Non-clinical Safety Evaluation of Biotherapeutics – Challenges, Opportunities and new Insights

Non-clinical Safety Evaluation of Biotherapeutics – Challenges, Opportunities and new Insights

Accepted Manuscript Non-clinical safety evaluation of biotherapeutics – Challenges, opportunities and new insights Guenter Blaich, Andreas Baumann, Sv...

2MB Sizes 0 Downloads 8 Views

Accepted Manuscript Non-clinical safety evaluation of biotherapeutics – Challenges, opportunities and new insights Guenter Blaich, Andreas Baumann, Sven Kronenberg, Lolke de Haan, Peter Ulrich, Wolfgang F. Richter, Jay Tibbitts, Simon Chivers, Edit Tarcsa, Robert Caldwell, Flavio Crameri PII:

S0273-2300(16)30242-2

DOI:

10.1016/j.yrtph.2016.08.012

Reference:

YRTPH 3654

To appear in:

Regulatory Toxicology and Pharmacology

Received Date: 20 August 2016 Accepted Date: 25 August 2016

Please cite this article as: Blaich, G., Baumann, A., Kronenberg, S., de Haan, L., Ulrich, P., Richter, W.F., Tibbitts, J., Chivers, S., Tarcsa, E., Caldwell, R., Crameri, F., Non-clinical safety evaluation of biotherapeutics – Challenges, opportunities and new insights, Regulatory Toxicology and Pharmacology (2016), doi: 10.1016/j.yrtph.2016.08.012. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Workshop report

2

Non-Clinical Safety Evaluation of Biotherapeutics – Challenges, Opportunities and New

3

Insights

4

Guenter Blaich a, Andreas Baumann b, Sven Kronenberg c, Lolke de Haan d, Peter Ulrich e, Wolfgang

5

F. Richter c, Jay Tibbitts f, Simon Chivers g, Edit Tarcsa h, Robert Caldwell i, Flavio Crameri c

6

a

7

c

8

Sciences, Roche Innovation Center Basel, Switzerland; dMedImmune, Cambridge,

9

UK; eNovartis Pharma, Basel, Switzerland; fUCB Celltech, Slough, UK; gADC

10

Therapeutics, SA; hAbbVie BioResearch, Worcester, USA.; iAbbVie Inc., North

11

Chicago, USA.

RI PT

1

AbbVie GmbH, Ludwigshafen, Germany; bBayer Pharma AG, Berlin, Germany;

M AN U

SC

Roche Pharmaceutical Research and Early Development, Pharmaceutical

12 Abstract

14

New challenges and opportunities in nonclinical safety testing of biotherapeutics were presented and

15

discussed at the 5th European BioSafe Annual General Membership meeting in November 2015 in

16

Ludwigshafen. This article summarizes the presentations and discussions from both the main and the

17

breakout sessions.

18

The following topics were covered in six main sessions:

AC C

EP

TE D

13

19

(i)

20

(ii)

Unexpected side effects of biotherapeutics

21

(iii)

Safety testing of cell and gene therapies including vector safety and integration site

22 23

analysis and setting up a GLP facility in this field (iv)

24 25

Challenges around use of PEGylated biologics, results of BioSafe survey

Immunogenicity and PKPD including immunogenicity prediction or methodologies to prevent induction of anti-drug antibodies

(v)

Current approaches applied to antibody drug conjugate (ADC) development Page 1 of 46

ACCEPTED MANUSCRIPT 26

(vi)

27

Approaches for non-clinical safety testing of human-selective biologics, including use of transgenic animals and MABEL

The following questions were discussed across 4 breakout sessions (i-iv) and a case-study based

29

general discussion (v):

30

(i)

RI PT

28

Do bi-specifics offer nonclinical and clinical development advantages over combinations

31

? (ii)

Is the paradigm “SC is more immunogenic than IV” correct ” ?

33

(iii)

Why do we need a 6-month toxicology study for a monoclonal antibody (mAb) when we

35

already have a 13-week toxicology study ? (iv)

36 37

How do we investigate immune complex formation pre-clinically and does it translate

M AN U

34

clinically ? (v)

SC

32

Adversity in nonclinical safety studies: STP and ESTP opinion, your opinion needed ?

38

Key words: Biotherapeutics, PEGylated biologics, gene and cell therapy, pharmacokinetics,

40

nonclinical safety, antibody-drug conjugate, immunogenicity, cross-reactive species.

41

Introduction

42

BioSafe is the Preclinical Safety expert group of the Biotechnology Industry Organization (BIO), with

43

a mission to identify and respond to key scientific and regulatory issues and challenges related to the

44

preclinical safety evaluation of biopharmaceuticals. In addition to the Annual BioSafe General

45

Membership Meeting in the US, the 5th Annual BioSafe European General Membership meeting was

46

hosted by AbbVie on November 4-5, 2015 in Ludwigshafen, Germany. Participants included 125

47

scientists from the biopharmaceutical industry, small biotechnology companies and contract research

48

organisations primarily from Europe and the UK, but also from the US, and representing expertise in

49

pharmacology, toxicology, pathology, pharmacokinetics and bioanalytics. The focus of the meeting

50

was to share experiences and insights into nonclinical safety assessment of biologics including

AC C

EP

TE D

39

Page 2 of 46

ACCEPTED MANUSCRIPT monoclonal antibodies, recombinant proteins, and gene and cell therapies. The meeting covered

52

various nonclinical safety topics including PEGylated biologics, unexpected side effects, gene and

53

cell therapy, immunogenicity and PKPD, antibody drug conjugates (ADCs), and nonclinical safety

54

strategies for human-selective biologics. In each session, case examples were presented followed by

55

podium discussions. For the first time, a breakout session was introduced for discussion of four “hot

56

topics” in smaller groups, with the feedback being presented to all attendees afterwards by the

57

breakout session leads. Finally, a group discussion based on several case studies was stimulated

58

around a fifth “hot topic”, that of adversity in nonclinical safety studies and determination of the

59

NOAEL.

60

Session 1: Challenges in developing PEGylated Biologics

61

Andreas Baumann (Bayer) and Jenny Sims (Integrated Biologix) co-chaired the first main session

62

which focused on the nonclinical development of PEGylated biologics. Conjugation of Polyethylene

63

glycol (PEG) moieties to various types of biologics has become a standard method to improve their

64

pharmacokinetic properties, specifically prolonging plasma half-life (Jevsevar et al, 2010; Nakaoka et

65

al, 1997; Yang et al, 2004). PEG molecules themselves have been shown to be non toxic, non

66

immunogenic, hydrophilic, uncharged and non degradable polymers (Webster et al, 2009). For many

67

protein-conjugates, the primary mechanism to increase plasma half-life is through the reduction in the

68

rate of clearance by the kidneys (Mehvar, 2000; Yamaoka et al, 1995; Baumann et al, 2014; Webster

69

et al, 2009). Other mechanisms include shielding epitopes/surfaces of the target drug resulting in

70

lower visibility to the immune system, and protecting the protein from proteolysis.

71

comprehensive review of current experiences in non-clinical safety and pharmacokinetics of

72

PEGylated biotherapeutics, see references Ivens et al, 2015 and Baumann et al, 2014. Twelve

73

PEGylated biologics have received regulatory approval so far based on balanced risk/benefit

74

evaluations. Cellular vacuolation has been seen in toxicology studies in about half of these products,

75

however without any functional changes for organs and tissues (Rudmann et al, 2013; Ivens et al,

76

2015, Turecek et al, 2016).

AC C

EP

TE D

M AN U

SC

RI PT

51

Page 3 of 46

For a

ACCEPTED MANUSCRIPT Jennifer Sims (Integrated Biologix) presented results on a pharmaceutical industry survey on

78

nonclinical safety testing of PEGylated proteins, which was conducted by BioSafe in 2013 and

79

published recently in more detail (Ivens et al, 2015). Ten companies provided information on 17

80

PEGylated biopharmaceuticals including receptor or protein targeting products with molecular

81

weights ranging from 25 kDa to 350 kDa and currently in various stages of nonclinical and clinical

82

development. The molecular weight of the linear or branched PEG moieties ranged from 20 kDa to

83

60 kDa. PEG-related tissue vacuolation was seen in toxicology studies for 10 of the 17 compounds in

84

the survey. No other-PEG related effects were reported. For 5 of the 17 products there was some

85

evidence for an impact on disposition related to receptor-mediated uptake of the PEGylated protein.

