From: John Philo <jphilo@mailway.com>
  To  : Hiroshi Fujita <lvl79@mbox.kyoto-inet.or.jp>
  Date: Sat, 1 May 1999 10:15:01 -0700

RE: On the determination of D

If you'll permit another few cents worth...

I think we all agree that sample heterogeneity can easily turn the D
values derived from sedimentation boundary widths into pure fiction.
Indeed, the fact that the D values derived from velocity experiments
are strongly influenced by sample heterogeneity is precisely the
reason why this is a good approach to testing for heterogeneity, and I
believe that is exactly why Houphouet's reviewer (who is undoubtedly
reading all these messages) suggested doing this.

However, with all due respect to Dr. Fujita, I cannot agree with some
of his other points.

Certainly dynamic light scattering is a useful tool, and indeed I have
used both DLS and the sedimentation boundary fitting method on the
same samples. When you have a homogeneous sample, and work at
concentrations where non-ideality is not significant, both DLS and
boundary fitting give you the same value of D within 2-3%.

What I think Dr. Fujita perhaps does not realize is that we are now
running velocity experiments at far lower concentrations than were
usually used in the 60's and 70's (e.g. I routinely run velocity at
100 micrograms/ml), and thus non-ideality effects are far less
significant than they were in the past.

On the other hand, the same often does not hold true for dynamic light
scattering.  Many DLS instruments, such as the currently popular
Protein Solutions instrument, require concentrations of several mg/ml
(or more), well into the range where non-ideality is significant.
Even with a blazing 4 watt argon-ion laser Malvern setup like I had at
Amgen, for proteins in the 20-40 kDa range I needed 1 mg/ml or more to
get good DLS data.

Furthermore, while the fact that the signal intensities in DLS are
proportional to mass gives it great sensitivity to aggregates, that
sensitivity can actually be a drawback in characterizing a sample that
contains a small amount of aggregate. For such a sample the physical
separation in the centrifuge, combined with global analysis of many
scans, can permit extraction of a D value for the native state species
with fairly good accuracy, despite the presence of some aggregate. In
the DLS the aggregate will contribute proportionally far more to the
raw data, and extraction of the D for the native state based strictly
on mathematical separation will be a very ill-conditioned problem.

Certainly if you have access to a DLS you should use it, and Dr.
Fujita is correct that it is a fairly quick measurement. Overall,
though, using DLS does not save any time since you still need to do
the velocity experiment (and the D from that comes along "free").

I would also like to point out that one problem in combining the
DLS-derived D with the sedimentation coefficient is the issue of
sample temperature accuracy. Knowing the true temperature of your DLS
samples is usually fairly straightforward; knowing the true
temperature in the centrifuge (especially as you get farther from room
temperature) is problematic. If the D value has an accuracy of 2% then
you need temperature errors of less than 0.5 degree to avoid
compromising the data (and the XL-A/I calibration is certainly no
where near that good). When both s and D come from the same
experiment, temperature accuracy is not an issue.

John Philo

-----Original Message-----
From: Hiroshi Fujita [lvl79@mbox.kyoto-inet.or.jp]">mailto:lvl79@mbox.kyoto-inet.or.jp]
Sent: Friday, April 30, 1999 7:29 PM
To: rasmb@bbri.harvard.edu
Subject: On the determination of D


Dear RASMBers:
I am quite interested in seeing the discussions going among our
members on the determination
of D from sedimentation velocity experiments. Although I know little
about recent progress in
ultracentrifugal analysis, it seems that the UC people of my
generation well recognized
how risky it is to try this determintion for macromolecular
solutions,　because boundary
spreading is so sensitive to heterogeneity and the concentration
dependence of s. Thus, since
the early 1970s, the dynamic light scattering (DLS) has replaced the
ultracentrifuge and also
the classic diffusiometer as the convenient and standard method for D
of polymer as well as
protein solutions.  I strongly suggest protein physical chemistry to
get familar with DLS
in both theory and experiment.
Even with the D data from DLS the application of the Svedberg relation
(for the calculation of
M ) to systems whose homogeneity is doubtful is risky because this
relation is valid only for
strictly binary solutions (its form for multi-component systems is
quite complex and does not
allow easy application).  Such D data give information equivalent to
the s data from　the
ultrcentrifuge more rapidly and hence more conveniently. I wish the UC
users of the younger
generation to learn the historical developments of sedimentation
analysis since the Uppsala
days.
Sincerely yours,
Hiroshi Fujita

Hiroshi Fujita
35 Shimotakedono-Cho, Shichiku, Kita-Ku
Kyoto,　Japan
Phone/Fax：075-491-2061
e-mail:lvl79@mbox.kyoto-inet.or.jp