Philo, J.S., poster, Higher Order Structure Meeting, 2013

1. Copyright © 2013 Alliance Protein Laboratories, Inc.
One of our clients observed unusual behavior during size-
exclusion chromatography (SEC) analysis of a monoclonal
antibody (mAb-X), which suggested it might exhibit
substantial reversible self-association in the SEC mobile
phase at rather low protein concentrations (below 1
mg/mL).
Some initial sedimentation velocity experiments were done
at concentrations from 0.2 to 2.0 mg/mL. The systematic
increase in the weight-average sedimentation coefficient,
sw, with concentration (see graph below) is definitive proof
of reversible association.
Does mAb‐X reversibly self‐associate?
CG-MALS combines a classical light scattering detector with
computer-controlled syringe pumps, degassers, and filters.
The pumps can mix together combinations of buffer and up
to two different protein solutions to create a stepped series
of different concentrations and/or mixing ratios, and then
supply those mixed solutions to the scattering detector, and
then to a concentration detector (RI or UV) in series.
The graph below shows a CG-MALS experiment where a
solution of BSA at ~8 mg/mL in DPBS was diluted to zero
concentration in six equal steps. We see an initial build-up
of the scattering and RI signals to a plateau level as the flow
system is saturated with the BSA stock, and then a series of
six downward stair-steps as that stock is mixed with
increasing amounts of buffer.
Introduction to CG‐MALS: Analysis of BSA
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
0.0
0.2
0.4
0.6
0.8
monomer
dimer
tetramer
speciesconcentration(mg/mL)
total protein concentration (mg/mL)
individual species concentrations from the best-fit model
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
0
2
4
6
8
10
12
14
16
best fit of mAb-X CG-MALS data
LSsignal(arb.units)
total protein concentration (mg/mL)
experiment
monomer ↔ dimer ↔ tetramer fit
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
0
50
100
150
200
250
300
350
400apparentmolarmass(kDa)
total protein concentration (mg/mL)
combined data from three 10-step dilution experiments
0 1 2 3 4 5 6 7 8
60
61
62
63
64
65
66
67
68
69
70
apparentmolarmass(kDa)
BSA concentration (mg/mL)
experimental (3 angles)
virial expansion fit
extrapolated true mass 66.1 kDa
slope gives second virial
coefficient = 7.9 x 10-5
mol·mL/g
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
0.0
0.5
1.0
0.2 mg/mL; sw
= 7.61 S
0.5 mg/mL; sw
= 7.99 S
2.0 mg/mL; sw
= 9.44 S
normalizedc(s)[perSvedberg]
sedimentation coefficient, s20,w
[Svedbergs]
SV-AUC analysis of mAb-X at 3 concentrations in SEC mobile phase
John Philo, Alliance Protein Laboratories, San Diego, CA
Reversible Association of a Monoclonal Antibody Studied by Concentration‐
Gradient Light Scattering (CG‐MALS) and Sedimentation Velocity (SV‐AUC)
As the concentration increases new peaks are resolved, and
the peak at 10.7 S seen in the sample at 2.0 mg/mL is
sedimenting at a rate close to that expected for a dimer.
However the continuous shifting of the peak positions
indicates that these peaks represent dynamic mixtures of
different oligomers rather than single species. Further, a
reversible monomer ↔ dimer system normally does not
resolve into two peaks, so the peak seen at 10.7 S actually
suggests that reversible oligomers larger than dimer are also
present.
In summary, these initial SV experiments clearly established
that substantial reversible association does occur in this
concentration range, but at that time we did not attempt to
characterize the association stoichiometry or strength.
Our new ‘Calypso’ CG-MALS system has allowed us to re-
investigate the association of mAb-X in a more detailed and
quantitative way. The graph below shows the raw data for a
10-step downward dilution from a starting concentration of
~1.5 mg/mL (with UV concentration detection at 297 nm in
this case), obtained ~10 days ago.
Clearly the scattering and concentration traces are much
more divergent than for the BSA data (and in the opposite
direction). Here we see that the initial downward LS steps
are much larger than the later ones, which means the
effective molar mass is falling as the concentration falls.
Another important observation is that the light scattering
reaches a plateau quickly after each dilution, with no
evidence of slow dissociation of reversible oligomers.
Recent mAb‐X results from CG‐MALS
The drop in scattering intensity with each drop in
concentration can then used to calculate an apparent molar
mass at that concentration, as shown in the graph below.
The slight drop in apparent mass as the concentration
increases is due to repulsive solution non-ideality
(“molecular crowding”) effects. Fitting to a straight line
gives the expected 66 kDa mass at zero concentration and
also the second virial coefficient.
Two additional similar titrations were done starting from
~0.5 and ~0.2 mg/mL, giving 30 total concentration points.
The graph below summarizes all those data as an apparent
molar mass vs. concentration plot. This plot shows that near
1.5 mg/mL the weight-average mass exceeds that of dimer,
and hence oligomers at least as large as trimer must be
present.
Attempts were made to fit the CG-MALS data to both non-
specific indefinite association (isosdesmic) models and to
specific assembly stoichiometries. A monomer ↔ dimer ↔
tetramer model gives a reasonably good fit (summarized in
the graphs below) and returns dimer → monomer and
tetramer → dimer dissociation constants of 15 μM (4.5
mg/mL) and 3.5 μM (2.1 mg/mL), respectively.
Fitting the CG‐MALS data to assembly models
Because the CG-MALS results imply there is a specific
assembly pathway, an attempt to globally fit the existing SV
data at 3 concentrations seemed worthwhile. A monomer ↔
dimer ↔ tetramer model does provide a reasonably good fit,
returning similar dissociation constants and sedimentation
coefficients that are hydrodynamically reasonable.
This assembly model works for the SV‐AUC data too
• data interpretation at high concentrations is much easier
for CG-MALS because hydrodynamic non-ideality effects are
much stronger and more complex than thermodynamic ones
• SV data depends on molecular shape as well as mass; CG-
MALS is independent of shape
• SV requires an order of magnitude less sample volume
• SV data sets are much more information-rich, but only 3-7
samples/day can be run
• fitting SV data to association models requires more fitting
parameters because the sedimentation coefficients for
different species cannot be accurately predicted
Advantages & drawbacks of CG‐MALS vs. SV‐AUC
Note that a high quality fit is not expected for the 2 mg/mL data because the
model cannot account for the large hydrodynamic solution non-ideality at that
concentration. The dissociation constants also may differ from the CG-MALS
values due to differences in temperature and buffer composition.
500
0
50
100
150
200
250
300
350
400
450
8.12
0
1
2
3
4
5
6
7
Slice
14000.000 200.0 400.0 600.0 800.0 1000 1200
Plateaus
RI
Det. 1
Det. 2
Det. 3
AnalysisSignals
1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.73
0
0.1
0.2
0.3
0.4
0.5
0.6
Slice #
45000.000 500.0 1000 1500 2000 2500 3000 3500 4000
Det. 2
Det. 3
Plateaus
UV297
AnalysisSignals