template for electronic submission to acs journals · web view]urils as excipients in...
TRANSCRIPT
Interaction of a macrocycle with an aggregation-
prone region of a monoclonal antibody
Marcello Martinez Morales,†, ‡ Matja Zalar,§ Silvia Sonzini$, Alexander P. Golovanov,§
Christopher F. van der Walle,†* Jeremy P. Derrick ‡*
†Dosage Form Design & Development, AstraZeneca, Aaron Klug Building, Granta Park,
Cambridge, CB21 6GH, UK; $Advanced Drug Delivery, AstraZeneca, Aaron Klug Building,
Granta Park, Cambridge, CB21 6GH; ‡School of Biological Sciences, Manchester Academic
Health Science Centre, The University of Manchester, Manchester, M13 9PL, UK; §Manchester
Institute of Biotechnology and School of Chemistry, The University of Manchester, Manchester,
M1 7DN, UK
*, corresponding authors: Prof Jeremy Derrick, School of Biological Sciences, Manchester
Academic Health Science Centre, The University of Manchester, Manchester, M13 9PL, UK
Tel: +44161 64207; Email: [email protected], Dr Christopher F. van der Walle,
Dosage Form Design & Development, AstraZeneca, Aaron Klug Building, Granta Park,
Cambridge, CB21 6GH, UK Tel: +441223 898240 ; Email: [email protected].
1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
ABSTRACT
Colloidal stability is among the key challenges the pharmaceutical industry faces during
production and manufacturing of protein therapeutics. Self-association and aggregation processes
can not only impair therapeutic efficacy but also induce immunogenic responses in patients.
Aggregation-prone regions (APRs) consisting of hydrophobic patches are commonly identified
as the source for colloidal instability and rational strategies to mitigate aggregation propensity
often require genetic engineering to eliminate hydrophobic amino acid residues. Here, we
investigate cucurbit[7]uril (CB[7]), a water soluble macrocycle able to form host-guest
complexes with aromatic amino acid residues, as a potential excipient to mitigate protein
aggregation propensity. Two monoclonal antibodies (mAbs), one harbouring an APR and one
lacking an APR, were first assessed for their colloidal stability (measured as the translational
diffusion coefficient) in the presence and absence of CB[7] using dynamic light scattering
(DLS). Due to the presence of a tryptophan residue within the APR, we were able to monitor
changes in intrinsic fluorescence in response to increasing concentrations of CB[7]. Isothermal
titration calorimetry (ITC) and NMR spectroscopy were then used to characterize the putative
host-guest interaction. Our results suggest a stabilizing effect of CB[7] on the aggregation-prone
mAb, due to the specific interaction of CB[7] with aromatic amino acid residues located within
the APR. This provides a starting point for exploring CB[7] as a candidate excipient for the
formulation of aggregation prone mAbs.
Keywords: cucurbit[7]uril, binding affinity, host-guest chemistry, formulation, excipients
2
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
3
40
INTRODUCTION
An understanding of the self-association and aggregation processes during the formulation of a
therapeutic protein is required in the context of Chemistry, Manufacturing, and Controls (CMC),
because they may otherwise impair the shelf-life and efficacy of the drug product1. The
aggregation process in monoclonal antibodies (mAbs) is commonly driven by protein-protein
interactions of hydrophobic patches, denoted aggregation prone regions (APRs). APRs can
stabilize the native fold through intramolecular interactions (structural APRs) but can also be the
cause of unwanted intermolecular interactions when surface-exposed (critical APRs)2. The latter
are reported to be rich in aromatic amino acid content and can be found primarily on variable
chains in mAb complementarity-determining regions (CDRs)3,4.
One strategy to mitigate aggregation propensity involves re-engineering APRs in order to
eliminate aggregation-driving residues5,6. Aromatic amino acid residues in CDRs, play an
important role in antigen recognition, however; mutation thereof requires a trade-off between
decreasing binding affinity to the target epitope and reducing aggregation7,8. Another strategy
involves using pharmaceutical excipients to weaken the protein-protein interactions responsible
for self-association. There are a variety of excipients commonly used in the formulation of
protein drugs including different sugars, amino acids, surfactants, preservatives and
antioxidants9. Agitation-induced aggregation, for example, can be inhibited by addition of
surfactants such as polysorbates or cyclodextrin (CD)-based derivatives which compete for space
at the air-liquid interface where proteins can form partially unfolded intermediates that are prone
to aggregate10,11. In a different approach, in silico screening can be used to identify the interaction
potential between existing excipients and APRs in a model protein: Kale and Akamanchi have
4
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
described how attachment of a fatty acid moiety led to the generation of a multipurpose excipient
which combines surfactant properties with the ability to disrupt APR-mediated self-association12.
Cucurbit[n]urils (CB[n]s) are a family of macrocycles consisting of glycouril units connected
via methylene bridges which, similar to CDs, display a hydrophobic cavity able to harbour guest
molecules (Fig. 1A). In contrast to CDs, however, they exhibit electronegative portals which
enable additional electrostatic interactions at the cavity rims13. Cucurbit[7]uril (CB[7]), the seven
membered macrocycle, is the most water-soluble among the CB[n]s series and is known to form
host-guest complexes with aromatic amino acid residues (Fig. 1B)14. Measurements of binding
constants of CB[7] recorded optimal affinity for Trp (Kd=10-3M), compared with Tyr and Phe
(Kd=10-5M and 10-6 respectively)15,16. The best studied guest residues located on biological
molecules are N-terminal Phe residues, whose binding affinity is potentiated by the positively
charged amino group which is able to form ion-dipole interactions with the electronegative
carbonyl-lined rim of CB[7] (Fig. 1C)17,18. While CB[7] is currently not regulated for use in
pharmaceutical formulations, a review addressing the toxicity and safety of CB[7] concludes
relative safety of CB[7] with no significant short-term effects as observed in mouse models.
