where fu and fb are the fractions of the
unbound and bound protein, respectively, and δu and
δb are the chemical shift values of the unbound and
bound states of the protein. The linewidths were largely free of
exchange broadening and had ligand-dependent chemical shift values
similar to those without ligands. This was indicative of ‘fast-exchange’
behavior for protein-ligand interactions. The changes in combined
(1H and 15N) chemical shift
(∆δNH) were then calculated using equation 1.
An important first step towards binding analysis was to evaluate the
CSPs for all the residues in the protein sequence. Accordingly, the
plots of the observed CSPs at the highest concentrations of the SP
Sepharose, Capto MMC and Nuvia cPrime ligands are presented in Figures
4a, b & c, respectively (note: the Fc residues 1-228 correspond to the
IgG1 residues 220-447).The secondary structural elements of the
FC sequence are also shown at the bottom of Figure 4a.
The data were analyzed using the statistical method described by
Schumann et al. (Schumann et al., 2007) to determine the corrected
standard deviation σ0. The residues with CSPs smaller
than σ0 were grouped as non-interacting (noise) and are
presented as grey bars in the Figures. Residues exhibiting large CSPs
are represented in red bars while residues with intermediate changes are
shown in a red-grey color gradient. As can be seen in Figure 4a, for
interactions of the SP Sepharose ligand to the FC,
residues that experienced large changes in CSPs were spread across the
entire protein sequence and did not group into a specific region. In
addition, residues that exhibited intermediate changes were not
localized around residues that showed large CSPs. In contrast to the
single mode CEX ligand results, both the large and intermediate CSPs
associated with the binding of MM CEX ligands were clearly clustered in
specific regions in the FC sequence (Figures 4b & c).
Interestingly, the residues experiencing these CSPs were primarily
located in clusters associated with more flexible regions of the
protein. For example, by comparing these data with the secondary
structural elements of the FC, it can be seen that the
clusters observed near residues 29-38 and 87-98 (Figures 4b & c) were
associated with either loops or a combination of loops and α-helices.
While the CSPs for the CEX and MM CEX systems were quite different, the
changes within the MM CEX systems for Capto MMC and Nuvia cPrime were
very similar. This was quantified by the Pearson correlation coefficient
which was calculated to be 0.15 ± 0.03 for the CEX and each of the MM
CEX systems and 0.85 for the Capto MMC and Nuvia cPrime systems (Figure
S1).
To further visualize these ligand-induced changes in combined chemical
shift, the CSPs for the residues were visualized on the protein surface
using the PyMol viewer. A cartoon and a surface representation of these
results for the three ligands studied are presented in Figures 5a and 6,
respectively. To facilitate the discussion, the color schemes used in
these figures are the same employed in Figure 4. As can be seen in
Figure 5a, for interactions of the SP Sepharose ligand with the
FC, residues that exhibited large and intermediate CSPs
were spread across the entire protein surface. Further, as expected for
a cation exchange ligand such as SP Sepharose, the residues that
experienced CSPs were often associated with positively charged (blue
colored) regions on the electrostatic potential (EP) map of the
FC shown in Figure 5b.
In contrast, the results for binding of the Capto MMC and Nuvia cPrime
ligands with the FC, indicated a more focused
interaction region as highlighted by the dotted blue ellipses of Figures
6a and b, respectively. Interestingly, the residues that interacted with
the MM ligands and were present in the flexible loops and alpha helices
(as discussed above for Figure 4) were located in the interface of the
CH2 and CH3 domains. Further, residues
that experienced ligand-induced CSPs were mostly identified to be
positively charged and/or aliphatic or aromatic in nature. In order to
examine this in more detail, the focused interaction region for the MM
ligands was also indicated on the EP and SAP maps presented in Figure 5b
and c. As can be seen in the figures, this region was associated with
both positive EP and some hydrophobicity which is expected for the
interactions of multimodal ligands (Karkov, Woo, Krogh, Ahmadian, &
Cramer, 2015).
Interpretation of the CSP data does come with a few caveats. For
example, residues that are in close proximity of those directly
interacting with the ligands may also undergo significant changes in
chemical shift. In addition, residues interacting with a ligand may not
experience significant changes in electronic environment and thus may
not have a measurable CSP. In order to address these issues, the changes
in combined chemical shift during the titration experiments were
examined in more detail.
A few representative spectral overlays for the binding of the Nuvia
cPrime ligand to FC are presented in Figure 7 and the
corresponding CSPs as a function of ligand concentration are shown in
Figure S2. As can be seen in the figures, different trajectories of the
CSPs were observed with increasing ligand concentration for each of the
residues. For example, residue Asp 182 (Figures 7a and S2a) did not
exhibit any significant shift with increasing ligand concentration,
indicating that it likely did not participate in binding. For residues
that did show measurable CSPs, we observed primarily linear migrations
with both saturated and unsaturated behavior. As can be seen in Figures
7b and S2b, Val 178 exhibited linear migration without saturation, which
could be due to non-specific and/or weak interactions of the ligand with
this residue. On the other hand, for His 66 (Figures 7c & S2c), a
saturating linear trajectory was observed which was indicative of a
simple two-state binding behavior that could be readily fit to the
Langmuir equation (equation 2) to determine the binding dissociation
constant (KD).
