this and didn't want to throw up on my keyboard, so I'll leave it here for others to find and read. 

Every author on this review, btw, appears to be an employee of SomaLogic. 

Reminder that misleading mouse studies waste medical resources

I'm great Mr. Mouse, thank you for asking. So rather than share any annoyance about what I read over my lunch, we're going to remind people about these 10 year old findings! 

Look, there are places for animal model data. There really are. Core metabolism is just like ours with few changes. However, if you're using mice as a neurological model and testing something that impacts higher level functioning in humans - using mice, which absolutely do not have the pathways in question - you're just being silly. I linked the editorial, because no one has to read much past the fact that 0 ALS drugs that worked in mice had any level of efficacy in humans.

Rapid one-pot workflow for phosphoproteomics prep (and DDM coating procedure)!

 

One issue with phosphoproteomics enrichment methods is the large required sample input. There are other issues, such as the mod is intrinsically transient by evolutionary design, but this group totally tackled the first one. 

I figured that anything sounding this complicated would require me to wait until a commercial kit was available, but the set up sounds surprisingly (shockingly?) simple. The paper is open access, so you can get that part list yourself if you are interested.

Another reason to check out this paper is that they coat their low-bind plastic in n-Dodecyl-Beta-Maltoside (DDM) and show that helps a lot. If you're interested in just reducing your surface loss binding stuff but possibly also worried about the longitudinal effects of a recently described detergent blasting into your mass spectrometer, here is a process to coat and discard the active solution! 

41,000 (forty-one THOUSAND) human plasma proteomics samples!

 

41,000 plasma proteomic samples?

Clinical data linking 218 different diseases across these people? 

Yes, this was O-link explore, so just under 3,000 proteins probed (which, reminder, does NOT mean measured, it means they were theoretically detectable because there are probes for the protein).

This is still exciting.

LESS exciting.....

Single NUCLEI proteomics (and identification of cancer mutations in single cells!)

 

I'll start with the figure above because it's early in the paper. Super exciting to me personally because of my personal research interests. These peptides are hard to identify in bulk. This group is doing them in single human cells.

How? They're building libraries for the software that we like to run library free the old fashioned way. They're capturing the specifics of their instrument as well as the unique characteristics of how peptides at super low intensity/concentration tend to behave. So this is all label free proteomics by data independent acquisition used to resolve both human mutations AND PTMs in single human cells. Wow, right? 

THEN this group broke through to the next level of what single cell proteomics (SCP) needs to do. THEY ANALYZED SINGLE NUCLEI. For context, if you look at what a large percentage of "single cell seq / scSeq" is actually analyzing, it is actually nuclei harvested from fixed tissues. The nucleus is pretty tough and it can often be separated from materials where the rest of the cell won't be recoverable.

Things like fixed tissue. And we have LOTS of that.

They only analyzed around 100 single nuclei here, but I honestly thought we'd see nuclei 5 years from now. When I first saw these results around the end of 2023, I couldn't believe it. It should be noted that we've recently saw a new preprint that did a few thousand single nuclei, so I was waaaay off in my estimates. Super super cool new paper. 

Struggling with protein acylation? You should try Acyl(S-) Trapping!

 

If you are a rational human being and you want to stay one, you should probably just forget that proteins reversibly acylate all the time. Most notably, acylations tend to occur to drive intracellular spatial stuff. Like your protein gets a terrible acylation on it that helps it migrate to the membrane where it now has activity while it's tethered there. 

Typically to measure these awful things you start with a huge input then do some enrichment and cleave off your enrichment tag or something. Other times you over-express your protein in a system where the acylation is forced to occur but it has no biological function. Force that KRAS to go to the E.coli membrane expressed at completely nonphysiologically relevant concentrations, because that will help you characterize those mods while learning absolutely nothing else about what that system does. 

OR you can and should do this? 

