[RASMB] density matching ASB-14
Tina Daviter
t.daviter at mail.cryst.bbk.ac.uk
Wed Mar 10 15:10:24 PST 2010
Dear all,
apologies for taking so long to thank you all for incredibly helpful
replies to my post in January. I have some rather practical questions
remaining and attempted a summary on approaches to SV experiments with
detergent-solubilized proteins. I hope answers and comments might be
shorter than this post (which got far longer than intended).
It seems there are lots of approaches eliminating errors introduced by
the presence of detergents, but that apart from SE or
nanodiscs/copolymers none eliminate all errors.
Nevertheless, I would like to get a better idea how large some of these
errors can be, and how accurately people normally work, so I'd
appreciate some worst-case reports to relate to (i.e. has anyone ever
mistaken a dimer for a monomer based on these or other errors).
How large is the error introduced through omitting the detergent from
Sednterp-calculation of the viscosity? How large is the difference of
viscosity between detergent-containing and detergent-free buffers?
(Thanks, Dave, for the offer to measure!)
I have once measured the density of a buffer and protein, to calculate
vbar of the protein (glycosylated in that case), something I'd have to
do for protein-detergent complexes. Buffer density was 1.005745 g/ml
measured, but Sednterp-calculated was 1.0047 g/ml. I used the measured
value to calculate the density increment, but relied on the
Sednterp-calculated value for SV data analysis. Basically, I trusted the
relative, but not absolute values of the measurement. I did not make the
buffer myself, it seems a long way off, could it be from not accounting
for e.g. NaOH added to Tris-Cl in the calculation?
How accurate is it to use IF optics (or Abs, if absorbent) in an SV
experiment to determine the amount of bound detergent to a protein
sample of known concentration? Do I take the number of fringes in the
plateau of processed (noise subtracted) data? Or do I literally count
the number of fringes in the boundary if I can see them?
In a dynamic equilibrium of -say- monomer-dimer, how large is the error
introduced by the assumption that the ratio of bound detergent/protein
is the same for both (in not-density matched experiments)? Fitting with
a global vbar might not be appropriate.
What I would like to hear is if SV is ever going to be trustable to
examine the oligomeric state or simple equilibria, if all this is
inaccurate from the outset (worst case). Does it merely give good
starting guesses for SE and serve to see if the sample is clean?
Now here is the summary of SV with protein-detergent complexes, most is
taken from Salvay et al (JBiolPhys 2007, Ebel group) and Butler and Tate
(in Scott, Harding, Rowe book of 2005). I hope there are no big
misunderstandings.
The problem is that the amount of detergent bound to a membrane protein
is unknown. In short, the main consequences are that (1) the vbar to use
for data analysis is unknown. Any resulting MW would have errors due to
that, and also (2) due to the wrong f/f0. Any resulting (inaccurate) MW
would (3) not be interpretable anyway as it contains the contribution of
bound detergent.
Solutions for (1):
- If the density of the detergent itself is matched with that of
solvent, the Sednterp-calculated vbar for the protein can be used.
- The apparent(?) vbar (of protein-detergent complex) (or of detergent
on its own if no data exist for density matching) can be measured by SV
experiments in different h20/d20 mixing ratios by extrapolation of s to s=0.
- The apparent vbar can also be determined by measuring density
increments in a densitometer.
- The apparent vbar could be calculated through summation if the amount
of bound detergent (and/or lipid) and the density of the detergent is known.
I understand that the apparent vbar of the PD complex gives the correct
s for a protein-detergent complex, but that problem (2) persists, as the
f/f0 is for the complex and hence introduces an error to the MW
calculation. Problem (3) persists as well, i.e. MW contains that of
bound detergent.
Only solution to problem (2): do SE instead. This would also solve
inaccuracies of SV data analysis due to unknown viscosity.
