[proteomics] High Throughput Protein Identification by 2DGE-MALDI MS: The Reality Versus The Dream

  • From: Mavi Gozler <mavigozler@xxxxxxxxx>
  • To: proteomics@xxxxxxxxxxxxx
  • Date: Thu, 8 Feb 2007 23:04:50 -0800 (PST)

If you are thinking about building a proteomics laboratory or environment, or
writing your proposal, you will probably want to focus all your efforts from
and put as a low budget priority the approach of identifying proteins by 
two-dimensional gel electrophoresis and MALDI MS of peptides extracted from
in-gel digestion.

It's labor-intensive, and often the percentage of good results do not live up
to expectations.

At one meeting I attended, one speaker who made use of a proteomics expert in
his study told me that his expert does not even bother with 2DGE-MALDI MS-based
identifications.  He devotes his time completely to LC-MS/MS work, probably
using stable isotope labeling for when he wants to get quantitative.

We are just sort of starting out doing expression proteomics work and wanted to
direct a great deal of effort to 2DGE-MALDI MS approaches to "high throughput"
protein identification.  After actually doing it, I can see how one can sour on
the idea of using 2DGE-MALDI MS.  It is attractive initially, since getting 2D
gels to look good takes a shorter amount of time, and the use of large gels and
narrower pI range strips gives one the illusion of protein purity in a single

But with and after the gel excision, things can get complicated.

First there is the robotic spot cutter we use.  Without our particular robotics
spot cutter, just sitting there for any length of time, it can lose its spatial
calibration for reasons only Bio-Rad knows.  Calibration is also complicated
and involved and expensive:  it seems we went through a half dozen large sheets
of PVDF (yes!) to get the instrument calibrated at one time (moreover the
camera was not properly placed).  

Although not given as an instruction by Bio-Rad, prior to any spot cutting
task, the user is well-advised to keep and use scrap gels to test the spatial
calibration of the cutter.  Before we realized this difficulty, we had lost one
set of valuable gels because of wrong assumptions involving the spatial

As necessary controls for the cutting and in-gel digestion process, one must
ALWAYS include control proteins (one whose identity is certain) cut from the
gels.  Naturally we use the molecular weight markers used for the second
dimension.  We designate manual cutting of several spots:  from a single band
we cut one spot, and then from the same band 3 or 5 spots, to gauge the
sensitivity of the process.

Then we go to the in-gel trypsin digestion.  We have a MassPREP liquid handling
station that also works better in theory than in practice.  The thing is
supposed to run overnight while the scientist is away, but even after the deck
calibration, the tip and lid pickup tests, and other preparations, it seems
something ALWAYS goes wrong with it.  So we have abandoned the overnight
all-in-one process and divided digestion into 2 parts in 2 days, done while the
scientist is on duty/at work, to make sure the method moves along.  The first
part does the reductive alkylation up to the drying of the gel prior to trypsin
addition.  The 2nd part is the addition of trypsin all the way through to
extraction.  Why don't we do this manually instead of using the MassPREP? 
Because, as error-prone as the MassPREP is, whenever a human is involved in
pipetting solutions to massive numbers of wells on two or more multi-well
plates, the likelihood of error or well contamination by keratins is greater. 
And so we try to automate as much as possible.

One of our problems in the laboratory in which we work is consumables
availability and delivery.  We work in an environment where money is more
valuable than time---which probably means we should not be operating a
proteomics laboratory in the first place!!  For instance, the people who bought
the MassPREP did not buy the MALDI spotting option, and no one ordered
ZipTips---which EVERYONE seems to be using---because of the expense.  Yes, we
have no idea how much ZipTips improve the MALDI signal because of the
tight-fistedness of operations around here.

So after the in-gel peptide extraction is done, we go to MALDI spotting by
hand.  Because this is a human activity, one has to be on their toes.  I have
described a method where the technician does not try to spot the whole plate in
one sitting.  We first transfer the extracted peptides in their entire volume
to narrow conical multiwell plates (MassPREP recommends ABgene 080 plates, and
that is smart).  We then dry this down with the vacuum centrifuge (SpeedVac).