This clever use of both a modified suspension trap and what/when/how you add isobaric tags allowed this group to characterize protein acylation in a complex system starting from as little as 20 ug of protein. It's a super cool new approach to get at very releavant protein modifications that are very very tough to do otherwise. 100% recommended. 

ARE WE THERE YET?!? What does single cell proteomics need to do next?

 

Just leaving this here before I run out the door! Great insight (perspective, even?) from one of our most forward thinking protein informatics groups. Cool that we can do this stuff, but is it read to help people yet? Why not? Totally worth a read. (Open access)

Proteomic analysis of a super promising new active RAS inhibiting drug!

Y'all, KRAS small molecule drugs are sort of my jam and I have exactly zero guesses of how this drug could possibly work. My calendar is packed today, though, and I'm going to drop this here without looking it up.

Here is the idea, though: KRAS by itself is generally not bad. The problem is that the active sites get mutated and then the stupid new version of the protein stays active all the time. As KRAS and it's cousins NRAS and HRAS sit on top of pro-proliferation pathways, having them active all the time is a terrible idea.

This new drug, currently RMS-7977 (the name generally changes if it has success in human trials, you can guess by the designation that it's probably not there yet, so a healthy grain of skepticism should be involved here for this exact molecule) appears to only inhibit the active sites. 

Imagine a KRASG12C mutation where you've got a cysteine very actively holding onto GTP so that it is activating those proliferation pathways. You've got too much active stuff around. This drug doesn't block the GTP site, it blocks the activity of the active GTP bound protein. 

What's surprising is that there is a very different shape for a G12D mutated pocket and a G12V or G12R, etc., but this thing doesn't care! 

What's funny is that they did some awesome proteomics and it didn't really make the paper. TMT proteomics and phosphoproteomics on an Exploris 480 using the turbo TMT mode. 

If you're as interested as I am, the files are here. There appears to be a second repository, but there is a typo in the paper so I can't find it without some digging. 

Really really cool stuff. And some of the early small molecule RAS inhibitors aren't really doing well in the clinic at all right now, despite being approved for use. However, the molecules based on those are rapidly evolving and each one is better than the last. Even if this one doesn't go forward, the fact that you can inhibit a bunch of deleterious mutations with a single drug is a super promising development!! 

Save the date! ABRF 2025 - Las Vegas March 23-26!

 

Today we'll talk about why you have the coolest person in your field head your academic associations. 

I'm of course talking about -- 

THE Dr. Sue Weintraub  -who may have had nothing at all to do with the fact that the Association of Biomolecular Research Facilities is meeting in 2025 in Las Vegas, but I'm going to pretend that I have inside knowledge that it was entirely her idea. 

Thanks Sue! You can find out more and register here! 

I haven't been to an ABRF meeting in a few years. I did a talk on single cell proteomics in 2022, maybe, because I haven't been in a core facility in a few years. I'm a huge fan of the organization and the conference. 

If you are an early career researcher (ECR) and you are working on building up your credibility and exposure ABRF can be an amazing thing to be part of. Look, we already know who is going to be headlining ASMS next year, right? Same dudes that founded the whole thing in the 1960s. ABRF heavily promotes ECRs because it is great for everyone. Core labs sometimes don't get to do a lot of method development because they're too busy applying established methods for the hundreds of customers necessary to keep their lights on. You come in from you academic lab with the techniques you developed and you can get a chance to present those new methods to the lab pros who may have a very different angle than you've considered. Someone with 20 years of applying proteomics techniques can be an unbelievable untapped resource for new insights and ABRF is where you get to hang out with them.

I've never gambled in my life but I've gotten to climb in the Red Rock Canyon just outside of Vegas.... that might still be my profile picture for this blog.....

Obviously you go for the science, but it never hurts when there is fun stuff to do before it starts or a day or two after all the science has wrapped up.

Using a carrier proteome approach to "amplify" pericardial fluid proteins!

