[RASMB] Can DCDT+ be used to detect small change in protein conformation? What is its limit?

[John Philo]

Chris (and all),

You've caught me at a bad time with this query, as I am leaving tomorrow
morning for a month-long trip to China (and haven't finished packing yet).

It's unfortunately not very clear to me exactly how you are fitting the
data---are you always fitting as one species, are you always fitting the
entire range of sedimentation coefficients or just a window around the main
peak, etc? I would need to see the saved .fit files to really see what you
have done. Your documents also don't list the error bars on the numbers so
it's hard to evaluate whether the differences are even significant from that
point of view.

One thing I would be concerned about is that it seems likely you are
including too many scans in the computation. Twenty files is a lot for
absorbance data, and indeed given the scan numbers like 10-29 or 20-39 it
sounds like you are including a large fraction of the run, which sounds like
it may be too many even for such a low molecular weight protein.

Part of what you are asking appears to be a technical question about the
DCDT method and why the apparent sedimentation coefficient varies depending
on when you analyze the data during the run. This is actually a (hopefully)
well known phenomenon, especially for lower molecular weight proteins.

I will also point out that according to Walt Stafford's last statement on
the issue (see his RASMB e-mail at
(http://server1.bbri.org/rasmb_archives/1996/23.html) he says "Below about
15-17kDa it is nearly impossible to get the boundary to clear the meniscus
and so it is recommended that the time derivative method not be used below
that range". I think that is a bit extreme, but if you are working with a 13
kDa protein you will definitely have systematic errors at some level---they
are unavoidable, and documented in my paper that is the basic reference for
the program [Philo, J. S. (2000). A method for directly fitting the time
derivative of sedimentation velocity data and an alternative algorithm for
calculating sedimentation coefficient distribution functions. Anal. Biochem.
279, 151-163].

The situation and basic source of the errors is slightly different depending
on whether one is fitting to the g(s*) data or to the dc/dt data, with the
dc/dt fits usually having somewhat smaller errors on the sedimentation
coefficient. Either way, even if you are keeping the same total number of
scans in the analysis, as you move that group of scans earlier and later in
the run you will be varying the degree of peak broadening arising from the
broad time span, and thus the systematic error (see Fig. 6 in that paper).

So the systematic shift in sedimentation coefficient as you analyze scans
from different times in the run is, at least in part, an expected result
based on the inherent approximations of this method. On top of that, if (as
is likely the case) you have more than one species in your sample, then as
you analyze the data later and later in the run you are 'losing' the faster
sedimenting components.

You can of course do a numerical simulation of your experiment to easily
evaluate the degree to which this shift is expected (and I would recommend
doing that). Other sources of systematic shifts in sedimentation coefficient
as the run proceeds are errors in the meniscus position, or temperature
changes during the run.

So, bottom line, yes, it is possible to detect small changes in protein
conformation with this approach, but for proteins of this size you must be
very careful and must compare data at similar times in the run (and of
course at a time when all the relevant species are still entirely in the
cell). Walter's difference sedimentation suggestion is of course another
good approach.

To give you an idea of the precision and reproducibility that is possible,
at least with larger proteins, and if one is careful, here are some of my
results from DCDT+ for the same protein measured 4 times over 18 months:
                6.444 +/- 0.005 S
                6.448 +/- 0.005 S
                6.450 +/- 0.005 S
                6.450 +/- 0.005 S

Overall, though, I always say that the most accurate sedimentation
coefficients will never come from anyone's DCDT implementation. The most
accurate values come from whole boundary approaches, like my SVEDBERG or
Peter Schuck's SEDFIT or Joachim Behlke's LAMM etc., because those
approaches use data from the whole run, not just a small portion.

With regard to the broader issue of whether this approach is a good one for
your experiments, I'm not sure of this but apparently you are always
analyzing the data as a single species. That may not always be an
appropriate model for these samples, and if it isn't then the results are
automatically at best suspect and at worst invalid. You seem to be assuming
that the protein is always a monomer, but isn't it likely that your aged
sample can form aggregates, not just change the conformation of the monomer?
And don't your refolded samples potentially contain two species, a folded
and an unfolded form (and maybe aggregates too, for that matter).

You may want to consider whether what you really want to report is an
overall weight-average sedimentation coefficient for the sample, rather than
the value from a fit as a particular species (which will not necessarily
give you a thermodynamically valid weight average if you have multiple
conformations within that 'single species').

We can perhaps discuss this further after my return.

John

  -----Original Message-----
  From: rasmb-admin at rasmb-email.bbri.org
[mailto:rasmb-admin at rasmb-email.bbri.org]On Behalf Of Chin, Christopher
  Sent: Friday, September 20, 2002 11:55 AM
  To: John Phillo (E-mail)
  Cc: Rasmb (E-mail)
  Subject: [RASMB] Can DCDT+ be used to detect small change in protein
conformation? What is its limit?


  To John, all DCDT+ users and all interested colleagues,

   John, I would like to have your comment in the following three
experiments using RNAseA as a model and DCDT+ as the method to do data
analysis.

  Experiment 1:  The purpose is to check the stability of RNAseA. The
comparison is made between the protein left at room temperature for a couple
of days and that storage in the freezer.
  Experiment 2:  Question: Can the denatured RNaseA (has 4 s-s bond) in
GuHCl be refolded back to its native state in the absence of reducing agent?
  Experiment 3:  Question: Can the denatured RNaseA (has 4 s-s bond) in
GuHCl be refolded back to its native state in the presence of reducing agent
TBP (Tri-n-butylphosphine).

  The major issue here is how I can differentiate the DS difference due to
scan chosen from the DS difference due to real experimental finding (between
the control and the sample).

  I am pleased to say that, using the criteria that I have stated, e.g.
maximized or magnified the DS difference due to scan chosen (use a smaller
number for denominator when converted to %), minimized or play down the DS
difference between the control and the sample (use a larger number for
denominator when converted to % ), the results are pretty much I have
expected.  Do you think I try to push the instrument limit too far to obtain
these results by manipulating the treatment of data?

  s20,w values reported here are the results of fitting dcdt.

  Thank you in advance for your comment or advice.

  Chris

  --------------------------------------------------------------------------
  Christopher Chin
  Manager, XLA-Analytical Ultracentrifugation facility
  Sealy Center for Structural Biology
  HBC&G, 5.134 MRB.UTMB, Galveston,TX 77555-1055
  cchin at utmb.edu, 409-772-1693, efax 708-585-1920
  --------------------------------------------------------------------------
-


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.rasmb.org/pipermail/rasmb-rasmb.org/attachments/20020923/5afc623a/attachment.htm>