Let's see... a typical mammalian cell contains about 250pg of protein, distributed over, perhaps 10,000 proteins.

Let's assume an 'average' molecular weight for the proteins of, say, 50kDa.

250pg = 5fmol of protein

5fmol = 3,000 million molecules, which sounds like a lot, but remember, that is distributed over 10,000 proteins, even without calculating their further subdivision into post-translational space. This is, on average, 0.5amol of each. In my opinion this average value is below the reach of any MS system, including QqQ systems, when the analysis is conducted in a 'dirty' cellular background.

If you were lucky, the asymmetry in protein expression (perhaps four or five orders of magnitude) might mean that the proteins in which you are interested might be at the top of the pile, and you might be able to analyse them. But every protein above the average demands that others be below the average.

And all of this is theoretical, before you embark on the handling of a single cell preparation.

So, in my opinion, single cell proteomics in any global sense is not currently possible, by MS. Optical methods can give single molecule sensitivity, of course. And, immunoassays are more sensitive than MS proteomics.

Nikhil, please check the work of Jonathan Sweedler at U Illinios - Urbana/Champaign. He has done that stuff. Check doi:10.1039/c3an01010d, doi:10.1038/npp.2013.145, doi:10.1038/nprot.2013.035 or doi:10.1021/ac302283u. The latter two are also availble as a free article.

Sweedler group did the pioneer work, but only focused on several neuropeptides and not on the proteome level. And the cells they used are big, not on the 10-20 micrometer level.

I think the modern MS has the ability to reach the sensitivity, but the main issue is how to get the small amount of materials to mass spectrometer. If you use bottom-up strategy, how to extract, digest the single cell proteins; If you use top-down, the issue will still be the sensitivity and also the protein loss. In a word, the main challenging is sample preparation.

But I believe I will be better in the near future.

It's not easy. First of all, do you have experience with proteomics in general, i.e., with sufficient sample-sizes that sensitivity is not an issue? Do you have access to nanospray, and to colleagues who are experienced in microfluidics and flow cytometry? Do you have experience with isolating - and characterizing - single cells? Basically, what you're proposing is very challenging, and is probably not the best place to start. With that being said, however, best of fortune to you! You might want to take a look at these: http://www.sciencedirect.com/science/article/pii/S0958166913005843; http://www.ncbi.nlm.nih.gov/pubmed/21347466; http://www.pnas.org/content/109/2/419.abstract.

(Did you try googling 'single-cell proteomics'?)

But I believe it will be better in the near future.

Dear John, Liangliang Sun, and Marcus, Many thanks for your reply.

Markus, because I am thinking of doing it either on brain slices or isolated primary neurons. Those are quite helpful.

I agree with you LIangliang about the sample preparation being the main challenge.

Since the sensitivity of the mass spec we have is 100 attomol, in theory even if i manage to load everything from one cell,(correct me if i am wrong), We would end up sequencing only the most abundant proteins. (actins, tubulins etc)

I might have to try multiple cells of the same type to increase the amount of protein. (of course it doesnt stay single cell proteomics anymore). :P

Dear John, Yes we do have access to nanospray ESI-MS etc. My question was more pertaining to Mass spectrometry based approach. The articles you refer to me are not exactly mass spectrometry based, although "Single cell proteomics" is the in title, its not REAL MS based proteomics. (of course I tried googling it before posting it here, and I found the same 3 articles you have sent me)

I think for single cell proteomics, the traditional MS/MS based peptide identification will not very successful at present, and the accurate mass and time tag based strategy (no MS/MS) will be more useful. In addition, the target proteomics way will be more straight forward, due to higher sensitivity.

An alternative approach could be proteomics imaging (depending on what you want to find in the end of course) using SIMS. These guys did single cell/sub-cellular imaging in a very nice fashion I must say: http://www.sciencedirect.com/science/article/pii/S0169433206005332

Perhaps laser-microdissection and a very narrow (30µm ID? monolithic?) column coupled to MS might give some data.

We did this with FACS-sorted cells of Arabidopsis roots: Lan P, Li WF, Lin W-D, Schmidt W. 2013. Mapping gene activity in Arabidopsis root hairs. Genome Biology 14(6):R67

Dear Liang, Thanks for your input. I agree with you that a targeted strategy is likely to be more successful, but since I plan to do a more discovery based experiment, an unbiased study is necessary. But I went through some of your papers and the techniques you have been developing are quite nice. Have you thought of incorporating SWATH analysis as well?