86

For the remaining 12 products it is not known whether there was receptor-mediated uptake, or

87

whether it was not evaluated. For 7 of 17 products no specific studies were conducted to assess the

88

disposition of the PEG moiety, while information on pharmacokinetics and/or disposition was

89

gathered from 10 products. For 11 products there no specific assays for measurement of anti-PEG

90

antibodies were developed or employed in nonclinical studies.

91

antibodies were detected in a single dose clinical study and others stated that clinical assays were

92

available or planned.

93

developed for 6 compounds, but either no positive responses were observed or there was little impact

94

on the interpretation of the study. One company stated that the development of IgM and IgG anti-

95

PEG antibodies was observed in mice and macaques, but this was not considered to have any impact

96

on safety considerations in humans.

97

Toxicity studies included in the survey were repeated IV or SC dose studies in mice, rats, dogs, and

98

cynomolgus monkeys and ranged from 2 to 52-weeks duration. The dosing frequency varied from

99

daily to weekly. For one product both the IV and intravitreal route of administration was used in

100

toxicology studies. No cellular vacuolation was observed for 7 of the 17 compounds. For these

101

compounds, the dose levels used were relatively low and duration of dosing was generally short (up to

102

4-5 weeks). For 2 compounds studies of 26-39 weeks in duration were conducted. None of the 4

103

products with a single 20 kDa conjugated PEG showed evidence of vacuolation. In general, the

M AN U

SC

RI PT

77

TE D

For 1 compound no anti-PEG

AC C

EP

Assays for detection of anti-PEG antibodies in nonclinical studies were

Page 4 of 46

ACCEPTED MANUSCRIPT maximum study duration for these 4 products was only 5 weeks, although for one product there was a

105

26-week study in cynomolgus monkeys. However, cellular vacuolation was observed in some of the

106

4-week studies with 30 to 40 kDa PEG products.

107

One of the survey questions related to whether or not any special methodology was used to evaluate

108

potential adverse effects associated with PEG accumulation and/or vacuolation (e.g. IHC, electron

109

microscopy, functional evaluations such as nerve conduction velocity measurements, or in vitro

110

studies). No special investigations other than IHC were conducted for 16 of 17 compounds.

111

In the next presentation, Hanne Kjær Offenberg and Inga Bjørnsdottir (Novo Nordisk) presented

112

on the nonclinical safety and DMPK of N9-GP and N8-GP, 40 kDa pegylated human recombinant

113

coagulation factors IX- and VIII, respectively. Nonclinical safety testing of new therapeutic drugs

114

typically includes chronic toxicity testing of up to 26 or 52 weeks duration to support chronic dosing

115

in the clinic. However, human coagulations factors are highly immunogenic in animals resulting in an

116

anti-drug antibody response which reduces exposure considerably. Therefore, an alternative animal

117

model involving immuno-deficient athymic rats (Rowett Nude), which are devoid of T-cells and do

118

not mount T cell-dependent antibody responses, was used in 26 week studies for both compounds. In

119

general both N8-GP and N9-GP were very well tolerated in these studies. The choroid plexus

120

epithelial cells (CPE) have been described to develop vacuoles after dosing with PEGylated biologics

121

(Ivens et al. 2015). Therefore, the choroid plexus was evaluated for presence for PEG. Using IHC

122

staining, no PEG was detected in the CPE cells for N8-GP, whereas for N9-GP, PEG was detected in

123

the CPE cells at all dose levels. PEG did not induce vacuole formation in any tissue in any of the

124

studies. In the absence of any degenerative or inflammatory findings, presence of PEG in the CPE

125

cells is not considered adverse. The No-Observed-Adverse-Effect Level for N9-GP or N8-GP in

126

these studies was considered to be 1200 IU/kg every fifth day or 1200 IU/kg every fourth day,

127

respectively. The immune-deficient rat model proved suitable for a chronic study of 26 weeks

128

duration and could be considered in cases where a model is needed for the chronic testing of human

129

proteins which are highly immunogenic in animals (Offenberg et al, 2015).

AC C

EP

TE D

M AN U

SC

RI PT

104

Page 5 of 46

ACCEPTED MANUSCRIPT 130

The biologic fate of the [3H]PEG-moiety incorporated into N8-GP and N9-GP was evaluated in a

131

single dose IV study in rats.

132

investigated to assess if pharmacokinetics was dose-dependent (dose range: 0.6 mg/kg and up to 200

133

mg/kg). For all compounds, plasma pharmacokinetics, distribution and excretion pathways were

134

investigated, based on total radioactivity measurements. The plasma concentration of the intact N8-

135

GP or N9-GP conjugate was also measured using an antibody-based analysis method. After single IV

136

administration to rats, [³H]N8-GP, [³H]N9-GP as well as [³H]PEG were shown to be widely

137

distributed, mainly in highly vascularized tissues, with the lowest levels of radioactivity found in the

138

CNS. The protein part of N9-GP/N8-GP was degraded over time, with the PEG moiety circulating in

139

plasma. The PEG was subsequently excreted as 40 kDa PEG and smaller sized PEGs. Biphasic

140

elimination of the PEG moiety was observed from plasma with more than 50% of PEG excreted into

141

urine and faeces within the 1st phase (t½, 1st phase ~2-3 days in rats, t½, 2nd phase ~ 13-18 days in

142

rats). The [³H]PEG and metabolites were eliminated mainly via the kidney into urine but also via the

143

liver into feces, with a larger fraction found in the feces for [³H]N8-GP and [³H]N9-GP compared to

144

[³H]PEG. Elimination of the 40 kDa PEG-moiety was shown to be dose-dependent with faster

145

elimination at lower dose levels.

146

Finally, Andreas Baumann (Bayer) presented on the nonclinical safety and PK of a site-specific

147

PEGylated-Factor VIII conjugate (BAY 94-9027). BAY 94-9027, a B-domain–deleted recombinant

148

FVIII with one 60 kDa polyethylene glycol (PEG; 2 × 30 kDa branched) attached via a linker to

149

cysteine 1804, is in clinical development for acute and prophylactic treatment of patients with

150

hemophilia A. In vivo toxicology studies with BAY 94-9027 or PEG-60 alone involving single and

151

repeated administration were conducted. Specifically, in a repeat-dose toxicity study, male Sprague

152

Dawley rats were given PEG-60 IV at dose levels of up to 11.0 mg/kg every other day for 4 weeks

153

which was in the range of the total expected cumulative life-time dose of PEG-60 in humans given

154

BAY 94-9027. No adverse effects or histopathological changes were observed in these studies (Ivens

155

et al, 2013). Whilst there is no clinical evidence to suggest that administration of PEGylated proteins

156

is associated with long-term safety issues, questions have been raised regarding the pharmacokinetics

AC C

EP

TE D

M AN U

SC

RI PT

In addition, the biologic fate of a 40 kDa [³H]PEG-moiety was

Page 6 of 46

ACCEPTED MANUSCRIPT 157

(PK) including metabolism and distribution, in particular for PEGylated proteins used chronically

158

and/or in the pediatric population.

159

distribution of residual radioactivity (full length PEG-60 and radioactive metabolites) on Days 7–168

160

after single IV administration of [14C] BAY 1025662, the cysteine-linker-PEG 60 kDa component of

161

the molecule containing a covalently linked [14C] label in the linker, in male Wistar rats.

162

Concentrations and amounts of radioactivity in organs and tissues were measured using whole body

163

autoradiography 6 months after administration. Between Days 7–168 post-injection, radioactive

164

residues in the body decreased steadily from 22.5% of the dose to 1.8%. A heterogeneous distribution

165

of radioactive residues was observed, with apparent preferential distribution of radioactivity to

166

endocrine and exocrine glands and lymphatic organs. No residual radioactivity was detected in brain

167

and adipose tissue. Elimination was fairly complete by the end of the study, with a total recovery of

168

radioactivity of 94% at Day 168. Terminal elimination was slow, but there was no indication of PEG

169

retention in rats.

170

Session 2: Unexpected side effects with biotherapeutics

171

In the second main session, chaired by Sven Kronenberg (Roche) and Lolke de Haan

172

(MedImmune), case examples describing the utility of nonclinical safety studies to detect unexpected

173

side effects of biologics were presented.

174

In the first presentation, Lolke de Haan, Simon Henderson and Mary McFarlane (All

175

MedImmune) presented a retrospective portfolio-wide analysis of the utility of early, non-GLP, dose-

176

range finding toxicity studies in the development of biologics. The analysis encompassed 53 drug

177

candidates generated in support of the MedImmune drug portfolio between 2008-2014, with

178

modalities including monoclonal antibodies, bispecifics, fusion proteins, antibody-drug conjugates,

179

therapeutic proteins and peptides. Overall, the analysis showed that with 7/37 drug candidates that

180

were tested in non-GLP toxicity studies, toxicity was observed, which for 4/37 lead to termination of

181

further drug candidate development. Other factors influenced drug candidate attrition to a greater

182

extent, with undesirable PK/PD properties, lack of disease rationale and/or portfolio fit being the

AC C

EP

TE D

M AN U

SC

RI PT

The aim of the presented PK study was to investigate the

Page 7 of 46

ACCEPTED MANUSCRIPT reasons for attrition for 10/37 candidates. Of the 27 drug candidates that were taken forward in to

184

Good Laboratory Practice (GLP) toxicity studies, with 5 candidates new toxicities were identified,

185

with 4/5 of these toxicities becoming apparent after subchronic dosing, and development of 2/5

186

candidates being halted because of these toxicities. Importantly, when toxicity was observed, it was

187

consistent with the mode of action or pharmacology associated with the drug candidate. Furthermore,

188

it appeared that the impact of conducting non-GLP dose-range finding toxicity studies on candidate

189

drug progression was limited, with the majority of the observed toxicities in non-GLP studies either

190

being acceptable based on safety margins, the intended clinical indication or the anticipated duration

191

or frequency of clinical dosing. In addition, given that additional toxicities observed in GLP studies

192

typically only became apparent in studies involving subchronic dosing, non-GLP dose range finding

193

studies would be expected to have limited impact on progression the progression of these candidates

194

to GLP toxicity studies. Based on this analysis, MedImmune has currently ceased to conduct non-

195

GLP dose-range finding toxicity studies as a default, and conducts such studies only on a case-by-case

196

basis, when there is a clear cause for concern based on modality (e.g. antibody-drug conjugates, which

197

are inherently toxic) or based on an extensive review of the target biology and the likelihood of target

198

modulation resulting in exaggerated pharmacology and toxicity in a naive animal model.

199

In the second presentation, Sven Kronenberg (Roche) provided case examples of unexpected side

200

effects with biotherapeutics that eventually contributed to the program termination. The first case

201

example summarized findings from a single dose PK study in cynomolgus monkeys with a chimeric

202

IgG4 mAb with abolished effector function targeting a receptor of the tetraspanin protein family. The

203

antibody target is highly expressed by hepatocytes, endothelial and epithelial cells in various tissues

204

and on hematopoietic cells. There were several key challenges with respect to the disposition and

205

safety of the molecule: The ubiquitous receptor expression led to target-mediated drug disposition at

206

lower doses. The huge antigen sink resulted in an unexpected and more than dose-proportional

207

increase in exposure, i.e. a 3-fold increase of dose led to a 100-fold increase in drug concentration,

208

making nonclinical safety testing and potential clinical studies challenging. Furthermore, safety-wise,

209

two monkeys in the single-dose PK study became moribund within 2 hrs resulting in euthanasia of

AC C

EP

TE D

M AN U

SC

RI PT

183

Page 8 of 46

ACCEPTED MANUSCRIPT these animals; the monkeys showed symptoms of anaphylactoid shock and vascular leakage including

211

edema and release of cytokines (IL-6, IL-2, MCP-1). There were no compelling in vitro alerts (on

212

mast cell degranulation, vascular permeability, complement activation) except for cytokine release

213

from whole blood. Cytokine release probably contributed to the fatalities, however, based on cell-

214

based assays, no agonistic activity of the mAb could be demonstrated. Rather, the risk for cytokine

215

release could be mitigated by modifications outside of the antigen-binding site of the mAb. In

216

conclusion, while the exact mechanism of the fatalities could not be elucidated, modification of the Fc

217

part reduced the risk of cytokine release but did not solve the steep “all or nothing” PK. In a 2nd case

218

example, results from nonclinical safety testing with one of several pegylated peptides tested at Roche

219

were presented. The peptide consisted of a 30 kDa polyethylene glycol (PEG) moiety and was dosed

220

IV or SC to rats and cynomolgus monkeys. Accumulation of the PEG moiety in neurons of the

221

hypothalamus and spinal cord was observed in 4-week rat and monkey toxicity studies. While these

222

findings were also observed in the rat at the end of a 12-week recovery phase, no clinical signs were

223

associated with the vacuolation. Neuronal uptake of PEG into neurons may have occurred via a

224

specific transporter in the brain. Since neurons have only slow or no renewal, a 1-year toxicity study

225

with weekly IV dosing was conducted in the monkey to investigate the potential for these effects to

226

result in impairment of neuronal function. At the end of this study, PEG was present in all dose

227

groups in all animals, including those of the 12-week recovery groups.

228

vacuolation was evident in several organs and tissues, including the neurons of CNS and spinal cord.

229

A reversible deficit in sural sensory and peroneal motor conduction velocities in animals at the high

230

dose was noted in week 24 only, but not at the other time points. This was inconsistent with a

231

progressive, degenerative peripheral neuropathy. Whether this finding at the high-dose at a single time

232

point was spurious, related to accumulation of PEG, or a potential consequence of target engagement

233

in the nervous system remains unclear. There were no electrophysiological effects seen in the CNS,

234

nor were any neurobehavioral changes observed.

235

Finally, Andreas Popp (AbbVie) reported on side effects of a mAb leading to the identification of a

236

new clinical indication. Humanized mAbs ABT-207 and h5F9-AM8 directed against the Repulsive

Presence of PEG and

AC C

EP

TE D

M AN U

SC

RI PT

210

Page 9 of 46

ACCEPTED MANUSCRIPT Guidance Molecule a (RGMa) and with cross reactivity to hemojuvelin (RGMc) were described. In

238

healthy animals (mice, rats, cynomolgus monkeys) both mAbs increased serum iron and decreased

239

serum iron binding capacity. Histopathologically, iron was released from spleen macrophages

240

(reduced iron in the spleen) and accumulates in periportal hepatocytes. The observed effects were

241

dose-dependent and correlated negatively with hepcidin expression. Administration of a 20 mg/kg

242

single dose of the more potent h5F9-AM8 antibody to female rats resulted in significantly reduced

243

serum hepcidin levels compared to control, which lasted 6 weeks. In toxicology studies performed in

244

rats and cynomolgus monkeys for up to 13 weeks with ABT-207, no adverse effects were observed, in

245

particular no degeneration or inflammation due to the iron accumulation in the liver. The hepcidin-

246

related effects on iron homeostasis showed complete or partial recovery after a 12 week treatment free

247

period.

248

In addition, ABT-207 and h5F9-AM8 were tested in animal models of chronically and genetically

249

induced anemia and chronic inflammation. ABT-207 and h5F9-AM8 corrected anemia and shifted

250

macrophage population in this inflammation model significantly from M1 to M2 by lowering hepcidin

251

expression in the liver and serum hepcidin levels. In conclusion, targeting RGMc as a new concept to

252

reduce hepcidin expression could be of high value for the causal treatment of conditions of chronic

253

anemia.

254

Session 3: Gene and Cell therapy

255

The third main session of the meeting was chaired by Peter Ulrich (Novartis). First a short update

256

on hot topics in the field of Advanced Therapy Medicinal Products was provided. This introduction

257

was followed by two presentations on vector safety and integration site analysis by Manfred Schmidt

258

and about the set-up of a GLP facility in the Gene Therapy field by Patrizia Cristofori.

259

Ulrich summarized regulatory information from the meeting of the European Society for Gene and

260

Cell Therapy (ESGCT) in Helsinki, September 2015. Paula Salmikangas, chair of the EMA

261

Committee for Advanced Therapies (CAT), presented the CAT view on several ATMP topics in a

262

regulatory session. She especially focused on the ongoing public consultation of Regulation (EC) No

AC C

EP

TE D

M AN U

SC

RI PT

237

Page 10 of 46

ACCEPTED MANUSCRIPT 1394/2007 “Good Manufacturing Practice for Advanced Therapy Medicinal Products”. A summary

264

was published in Dec 2015 and can be downloaded from the webpage of the European Commission

265

(European Commission>DG Health and Food Safety>Public health>Medicinal products for human

266

use>Advanced Therapies>Developments). Another topic was the new clinical trial regulation (EC)

267

536/2014, which asks for GLP compliance of pivotal non-clinical safety studies supporting first-in-

268

man (FiM) clinical trials. The regulation would also include ATMP non-clinical studies, which raised

269

many questions about the availability and the validation status of animal models in the heterogeneous

270

ATMP area. Salmikangas pointed out that for ATMP a GLP-compliant pivotal non-clinical safety

271

study may not be mandatory for FiM trials, however for Marketing Authorization Application (MAA)

272

such a study may be required. For ATMPs without the possibility to utilize validated animal models in

273

a GLP environment, the CAT would recommend to follow the instructions provided by the risk-based

274

approach guideline (EMA/CAT/CPWP/686637/2011). In December 2015 the CAT position on the

275

Application of GLP principles to ATMPs was adopted via a written procedure, which was forwarded

276

to the national health authorities (see CAT meeting minutes 12/2015 on www.ema.europe.eu). With

277

respect to hospital exemption (article 28 in EC 1394/2007) Salmikangas pointed out that the CAT

278

favors revision of the respective regulation by the European Commission in order to provide more

279

security for ATMP developing firms.

280

The recent developments in integration site analysis of gene therapy vectors and genome editing

281

approaches were also discussed (see also presentation by Manfred Schmidt below). The identification

282

of off-target integration sites by genome editing methods like CRISPR or TALENs was subject of

283

publications in 2015 as well as discussion in the ESGCT meeting (Helsinki 2015). All in all, the

284

conclusion from these recent investigations and method validations suggest that each engineered

285

nuclease has to be individually evaluated for their potential off-target activity and the conditions

286

influencing this potential (Gabriel et al., 2015). Nevertheless, the same authors see a major

287

methodological advancement of sensitive, genome-wide detection of nuclease activity.

288

Another

289

immunocompromised mouse model, where NSG mice were genetically engineered with human

AC C

EP

TE D

M AN U

SC

RI PT

263

highlight from the ESGCT

2015

meeting was the presentation

Page 11 of 46

of a

new

ACCEPTED MANUSCRIPT transgenes coding for IL-3, GM-CSF and SCF. The resulting mouse strain, also called NSG-SGM3

291

(JAX Datasheet – 013062), shows less xeno-induced graft-versus-host-disease (GvHD), when

292

reconstituted with human haemopoietic stem cells (HSC). In addition, these mice, when reconstituted

293

with HSC, develop human monocytes. Norelli et al. (2015) reported that mice reconstituted with

294

human HSC and with CD44v6+ Acute Lymphoblastic Leukemia (ALL) semi-cell line, developed

295

features consistent with a cytokine release syndrome after infusion with human autologous T cells

296

genetically modified with a CD28-endocostimulated CD44v6 CAR. Leukemia clearance associated

297

with transient malaise, high fevers and weight loss, appeared with CAR deletion of monocytes.

298

Increased serum levels of IL-6 and IFNγ in relation to clinical observations were measured with

299

monocytes as the main source of IL-6. The presented data suggest that this model may be suitable for

300

assessing CRS potential of CARTs in a non-clinical setting.

301

Viral vectors (Manfred Schmidt, NCT) have shown their efficacy in clinical studies of rare diseases.

302

However, vector-mediated insertional side effects and malignant transformation have occurred in

303

individual patients. These observations have prompted new vector designs, such as self-inactivating

304

(SIN) LTR configurations and tissue-specific promoters that may improve safe and efficient viral

305

vector-mediated gene-correction in patients. Now, with the advent of mammalian whole genome

306

sequence availability and next generation sequencing (NGS), large-scale identification of vector

307

integration profiles and vector persistence studies over time have become an essential part of

308

pharmacokinetic studies for a gene therapeutics.

309

mediated PCR) coupled to NGS technology for vector integration site analysis is subject to

310

continuous further development. The current state-of-the-art and future perspectives were presented,

311

which includes new technologies like capture-based quantitative sequencing of viral vector sequences

312

and integration sites as well as associated bioinformatical data mining tool suites.

313

Patrizia Cristofori (GSK) presented the GLP Principles in Gene Therapy and the Experience of

314

Academic facilities GLP certified (HSR-TIGET). Gene therapy (GT) is now emerging as a medical

315

reality with clinical efficacy demonstrated in a number of gene therapy trials and an increasing

316

number of products entering phase II and III trials each year. For the promise and potential of a gene

The current LAM-PCR (linear amplification-

AC C

EP

TE D

M AN U

SC

RI PT

290

Page 12 of 46

ACCEPTED MANUSCRIPT therapy medicinal product (GTMP) to be fully realized it is important to address regulatory

318

expectations. Guidelines for GTMP progression in clinical trials and marketing authorizations are

319

available to facilitate a harmonized approach in the EU and US. Non-clinical studies have the

320

primary objective of providing sufficient information for a proper risk assessment for the product’s

321

use in human subjects. The paradigm described in ICH M3 for safety evaluation of conventional

322

pharmaceuticals is not always appropriate or relevant to GTMPs. Non-clinical studies should be

323

designed on a case-by-case basis, understanding the relevant aspects of the science underpinning that

324

product and need for specific expertise beyond the traditional pharmaceutical field. The combination

325

of expertise of personnel trained in genetic therapy research, experimental pathology, safety

326

assessment and quality assurance has enabled setting up a GLP Test Facility in an academic

327

environment. The main objective was to i) maximise the early collection of proof-of concept data and

328

provide adequate preclinical information to support the safety and the scientific basis for the

329

administration of an investigational product in the target patient population, ii) provide data of high

330

regulatory standard with assurance of scientific integrity, validity and reliability, and iii) minimise the

331

use of animals (3Rs). Challenges to be addressed in designing non-standard study to investigate in

332

vivo fate of genetically modified cells (biodistribution) and toxicological and tumorigenicity potential

333

were discussed. Discussion focused on the definition of test item and its characterisation, the choice of

334

test system (animal model), sample traceability, duration of treatment and endpoints to be measured.

335

The key role of Quality Assurance (QA) was highlighted. QA must be aware of all non-standard

336

aspects and challenges of GT preclinical studies and knowledgeable in technical procedures and

337

methods working closely with technical personnel to share criticality and best practices. In

338

conclusion, due to the complexity of GTMPs, non-clinical safety testing can vary considerably and is

339

recommended to be handled on a case-by-case basis, understanding the product and areas of non-

340

compliance.

341

Session 4: Immunogenicity and PKPD – from Prediction to Interpretation

342

Formation of anti-drug antibodies (ADA) is one of the hallmarks of an immune response against

343

biotherapeutics. ADA may affect safety, efficacy and pharmacokinetics of a biotherapeutic. Thus,

AC C

EP

TE D

M AN U

SC

RI PT

317

Page 13 of 46

ACCEPTED MANUSCRIPT ADA often change the pharmacokinetic – pharmacodynamics (PKPD) of biotherapeutics. Therefore,

345

presence of ADA needs to be considered in the interpretation of PKPD results. The fourth main

346

session, chaired by Wolfgang Richter (Roche) and Jay Tibbitts (UCB), gave overviews on PKPD

347

aspects of immunogenicity in toxicology studies, the immunogenicity of bispecific antibodies in

348

monkeys, and the current knowledge on in vitro tools for immunogenicity prediction. In addition,

349

approaches were discussed how to prevent immunogenicity in animal studies.

350

Olivier Petricoul (Novartis) presented the results of a survey conducted by the BioSafe PK/PD

351

working group on the analysis and reporting of toxicokinetic (TK) parameters. Toxicokinetic analysis

352

is a requirement for toxicology studies and, occasionally, in repeat-dose studies the test article

353

exposure in individual animals will change over the course of the study with increasing variability in

354

exposure in the TK assessment at the end of the study compared to Day 1 of the study. This

355

phenomenon may be due to the impact of ADA. The toxicokinetic analyst is then left to determine

356

how the TK data be reported as the effects of ADA formation on toxicokinetic/pharmacodynamic

357

parameters should be considered when interpreting the data (ICH S6).

358

It was highlighted that one of the main objectives of TK analysis is to show the validity of the

359

toxicology study, i.e. to confirm drug exposure at the start and end of the dosing period. Estimates of

360

the maximum observed concentration (Cmax) and area under the concentration-time curve (AUC) are

361

the most commonly generated TK parameters and usually the exposure multiple relative to human

362

exposure is based on the relevant mean AUC and Cmax values obtained in toxicology studies. As

363

ADA may impact on exposure (often a reduction, but in certain cases a prolonged exposure can be

364

observed) and consequently the estimates of Cmax and AUC, careful consideration should be given to

365

how mean values of these parameters are derived. Reporting of mean (and other summary statistics)

366

TK parameters in the context of ADA can be done in several ways. The mean TK parameters could

367

be calculated based on all animals independent of their ADA status. Alternatively, the mean TK

368

parameters of only ADA negative animals could be provided (Thway et al, 2013). The former could

369

be considered a more conservative estimate of drug exposure for the purposes of exposure margin

AC C

EP

TE D

M AN U

SC

RI PT

344

Page 14 of 46

ACCEPTED MANUSCRIPT calculations. Another approach would be to use TK parameters derived from Day 1 dosing (or a

371

similar early dosing period), where usually there is no impact of ADA.

372

A total of nine companies provided answers to the survey questions on how they calculate and report

373

TK parameters. All companies calculate TK parameters for all animals, regardless of ADA status.

374

Seven companies derive summary statistics using all animals, but four report, in addition, means for

375

ADA negative animals while the other three derive means for ADA negative animals only when ADA

376

appear to have affected the exposure. One company provides only one mean of TK parameters,

377

excluding ADA positive animals from the summary statistics when clear effects on exposure are

378

observed. Finally one company derives summary statistics using all animals, independent of the ADA

379

status. One example of a monoclonal antibody targeting a cell surface receptor was given to highlight

380

the effect of ADA on the TK (reduced exposure) associated with loss of PD effect. It shows the

381

importance to integrate the ADA status into the study summary to provide the appropriate

382

interpretation of the TK and PD results.

383

As noted above, one of the key elements of TK analysis is to confirm drug exposure in the study; and

384

as pointed out during the discussion on TK calculations some investigators choose to exclude animals

385

from summary TK parameters based on the degree of impact ADA has on exposure. Thus, one of the

386

major difficulties becomes how to define the «impact on exposure» of ADA? Several options were

387

discussed on how exclusion of ADA positive animals from mean TK parameter could be done, such

388

as scientific judgement, using a defined threshold (e.g. based on AUC decrease or increase from day

389

1), to report the median instead of the mean or to exclude time-points from profiles after animals

390

become ADA positive. Clearly, no established or mutually agreed-upon strategy exists for calculating

391

and reporting TK parameters in the presence of ADA. Good scientific judgement, clear explanation

392

and rationalization of the strategy used, and thoughtful consideration of the impact of the chosen

393

strategy on the exposure multiples and starting dose in humans are critical to appropriate risk

394

assessments under these circumstances.

AC C

EP

TE D

M AN U

SC

RI PT

370

Page 15 of 46

ACCEPTED MANUSCRIPT 395

Finally, there was a discussion regarding the exposure data necessary to support a toxicology study

396

versus those required to provide suitable PK data. The conclusion was that both PK (for dose

397

predictions and modeling) and TK (to establish an exposure-toxicity relationship or an exposure

398

multiple) can be simultaneously obtained from toxicology studies.

399

employing the strategies described above to derive mean TK parameters for exposure multiples and

400

using data from ADA negative/not impacted by ADA for PK evaluations.

401

Eric Stefanich (Genentech) presented preclinical data assessing the immunogenicity of bispecific

402

therapeutic antibodies in cynomolgus monkeys. Some of the potential differences between bispecific

403

and monospecific antibodies that may contribute to differential immunogenicity were discussed

404

including increased “foreignness” in bispecific antibodies due to two engineered complementarity

405

determining regions (CDRs); unique product-related variants; the formation of different sizes of

406

circulating immune complexes with ADA; and asymmetry in the fragment antigen-binding (Fab)

407

region (e.g., charge differences) that could impact tertiary orientation at the hinge region. The

408

emerging data suggest there may be increased immunogenicity (both incidence and titer) with some

409

bispecific MAbs compared to what is typically encountered for monospecific MAbs. However there

410

are limited case studies for bispecific antibodies to date. The observed magnitude of ADA can impact

411

the ability to conduct long-term preclinical safety studies in monkeys.

412

immunogenic epitopes from cynomolgus monkey studies are mostly associated with CDRs of the

413

bispecific antibodies. Based on limitations of existing immunogenicity risk-prediction tools, we may

414

need to rely on clinical trial results to understand clinical risk.

415

Sebastian Spindeldreher (Novartis) presented four case studies to exemplify how currently

416

available tools and assays can be used to assess the immunogenicity risk of biotherapeutics. Before

417

introducing in vitro tools for immunogenicity risk assessment methods Sebastian briefly introduced

418

ABIRISK (Anti-Biopharmaceutical Immunization: Prediction and analysis of clinical relevance to

419

minimize the RISK), which is a project of the Innovative Medicines Initiative (IMI) of the European

420

Union.

421

research groups and clinical centers. It is aimed at improving the understanding of patient- and

Where assessed, the

AC C

EP

TE D

M AN U

SC

RI PT

This can be achieved by

ABIRISK is a public-private partnership between pharmaceutical companies, academic

Page 16 of 46

ACCEPTED MANUSCRIPT compound-related factors influencing immunogenicity in order to identify clinical biomarkers for

423

immunogenicity and to validate available, and develop new, assay tools to assess the immunogenicity

424

risk for a given compound.

425

In vitro tools have been developed that may allow immunogenicity risk assessment. One of these

426

tools is MHC-associated peptide proteomics (MAPPS), a technology that identifies sequences of

427

peptides bound to HLA-DR, i.e. potential T cell epitopes. Other commonly used in vitro tools are T

428

cell activation assays, which are applied in many different variations. More recently, dendritic cell

429

activation assays have been developed to understand the potential effects on the innate immune

430

system.

431

The first case study showed that the number of potential T cell epitopes identified with MAPPs and

432

the response rate in T cell activation assays conducted with several marketed monoclonal antibodies

433

matched roughly the ranking based on clinical immunogenicity incidence (Karle et al, 2016).

434

However, one of the key issues for clinical validation of immunogenicity prediction is the lack of

435

reliable clinical immunogenicity data that can be compared across studies. Published data are based

436

on different immunogenicity assays with different sensitivities and drug tolerances, different sampling

437

time points, different diseases and disease states, different concomitant treatments and potentially

438

further differences. A solid comparison of in vitro and clinical data will only be possible with reliable

439

clinical immunogenicity data that can be compared across studies.

440

In vitro tools have been developed that may allow immunogenicity risk assessment. One of these

441

tools is MHC-associated peptide proteomics (MAPPS), a technology that identifies sequences of

442

peptides bound to HLA-DR, i.e. potential T cell epitopes. Other commonly used in vitro tools are T

443

cell activation assays, which are applied in many different variations. More recently, dendritic cell

444

activation assays have been developed to understand the potential effects on the innate immune

445

system.

446

The second case study focused on unpublished data from the ABIRISK consortium

447

[www.abirisk.eu]. The work conducted in the labs of Bernard Maillère (CEA) and Anette Karle

AC C

EP

TE D

M AN U

SC

RI PT

422

Page 17 of 46

ACCEPTED MANUSCRIPT (Novartis) showed good agreement between HLA-DR-presented peptides identified by MAPPs, T cell

449

epitopes identified from drug-naïve healthy blood donors and T cell epitopes of memory T cells

450

recovered from patients who were treated with and developed immunogenicity to the tested

451

biotherapeutics. Those data nicely confirm that the biology of patients who develop ADA against

452

biotherapeutics is mimicked in in vitro assays that are based on drug-naïve healthy subjects.

453

The third case study looked at the effects of protein aggregation on dendritic cell activation, antigen

454

presentation and T cell activation (Rombach-Riegraf et al, 2014). A correlation was found between

455

the amount of subvisible particles as well as the protein amount contained in these particles and the

456

amount of peptides presented in the context of HLA-DR. This data was confirmed with four different

457

marketed monoclonal antibodies that were studied in the context of ABIRISK.

458

monoclonal antibodies reacted very differently to physical stress, leading to different extent of

459

aggregation, which was again linked to increased antigen presentation as determined by MAPPs.

460

The last case study indicated that transgenic mice which express a mini-repertoire of human IgG1

461

antibodies can mount ADA responses to foreign antigens but are tolerant to native monomeric human

462

antibodies. Chemical modification by UV but not process-related or pH-induced oligomers could

463

induce ADA responses, which led to the hypothesis that neo-epitopes are required for the break of the

464

tolerance in these mice (Bessa et al, 2015).

465

During the discussion the audience largely agreed to the overall conclusion that it is currently not

466

possible to predict clinical immunogenicity incidence or outcome of immunogenicity. However,

467

promising mathematical models are being developed (Chen et al, 2014). Data from the “prediction

468

tools” should be considered as an assessment of the potential for immunogenicity but not as an

469

absolute prediction of immunogenicity incidence and outcome in human. However, the currently

470

available methods can be used for investigative and mechanistic studies and can help candidate

471

selection and internal decision making.

472

Wolfgang Richter (Roche) gave an overview on approaches to prevent immunogenicity in non-

473

clinical studies. The conduct of non-clinical studies with human or humanized biotherapeutics can be

These four

AC C

EP

TE D

M AN U

SC

RI PT

448

Page 18 of 46

ACCEPTED MANUSCRIPT 474

hampered by an immune response against the biotherapeutic. As previously noted, the formation of

475

ADA can substantially affect biotherapeutic exposure.

476

immune response and formation of ADA usually leads to an accelerated clearance and loss of

477

biotherapeutic exposure. However, for small biotherapeutics ADA may result in increased clearance

478

but, in some instances, may result instead in an apparent reduction in clearance (“sustaining ADA”)

479

due to slowing of the often fast endogenous renal elimination by virtue of binding of the small

480

biotherapeutic to a slowly cleared IgG ADA. Beyond the biotherapeutic exposure implications,

481

potential safety sequelae of immune responses include deposition of immune complexes in some

482

organs (eg. lungs, kidney). Re-dosing after formation of ADA may lead to anaphylaxis and acute

483

death (Hattori et al, 2007). Therefore, conduct of long term studies in mice may benefit from

484

suppression of an immune response. Ways to prevent immunogenicity include general suppression of

485

the immune system or antigen-specific tolerance induction. General suppression of the immune

486

response via pre-treatment with an anti-CD4 antibody was successfully used to conduct long-term

487

pharmacology studies in mice with the human monoclonal antibody gantenerumab (Bohrmann et al,

488

2012). The anti-CD4 treatment acted by transient depletion of peripheral CD4+ T-helper cells one

489

day before the start of gantenerumab administration. Alternative approaches include e.g. treatment

490

with the immunosuppressant mycophenolate mofetil.

491

antigen-specific tolerance induction keeps the immune system fully functional. Antigen-specific

492

immune tolerance may be induced by administering a high loading dose of a biotherapeutic followed

493

by administration of a lower dose during the test period. High dose tolerance induction was used to

494

enable a low dose group in a mice juvenile toxicity study with the antibody MR16-1 without ADA

495

formation or ADA-related side effects (Sakurai et al, 2013). Immune tolerance can also be induced

496

by exposing neonates to the tolerogen. In a study with adalimumab in mice, neonatal immune

497

tolerance was induced by dosing of mother mice after delivery (Piccand et al, 2016). Adalimumab

498

was transferred to the suckling pups via milk, thus providing antigen that induced tolerance. After

499

reaching adulthood at 8 weeks of age the offspring was dosed with adalimumab. Offspring exposed

500

to adalimumab the first days after birth showed no indication for an immune response, while control

By contrast to the previous approaches,

AC C

EP

TE D

M AN U

SC

RI PT

When the biotherapeutic is an IgG, the

Page 19 of 46

ACCEPTED MANUSCRIPT 501

mice showed an immune-mediated accelerated adalimumab clearance starting 7 days after dosing and

502

formation of ADA.

503

Overall, the session showed multiple aspects of immunogenicity and its consequences for the PKPD

504

and safety assessment of biotherapeutics.

505

biotherapeutic exposure is an important component in the TK and PK assessment. Novel formats may

506

differ in their immunogenicity from canonical monoclonal antibodies. In vitro tools for

507

immunogenicity are useful for investigative and mechanistic studies and can help candidate selection

508

and internal decision making. Various approaches for immune tolerance induction have been reported

509

to enable long term studies in animals without immune response. Immune tolerance induction has

510

been mainly applied in mouse studies, while experience in other species is limited or lacking.

511

Session 5: Antibody Drug Conjugates (ADCs)

512

The fifth main session was chaired by Simon Chivers (ADC Therapeutics SA) and Edit Tarcsa

513

(AbbVie), and covered the current approaches applied to antibody drug conjugate (ADC)

514

development. ADCs comprise of a monoclonal antibody and a cytotoxic payload attached via a

515

chemical linker, designed with the notion that the specificity of the antibody will lead to exclusive

516

delivery of the toxic payload to the cancer cells expressing the antigen on the cell surface, at the same

517

time eliminating systemic toxicities typically observed when administering the unconjugated payload.

518

Most ADCs currently in development have been generated by randomly attaching the payload to

519

either lysine residues or to reduced interchain cysteine in the antibody. These processes generate

520

heterogeneous mixtures with varying drug to antibody ratios (DAR), where the different forms may

521

contribute differentially to efficacy and/or toxicity. Homogenous DAR ADC’s are also possible

522

through the introduction of engineered cysteins, such as with the Thiomab technology. In addition,

523

some ADCs are not stable in vivo, some may be prone to toxin loss via proteolytic cleavage of the

524

linker or unwanted migration of the payload to plasma proteins. Evidence is emerging that in addition

525

to targeted delivery of the payload to tumor cells, ADCs are taken up via a variety of mechanisms

526

(pinocytosis, phagocytosis, FcR-mediated, or pattern recognition receptor-mediated uptake, etc), often

AC C

EP

TE D

M AN U

SC

RI PT

Knowledge of ADA formation and its impact on

Page 20 of 46

ACCEPTED MANUSCRIPT 527

the ‘normal’ clearance route for antibodies that might contribute to the non-target related toxicities

528

observed in the clinic. Thus understanding the composition and in vivo fate of ADCs can be critical to

529

optimize their properties.

530

development, it is of outmost importance to understand how to optimize the therapeutic margin of

531

these complex therapeutic modalities.

532

The first presentation by Stephanie Fischmann (AbbVie) talked about what should be measured for

533

ADCs – examples and experiences. Today, 3-4 bioanalytical assays are required to support the

534

development of ADCs. Typically, un-conjugated or “free or toxin”, “total antibody” and “conjugated

535

antibody” concentration are used for this purpose. The toxin load of the antibody can also be

536

represented by the measuring the “conjugated toxin” levels. The latter is very valuable if cleavable

537

linkers were chosen as this enables the use of generic methods. From a bioanalytical perspective

538

Ligand Binding Assays (LBA) were the “state-of-the-art” platform over the years for any “total

539

antibody” or “conjugated antibody” assays run in support of an ADC projects. Meanwhile, more and

540

more Mass spectrometry based methods have been developed, i.e. High Resolution Mass

541

spectrometry, where the drug-antibody ratio can be elucidated either by way of full-size ADC

542

construct molecular weight measurement or after reduction for light chain and heavy chain and

543

determining their molecular weight separately. Hybrid LC-MS/MS assays are also evolving e.g. for

544

determination of conjugated toxin levels. Here, affinity pull-down followed by enzymatic cleavage of

545

the linker facilitates LC-MS/MS quantification of the released toxin as a small molecule. Another

546

approach follows unique peptides identified to represent total antibody and conjugated antibody levels

547

after tryptic digestion. Mass spectroscopy based methods in general offer more detailed information

548

on the ADC backbone (i.e. number and position of toxins) and on the potential biotransformation of

549

the payload. However, MS-based hybrid assays are more tedious and can be limited in throughput or

550

sensitivity. Hence, from a bioanalytical and scientific perspective LBA can serve as the method

551

platform of choice throughout the whole life cycle of an ADC project. However, additional mass spec

552

technology based methods are beneficial to elucidate the in vivo fate of the ADC constructs with more

553

granularity (esp. early stage). Today, many different factors influence the choice of assays in support

AC C

EP

TE D

M AN U

SC

RI PT

With 2 ADCs currently approved and about 40 more in clinical

Page 21 of 46

ACCEPTED MANUSCRIPT of an ADC, including the development stage of the project, reagents availability, assay sensitivity,

555

assay development time and cost (cost per validation and cost per sample). For such reasons generic

556

platforms are often favorable at early stage but replaced by more specific assays later. Last, but not

557

least attempts are being made to simplify the bioanalytical efforts (i.e. the number of PK assays

558

undertaken in support of ADC projects). For example, multiplex technologies on LBA platforms and

559

multi-analyte strategies on LC-MS/MS platform are being evaluated.

560

In order to elucidate the immunogenicity of complex constructs like ADCs, all potential immunogenic

561

epitopes must be evaluated. Currently, a multistep strategy is recommended where after positive

562

screening assays the samples with potential ADAs are also tested in individual spike in experiments.

563

This so called “domain characterization” assay strategy employs unconjugated antibody spike as well

564

as full ADC spikes or spike-in experiments where only specific moieties of the ADC are spiked in,

565

thus the potential immunogenic epitopes can be identified.

566

Next, Matthias Machacek (LYO-X) presented on the PK, PD and toxicology modeling of IgG1

567

based ADCs and critical success factors. For ADCs three different PK levels can be identified: whole

568

body distribution, tumor penetration and uptake into the cell. Each of these three levels is critical for

569

successful toxin delivery to the tumor with a wide therapeutic margin. On a whole body level, it is the

570

IgG1 typical distribution and elimination processes that are relevant. IgG1 molecules remain

571

circulating in blood and interstitium until their uptake by cells and subsequent metabolism or

572

recycling via the FcRn mechanism. Thus, eventually all of the attached toxin will be delivered to

573

tissues that are active in IgG1 catabolism after internalization. For ADCs the IgG1 specificity to a

574

tumor target, a tumor specific cell surface molecule, will divert some of the ADC from the normal

575

clearance pathway. ADC binding to the tumor target will depend on the affinity and tumor target

576

abundance. Next, the tumor target bound ADC is internalized delivering the toxin into the cell.

577

Internalization depends on tumor target expression and the internalization rate. A back of the envelope

578

calculation showed that for a cell expressing 20’000 tumor targets and with a receptor internalization

579

half-life of about 2 hours intracellular toxin concentrations in the order of 10 nM are rapidly achieved.

580

A live ADC PK model simulation was then shown to illustrate the discussed concepts. The simulation

AC C

EP

TE D

M AN U

SC

RI PT

554

Page 22 of 46

ACCEPTED MANUSCRIPT demonstrated the impact of dose on how much toxin is delivered to the tumor versus to other tissues.

582

For high doses the fraction delivered to the tumor versus tissues was small. This was explained by the

583

saturation of the tumor up-take capacity in which case a large fraction of the toxin is delivered to IgG1

584

catabolizing tissues. By lowering the dose, the fraction delivered to the tumor could successively be

585

increased up to 80% for that specific case. Thus, for an optimal clinical dosing strategy the tumor

586

uptake capacity should be considered and the dose and regimen accordingly adjusted. Understanding

587

mechanistically and quantitatively the tumor target properties, ADC distribution and clearance as well

588

as the information drawn from the clinical PK profiles all help in better assessing the tumor uptake

589

capacity and hence the optimal dose and regimen.

590

Next, Andreas Pahl (Heidelberg Pharma) gave a talk about the toxicology of amanitin-based

591

ADCs. Payloads of today’s ADCs are exclusively based on compounds acting on microtubules or

592

DNA, are suffering from various limitations. New generations of payloads are entering the field

593

including Heidelberg Pharma’s amanitin, a highly effective inhibitor of the eukaryotic RNA

594

Polymerase II. Due to its unique mode of action and its hydrophilic nature this toxin differs

595

significantly from well-known payloads. Amanitin is the toxin of the green death cap mushroom.

596

Poisoning by amanitin leads to liver failure and subsequent death depending on the amount of toxin

597

ingested. Due to its hydrophilic nature amanitin cannot penetrate cell membranes passively. The

598

reason for its liver toxicity is a specific transporter OATB1B3, which is exclusively expressed in liver

599

cells. In vitro studies demonstrated that amanitin conjugated antibodies are not substrates for this

600

transporter indicating that amanitin-ADCs are not toxic to the liver. This was further explored in a

601

toxicological study in cynomolgus monkeys designed as an explorative study. Escalating doses of

602

Trastuzumab-Amanitin-ADCs from 0.3 to 10 mg/kg were administered to groups of three female

603

cynomolgus monkeys. In line with the in vitro findings no significant elevations of liver enzymes

604

were observed up to doses of 10 mg/kg. No other signs of liver toxicity or histological changes of the

605

liver were observed. LDH was the most sensitive parameter from clinical chemistry. A transient

606

increase at day 7 was observed. This was accompanied by hematological changes, namely elevated

607

levels of neutrophils. Dose-limiting toxicities were skin and ocular toxicity. These DLTs can be

AC C

EP

TE D

M AN U

SC

RI PT

581

Page 23 of 46

ACCEPTED MANUSCRIPT explained by target biology. Her2, the target of trastuzumab, is expressed on various epithelial cells.

609

Differences to T-DM1 toxicological findings can be explained by amanitin’s ability to kill also non-

610

dividing cells, whereas T-DM1 is only killing dividing cells. Next, different linker and conjugation

611

strategies for amanitin were compared. The MTD and the therapeutic index were significantly higher

612

for amanitin-ADCs based on site-specific conjugation to trastuzumab-thiomabs as compared to

613

random lysine conjugated ADCs. The therapeutic index for the amanitin-trastuzumab-thiomabs was

614

around 20 based on murine effective doses and monkey tolerable doses bridged by the AUC of the

615

ADC. This favorable therapeutic index will enable this amanitin-linker technology to move forward to

616

development.

617

In the final presentation for this session, Steve Alley (Seattle Genetics) described biotransformations

618

of auristatin ADCs. Biotransformations can be the intended effect of ADCs, such as release of

619

conjugated drug in the lysosomes of antigen positive cells, as well as amino acid modifications, loss

620

of drug-linker, differential clearance, or drug metabolism in the liver. As an example of the intent of

621

targeted delivery, the release of MMAE from brentuximab vedotin was demonstrated in vitro (Okeley

622

et al, 2010) and in mouse xenograft models. ADC modifications that occur in plasma were

623

demonstrated using native mass spectrometry of affinity-purified ADCs, showing that loss of drug-

624

linker as well as intact clearance of highly loaded ADCs contribute to the change in drug-load

625

distribution over time in vivo. Analyzing the samples under denaturing conditions allowed additional

626

modifications such as maleimide hydrolysis in the linker and cysteinylation of antibody cysteines to

627

be observed (Hengel et al, 2014). New drug-linker designs can lead to ADCs where loss of drug-

628

linker and differential clearance in plasma can be substantially reduced (Lyon et al, 2014). Finally,

629

MMAE released from brentuximab vedotin is excreted largely unchanged in feces, although several

630

metabolites due to demethylation, dehydrogenation, amide hydrolysis, and oxidation were observed

631

(Han et al, 2013).

632

Session 6: Lack of cross-reactive species – path forward to support FIH and clinical

633

development

AC C

EP

TE D

M AN U

SC

RI PT

608

Page 24 of 46

ACCEPTED MANUSCRIPT 634

The sixth and final main session was chaired by Rob Caldwell (AbbVie) and Flavio Crameri

635

(Roche).