However, long-term side effects are yet to be investigated in future studies19.
Here we report an investigation into the interaction of CB[7] with PPI-12, an aggregation
prone mAb containing an APR formed by Trp30 and Phe31, located in the variable heavy chain
of CDR1 (VHCDR1), and Leu56, located in the variable heavy chain of CDR2 (VHCDR2), in
order to understand the molecular basis for its function as an excipient for this mAb. Previous
work showed that a variant of PPI-12, PPI-12m, is more physically stable following substitution
of Ser, Thr and Thr for Trp30, Phe31 and Leu56, respectively (Fig. 2)6. To date, relatively few
studies have been conducted into complex formation of CB[7] with interchain amino acid
5
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
residues in peptides and proteins18,20–23. Our results provide evidence that CB[7] improves the
physical stability of PPI-12 by binding to residues located in the APR and suggest that other
mAbs could be stabilised in a similar fashion, by disrupting protein-protein interactions
responsible for self-association.
MATERIALS AND METHODS
Materials and sample preparation
PPI-12 and PPI-12m were provided by AstraZeneca PLC., Cambridge, UK. PPI-12 was
provided as a 17.4 mg/mL solution in 25 mM histidine, 240 mM sucrose at pH 6.0, whereas PPI-
12m was provided as a 21.1 mg/mL solution in 20 mM histidine, 240 mM sucrose at pH 6.0.
Sodium phosphate monobasic and sodium phosphate dibasic were purchased from J.T. Baker
(Avantor Performance Materials B.V., Arnhem, Netherlands). Cucurbit[7]uril was purchased
from Strem Chemicals UK (Cambridge, UK). All aqueous solutions were prepared with
deionized water (18.2 MΩ, Milli-Q system, Merck-Millipore, Watford, UK). A 10 mM sodium
phosphate buffer at pH 7.4 was prepared by combining 0.275% NaH2PO4·H2O with 2.147%
Na2HPO4·7H2O and adding 1000 mL Milli-Q deionized water. Potassium iodide (KI) for
fluorescence quenching experiments was purchased from Sigma-Aldrich (St Louis, MO, USA).
For NMR experiments, a 10 mM sodium phosphate buffer at pH 6.0 was prepared using 1.189%
Na2HPO4·7H2O with 0.375% NaH2PO4·H2O and 1000 mL Milli-Q deionized water. A 5 mM
CB[7] stock solution was prepared in each of the sodium phosphate buffers. Exchanging buffers
of both mAbs was achieved by exhaustive dialysis (three buffer exchanges, each under gentle
stirring for ~8 hours following the manufacturer’s instructions) using Slide-A-Lyzer dialysis
cassettes (10 kDa MWCO, Thermo Fisher Scientific, UK). Protein concentrations after dialysis
6
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
were measured using a NanoDrop (Thermo Fisher Scientific, UK) UV/vis spectrophotometer
(extinction coefficients used were 1.61 and 1.54 mg/mL-1cm-1 for PPI-12 and PPI-12m,
respectively) and diluted to the desired sample concentration for measurements. Samples for
ITC, fluorescence spectroscopy and DLS were filtered using a 0.22 µm Millex polyethersulfone
membrane filter (Merck-Millipore, Watford, UK) prior to measurement.
Dynamic Light Scattering (DLS)
Translational mutual diffusion coefficients (DtDLS) and hydrodynamic radii for both mAbs were
determined using a DynaPro Plate Reader II (Wyatt Technology UK Ltd, Haverhill, UK). For
each mAb, three replicates with increasing concentrations of CB[7] (0, 0.5, 1, 1.5 and 2 mM) at 7
concentrations (1, 2, 4, 6, 8, 10 and 12 mg/mL) were prepared. 30 μL of each sample were
loaded in triplicates into a Corning#4681 384-well tissue culture treated cyclic olefin co-polymer
plate (Merck Sigma-Aldrich, Poole, UK). All plates were centrifuged prior to measurement to
remove air bubbles introduced by sample loading for 2 min at 2000 rpm using a Heraeus
Megafuge 11R centrifuge (Thermo Fisher Scientific, UK). 10 measurements with an acquisition
time of 5 s were taken for each well at 25°C and a laser wavelength of 820.170 nm. Parameters
reported here represent the average of three replicates.
Fluorescence Spectroscopy
Changes in intrinsic fluorescence in response to increasing concentrations of CB[7] were
monitored using a Hitachi F-7000 fluorescence spectrophotometer (Hitachi High-Technologies,
Maidenhead, UK). Increasing aliquots from a CB[7] stock solution were added to antibody
samples of 10 mg/mL. Three 100 μL replicates of each sample were loaded into a Corning#3993
7
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
96-well half area polystyrene plate (Merck Sigma-Aldrich, Poole, UK). Samples were excited at
292 nm (selective for Trp) and emission was recorded between 300 and 400 nm with an
excitation and emission slit width of 5 nm and a scan speed of 1200 nm/min. Plotted data
reported here represent the average of three replicates. Fluorescence quenching experiments were
conducted for PPI-12 and PPI-12m in presence and absence of 2 mM CB[7] using potassium
iodide as a quencher. Stern-Volmer constants (K SV) were calculated using the Stern-Volmer
relation:
I 0
I=1+K SV × ¿
¿
with I 0 representing fluorescence intensity in the absence of iodide,I representing fluorescence
intensity at ¿ and ¿ representing the iodide concentration. Stern-Volmer plots were obtained
using linear regression analysis implemented in GraphPad Prism 8.0 (GraphPad Software, Inc.,
San Diego, CA, USA).
Isothermal Titration Calorimetry (ITC)
Calorimetric titrations were performed using a Microcal Auto-ITC200 (Malvern Panalytical Ltd.,
Malvern, UK). PPI-12 was concentrated following dialysis using an Amicon Ultra-4 centrifugal
filter device (10 kDa MWCO; Merck-Millipore, Watford, UK). CB[7] was dissolved in the bulk
buffer solution of the last dialysis to eliminate buffer mismatch. Antibody samples loaded into
the measurement cell were prepared at a concentration of 83.5 μM, while the titrant of CB[7]
was prepared to a concentration of 2 mM. Before each titration, the measurement cell was pre-
rinsed with buffer. After an initial injection of 0.4 μL, 19 injections of 2 μL each were
sequentially titrated into the sample solution spaced by 90 s intervals between each injection.
8
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
Titration was performed at 25°C and 750 rpm stirring speed. Obtained thermograms were
processed and fitted with Origin Data Analysis 7.21 (Malvern Panalytical Ltd, Malvern, UK).
Integrated heat resulting from the first injection of 0.4 μL was excluded from analysis. Integrated
heat values from a titration of CB[7] into buffer were subtracted from experiments using point by
point subtraction.
Nuclear Magnetic Resonance Spectroscopy (NMR)
All NMR experiments were recorded at 25 °C on a Bruker 800 MHz Avance III spectrometer
equipped with a 5 mm triple resonance TCI cryoprobe and temperature control unit (Bruker UK
Ltd, Coventry, UK). All spectra were acquired and processed using Topspin 3.5 and further
analysed using Dynamics Center 2.5.1 (Bruker UK Ltd, Coventry, UK). All antibody samples
were prepared at 10 mg/mL in a 10 mM sodium phosphate buffer, pH 6.0 as described above.
1H-NMR spectra were recorded using standard Bruker zgesgp pulse program. Translational
diffusion coefficients (DtNMR) were measured by Bruker’s standard pulse program stebpgp1s19
with additional water signal presaturation during the relaxation delay. Diffusion time (Δ) and
gradient length (δ) were optimised for CB[7] in the absence of mAbs according to a standard
Bruker protocol and set to 49.9 and 4 ms, respectively. The diffusion spectra were recorded with
16 scans and 32 linear gradient strengths steps using 2-98% of maximal gradient intensity (48
G/cm). DtNMR values were derived by fitting the signal decay as a function of gradient strength to
the Stejskal-Tanner equation24 in Dynamics Center 2.5.1. The errors in DtNMR were estimated
using Monte Carlo simulations as implemented in Dynamics Center 2.5.124. For STD NMR
titration studies a standard Bruker’s pulse sequence stddiffesgp.3 was used with 25 ms TOCSY
spin-lock filter to eliminate protein signals and 2 s of protein signal saturation. On- and off-
9
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
resonance irradiation frequencies were set to -0.34 and 30 ppm, respectively. STD amplification
factors (AFSTD) were calculated as:
AFSTD=I off−I on
I offx [ L][P ]
(2 )
where I off and I on denote signal intensity in the spectra with off- and on- resonance saturation,
respectively, while [ L ] and [ P ] represent concentrations of CB[7] and mAbs, respectively.
Fraction of signal loss parameter Fobs/exp was calculated as:
Fobs/exp=I obs
Imaxx [CB [7 ]¿¿max ]
[CB [7 ]]¿ (3 )
where I obs denotes measured CB[7] signal intensity in the spectra at a chosen concentration
[CB[7]] and I max is the signal intensity of CB[7] at the highest concentration of [CB[7]max] used
during titration experiments where CB[7] is overwhelmingly in the free state. Reduced values of
this Fobs/exp parameter below 1 reflect CB[7] signal loss due to broadening and interaction with
mAbs.
RESULTS
CB[7] reduces concentration-dependent self-association of PPI-12
Dynamic light scattering measurements were performed to monitor the mutual diffusion
coefficient (DtDLS) of PPI-12 and PPI-12m in the absence of CB[7] and at increasing
concentrations of CB[7] (Fig. 3). This plate-based experiment was chosen to quickly examine the
impact of CB[7] on colloidal stability. The value of DtDLSfor PPI-12 decreases steadily as
antibody concentration is increased: its initial value at 1 mg/mL is reduced threefold when
reaching 12 mg/mL. Similarly, DtDLSdecreases in value with increasing antibody concentrations
10
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
upon addition of CB[7]. However, the decrease is less pronounced and reaches more than double
the value of PPI-12 without CB[7] at 12 mg/mL (Fig. 3A). By contrast, DtDLSfor PPI-12m in the
presence of up to 2 mM CB[7] does not show significant differences from the values obtained
without CB[7] (Fig. 3B). The dilute conditions of the experiment allow correlation of DtDLSand
the hydrodynamic radius (rh) through the Stokes-Einstein equation:
DtDLS= kT
6 πηr h
(4 )
with krepresenting Boltzmann’s constant, T the absolute temperature in Kelvin and η the
solution viscosity. Calculated apparent hydrodynamic radii range from 7.2 to 20.4 nm (1 mg/mL
and 12 mg/mL, respectively) for PPI-12, indicating the presence of dimer and oligomer species,
consistent with an earlier report of PPI-12 oligomer formation at concentrations above 1
mg/mL6. Addition of up to 2 mM CB[7] reduces the apparent hydrodynamic radii of PPI-12 to
6.8 and 9.6 nm (1 mg/mL and 12 mg/mL, respectively), consistent with a lower level of dimer
and oligomer species in the presence of CB[7] (Fig. S1A).
Values for the diffusion interaction parameter (k D) for PPI-12 and PPI-12m were derived from
linear regression of DtDLS versus antibody concentration and are listed in Table 1. All k D values
are negative indicating attractive protein-protein interactions. Addition of CB[7] results in an
increase of k D values for PPI-12 from -70.9 mL/mg in absence of CB[7] to -24.8 mL/mg at 2
mM CB[7]. Addition of CB[7] to PPI-12 shows a small decrease from -19.4 mL/mg in absence
of CB[7] to -14.9 mL/mg in presence of 2 mM CB[7]. Plots of Dc/Do versus mAb concentration
for PPI-12 and PPI-12m are consistent with the change in k D versus CB[7] concentration (Fig.
S1 C&D).
11
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
Overall, the data obtained from the DLS measurements show a concentration-dependent
increase of the apparent hydrodynamic radius of PPI-12 in the presence and absence of CB[7].
However, addition of CB[7] results in significant reduction of the apparent hydrodynamic radius
at 12 mg/mL, compared to PPI-12 without CB[7]. In contrast, hydrodynamic radii obtained for
PPI-12m in the presence and absence of CB[7] did not differ significantly (Fig. S1B). Calculated
k D values indicate attractive protein-protein interactions that decrease as a function of CB[7]
concentration for PPI-12 while k D values for PPI-12m are only slightly affected. These results
suggest that the interaction between CB[7] and the APR reduces aggregation of PPI-12.
Study of CB[7] binding using intrinsic Trp fluorescence
We used intrinsic Trp fluorescence, which is highly sensitive to local changes in environment25,
as a complementary technique to analyse comparative changes in PPI-12 and PPI-12m in
response to CB[7]. Figures S2 A&B show the emission spectra (λex=292 nm) of PPI-12 and PPI-
12m with increasing concentrations of CB[7]. The PPI-12 Trp emission spectrum peak decreases
with increasing concentrations of CB[7] (Fig. S2C, 13% reduction at 2mM). A much smaller
decrease is observed for PPI-12m (Fig. S2D; 5% reduction at 2mM). The observed loss of
fluorescence intensity could be interpreted as increased solvent exposure of Trp residues in
response to CB[7] (there are 10 in the heavy chain and 5 in the light chain of PPI-12). Complex
formation of CB[7] with N-terminal Phe residues is known to be accompanied by local unfolding
of adjacent peptide secondary structure in some circumstances17. However, there seems to be
little overall change in the intrinsic fluorescence of PPI-12m with CB[7] concentration (Fig
S2D), suggesting that CB[7]-mediated global unfolding of PPI-12m or PPI-12 is unlikely. This is
confirmed by the fact that 2mM CB[7] has no significant effect on the antigen binding affinity of
12
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
either mAb (see below and Fig. S5). To investigate further, fluorescence quenching experiments
were conducted to monitor changes in K SV (the Stern-Volmer constant) for both mAbs in the
absence and presence of CB[7] (Fig. 4 and Table 2)26,27. PPI-12 had a higher KSV in the presence
of CB[7], indicative of higher Trp solvent accessibility in the presence of the excipient. By
contrast, KSV values for PPI-12m did not vary significantly when CB[7] was added. These
observations are consistent with a model where, for PPI-12, inhibition of protein-protein
interactions by CB[7]-Phe31 complex formation results in an increase in solvent exposure for the
adjacent Trp30 through a reduction in oligomer formation. Consistent with this proposition,
chemical cross-linking and docking simulations have revealed that the APR forms part of the
binding surface6. This effect is absent for PPI-12m, due to a combination of its weaker affinity
for CB[7] and lower propensity for oligomerization.
Analysis of binding of CB[7] to PPI-12 and PPI-12m by isothermal titration calorimetry
We sought to quantify the binding affinity of CB[7] for both mAbs. Thermograms obtained by
ITC yield different binding isotherms for the titration of CB[7] with PPI-12 and PPI-12m (Fig.
S3 A&B). In both cases, a change in heat is observed upon titration with CB[7] which decreases
over the course of the titration suggesting interaction of CB[7] with both PPI-12 and PPI-12m.
Nevertheless, the integrated heat signals do not reach the baseline equilibrium during the
experiments and limit the ability to unequivocally determine the exact number of all possible
binding sites. Evaluating PPI-12 first and assuming that CB[7] interacts with both APRs present
(one on each VH chain), thermograms were fitted with a sequential binding model with N=3
apparent binding sites (the sequential binding model takes independent as well as dependent
binding sites into account). The first two apparent binding sites are assumed to consist of a
13
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
specific interaction on each APR and the last apparent binding site is assumed to be caused by
non-specific interactions present in both PPI-12 and PPI-12m. Consequently, the thermograms
obtained for titration of PPI-12m with CB[7] were fitted with a one set of sites binding model,
fixing the stoichiometry to 1. Apparent binding affinities (Ka), which in this case reflect
interactions at several binding sites and changes in enthalpy (ΔH) extracted for PPI-12 and PPI-
12m, are listed in Table 3. The presence of several low energy non-specific interactions in both
mAbs together with the solubility limit of CB[7] prevent saturation of the binding process used
in the experiments presented here; whilst using lower PPI-12 concentrations resulted in very low
heat signals that were not suited for any of the fitting models available.
In summary, our results suggest CB[7] interacted with both mAbs: a change in heat could be
detected for both titration experiments. For the first two apparent binding sites the binding
affinities of CB[7] with PPI-12 are similar and suggest a higher apparent binding affinity
compared to the third apparent binding site which appears to be present in both variants, PPI-12
and PPI-12m, with a similar binding affinity (Table 3). Although, the limitations of ITC to detect
very weak interactions coupled with limitations in sample and CB[7] solubility impeded full
characterization of the number and affinity of all possible binding sites, the results suggest that
the PPI-12m mutant has lost at least two apparent binding sites, compared with PPI-12.