Dear Magnus, Proteomics Imaging is an exciting area, unfortunately we lack the setup to do it. Once again, I would like to do a hypothesis free experiment. Just characterize as many proteins that I can from the cytoplasm of a single cell (or many cells of the same type). So its not necessary for my case. Its interesting none the less.

Dear Stefanie, Many thanks for your reply. What do you mean by a very narrow column? By the way, As far as I know from the laser capture microdissection setups that we have, It is rather impossible to get a single cell using it without destroying the surrounding tissue and hence 'contaminating' the single cell sample.

Everyone else, many many thanks for your inputs. I hope to set up the experiment/s sometime soon and will keep you posted. I like this forum that people from very different fields come together to address a very difficult problem (I would call it ideal science). :)

Dear Nikhil, I think SWATH is a good way to go since it will benefit low abundant proteins. But I am worrying about the quality of generated tandem spectra from the tiny amounts of sample, especially low abundant ones. I prefer to generate a home-made database based on bulk analysis to approachable proteome list and corresponding peptide information. Because for single cell analysis you tend to identify high abundant proteins (peptide), for the database searching strategy, it can be different from the traditional shot-gun proteomics.

In addition, the sensitivity of separation and detection system is also important. Like Stefanie said, the narrow separation column (10-20 micrometer ID reversed phase column, Anal. Chem. 2004, 76, 144-154; Anal. Chem. 2007, 79, 6174-6181) with 10-20 nL/min level flow rate will improve the ionization efficiency a lot, which will be very useful for single cell analysis.

Recently, we found capillary electrophoresis-mass spectrometry is very useful for high sensitive proteome analysis ( DOI: 10.1002/anie.201308139 ).

Hope it can help.

Dear Nikhil, did you look for "cell MALDI imaging"? I don't think there will be already papers about one single cell, but this approach could be useful for MS-based cell proteome analysis.

Also, watch out for contamination, specially working with such small amounts of proteins...

Let's see... a typical mammalian cell contains about 250pg of protein, distributed over, perhaps 10,000 proteins.

Let's assume an 'average' molecular weight for the proteins of, say, 50kDa.

250pg = 5fmol of protein

5fmol = 3,000 million molecules, which sounds like a lot, but remember, that is distributed over 10,000 proteins, even without calculating their further subdivision into post-translational space. This is, on average, 0.5amol of each. In my opinion this average value is below the reach of any MS system, including QqQ systems, when the analysis is conducted in a 'dirty' cellular background.

If you were lucky, the asymmetry in protein expression (perhaps four or five orders of magnitude) might mean that the proteins in which you are interested might be at the top of the pile, and you might be able to analyse them. But every protein above the average demands that others be below the average.

And all of this is theoretical, before you embark on the handling of a single cell preparation.

So, in my opinion, single cell proteomics in any global sense is not currently possible, by MS. Optical methods can give single molecule sensitivity, of course. And, immunoassays are more sensitive than MS proteomics.

I think one needs to think through whey exactly one would want to "do" single cell proteomics. Recall that single cell mRNA sequencing allows you to inventory and count pretty much all transcripts. That gives you a quantified sense of gene "expression". Of course you will first need to work out handling conditions so that the original information content is preserved, for the result to make any sense. To the extent translational control and protein turnover is expected to meaningfully vary form cell to cell and is important for your questions, then of course measuring specific protein concentration would be of interest. That may not be a common question though. Likewise for those questions where post translational modifications matter, it would also be of interest to measure the specific modified protein concentration. But in neither of these types of questions "proteomics" would seem optimal, since it would be expected to identify a certain entity but would not provide a very accurate quantification. It would seem more appropriate to work out sensitive and specific protein specific assays, of which there are many different types. This would require to first have worked out a meaningful and addressable scientific question based on a testable hypothesis. But then again there is really no way around that if you want meaningful results. "Discovery based" or "unbiased" screening studies are of course not unbiased, since that are built on a hypothesis or bias that is not clearly stated, such as that abundant analyses matter more that rare ones, et.c. unwarranted assumptions and are also not designed to clearly discriminate between, or dissect, alternative possibilities.

Agreed with Dr. Beynon.

I think it has been done using on-line cell trapping/digestion in nanocapillary configuration. I can measure into the low attamoles of peptides on an Orbitrap (probably lower still in a targeted mode) so you have a couple of orders of magnitude to work with still.

I'd be impressed if you can routinely get to low attomol (for me, this is less than 10) in a complex proteome, with a S/N ratio for quantification of 10:1. We need zeptomol instruments!