 

The importance of this new paper can be better appreciated after some quick google searches. Something like "what is pericardial fluid and how do you extract it" is a good start.

You use imaging based approaches to guide a microbiopsy needle into SOMEONE'S BEATING HEART and you remove some fluid. Ideally there is NEITHER A LOT OF SPARE FLUID not DO YOU WANT TO TAKE A LOT OF EXTRA SAMPLE OUT. 

You are - quite hopefully - very sample limited here. So what this group did was TMT label the peptides from a pool of some fluids to use it as their "carrier" and then TMT labeled individual fluids. We know that too much carrier messes with quan, so they carefully optimize the levels to get a lot of IDs but without compromising their quantification accuracy by too much. 

I am required by a new blog rule to point out (thank you reviewer #1 for a paper just accepted today (wooo!) that using a labeled sample to increase your odds of detecting samples in a body fluid is the premise behind the TMTCalibrator, which pre-dates the preprint of SCoPE-MS by about a year. 

As you might be able to tell from the aforementioned "googling", I am no expert in pericardial fluids and I didn't have time to become one on possibly the hottest single day I've ever personally experiences (heat index of 116F? On an isolated wooded mountain in Pennsyltucky?) poor day to mow my grass. I might have broke my brain. But these people are and they seem very happy with the biological applications of their large relative increase in numbers of identifications. 

I'm happy for them, because this is a cool approach and - if I was going to have pericardial fluid removed for some nerds to do proteomics - I'd want them to take as little as possible. I

Equi-CP - Semitargeted quantification of drug treated single cell (and other things!)

 

I really like this protocol because it is so very straight forward (after the sample prep...ugh...) and we took a swing at something similar a while back and had absolutely no luck whatsoever (on a different instrument platform).  Here is a very clear, well written, protocol to show you how to do it. 

It should be open since I don't have an affiliation right this second and I can read it. 

Here is the idea, though. When you use TMTCalibrator/SCoPE-MS/iBASIL/BOOST. You generally start with a whole proteome that you label with one isobaric tag. Then you mix that with isobaric tagged extremely low level or single cell samples, right?

In Equine Clostridium perfringens (Equi-CP, and to be fair, I didn't read the part where I was supposed to learn where the name came from. You take synthetic peptides and label those with one channel, clean them up and then spike those in your labeled single cells.

Then this group just does plain old DDA on a nice previous generation Orbitrap. It goes along just doing MS1 scans until it hits your synthetic peptides, fragments those and - boom - you have single cell data. I also like how they pseudo-randomize their samples with the MANTIS and (had we seen this earlier) Dr. Colten Eberhard's spring working out how to pseudorandomize 4,000 mouse brain cells from 6 different mice might have been a little less stressful for him to sort out because they ultimately came to very similar methods. (Though Colten had a x6 matrix, not a x2 and the guy with career defining dysgraphia/dyslexia typing this post was like "omg, there is no way in this world I can help you at all" (something he clearly knew after 4 years). 

Super cool, though, right? What we tried to do was take all the peptides from the RAS pathway kits from Thermo and then label those with one channel. Honestly, I still don't know if we were just way too ambitious / outside of our dynamic range or if the cells just didn't tag well in that batch. Either way, when I do it again, I'm using this protocol verbatim. 

More single cell histone PTMs (epiproteomics!) label free, targeted WITH derivatization!

 

I've whined about this before, but I went to an epigenetics meeting last year and had a lightning talk (and awesomely lucky poster location by the coffee) and pitched single cell histone PTMs ("come talk to me by the coffee"). 

No one did. 

For real, every single single cell proteomics dataset I've analyzed (and I've looked at just about everyone's RAW data - except the people who don't make theirs public even after I request it - and you know who you are and you suck) - has histone PTMs in it. 

