Friday, November 17, 2017

What is the best digestion method for ultralow levels of protein?

I don't have to tell anyone doing proteomics that the demands on us are getting harder as time goes by. We're seeing a lot less "how many proteins can you find in this unlimited amount of cell culture I have here" and more -- "what PTM changes the most in the membrane prep of these 4,000 cells my grad student spent 11 years isolating from the earlobes of these 2 gnat species."

This new study takes aim the latter problems -- how do different sample prep techniques compare when we've got incredibly low amounts of sample to work with?

The comparison is my favorite prep method, FASP vs. solid phase single pot enhanced sample prep (SP3) vs digestion in a stage tip (iST).

Full disclaimer here -- I've never personally tried the SP3 or iST -- I'm not even sure if I've heard of the latter.

SP3 was originally described here, in case you're interested (they validated it by digesting a single drosophila embryo. I'm no expert on the subject, but that sounds like it might be comparable in size to 4,000 gnat earlobes)

iST was introduced in this Nature Methods paper, and now I know what they're talking about. We've seen this in more than a few papers over the years. I think the abbreviation was throwing me off.

These authors start with low amounts of protein and/or cell counts for all 3 methods. I'm talking low --their high point is 20ug of starting material!

They compare the number of peptides identified after using each digestion methodology and, perhaps more importantly (at least to me) the reproducibility of the peptides ID'ed.

Most of this is given away in the abstract so I don't feel bad telling you about it here.

At "high" load (20ug of protein is high load these days, LOL!) FASP is right there with the other 2 methods. Both in peptides ID'ed and it looks like it has the best CV -- though iST is right there with it.

However -- as you drop down in starting amount, the effectiveness of FASP drops faster than the resale value of a BMW i8 after Tesla's new roadster was revealed. (Maybe not that fast)

SP3 looks good in terms of peptides ID'ed, but the CV gets wonky as the load drops. The clear winner in both categories is the iST methodology.

The authors go on to validate this by flow sorting some sells (!!! awesome !!!) to just a few thousand and reveal that the sorting still left a heterogeneous mixture of whatever cells there are.

I want to give a big shoutout to Dr. M for sending me this ultracool paper. It is quite seriously my favorite thing I read this week.

Oh, and all the RAW data has been uploaded to ProteomeXchange here via PRIDE (PXD006760).

ERROR/EDIT:  It has been brought to my attention by some readers that I completely misinterpreted the results of this paper! Looking at it, I definitely concur. I did not interpret the main figures and -- really -- the central findings of this paper correctly. Swap what I said about SP3 and iST in terms of peptide IDs and you're on the right track.

I'm going to leave my mistakes here for the sake of posterity -- if you'd like an accurate description of the results of this paper, please see the comments on this post!


  1. I frequently read your blog, and generally enjoy the content, so please do not take this as reflection on that. I have no stake in this paper, I just think you misrepresented the findings. SP3 was the clear winner here in the low ug range:

    Whereas the number identified proteins remained in the same range for SP3 independent of the amount of starting material (20–1 μg), a small drop in detectable proteins could be observed for iST comparing the 20 μg and the 1 μg sample. On average, around 23 500 peptides and 3020 proteins could be identified with iST in 1 μg of HeLa lysate and approximately 27 000 peptides corresponding to 3300 proteins with SP3. Even in the low μg range, both methods displayed high quantitative reproducibility between technical replicates. However, SP3 (R2 = 0.97 (1 μg)) slightly outperformed iST (R2 = 0.93 (1 μg)). In addition, precision was high for both protocols with median CVs of 8.5% (SP3, no normalization) and 7.4% (iST, no normalization) (Figure 2B).

    8.5% vs. 7.4% is wonky? I'd personally take a 3500 peptide number increase, with a sacrifice of 1.1% CV any day.

    Even in the macrophage data:

    In line with the previous experiments, both SP3 and iST show high precision with median CVs of 8.1% (SP3) and 7.7% (iST) after normalization and a high correlation of label-free quantification data between experimental replicates (R2 = 0.93–0.96, Figure S8).

    In terms of IDs with the macrophages:

    SP3 showed the best results identifying on average 3152 proteins and 22 202 peptides. Whereas on average 2343 proteins and 17 674 peptides were still detectable with iST, we found that FASP is not feasible for low numbers of FACS sorted cells.

    So, although the CV with SP3 is higher by 0.4%, the number of peptide hits is >4500 higher. All of this leading to:

    Since SP3 outperformed both FASP and iST when analyzing low numbers of macrophages, we applied SP3 for the ex vivo characterization of TAMs that were isolated by FACS

    Not sure why you can state that 'the clear winner is iST', when in the paper they state the clear winner is SP3.

  2. Dear Ben,

    it's actually the SP3 that comes out as the winner in that paper.