NMR Spectroscopic analysis of the bound state of CB[7]
NMR offers a powerful tool to examine the molecular interaction between CB[7] and PPI-12 and
provides additional information on binding and diffusion in solution. Earlier studies using NMR
spectroscopy to investigate complex formation of CB[7] monitored chemical shift changes in the
guest molecule, rather than CB[7] 28,29. This requires 15N-labelled material which subsequently
14
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
enables determination of the binding site and calculation of Kd values 30. As 15N-labelled protein
samples were unavailable for PPI-12, we have utilised ligand-observed NMR methods to
characterize interactions between CB[7] and both mAbs. In the first instance, we recorded the
1H-NMR spectra of CB[7] and confirmed assignment of the resonances as previously reported
for this macrocycle (Fig. S4)31,32. 1H-NMR spectra were then collected for both mAb mixtures
over a range of CB[7]:mAb concentration ratios (Fig. 5 A&B). In the absence of interaction
between ligand and mAb, we would expect a linear increase in CB[7] signals intensities as
concentration of CB[7] increases, meaning that the value of the signal loss parameter Fobs/exp will
be close to 1. However, deviation from linear behaviour due to signal broadening indicative of
interaction between CB[7] and mAbs has been observed during titration series (Fig. 5C). In case
of PPI-12 the Fobs/exp remains low, well below 1, until all binding sites for CB[7] on the mAb are
occupied. This behaviour is much less pronounced for PPI-12m (Fig. 5D) suggesting that there
are less binding sites available for binding. Moreover, chemical shift perturbations of H2 and H3
CB[7] signals relative to the free CB[7] state have been observed for PPI-12 but not for PPI-12m
(Fig. 5 E&F). This suggests that the H2 and H3 nuclei are sensitive to interactions with mAbs,
and also that PPI-12 contains more binding sites for CB[7] than PPI-12m. A comparison with the
equivalent experiment carried out on PPI-12m shows a much weaker effect for the mutant; we
therefore infer that PPI-12m has less binding sites for CB[7], presumably due to the triple
mutation.
DOSY experiments were conducted to determine the translational diffusion coefficient (DtNMR)
of CB[7] for both mAbs at different CB[7]:mAb ratios (Fig. 6A, B). Upon binding to a
macromolecule, translational movement of CB[7] throughout the solution is expected to be
slowed down which is reflected in a lower DtNMR
33. The Dt
NMR values of CB[7] are significantly
15
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
decreased in the presence of PPI-12 when CB[7] is in moderate excess, 4:1 or 10:1
(CB[7]:mAb), but at higher 20:1 ratio, the freely diffusing CB[7] becomes overwhelmingly
dominant species (Fig. 6C). The DtNMR of CB[7] in presence of PPI-12m decrease only slightly at
4:1 and 10:1 excess, compared with free CB[7] (Fig. 6C). The slowing down of molecular
diffusion of CB[7] in the presence of nearly-equimolar PPI-12, but not PPI-12m, can be
explained by the interaction of CB[7] with PPI-12 but not PPI-12m at a number of binding sites.
Saturation transfer difference NMR Spectroscopy (STD-NMR) was performed at increasing
molar ratios of CB[7]:mAb. Saturation difference amplification factors (AFSTD) were calculated
and plotted as a function of CB[7]:mAb ratio (Fig 7). The data shows that significantly stronger
STD effect is achieved for PPI-12 than for PPI-12m at lower CB[7]:mAb ratios, suggesting more
interaction for CB[7] with PPI-12, with the level of saturation transfer for PPI-12 reaching a
maximum at 10-fold excess of CB[7], presumably near the optimal point where the number of
available binding sites is equal to the number of ligand molecules (Fig. 7A). Once CB[7]
becomes strongly in excess (at 20:1 ratio), the efficiency of saturation transfer drops. The ratio-
dependent increase in AFSTD in presence of PPI-12m is much less pronounced (Fig. 7B), likely
due to less binding sites on PPI-12m available for CB[7] interactions. These results suggest that
PPI-12 has more binding sites available for CB[7] than PPI-12m.
In summary, the NMR experiments provide further evidence that PPI-12 has more binding
sites for CB[7] than PPI-12m which translates into apparent tighter binding. It is likely that the
mutated residues in the APR region are responsible for these differences and therefore form part
of CB[7] binding sites.
DISCUSSION
16
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
A common approach taken to stabilize mAbs relies on high-throughput screening techniques
covering various solution conditions to identify a suitable formulation. A less empirically driven
approach can be taken if the underlying interactions responsible for self-association are identified
and subsequently targeted to reduce self-association. Methods to identify aggregation-driving
residues in proteins are continuously improving and can help to offer a rationale for their specific
targeting using excipients. While CB[7] complexation of N-terminal amino acid residues, to Phe
in particular, has been extensively studied, binding of CB[7] to proteins lacking a N-terminal Phe
residue has only been reported on sparsely17,34.
In the context of protein drug formulation, CB[7] has been shown to inhibit fibrillation of
insulin through binding to N-terminal and interchain Phe residues21. Intermolecular interactions
in the fibrillation process comprise π-stacking interactions and result in fibrils with β-sheet
dominant secondary structures35. Despite yielding structurally more complex and less ordered
aggregates, the self-association processes in proteins can also be driven by hydrophobic
interactions comparable to those observed during fibrillation36. The role of hydrophobic
interactions for self-association processes is highlighted in a report where addition of aromatic
amino acids Phe and Trp, albeit not fully preventing oligomerization, reduce the degree of
oligomer formation, an effect ascribed to preferential interaction of hydrophobic patches with
each of the added amino acids37. Similar to the study involving cyclodextrins mentioned above,
inhibition of agitation- induced aggregation of a mAb was successfully achieved through CB[7]-
mediated, non-covalent PEGylation at the N-terminus38. However, CB[7] did not have a direct
effect on the stabilization of the mAb but was, rather, utilized to attach PEG to the mAb.
Results obtained by DLS indicate a stabilizing effect of CB[7] on the aggregation-prone mAb,
PPI-12. Although the DLS data cannot directly identify the precise binding sites of CB[7], the
17
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
fact that the excipient has small effects on DtDLS for PPI-12m suggests that one or more of the
residues constituting the APR (Trp30, Phe31 and Leu56) may be the CB[7] binding sites.