If you really want to do histone PTMs right, though, you need to do something about all those basic residues. The gold standard is propiprolypropylianation (spelling?) where you derivatize every lysine that is unmodified and then you digest. If you've never done it, it sucks. You can't buy the reagents. You have to buy some reagents and then create the active (reactive) form and then you add it to your sample before it expires. Not fun and I didn't think - for even a single second - that the reaction could be controlled well enough to do it for single cells.

I was wrong. 

This group did some super high precision work on the CellenOne and not only isolated single cells well and lysed them in tiny volumes, but they also dumped that live reactant into the single cell lysates. Derivatizing their lysines from single cells. When I said that I see histone PTMs in every single cell data set, it's not hundreds of them. It's the easy ones. K27/K28 (depending on whether you count the M, which UniProt does) and K80 are friendly and tryptic. Unfortunately, there is also a lot of homology in all the histones so finding a modification site is generally easy at a peptide level....tryptically digested it is tough to figure out what histone it came from. If you propionylate (spelling?) then you get a longer sequence that can help break those populations up and assign them to the original correct proteoform.

nSWATH - more success with small fast DIA windows!

 

Now that we've got multiple instruments breaking the 100 scans/second limit, we're starting to see some promising old ideas coming back. One is the data independent acquisition (DIA, or in this case, the trademarked "SWATH") running with isolation widths we'd normally see for data DEpendent (DDA). To get enough scans across the peak, this group compromised a little on their mass range limit but sees improvements in peptides/proteins and %CVs in a 2 proteome digest standard. 

Mapping the human hematopoietic stem and progenitor cell hierarchy with single cell proteomics!

Holy cow. I think it might be time to stop saying things like "the emerging field of single cell proteomics". I think it's emerged and is the tasty form of the cicada (where are those things, btw? I thought the US was going to be buried in billions of them this summer...I've seen nothing but the evil spotted lanternfly).

This study might be a new high water mark for more than a few reasons. 

The first is that this is just a really smart model to apply the technology to, right? Take the stem cells and force them to differentiate out. The single cell proteomics (SCP)  method in question appears to be Reticle, which makes sense for 2,500 single cells.  

The second is - wow - the integration of these data is just top notch. I've tried doing this as well and people have let me publish the results - to see it done THIS well, is just fantastic. 

Third, and most important, is that they actually learned something from this IN VIVO analysis of these cells. They took bone marrow cells from 6 patients (!!) and cleverly sorted their control (stem cell/progenitors? not my field and just trying to follow along) as well as randomly sorted cells that have a differentiator marker on the cell surface. (FACs machine, just take anything within this great big gate and randomly deposit them, I don't care what size they are or how they autofluoresce over here, just catch them arbitrarily and drop them in wells).

Really really super cool study and something that should get any immunologist interested in what you can do with our technology. Though.....I'd personally be quick to warn them that integrating the data this well might take a team like this who has been steadily improving study after study over the last 4+ years. 

1,220 FFPE Tissues analyzed at >4,000 protein depth/sample!

 

Hell yes. Now we're talking. Gotta run, but you should 100% check this one out! 

10,000 proteins in aggregate quantified across an invaluable sample library. What an outstanding resource for the future and a demonstration of what proteomics is capable of today! 

Proteomics discovers first candidate diagnostic markers for disease in seals!

 

I take a couple of weeks off to go on a bunch of interviews and someone graffiti'ed the front of my favorite journal? 

It turns out that seals have been increasingly suffering from a gross disease called domoic acid toxicosis (link) and there are no diagnostics. Currently it sounds like you've got a sick acting seal or a dead seal and someone guesses. 

This group fixed that with proteomics of their CSF! (Cerebral spinal fluid)

They identify a list of proteins differential in seals that are suffering from symptoms and an additional list of proteins that can determine which ones have the chronic DAT. 

BOOM. Proteomics steps in to a disease I bet almost no one has heard of and comes up with diagnostics. I imagine that it isn't entirely without challenges....like where do you get CSF out of a seal...but it's still a promising use of the magical magnets in our labs. 

Moving on!