Ben Orsburn received his PhD at Virginia Tech (VA, USA) by applying mass spectrometry to solve biological problems. He has held positions at Johns Hopkins (MD, USA), Thermo Fisher Scientific (CA, USA), NIAID (MD, USA) and the National Cancer Institute (MD, USA). He has spent time developing LC–MS methods in over 150 labs around the world and utilizes these skills as the founder of LCMSmethods.org, a volunteer organization that provides validated instrument methods to researchers in an increasingly broad range of application areas.
This exclusive editorial explores the very recent and exciting developments in the field of single cell proteomics. Is it time to realize the potential of SCoPE-MS?
Over the holidays I started one of my favorite chores, plotting my conference schedule for 2020. Proteomics has grown rapidly over the last decade and keeping up with the newest advances in hardware, methods and software is a constant requirement to delay obsolescence. To assist, there are annual proteomics meetings in nearly every country and in the US it is closer to an impressive and impactful meeting somewhere every month or two. Regardless of where I end up, one thing is certain, single cell proteomics will be a major topic on stage and over coffee (and possibly over a beer or two).
Single cell proteomics has been a dream for practitioners of the ion selection arts since John Fenn’s lab first described the ionization of proteins and peptides in the 1990s. The whole term ‘proteomics’ can be described the same way, as the name implies a promise we can only recently claim to have fulfilled – collecting at least a little data from all the proteins an organism produces. Where genomics has been measuring every gene or transcript present for a decade or two, the ability to measure the true proteome didn’t become a reality until about 3 years ago. The gap between the first human genome and proteome draft maps? Over a decade. The trick, it turns out, was the successful coupling of three components: A mass spectrometer that was just fast enough, chromatography that was just sophisticated enough and – while often overlooked – data processing algorithms that were just good enough to pull it all together. Improvements in sample handling and protein extraction haven’t hurt, either. I have my own biases, but I consider June 7, 2017 the day proteomics became a reality, when ‘An optimized shotgun strategy for the rapid generation of comprehensive human proteomes’ was published in Cell Systems . The punchline? Near-theoretically complete coverage of the human proteome in around the same amount of instrument time required to perform transcriptomics with RNA-Seq. The method is fast, relatively easy and didn’t even require the purchase of the most expensive mass spectrometer on earth. A fast, mid-tier benchtop system could do the job, realistically opening the proteome to any lab that really wants to measure it.
The downside to this, and to many recent proteomics methods, is the amount of material required for analysis. Sure, by directly measuring the proteins we’re closer to the phenotype than when we study the transcripts, but there is some cost to muddying the signal by combining molecules from 10 times the number of cells. And this is the real crux of proteomics technology, relative to genomics. Proteomics has always lacked the ability to amplify protein signal. Genetics, of course, hasn’t had that problem thanks to the polymerase chain reaction (PCR), which allows you to make what seems like an infinite amount of DNA out of any little bit you can find. I’ve seen the movies. I’m just waiting on the dinosaurs.
I’ll commit to another critical date here and that is January 24, 2017. Many things probably happened that day, but the most pertinent to this article was the posting of an understated article called: ‘Mass-spectrometry of single mammalian cells quantifies proteome heterogeneity during cell differentiation‘ on a preprint server . On this date – nearly 6 months before we saw realistic acquisition of complete human proteomes – the world got an open access text and hundreds of gigabytes of proof of the effective amplification of protein signal – enough amplification to allow, for the first time, proteomics of single human cells. Am I being dramatic? Sure. But this is a big deal. Are you wondering why wouldn’t I use this as the start date for the proteomics? Well, to be honest, it’s because my field, the proteomics field, didn’t really notice – at first.
I don’t know why a study this important was overlooked, but I have some ideas. Maybe it is because the method in question, Single Cell Proteomics by Mass Spectrometry (SCoPE-MS), seemed too good to be true? Maybe it didn’t help that the senior author was unknown and looked a little more like a rock star than your typical basement-dwelling ion chemists? As an aside, mass spectrometry used to require a lot of big magnets and those are easier to put in basements, hence the characteristic vitamin D deficiencies associated with practitioners of my field. I am sure that it didn’t help that this paper suggested the impossible was one of the first proteomics studies to appear in the controversial preprint server, BioRxiv. Whatever the reason, the work appeared to be mostly ignored and only appeared again in BMC Genome Biology nearly 18 months later . However, nearly ignored is not the same as ignored and this revolution has real grass roots, in large part because the technique makes a lot of sense. And, if you wanted to try it, it wouldn’t cost you very much money to give it a go.
Let’s back up – both genomics and proteomics techniques exist to allow the multiplexing of samples. The chemistry is very different, but the end results are similar, you get data from multiple samples at once. The most common in proteomics these days is the aptly named tandem mass tag (TMT). Today it is possible to label and combine up to 16 differently tagged TMT proteomes. This results in quantitative data on the proteins from all 16 samples simultaneously. TMT is used in some way in virtually every proteomics lab, due to the increase in throughput and corresponding decrease in cost per sample. While people might complain about the cost of the reagent, mass spectrometers are expensive to purchase and operate and the math increasingly favors the use of multiplexing.