Disruption of self-association of PPI-12 is accompanied by exposure of Trp30 within the APR,
consistent with our measurements by intrinsic fluorescence spectroscopy and quenching
experiments. Similar observations of a CB[7]-mediated decrease of intrinsic fluorescence have
been observed for short hydrophobic peptide chains with interchain Phe residues20.
ITC provides evidence that PPI-12 contains more apparent binding sites than PPI-12m, which
can be inferred as originating from the APRs present in each VH and provides some
quantification of binding affinity. Both Ka1 and Ka2 are in good agreement with binding affinities
reported for CB[7]-Phe complexation16,17, taking into account the fact that Phe31 is located
within an APR and presumably less accessible than would be the case at the N-terminus. The
third apparent binding site shows much weaker apparent affinity towards CB[7] and is present in
both PPI-12 and PPI-12m. The low aqueous solubility of CB[7] prevented the use of higher
concentrations. Additionally, NMR data supports the existence of more apparent binding sites
within PPI-12 which seem to have been lost in PPI-12m, providing further evidence of a specific
interaction between CB[7] and the residues within APR located on PPI-12. NMR signals
originating from CB[n] protons generally remain unaffected by formation of inclusion
complexes but are reported to change upon formation of exclusion complexes39. Therefore,
observed changes in chemical shifts for protons H2 and H3 for PPI-12 suggest formation of
exclusion complexes, likely originating from interaction of CB[7] with the increased available
surface of PPI-12 upon disruption of oligomer formation. While the binding affinities from ITC
experiments suggest the existence of Phe-CB[7] inclusion complexes, the NMR chemical shift
18
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
changes suggest the additional presence of exclusion complexes. Notably, the presence of CB[7]
did not affect the binding affinity of PPI-12 to its therapeutic target (Fig. S5).
Although mutagenesis from PPI-12 to PPI-12m might suggest an effective strategy to improve
colloidal stability, mutations can impair antigen binding affinity and effector functions, and may
also set the development of a candidate protein drug back to its engineering stage, which is
usually not an option considered at the formulation stage in industry. The use of excipients in
formulation therefore needs to be fully explored.
To the best of our knowledge, the direct use of CB[7] in the context of protein formulation has
not been investigated to date. Our study provides the first evidence of a stabilizing effect by
CB[7] through directly binding to aggregation-driving residues. The results provide evidence for
a specific interaction of CB[7] with aromatic residues located on the APR of a mAb, and
concomitant improvement in colloidal stability.
FIGURES
Figure 1. Structure and function of CB[7]. A) Chemical structure and 3D representation of
CB[7] with specific hydrogen atoms labelled and color coded (see Results sections). B)
Schematic showing complex formation between Phe and CB[7], highlighting ion-dipole
19
379
380
381
382
383
384
385
386
387
388
389
390
391392
393
394
395
396
397
interactions. C) Schematic showing complex formation between N-terminal Phe residue on a
protein and CB[7] highlighting the ion-dipole interaction at the N-terminus
Figure 2. Details of the APRs from homology models for PPI-12 and PPI-12m mAbs. A)
VHCDR1 of PPI-12. B) VHCDR1 of PPI-12m. Corresponding residues from the three mutations
between the two mAbs are highlighted and labelled.
20
398
399
400
401
402
403
404
405
406
407
408
409
410
411
Figure 3. Effect of CB[7] concentration on the translational diffusion coefficients (DtDLS) of both
mAbs measured by Dynamic Light Scattering. (A) PPI-12. (B) PPI-12m. Data points are the
mean of three individually prepared replicates.
21
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
Figure 4. Stern-Volmer Plots obtained from quenching experiments with iodide for PPI-12 and
PPI-12m (at 10 mg/mL) in absence and presence of 2 mM CB[7]. R-squared of the linear
regression analysis are shown in parentheses.
22
429
430
431
432
433
434
435
436
437
438
439
440
Figure 5. Differences in interaction of CB[7] with antibodies PPI-12 and PPI-12m studied by 1D
1H NMR spectroscopy. Overlay of characteristic regions of NMR spectra of CB[7] as a function
of CB[7]:mAb ratio for (A) PPI-12 and (B) PPI-12m. Samples were prepared in a 10 mM Na-
phosphate buffer pH 6.0 and experiments were performed at 25 °C The spectrum scaling was
normalized by the concentration of CB[7] present in the samples to highlight relative peak
broadening and shifts at different ratios. The individual CB[7]:mAb ratios and the color key are
23
441
442
443
444
445
446
447
depicted below spectra. The plots of observed CB[7] peak intensities presented as a fraction of
their expected value are shown for mixtures with (C) PPI-12 and (D) PPI-12m. The values of
CB[7] signal shifts upon changing CB[7]:mAb ratios are shown for (E) PPI-12 and (F) PPI-12m.
24
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
Figure 6. NMR-derived translational diffusion coefficients (DtNMR) of CB[7] upon addition of
mAbs. Overlay of 2D DOSY plots of free CB[7] (grey) and in the presence of (A) PPI-12 and
(B) PPI-12m at different molar ratios. (C) AveragedDtNMR values for CB[7] in the presence of
25
465
466
467
468
PPI-12 and PPI-12m at different ratios. Horizontal grey dashed line represents the DtNMR value of
free CB[7].
26
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
Figure 7. Saturation transfer difference amplification factors for PPI-12 (A) and PPI-12m (B) at
different CB[7]:mAb ratios.