The CSP data at different ligand concentrations for all the
FC residues were fit to equation 2 using Matlab R2019b
with a maximum fitting error of 10% in a 95% confidence interval. For
interactions of the SP Sepharose ligand with the FC,
residues that showed saturation behavior exhibited high
KD values (> 150 mM) which were indicative
of very weak interactions. Further, the SP Sepharose ligand was observed
to interact throughout the FC surface without clustering
into specific regions (results not shown).
Due to solubility constraints with the Capto MMC ligand (≤ 3.2 mM at pH
5), the KD determination with the multimodal ligands was
limited to the more soluble Nuvia cPrime ligand. The CSP data at
different Nuvia cPrime ligand concentrations for all the
FC residues were fit to equation 2 to determine the
residue specific KD values which are presented in Table
S1. Based on the maximum solubility of the Nuvia cPrime ligand (60 mM at
pH 5), only KD values below 40mM were able to be
accurately determined. As can be seen in the Table, the resulting
KD values were up to two orders of magnitude smaller
than those observed with the single mode SP Sepharose ligand. These
results qualitatively agree with the marked differences observed in the
elution salt concentrations, 0.12 and 0.63 M NaCl, for
FC in SP Sepharose and Nuvia cPrime chromatographic
systems, respectively.
The FC residues that exhibited saturation behavior for
interactions with the Nuvia cPrime ligand were color coded based on
their KD values and the resulting projections on the
protein surface are presented in Figure 8. As can be seen in the Figure,
residues interacting with a “high” binding affinity (1.2 ≤
KD ≤ 10 mM) were primarily located in the hinge region
(green ellipse) and near the interface of the CH2 and
CH3 domains (blue ellipse) (note: these regions are the
same as those in Figure 6 which were based on the CSPs). The strong
interactions with a cluster of positively charged (His 66, His 5, Lys 3
and Lys 55) and polar (Thr 4, Thr 6 and Cys 7) residues in the hinge
region is indicative of this being a preferred binding region for the
ligand. However, it is important to note that interactions observed near
the flexible hinge region may be due to direct participation of the
residues in ligand binding and/or NMR shifts due to ligand binding
induced conformational changes in that region of the FC.
Further, while the hinge region was identified by NMR as a preferred
binding region for the ligand, steric affects would likely impact ligand
interactions with the FC when present in an intact mAb
and/or Fc fusion proteins.
As was observed with the CSP data (Figure 6), the KDresults also indicated that the interface of the CH2 and
CH3 domains was an important preferred binding region
for interacting with the Nuvia cPrime ligand (Figure 8). Residues
involved in binding to the Nuvia cPrime ligand in this region are
indicated in Table S1 with a #. As can be seen in the table, this
contiguous MM ligand binding region was composed of positively charged
(His 214, Arg 36, His 210 and His 91), polar (Asn 215, Gln 219, Thr 31,
Thr 218, Thr 37, Gln 92, Ser 221, Ser 35 and Gly 166) and aliphatic (Leu
32, Ile 34, Ala 159, Val 160 and Met 33) residues. In this region,
residue 214H interacted with one of the lowest KD values
(3.94 ± 2.1 mM), indicating relatively strong binding to the ligand.
Interestingly, this histidine residue has also been shown to be one of
the most important residues for binding of the FCportion of IgG1 mAbs to Protein A resins (Idusogie et al., 2000; Moiani
et al., 2009). As can be seen in Table S1, a few negatively charged
residues (e.g. Glu 211 and Asp 93) also exhibited NMR shifts which may
be due to their proximity to the clusters of positively charged and
polar residues discussed above.
As was discussed above with the CSP results, this preferred binding
region at the interface of the CH2 and
CH3 domains (indicated by the blue ellipse) corresponded
to an overlapping region of positive EP and hydrophobicity (Figures 5b
and c). The fact that the KD results presented in Figure
8 are even more focused in this region lends further support to the
importance of both types of interactions in MM systems.
Molecular Dynamics Simulations
The NMR experiments described so far provide a window into
protein-multimodal ligand interactions on a molecular level. While NMR
experiments yield molecularly-detailed data, they are time- and
resource-intensive making it difficult to quickly apply these tools to a
wide range of systems. In order to more deeply investigate the
FC-ligand interactions at the molecular level and to
explore some of the docked conformations of the Nuvia cPrime ligands, we
carried out MD simulations. As described in the Methods section,
simulations were performed with multiple ligand copies in free solution
with the protein, and throughout the simulation ligands were free to
reversibly interact from the protein surface. The free energy of a
ligand binding to a given region of the protein surface was calculated
as: