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: jeremy.derrick@manchester.ac.uk, Dr Christopher F. van der Walle,

Dosage Form Design & Development, AstraZeneca, Aaron Klug Building, Granta Park,

Cambridge, CB21 6GH, UK Tel: +441223 898240 ; Email: wallec@medimmune.com.

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