27
494
495
496
497
498
499
500
501
Tables
Table 1 (D 0DLS)* and (k D)** of PPI-12 and PPI-12m at different CB[7] concentrations
PPI-12 PPI-12m
CB[7] conc. D0DLS( x 10-7) [cm2/s]
k D [mL/mg]
D0DLS( x 10-7) [cm2/s]
k D [mL/mg]
0 mM 2.75 - 70.9 (± 2.9) 4.18 - 19.4 (± 1.3)0.5 mM 2.76 - 71.1 (± 1.3) 4.15 - 19.2 (± 0.8)1 mM 2.83 - 59.2 (± 3.1) 4.25 - 21.4 (± 1.5)
1.5 mM 3.07 -35.8 (± 1.6) 4.21 - 17.9 (± 0.9)2 mM 3.04 - 24.8 (± 2.4) 4.15 - 14.9 (± 0.7)
*D0DLS : extrapolated mutual diffusion coefficients at 0 mg/mL mAb concentration; **k D :
diffusion interaction parameter
28
502
503
504
505
506
Table 2 Stern-Volmer constants (K SV ¿ obtained from quenching intrinsic Trp fluorescence with iodide
KSV (M-1)
PPI-12 1.51 (± 0.02)PPI-12 +
CB[7] 2.60 (± 0.03)
PPI-12m 1.21 (± 0.07)PPI-12m +
CB[7] 1.22 (± 0.03)
29
507
508
509
Table 3 Stoichiometry and thermodynamic parameters obtained from isothermal
titration calorimetry for PPI-12 and PPI-12m with CB[7]
Ka1 [M-1] ΔH1
[cal/mol] Ka2 [M-1] ΔH2
[cal/mol] Ka3 [M-1] ΔH3
[cal/mol]
PPI-12 · CB[7]*
1.6 ± 0.4 x 105
-3.9 ± 0.7 x 103
1.7 ± 0.5 x 104
-6.5 ± 0.8 x 103
1505 ± 96
-1.2 ± 0.3 x 104
PPI-12m · CB[7]* - - - - 2304 ±
122-2.3 ± 0.8
x 104
* PPI-12 binding was fitted with a sequential binding model fixing N=3; PPI-12m binding was fitted with a one set of sites model within Origin 7.0 (detailed description in Results)
30
510
511
512
Supporting Information
Figure S1, hydrodynamic radii (A, B) and plot of of Dc/Do (C, D) as a function of mAb
concentration and increasing concentrations of CB[7] as calculated from DLS experiments.
Figure S2, Intrinsic fluorescence emission spectra of PPI-12 (A), PPI-12m (B) and changes of
maximum fluorescence intensity of PPI-12 (C) and PPI-12m (D).
Figure S3, thermograms and binding isotherms obtained for PPI-12 and PPI-12m from ITC
experiments.
Figure S4, 1H-NMR spectra of CB[7] with corresponding assignment of CB[7] nuclei (inset).
Figure S5, Graph showing competitive binding of PPI-12 and PPI-12m in presence and absence
of 2 mM CB[7] to its therapeutic target NGF.
Table S1, IC50 Values obtained for PPI-12/PPI-12 and non-binding isotype control mAbs with
NGF in absence and presence of CB[7]
Author Contributions
The manuscript was written with contributions from all authors. All authors have given approval
to the final version of the manuscript.
ACKNOWLEDGMENT
We wish to thank Andreas Tosstorff from Ludwig-Maximilans-University Munich for providing
homology models for PPI-12 and PPI-12m. We would also like to thank Lisa Vinall and
Elizabeth England from AstraZeneca for providing guidance and materials to conduct the epitope
competition assay. This work is part of a project that has received funding from the European
31
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie
grant agreement No.675074.
REFERENCES
1. Manning, M. C., Chou, D. K., Murphy, B. M., Payne, R. W. & Katayama, D. S. Stability
of Protein Pharmaceuticals: An Update. Pharm. Res. 27, 544–575 (2010).
2. van der Kant, R. et al. Prediction and Reduction of the Aggregation of Monoclonal
Antibodies. J. Mol. Biol. 429, 1244–1261 (2017).
3. Wang, X., Das, T. K., Singh, S. K. & Kumar, S. Potential aggregation prone regions in
biotherapeutics: A survey of commercial monoclonal antibodies. MAbs 1, 254–267
(2009).
4. Wang, X., Singh, S. K. & Kumar, S. Potential aggregation-prone regions in
complementarity-determining regions of antibodies and their contribution towards antigen
recognition: A computational analysis. Pharm. Res. 27, 1512–1529 (2010).
5. Chennamsetty, N., Voynov, V., Kayser, V., Helk, B. & Trout, B. L. Design of therapeutic
proteins with enhanced stability. Proc. Natl. Acad. Sci. 106, 11937–11942 (2009).
6. Dobson, C. L. et al. Engineering the surface properties of a human monoclonal antibody
prevents self-association and rapid clearance in vivo. Sci. Rep. 6, 38644 (2016).
7. Peng, H.-P., Lee, K. H., Jian, J.-W. & Yang, A.-S. Origins of specificity and affinity in
antibody–protein interactions. Proc. Natl. Acad. Sci. 111, E2656–E2665 (2014).
32
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
8. Julian, M. C., Li, L., Garde, S., Wilen, R. & Tessier, P. M. Efficient affinity maturation of
antibody variable domains requires co-selection of compensatory mutations to maintain
thermodynamic stability. Sci. Rep. 7, 45259 (2017).
9. Kamerzell, T. J., Esfandiary, R., Joshi, S. B., Middaugh, C. R. & Volkin, D. B. Protein–
excipient interactions: Mechanisms and biophysical characterization applied to protein
formulation development. Adv. Drug Deliv. Rev. 63, 1118–1159 (2011).
10. Bam, N. B. et al. Tween protects recombinant human growth hormone against agitation-
induced damage via hydrophobic interactions. J. Pharm. Sci. 87, 1554–1559 (1998).
11. Serno, T., Carpenter, J. F., Randolph, T. W. & Winter, G. Inhibition of Agitation‐Induced
Aggregation of an IgG‐Antibody by Hydroxypropyl‐β‐Cyclodextrin. J. Pharm. Sci. 99,
1193–1206 (2010).
12. Kale, S. S. & Akamanchi, K. G. Trehalose Monooleate: A Potential Antiaggregation
Agent for Stabilization of Proteins. Mol. Pharm. 13, 4082–4093 (2016).