Solution to problem (3):
Determine the amount of bound detergent using IF (or Abs) optics and the
known parameter for proteins (ext coeff for Abs or 1 fringe per 0.33
mg/ml for IF (universal?)) and a measured one for detergent of known
concentration (from a separate SV experiment of detergent in buffer
against a buffer reference). This information could also be used to
calculate the vbar of the complex from calculated vbar of protein and
known vbar of detergent. I assume this could be taken from data
preliminarily analysed (with approximate vbar etc).
Again, thanks a lot for your previous comments. I'd be grateful for more
thoughts/experience on any of the above points.
Best wishes
Tina
Tom Laue wrote:
> Hi-
> There is a 3x lever arm for uncertainty in vbar with respect to M, so a
> 1% error in vbar results in ~3% error in molecular weight. The vbar of a
> protein can be calculated from its composition to ~1%. The uncertainty
> increases for shorter proteins and peptides (where any unusual
> composition, high charge, etc. becomes significant) and will be greater
> for highly charged proteins or modified proteins. With your 'composite'
> molecules, I'd guess that 10% is a fairly safe uncertainty.
> Best wishes,
> Tom
>
> Timothy Dafforn wrote:
>> Whoops forgot to also mention..
>> We have also done a fair amount of AUC SV of membrane proteins in
>> detergent..
>> Most of the time this is just to check the samples contain single
>> species (for the Xtalographers!)
>> However they always ask if I can give them an approximate mass to
>> ensure that, for example, if they expect a dimer it is in fact a dimer..
>> I have to say I am always cagey, but just how inaccurate are the
>> masses likely to be?
>> +/- 100%? Or +/- 10%...
>> Tim
>>
>> -----Original Message-----
>> From: Tom Laue [mailto:Tom.Laue at unh.edu] Sent: 26 January 2010 15:52
>> To: Timothy Dafforn
>> Cc: Tina Daviter; rasmb at server1.bbri.org
>> Subject: Re: [RASMB] density matching ASB-14
>>
>> Hi-
>> To a good first approximation, yes the masses will add. The only
>> caveat is that you are assuming the partial specific volumes of the
>> components (polyer, lipid and protein) are unchanged in the complex
>> from what they are uncomplexed. This assumption usually is valid to a
>> percent or so, so it is not a big problem.
>> What is the polymer? (Can you share more info?). It is wonderful to
>> finally have ways to study membrane proteins as soluble entities.
>> Best wishes,
>> Tom
>>
>> Timothy Dafforn wrote:
>>
>>> Hi All,
>>> First post here..
>>> Just responding to Toms comments on Lipodiscs..
>>> We have an analogous system that replaces the protein that stabilizes
>>> the disc with a simple polymer.
>>> We have run a load of SV of these with various membrane proteins in
>>> them and empirically it seems that if we do a simple sum along the
>>> lines of:
>>>
>>> mass of protein in disc = mass of disc with protein - mass of disc
>>>
>>> Then we get masses that match with what we expect for the protein.
>>>
>>> My question is, is this a reasonable approach, or am I missing some
>>> technical issue.
>>> Cheers
>>> Tim (A CD expert not an AUC expert!)
>>> -----Original Message-----
>>> From: rasmb-bounces at rasmb.bbri.org
>>> [mailto:rasmb-bounces at rasmb.bbri.org] On Behalf Of Tom Laue
>>> Sent: 09 January 2010 16:13
>>> To: Tina Daviter
>>> Cc: rasmb at server1.bbri.org
>>> Subject: Re: [RASMB] density matching ASB-14
>>>
>>> Hi-
>>> The benefits to doing density matching are marginal for sedimentation
>>> velocity, whereas they can be considerable for sedimentation
>>> equilibrium. Density matching the detergent for equilibrium allows
>>> the buoyant mass of the protein to be determined without regard to
>>> the extent of detergent binding. For velocity sedimentation, however,
>>> density matching does not remove the contribution of the bound
>>> detergent to the frictional coefficient.
>>> By varying the solvent density it is possible to determine the amount
>>> of bound detergent and the protein molar mass, but it will not
>>> account for the effect on f. Be aware, too, that solvent density
>>> matching using other solvent components (e.g. salts, sugars) may
>>> result in a significant shift in the protein's vbar if there is
>>> preferential solvation, leading to inaccuracies in interpretation.
