[RASMB] Chaotic Profiles

[H. Olin Spivey]

Friends,

    We have been trying to characterize a protein for a colleague that 
behaves unlike anything I have encountered with the ultracentrifuge. 
This protein is reported to have a monomer subunit of 116 kDa that 
undergoes self-association.  Sedimentation equilibrium experiments in 
the XL-A use the 6 sector centerpiece with a column height of about 
1.6 mm and are performed at speeds from 6,000 to 12,000 rpm. The 
buffer is 20 mM Tris, pH 8.0, 5 mM beta-ME, 50 mM ZnCl2, and 0.2 M 
NaCl.  Experiments are run at 10 deg. C.  Loading protein 
concentrations are 0.3 and 0.6 mg/ml giving initial A280 values near 
0.3 and 0.6, resp.  A280 values are averages of 16 points for each 
radial point.

    In all the experiments, the protein never reaches equilibrium even 
if we run for 60 hrs.  We don't perceive a decrease in signal 
amplitude over time so I don't think we are losing protein with time. 
Admittedly, we should be more quantitative in evaluating this, e.g., 
by integrating the signal.  However, the most puzzling thing is the 
random and large fluctuations of the profile during successive scans. 
After a large deviation they relax back to approximately the earlier 
smooth profile.  These random deviations persist throughout all runs 
lasting typically 36 hours, but on occasions 60 hrs.  The amplitudes, 
locations, and duration of these deviations are not reproducible from 
one occurrence to another, but they have an average periodicity of 
about 6 hours.  This pattern is not changed if we use FC43 silicon 
oil or not.

    The slow approach to equilibrium is not unexpected.  The protein 
sample on a size exclusion column separates into two peaks - the 
first peak corresponding roughly to twice the MW of the second one. 
This indicates a slow interconversion relative to the time of 
chromatography.  The chaotic deviations from smooth profiles is, 
however, inexplicable to me.  We have tested the obvious.  Control 
experiments with thyroglobulin at these same speeds and temperature 
behave normally.  Thyroglobulin has a similar MW (669 kDa) to the 
associated protein being studied.  Observing that SDS-PAGE bands look 
much sharper and straight with 0.3 M DTT, we have tried to use much 
higher concentrations of sulfhydryl reagents in these ultracentrifuge 
experiments. However, our first attempt with DTT caused an 
intolerably high 280 nm absorbance.  We have purchased TBP and may 
try this, but I am not hopeful.  Would interference optics avoid this 
problem with RSH reagents?  We don't have such optics, but we could 
send the sample elsewhere.

    Does anyone understand how these "chaotic" events occur?  Are 
experiments with higher concentrations of sulfhydryl reagents worth 
performing?  If so, is the interference optical system a better way 
to obtain these data?

    A cycle of these events can be seen at 
http://opbs.okstate.edu/~spivey/auc/chaos.html.  Since virtually all 
experiments show a large range of molecular weights, we analyze the 
profiles to obtain weight average molecular weights, M_subscript_w, 
at successive radial positions.  That is, we calculate M_subscript_w 
from the slopes of the ln(A280) vs. r^2 curve at each radius.  These 
are then plotted vs. A280, which is proportional to total protein 
concentration.  I have written the program to obtain these slopes and 
weight averages since I was not pleased with the Beckman program for 
these calculations.  In some cases the deviations are so large that 
MW averages decrease monotonically with increasing absorbance over 
the entire radial range!  Taking derivatives magnifies deviations and 
noise/signal ratios.  However, the same plots obtained from 
thyroglobulin data remain smooth and reproducible in time.  Also the 
primary data plots (A280 vs. r) show the deviations for the test 
protein albeit as much smaller deviations relative to the scale 
exhibited.  Such a primary plot for the 20 hr profile is given at the 
above site to puts things in the perspective we are more familiar 
with.  Looking at the latter, one can see that the random deviations 
in primary data don't need to be especially large to cause these 
effects.  But why are they present at all?

    Many thanks for your consideration and my apologies for the length 
of this letter.

Olin

-- 
H. Olin Spivey                       Phone: (405) 744-6192
Dept. Biochem. & Molec. Biology      Fax:   (405) 744-7799
246 NRC                              Email: OSpivey at Biochem.Okstate.Edu
Oklahoma State University
Stillwater, OK 74078-3035