Actually, zeptomol level can be reached with the modern mass spectrometer (i.e. Q-Exactive). DOI: 10.1002/anie.201308139

..but what is a zeptomole or two between friends? The new Thermo instruments do indeed have this level of sensitivity. Often the limitation in the system is the nanocapillary HPLC system components. I have some unique products which enable a user to approach theoretical limit of a nanospray nanocapillary system, developed in my lab here at Harvard. Contact me off line or visit www.esisourcesolutions.com for more details.

Liangliang - I didn't know that paper, and it makes interesting reading. 28pg is about 300 cells! But, only a couple of hundred proteins identified, and the quality of the analysis emphasises identification rather than quantification. The claim of zeptomol sensitivity is for a single, pure peptide not zeptomol quantities in the complex biological context of a cellular background

300 cells, and let's say a transcription factor present at 10 copies per cell (without any consideration of post-translational heterogeneity). That's 3,000 copies of the protein, and thus, 3,000 copies of each peptide = 5 zeptomol, at a signal to noise ratio of 10:1 demands a sensitivity of at least 0.5 zeptomol. I'm just not buying this idea of single cell quantitative proteomics yet. I hope to be corrected one day, but for me the numbers, and stochasticity will be a major problem.

The paper also mentions 400fg of E coli proteins - that is 4 cells! Only, three EFTu peptides - poss the most abundant protein in the cell. So, the error is huge, since 3 cells or 5 cells are the next error steps. That being said, still quite an achievement. But, there's no way you'd quantify from that.

John - I'd like to see some academic papers on your various solutions. Can you point me to them, please?

One last thought. It is not instrument sensitivity that matters. It is experiment sensitivity - and the desired S/N ratio.

Nikhil, you might like to attend our 'not for profit' meeting (www.proteommx.org) next March - a great, friendly, small meeting on quantitative proteomics (not single cell!).

I fully agree with Rob, it's not the instrument sensitivity which matters in this case at all. If you run a complex sample, memory effects in the whole system will be huge (actually BECAUSE of the high sensitivity of the current instruments!). Assuming a typical bottom-up experiment using 2-5 ug of protein material (I know it's no single cell, but that's what I have the numbers for) we routinely identify 3000-4500 proteins in a 4h LCMS acquisition on the QTOF. At least 2-3 blanks later we are still easily able to find several hundreds of them, BSA which we use as a QC sample at the 50 fmol level can often be identified many blanks later. That means the relevant factor here is the dynamic range of the sample you have to deal with - one or two proteins is easy, from 48 proteins (UPS2 standard) spanning a dynamic range of 6 orders of magnitude you will typically not find more than 4 levels, the lowest one represented essentially by one-hit wonders and only identified because you knew what you were looking for. A typical complex sample will give you maximum the top 30% of the proteome (around 12-1500 proteins for E. coli, max. 10000, typically 5-7000 proteins for human cells), independent of how much you inject. Thus it doesn't matter whether the instrument can see 1 zeptomol of a single protein/peptide - it won't see it next to a protein present at 1 fmol amount as this would mean spanning 5 orders of magnitude in a complex sample.

On top of this though, sensible quantitation means observing at least 3 peptides which I can be used for quantitation. With this requirement the accessible dynamic range will plummet significantly.

I agree with the challenges mentioned by Rob and Marcus. But I do not completely agree with them. First, single cell proteomics is not only the instrument sensitive issue. How to handle subnanogram amounts of complex protein sample to digest them and transfer them to MS is more challenging. Second, The technique development is step by step, is not a one-step thing. But only when you try to do it and improve it, it will improve gradually. The shot-gun proteomics is a good example. Ten years ago, only 100 proteins can be identified from single dimensional nanoLC-MS system. After ten years, with hard work of all the researchers, now 1D-LC-MS can generate thousands of proteins. So even now from single cell sample, you only can detect the most abundant proteins, after 10 years, I do not know what will happen, but it will be much better if you can continuously work on it.

Agree with Rob and Marcus regarding sensitivity of instruments for identification of pg of proteins. I agree with Liangliang, that only abundant proteins could be identified with the present technical status of the instruments. The proteins present in very low quantity as compared to that of abundant proteins, would be present in proportionately in too low concentration when we'll deal with let say ~100fg of protein present in single E.coli cell.

Low abundant proteins always needs to be enriched. I am sure according to the need of hour in proteomics is low abundant proteins, and in coming future the we'll have techniques/methods to tackle this issue.

People, It is very interesting to read the discussion. But, can you tell me what is the purpose of determining proteomics, particularly of a single cell, if you do not associate it to functions. I talked several times to the adherents of proteomics or genomics asking "simple" question: How you will interpret the fact that in a pathology you observe an increase (or decrease) of a gene expression or even corresponding changes in the protein content? Do these facts evidence an increase or decrease in the related functions? Nobody could tell me anything about functions. In fact, far to often increased expression of genes or protein contents speak about DECREASED function of a protein or metabolic pathway.

In our hands, we tried to circumvent ("in a way") the difficulty to study proteins in a single cell by using SINGLE TYPE of cells. Using our blood-brain barrier (BBB) model we studied the proteome of a homogeneous population of brain capillary endothelial cells. Even though, and using a panel of global proteomics approaches, only partial information was obtained on the proteome (approximately 800 proteins identified, 100 displaying differential abundance when compared to cells that have limited BBB functionality, and around 10 validated).

Dear Yannis, Many thanks for your reply. Do you have a reference for it? I think considering the difficulties in REAL single cell proteomics, I am considering using multiple cells of the same type.

Dear Nikhil, all our work is presented in my ResearchGate space. If necessary do not hesitate to contact me for more information or for papers that you cannot unload there.

Alexander - single cell proteomics has potential - it lets us address the fascinating idea of stochasticity in cellular systems - "all hepatocytes are equal, but some are more equal than others". Let's admit, and honestly, that we typically deal with an average mixture of cells, and if we start with tissues, an average mixture of different cells as well. I look forward to the day when sorted cells are pushed, one at a time, in the source of an MS system, and all 200 molecules of a transcription factor are passed, in a top down process, into the source of an MS for intelligent downstream processing in a lossless fashion - generating 200 events at the detector. That might work. Sadly, my gut feeling is that the interrogation will be optical, not MS based.

Oh well...

Dear Rob, That is a good answer. It holds even more for the brain. There are 100s of different cell types and even within the same cell type, some neurons are potentiated (or part of memory trace) and some not. if one could address these differences using single cell proteomics, it would be very beautiful. But becos neurons are highly polarized cells and many of the changes are very local, one might even have to go to subcellular proteomics. But, one step at a time..

But its important to answer questions like that of Alexander, the answer of course cannot be, "because we can".

Dear Rob and Nikhil, I appreciate your answers. First, I admit that to study a single cell is an enormously difficult task. However, I have reservations. First, what cell to study? A cell taken from the culture? But most cultured cells have nothing to do with the real life! Most cultured cells are transformed and far too often damaged with antibiotics. If you take a cell from the primary culture, you usually disrupt functional connections with other cells. Take a neuron, for example, it does not function by itself. It works only with conjunctions with the astroglial cell. Therefore, you cannot study normal functions, you can only study components of the cells, its proteomics. But again, we do not know the functions of most proteins. I agree we have to try. How to separate the differences in functions in different parts of the brain, and the differences in functions at different parts of a single cell? Mitochondria, for example, located in the neuron's body, are far different from those located in the presynaptic part (sending signal) and the postsynaptic part (receiving signal) of the same cell. I think we, researchers, have to be more specific regarding the goals.

Hello Nikhil,

I heard a talk this week of Bernd Bodenmiller from the ETH Zurich. He uses mass cytometry for single cell analysis. Therefore a flow cytometer is coupled to an ICP-MS. In brief the technique involves:

- Fixation of cells with formaldehyde

- Permeabilization of cells

- Labelling of target proteins with antibodies that are labelled with metal atoms (i.e. lanthanides that do not occur in cells, he said there would be 100 atoms per antibody)

- After labeling cells are separated by flow cytometry

- Cells are injected into the plasma

- the metall ions are ionized and detected by the mass spec

With different metall ions, you can do multiplexing (i.e. 34 different proteins). Bernd Bodenmiller said that the detection limit for a single protein would be around 100 copies per cell. However this strategy is a targeted approach. Maybe you look in the literature:

http://www.ncbi.nlm.nih.gov/pubmed/23918170

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3627543/

....

Particularly since you are in the Netherlands, I'd suggest contacting Jan Marijnissen at TU Delft who performed single cell proteomic measurements on E. coli microorganisms by analyzing them by mass spectrometry from the aerosol phase. If you need contact information for him, please feel free to contact me through my Research Gate account.

David, These papers from Marijnissen are very interesting but have I missed a single E coli? I still don't see how any MS system can pic up less than an amol in 'dirty' workflows - (by dirty, I mean biological :) ). Have I missed a paper.

1: Agostinho LL, Tamminga G, Yurteri CU, Brouwer SP, Fuchs EC, Marijnissen JC.

Morphology of water electrosprays in the simple-jet mode. Phys Rev E Stat Nonlin

Soft Matter Phys. 2012 Dec;86(6 Pt 2):066317. Epub 2012 Dec 20. PubMed PMID:

23368048.

2: Stowers MA, van Wuijckhuijse AL, Marijnissen JC, Kientz ChE, Ciach T.

Fluorescence preselection of bioaerosol for single-particle mass spectrometry.

Appl Opt. 2006 Nov 20;45(33):8531-6. PubMed PMID: 17086265.

3: Dekkers PJ, Tuinman IL, Marijnissen JC, Friedlander SK, Scarlett B. The

self-preserving size distribution theory. II. Comparison with experimental

results for Si and Si3N4 aerosols. J Colloid Interface Sci. 2002 Apr

15;248(2):306-14. PubMed PMID: 16290535.

4: Ijsebaert JC, Geerse KB, Marijnissen JC, Lammers JW, Zanen P.

Electro-hydrodynamic atomization of drug solutions for inhalation purposes. J

Appl Physiol (1985). 2001 Dec;91(6):2735-41. PubMed PMID: 11717241.

5: Moerman R, Frank J, Marijnissen JC, Schalkhammer TG, van Dedem GW.

Miniaturized electrospraying as a technique for the production of microarrays of

reproducible micrometer-sized protein spots. Anal Chem. 2001 May

15;73(10):2183-9. PubMed PMID: 11393839.

6: Stowers MA, van Wuijckhuijse AL, Marijnissen JC, Scarlett B, van Baar BL,

Kientz CE. Application of matrix-assisted laser desorption/ionization to on-line

aerosol time-of-flight mass spectrometry. Rapid Commun Mass Spectrom.

2000;14(10):829-33. PubMed PMID: 10825245.

To be honest, I'm not sure if they ever published the results. They were done for a DARPA-funded project that we were in friendly competition on about ten years ago. But I did see individual mass spectra of individual e. coli with repeatable high mass ions that could only be proteins and I'm sure that Marijnissen (or Stowers if he's there still) would be willing to discuss them. You're welcome to mention my name.

As for the sensitivity, it can become extraordinarily high when there is a very low background as when no surface is involved in the ionization. For a short time, our lab held the world record in protein sensitivity at roughly 10,000 units of gramicidin. Here's the reference:

Russel, S.C. et al, Achieving High Detection Sensitivity (14 zmol) of Biomolecular Ions in Bioaerosol Mass Spectrometry, Analytical Chemistry, 2005, 77, 4734-4741.

YOu can separate single cell. YOu can easily catch a single sperm and look at it under microscopy.

Hi All, Our simoa technology enable you to do single cell Proteomic analysis.

Let me know if you want to know more about it!

Amélie

Amelie, a brief description of how you do since cell proteomics here and some links would be helpful. Am I right in thinking that you'll need a specific antibody for each protein or each post-translational variant. Most of us would not describe that as single cell proteomics, but single cell targeted protein assays.

Hi Rob,

Yes you are right.

You can read the paper from Stéphanie Schubert following the link below.

http://pubs.acs.org/doi/abs/10.1021/acs.analchem.6b00146

thank you!

Having worked with clinical, i.e. medical, proteomics I'd say for the foreseeable future it is not possible to in a meaningful way work with single eukaryotic cell proteomics. It IS marginally possible to work with the exome, and that is pretty close.

Mass-spectrometry of single mammalian cells quantifies proteome heterogeneity during cell differentiation

http://biorxiv.org/content/early/2017/01/23/102681

Hi,

I am also having the same problem, but doing on plant system here they have rigid cell wall. Can anyone help me in this I was using microtome but the cell recovery is very less.

Found a related paper published last year. Mass-spectrometry of single mammalian cells quantifies proteome heterogeneity during cell differentiation. Bogdan Budnik, Ezra Levy, Nikolai Slavov. 2017, bioRxiv.

Reduce non-specific binding related sample loss is the key to realize single cell measurement, as current MS sensitivity is very good with sensitivity in single cell level.

We provide an approach toward single cell proteomics using microfluidic droplet technique:

https://www.nature.com/articles/s41467-018-03367-w

Please read papers:

1) Nature communications(2018) 9:882.

2) Angew. Chem. (2018) 130, 1-6