>>> Depending on what you use to match density, preferential solvation
>>> may result in substantial (5 - 10% error) in vbar. Since there is a
>>> 3-fold lever arm that results in a 3% error in M for every 1% error
>>> in vbar, there may be a 15-30% error in M if the preferential
>>> solvation is not accounted for.
>>> Judging from the detergent composition, I would anticipate the vbar
>>> for ASB-14 (my guess is ~0.75 ml/g) to be too low for solvent
>>> density matching using D2O, and that it will appear as 'extra
>>> protein' due to the nearness of its vbar to that of protein.
>>> We have done some experiments with nano-discs (essentially HDL belt
>>> protein + lipid). These are wonderful biophysical tools. The
>>> nanodiscs we used sedimented as a single boundary (~3.5 s) yielding
>>> the correct molecular weight. They were rock stable at 4 C for 5
>>> months (try that with liposomes!), did not stick to windows or
>>> centerpieces and showed no signs of aggregation upon resuspension
>>> after sedimentation. There are few proteins that are as well behaved.
>>> Furthermore, we were able to determine by sedimentation velocity that
>>> a protein integrated into the nanodiscs as a stable monomer or dimer
>>> or tetramer, with no evidence for trimer or interactions between
>>> nanodiscs. The results suggested that the protein may undergo a
>>> 1<>2<>4 reversible association, at least during the time when the
>>> nanodiscs were being synthesized. The nanodiscs could be used with
>>> interference, absorbance and fluorescence optics. Care had to be
>>> taken with absorbance data to use the correct extinction coefficient
>>> for the different protein/lipid complexes. Though we saw no evidence
>>> for it, one always needs to pay attention to the possibility of
>>> variation in the quantum yield with different assembly
>>> stoichiometries with the fluorescence optics. The 100-fold lower
>>> concentrations needed with the fluorescence optics were useful for
>>> saving the scarce protein.
>>> Best wishes,
>>> Tom Laue
>>>
>>>
>>> Tina Daviter wrote:
>>>
>>>> Dear All,
>>>>
>>>> I will have to run an SV experiment in the presence of the detergent
>>>> ASB-14 at 0.5%. It is a solid and I wondered if anyone has any
>>>> information on the density of it in solution so I can do density
>>>> matching? Or is it not that easy in this case?
>>>>
>>>> The aim of the experiment is to get the oligomeric state of a
>>>> membrane protein.
>>>>
>>>> If the density is unavailable, any suggestion on how to most easily
>>>> approach the problem would be highly welcome. I fear to determine
>>>> the density by running ASB-14 in H20/D20 mixes might fail as it does
>>>> not seem to absorb much at any wavelength (is interference sensitive
>>>> enough?).
>>>>
>>>> Also, I have not entirely understood how to account for the
>>>> influence detergents have on the viscosity. I should be able to
>>>> organise access to an Anton-Paar densitometer, but I don't think
>>>> there is a viscosimeter that I could use.
>>>>
>>>> Is SE inavoidable?
>>>>
>>>> For more information on the detergent, if that helps, here is the
>>>> full name and link to Sigma:
>>>> Amidosulfobetaine-14,3-[N,N-Dimethyl(3-myristoylaminopropyl)ammonio]propanesulfonate.
>>>>
>>>> http://www.sigmaaldrich.com/catalog/ProductDetail.do?D7=0&N5=SEARCH_CONCAT_PNO|BRAND_KEY&N4=A1346|SIGMA&N25=0&QS=ON&F=SPEC
>>>>
>>>>
>>>> Literature recommendations (preferably overviews and experimental
>>>> methods) are also welcome.
>>>>
>>>> Thank you very much in advance.
>>>>
>>>> Regards
>>>> Tina
>>>>
>>>>
>>>>
>>>
>>
>>
>
More information about the RASMB
mailing list