13. Lagona, J., Mukhopadhyay, P., Chakrabarti, S. & Isaacs, L. The Cucurbit[n]uril Family.
Angew. Chemie Int. Ed. 44, 4844–4870 (2005).
14. Lee, J. W., Samal, S., Selvapalam, N., Kim, H.-J. & Kim, K. Cucurbituril Homologues
and Derivatives: New Opportunities in Supramolecular Chemistry. Acc. Chem. Res. 36,
621–630 (2003).
15. Liu, S. et al. The Cucurbit[ n ]uril Family: Prime Components for Self-Sorting Systems. J.
Am. Chem. Soc. 127, 15959–15967 (2005).
16. Lee, J. W., Lee, H. H. L., Ko, Y. H., Kim, K. & Kim, H. I. Deciphering the Specific High-
33
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
Affinity Binding of Cucurbit[7]uril to Amino Acids in Water. J. Phys. Chem. B 119,
4628–4636 (2015).
17. Chinai, J. M. et al. Molecular Recognition of Insulin by a Synthetic Receptor. J. Am.
Chem. Soc 133, 8810–8813 (2011).
18. Li, W., Bockus, A. T., Vinciguerra, B., Isaacs, L. D. & Urbach, A. R. Predictive
Recognition of Native Proteins by Cucurbit[7]uril in a Complex Mixture. Chem.
Commun. 52, 8537–8540 (2016).
19. Wheate, N. J. & Limantoro, C. Cucurbit[ n ]urils as excipients in pharmaceutical dosage
forms. Supramol. Chem. 28, 849–856 (2016).
20. Sonzini, S. et al. Supramolecular dimerisation of middle-chain Phe pentapeptides via
CB[8] host–guest homoternary complex formation. Chem. Commun. 49, 8779 (2013).
21. Lee, H. H. et al. Supramolecular Inhibition of Amyloid Fibrillation by Cucurbit[7]uril.
Angew. Chemie Int. Ed. 53, 7461–7465 (2014).
22. de Almeida, N. E. C. et al. Opposing Effects of Cucurbit[7]uril and 1,2,3,4,6-Penta- O -
galloyl-β- d -glucopyranose on Amyloid β 25–35 Assembly. ACS Chem. Neurosci. 7,
218–226 (2016).
23. Guagnini, F. et al. Cucurbit[7]uril-Dimethyllysine Recognition in a Model Protein.
Angew. Chemie Int. Ed. 57, 7126–7130 (2018).
24. Sinnaeve, D. The Stejskal-Tanner equation generalized for any gradient shape-an
overview of most pulse sequences measuring free diffusion. Concepts Magn. Reson. Part
A 40A, 39–65 (2012).
34
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
25. Lakowicz, J. R. Protein Fluorescence. in Principles of Fluorescence Spectroscopy 529–
575 (Springer US, 2006). doi:10.1007/978-0-387-46312-4_16
26. Lakowicz, J. R. Quenching of Fluorescence. in Principles of Fluorescence Spectroscopy
277–330 (Springer US, 2006). doi:10.1007/978-0-387-46312-4_8
27. Lehrer, S. Solute perturbation of protein fluorescence. Quenching of the tryptophyl
fluorescence of model compounds and of lysozyme by iodide ion. Biochemistry 10, 3254–
3263 (1971).
28. Assaf, K. I., Alnajjar, M. A. & Nau, W. M. Supramolecular assemblies through host-guest
complexation between cucurbiturils and an amphiphilic guest molecule. Chem. Commun.
54, 1734–1737 (2018).
29. Reany, O., Li, A., Yefet, M., Gilson, M. K. & Keinan, E. Attractive Interactions between
Heteroallenes and the Cucurbituril Portal. J. Am. Chem. Soc. 139, 8138–8145 (2017).
30. Williamson, M. P. Using chemical shift perturbation to characterise ligand binding. Prog.
Nucl. Magn. Reson. Spectrosc. 73, 1–16 (2013).
31. Feng, K., Wu, L.-Z., Zhang, L.-P. & Tung, C.-H. Cucurbit[7]uril-included neutral
intramolecular charge-transfer ferrocene derivatives. Dalt. Trans. 3991 (2007).
doi:10.1039/b709221k
32. Yang, H., Tan, Y. & Wang, Y. Fabrication and properties of cucurbit[6]uril induced
thermo-responsive supramolecular hydrogels. Soft Matter 5, 3511 (2009).
33. Altieri, A. S., Hinton, D. P. & Byrd, R. A. Association of Biomolecular Systems via
Pulsed Field Gradient NMR Self-Diffusion Measurements. J. Am. Chem. Soc. 117, 7566–
35
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
7567 (1995).
34. Logsdon, L. A. & Urbach, A. R. Sequence-Specific Inhibition of a Nonspecific Protease.
J. Am. Chem. Soc. 135, 11414–11416 (2013).
35. Toyama, B. H. & Weissman, J. S. Amyloid structure: conformational diversity and
consequences. Annu. Rev. Biochem. 80, 557–585 (2011).
36. Holm, N. K. et al. Aggregation and fibrillation of bovine serum albumin. Biochim.
Biophys. Acta - Proteins Proteomics 1774, 1128–1138 (2007).
37. Esfandiary, R., Parupudi, A., Casas-Finet, J., Gadre, D. & Sathish, H. Mechanism of
Reversible Self-Association of a Monoclonal Antibody: Role of Electrostatic and
Hydrophobic Interactions. J. Pharm. Sci. 104, 577–586 (2015).
38. Webber, M. J. et al. Supramolecular PEGylation of biopharmaceuticals. Proc. Natl. Acad.
Sci. 113, 14189–14194 (2016).
39. Wang, W. et al. The chaotropic effect as an orthogonal assembly motif for multi-
responsive dodecaborate-cucurbituril supramolecular networks. Chem. Commun. 54,
2098–2101 (2018).
TOC
36
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633