We are then ready to spot.  As a standard solvent, we use 50% MeCN + 0.1% TFA. 
The technician adds about 2.5-3.0 ul of the solvent to the top of the well, and
only to ONE COLUMN of wells at a time (8 wells), because the 2.5-3.0 ul can
evaporate quickly.  The tech does a quick spin in the SpeedVac (because it has
a plate carrier) to bring the solvent down to the bottom of the conical well. 
The matrix (we used saturated or >20 mg/ml CHCA in the 50% MeCN + 0.1% TFA
solvent) is then spotted to the MALDI target plate:  the Gilson P2 pipet with
proper tip is set to about 0.8-0.9 ul.  Without changing tips used to spot
matrix, we immediately go into the well, use the tip to stir circularly in the
well, and pipet up and down a few times, and then take a sample out of the well
and pipet it into the matrix bubble already on the plate.  We do not even
bother to pipet up and down in a mixing action, as we probably lose matrix and
protein into the tip anyway.  The matrix seems to crystallize and form a nice
bed.  I advised people to do a spectral acquisition first on that spotting,
then if they want to do a relayering with EtOH or MeCN, or use one of the
cold-water washing techniques to improve signal-to-noise, they do that after
the first run acquisition.

As an alternative to manual spotting, I tried to automate the MALDI spotting
(albeit without ZipTip enrichment) by setting the 96-multiwell plate in an LC
Packings autosampler connected to an LC Packings Probot MALDI spotting tool,
but I could not get LC Packings to cooperate with me to get the software to
work correctly.  The autosampler is run by the Chromeleon software, and the
Probot is operated by a separate piece of software called Microcarrier, and the
user interface for both programs makes MathCAD understandable to a 6-year old. 
LC Packings was not very interested in the concept, since they probably thought
it bizarre that people would try to use their autosampler/Probot to do MALDI
spotting;  more shortsightedness here.

Finally we get to the spectral acquisition.  We use the Waters/Micromass
M@LDI-LR, an obsolete piece of equipment.  I have already gone on about the
problems with MassLynx software used to control the instrument, and to acquire
the spectrum, hopefully with accurate masses.  MassLynx is bug-free (not
really) compared to the bioinformatics nightmare that Waters/Micromass offers
to analyze the MALDI data and do protein identifications, ProteinLynx Global
Server.  I cannot say enough bad things about PLGS, although I have already
said plenty.  Waters/Micromass has not responded or replied to my laundry list
of things that are wrong with PLGS, and why no one should ever waste their
valuable time using it, even if it was handed to them without cost (for the
cost would be your time!).  They continue to believe in the usefulness of their
product, although I have put out a call to anyone who claims to use it with
good results.


Now I suppose that no one really said that expression proteomics work was the
proverbial walk in the park.

But tell me that my impressions of what is presented in the literature are

Go to any proteomics journal, or an article in which proteomics was used as a
tool to answer a question, and there is probably a multi-page table with a list
of 250-300 proteins reportedly identified, maybe with a figure of well
annotated gel (or gel detail) annotating the spot.  Have the authors stated
something like, "we detected 800 spots on the gel, cut out 500 of them, went to
MALDI and identified 450 of them."  What is that?  A 90% identification rate of
excised spots?  Did they even report the score and/or confidence
level/probability of identification?  Be cautious about that!  I believe some
are reporting the top-scoring hit off of MASCOT or whatever they are using, and
not bothering to tell you if it is indeed a high probability of identification.

I just want to warn you out there that if you are planning on using proteomics
as a "tool," then do not get the impression that this "tool" is the least of
your worries and frustrations doing your research.  It's really amazing
actually that many laboratories are offering the many techniques of proteomics
as a "service."   

For instance, if I have a 2D gel with 90 proteins I have designated for cutting
and identification, and I deliver the gel to the proteomics "service"
laboratory, do you not expect to see 90 high quality MALDI spectra with 90 high
confidence protein identifications in the report handed to you???  Or at least
say 3 in 4, say about 75 of them.  Right?

What if you got a report back that said, we got 35 high quality MALDI spectra
and only  15 identifications?   What would say to that?

Would you say that proteomics as a tool has fully matured?  Come into prime

Would you call it "high throughput"?  Or rather "high throughput in your

If anyone has actually paid for a proteomics service and been pleased with the
results, I invite them to respond to this post with the actual numbers, the
hard data.  I want to see for myself just how "high throughput" it is.

--- SMH

Sucker-punch spam with award-winning protection. 
Try the free Yahoo! Mail Beta.

Other related posts:

  • » [proteomics] High Throughput Protein Identification by 2DGE-MALDI MS: The Reality Versus The Dream