TMT works because the tags have the exact same mass, but when fragmented in tandem mass spectrometry (MS/MS) each sample reveals a reporter ion barcode of a different mass. What the Slavov lab realized that had escaped the rest of us, was that only one channel needed to have enough signal to be sequenced to allow that protein to be quantified in every sample. There are limitations here, but the math basically rounds out like this – you need 100 times more signal to sequence a peptide than you need to quantify it.
SCoPE-MS exploits this discrepancy. The method has evolved in both Slavov’s lab at Northeastern and rapidly in the labs of the first quiet adopters of the technique, but the core strategy remains about the same. Cells are separated by flow cytometry or fancy robots to end up with tiny wells of three different types. Control wells are used to determine the background by applying the barcode to empty buffer. The second type is the carrier channel. Opinions differ, but the best data I’ve seen uses 200 human cells for the carrier channel barcode. The carrier channel provides enough signal for the sequencing to occur. Every other channel is carefully used to label the proteins from a single human cell.
Carefully is another key word. The cell must be lysed, the proteins digested to peptides and the peptides labeled with TMT with the minimum sample loss possible. I have done this. It is difficult. Nikolai Slavov told me in a Skype call for this article that most of his students get good data on the 4th try. (His students are faster learners than me). Unsurprisingly, one of the first follow-up papers on SCoPE-MS detailed the use of automation to mitigate these challenges . If there has ever been a reason for using a precise robot in a proteomics lab, this is it, I promise.
If you get past this stage without losing the successfully digested and labeled peptides from your single cells, you combine all your labeled peptides and perform the mass spectrometry experiment as you would for any common TMT experiment, with one notable exception. The limiting factor in SCoPE-MS is still sensitivity and the peptide signal must be acquired for longer than usual, with the current magic number around 300 milli-seconds per peptide. While that might not seem like a lot, it is about 6-10x longer than the normal proteomics experiment with today’s best technology when you aren’t sample limited. There is something interesting here as well. In general, the most expensive mass spectrometers are the fastest and/or have the most features. When you are accumulating ions for relatively long times and don’t require additional capabilities, it makes a lot more sense to purchase an older and simpler mass spectrometer than you might normally. In addition to less expensive hardware, single cells don’t need much of the expensive TMT tag, which is typically purchased in aliquots of almost 1 milligram. This goes a long way when you are labeling nanograms of protein. The math will depend on a lot of factors, but we’re talking about a total cost per cell in the range of what you might spend on lunch. This is not a normal statement for -omics technology, where we’re used to thinking in units closer to new Porsches per study. I’m not even finished.
The final bit of good news is that the data processing is nearly the same as any standard experiment. Although the instrument output looks strange to the experienced eye, no special software is necessary. Every lab can process this data and can optimize with the hundreds of files that Slavov lab has made publicly available from the date of the very first preprint. You’ll need to think about the normalization a little more, but I promise you it is worth it when you’re looking at the relative quantification of hundreds of different proteins in each individual human cell.
Prior to the holiday break, I’d thought this would be where my article would end. However, as I was drafting this another study posted on BioRxiv that put the progress from the initial study into perspective. If you were thinking of this technology as a curiosity, but something that would have practical applications maybe 10 years down the road, the December 5, 2019 preprint titled ‘Single-cell mass-spectrometry quantifies the emergence of macrophage heterogeneity‘ might change your mind . As before, this title might not grab your attention, but this is the practical application of SCoPE-MS realized on a massive scale. The abstract will sum this work up better, but I’ll take a shot at it as well: 10 days of instrument time. Single cell proteomic data on >1,000 human cells. With a total depth of >2,700 total proteins evaluated.
This is real. This is today. And this is the single cell proteomics revolution.
 Bekker-Jensen DB, Kelstrup CD, Batth TS et al. An optimized shotgun strategy for the rapid generations of comprehensive human proteomes. Cell Syst. 4(6), 587–599 (2017).
 Budnik B, Levy E, Harmange G & Slavov N. Mass-spectrometry of single mammalian cells quantifies proteome heterogeneity during cell differentiation. BioRxiv. doi:10.1186/s13059-018-1547-5 (2018);(This article is a preprint and has not been certified by peer review).
 Budnik B, Levy E, Harmange G & Slavov N. SCoPE-MS: mass spectrometry of single mammalian cells quantifies proteome heterogeneity during cell differentiation. Genome Biol. 19, 161 (2018).
 Specht H, Harmange G, Perlman D et al. Automated sample preparation for high-throughput single-cell proteomics. BioRxiv. doi.org/10.1101/399774 (2018);(This article is a preprint and has not been certified by peer review).
 Specht H, Emmott E, Petelski AA et al. Single-cell mass-spectrometry quantifies the emergence of macrophage heterogeneity. BioRxiv. doi:10.1101/665307 (2019);(This article is a preprint and has not been certified by peer review).