636

justification as indicated by ICH S6(R1) Preclinical Safety Evaluation of Biotechnology-Derived

637

Pharmaceuticals. The guidance indicates that species selection should be based upon a weight-of-

638

evidence utilizing in vitro and/or in vivo biologic activity of biologic agent in toxicology species, and

639

comparable binding of biologic to human and preclinical species tissues based upon

640

immunohistochemical staining of tissue panels (e.g. tissue cross-reactivity experiments). The totality

641

of these datasets can guide selection of a rodent and non-rodent species. Given the unique selectivity

642

of biologic agents, the guidance acknowledges that this weight of evident approach may also result in

643

lack of relevancy for routine toxicology species. Toxicology studies in non-relevant species was

644

considered misleading and was discouraged. Alternative non-traditional toxicology test strategies

645

could employ use of homologous species-specific proteins as biologic surrogate, species-specific

646

surrogate monoclonal antibody, use of human clinical candidate in transgenic species expressing the

647

human target, or evaluation of preclinical species exhibiting the intended pharmacology (i.e. knockout

648

mice for mAb sequestering soluble target). All of these strategies would again be justified using

649

weight-of-evidence approach following generation of relevant in vitro and vivo datasets. Scientific

650

advances provide additional tools to justify relevant as well as irrelevant species for safety testing

651

including greater understanding of active (i.e. neonatal Fc receptors) placental transfer of novel

652

biologic therapeutic, enhanced understanding of embryology across human and preclinical species,

653

and more sensitive analytical techniques to justify relevant pharmacology and tissue distribution

654

across human and preclinical species. Each of these alternative models have scientific advantages

655

and/or disadvantages including pharmacologic relevance to human biologic, reference range datasets

656

to evaluate the toxicologic significance of observed effects, and potential for neutralizing anti-drug

657

antibodies which may preclude use of the preferred model/species. Finally, alternative strategies also

658

have resource needs which impact the timing and efficiency of the selected toxicology model

659

including cost of unique animal colony maintenance and generation of reference range datasets, and

660

generation of regulatory compliance for surrogate proteins. The session provided cases studies of

661

species justification for novel biologic therapeutics.

AC C

EP

TE D

M AN U

SC

RI PT

Rob Caldwell introduced the session by providing an overview of species selection

Page 25 of 46

ACCEPTED MANUSCRIPT Lolke de Haan (MedImmune) presented 2 case studies involving nonclinical safety testing of

663

human-selective monoclonal antibodies (mAbs) in transgenic mouse models. The first case study

664

described the challenges encountered in developing a transgenic mouse model for safety testing of

665

MEDI2843, a fully human monoclonal antibody directed against Toll-like receptor 4 (TLR4), the

666

innate immune receptor for bacterial lipopolysaccharide (LPS). MEDI2843 was isolated following

667

immunisation of a transgenic mouse strain expressing human Ig domains with human TLR4/MD-2

668

protein and/or TLR4/MD-2-expressing Human Endothelial Kidney (HEK) cells. None of the lead

669

panel of antibodies bound to rodent TLR4, and despite the fact that a number of lead antibodies bound

670

to cynomolgus monkey TLR4, none of these antibodies showed functional activity in cell-based

671

assays. By contrast, MEDI2843 was highly human selective and bound and inhibited functional

672

activity of human TLR4 in cell based assays very effectively.

673

pharmacologically relevant species for safety studies, attempts were made to generate human TLR4

674

transgenic mice. The approach followed involved deletion of the mouse TLR4 gene and insertion of a

675

mouse/human chimeric version of the TLR4 gene, comprised of gene fragments encoding the mouse

676

transmembrane and cytoplasmic TLR4 domains and the human TLR4 external domain. When this

677

gene construct was expressed recombinantly in a cell line, the resultant protein was shown to be fully

678

functional and mediate LPS signalling, with binding of MEDI2843 ablating LPS signalling. Given the

679

importance of MD-2 in LPS signalling, a second transgenic mouse was generated in which the mouse

680

MD-2 gene was replaced by the human MD-2 gene. A combined chimeric mouse/human TLR4 and

681

human MD-2 transgenic mouse strain could then be obtained by crossing these 2 transgenic strains.

682

Upon generation of the human/mouse chimeric TLR4 and combined human/mouse chimeric

683

TLR4/human MD-2 transgenic mice, bone marrow-derived macrophages were, however, shown to be

684

hyporesponsive to LPS. Furthermore, both transgenic mouse strains were hyporesponsive to LPS

685

challenge in vivo. Subsequent analyses showed that these defects were related to negligible

686

expression of both of the transgenic gene constructs. Hence, the transgenic animal approach failed to

687

produce a pharmacologically relevant transgenic mouse species that was suitable for pharmacology or

688

safety studies. In the second case study by Lolke de Haan, the development of a human C-X-C

689

chemokine receptor 2 (CXCR2) transgenic mouse model for safety testing of Hy29, a fully human

Given the apparent lack of

AC C

EP

TE D

M AN U

SC

RI PT

662

Page 26 of 46

ACCEPTED MANUSCRIPT monoclonal antibody directed against CXCR2 was presented. CXCR2 is a major chemotactic receptor

691

expressed on neutrophils with multiple chemokine ligands, including interleukin 8 (IL-8). Hy29 was

692

isolated after immunisation of a transgenic mouse strain expressing human Ig domains with human

693

CXCR2 overexpressing cells. Hy29 binds with high affinity to human CXCR2 and inhibits its

694

function in cell-based assays, whilst no or low affinity binding to rodent or cynomolgus monkey

695

CXCR2 was observed, and Hy29 did not inhibit rodent or cynomolgus monkey receptor function in

696

cell-based assays. In order to assess the nonclinical pharmacological activity and safety of Hy29, a

697

human CXCR2 transgenic mouse was generated in which the mouse CXCR2 gene was deleted or

698

disrupted and the human CXCR2 gene was inserted. Importantly, both in humans and mice CXCR2

699

has multiple ligands, not all of which are overlapping exactly with respect to function, and with some

700

ligands not having rodent counterparts in humans or vice versa.

701

precedence for the functionality of human CXCR2 transgenic mice, which provided additional

702

confidence in the approach (Mihara et al., 2005). Human CXCR2 transgenic mice were generated

703

successfully and functionality of the transgene was demonstrated in an intranasal LPS challenge

704

study, in which human CXCR2 transgenic mice showed similar lung neutrophil accumulation to wild

705

type mice 24 hours post challenge.

706

administration of Hy29 prior to LPS challenge in human CXCR2 transgenic mice, whilst the lung

707

neutrophil response to intranasal LPS was unchanged in wild type mice that received Hy29. The

708

human CXCR2 transgenic mouse model was then employed in a dose-range finding toxicity study, in

709

which animals received 3 doses of 4, 40 or 100 mg/kg Hy29 over a period of 14 days. In this study,

710

treatment with Hy29 was associated with dose-dependent reductions in circulating neutrophils as well

711

as dose-dependent increases in circulating mouse CXCR2 ligands, consistent with the anticipated

712

pharmacodynamic effects of Hy29.

713

anaphylactic reactions in individual animals following the second dose at 4 and 40 mg/kg Hy29, and

714

the third dose of 100 mg/kg Hy29. These effects correlated with a loss of exposure to Hy29 and

715

consequent pharmacodynamics effects, indicating that these reactions represented anti-drug antibody

716

(ADA)-mediated type III hypersensitivity reactions. Therefore, whilst the human CXCR2 mice

717

appeared fully functional and pharmacologically relevant, the immunogenicity of a fully human mAb

However, there is literature

TE D

M AN U

SC

RI PT

690

AC C

EP

Moreover, this response could be ablated by intravenous

However, treatment with Hy29 was also associated with

Page 27 of 46

ACCEPTED MANUSCRIPT to mice prevented repeat dosing beyond 10 days, limiting the utility of the transgenic mice as a

719

species for toxicity studies. Together, these 2 case studies demonstrated that whilst transgenic mice

720

represent an option for assessing nonclinical safety of mAbs, their generation and utility for repeat-

721

dose nonclinical safety studies should be carefully considered prior to selecting transgenic animals as

722

a potential way forward for nonclinical safety testing.

723

Lise Loberg (AbbVie) presented a case study on a mAb for a disease-specific target with no cross-

724

reactivity to normal animal species commonly used in toxicology studies, focusing on study design

725

considerations. Selection of the specific transgenic (Tg) model was based upon several factors

726

including commercial availability, understanding of the pharmacology of mAb X in the particular Tg

727

model, and tissue expression of the transgene (i.e., expression primarily in CNS). Another

728

consideration was the optimal age of animals at the start of the toxicology study, whereby the

729

transgene was expressed at sufficiently high levels for mAb X to bind target, but before the

730

degenerative nature of the disease affected survivability or other toxicity endpoints. Target protein

731

expression was variable in animals at 2 and 4 months of age, was consistent with age-dependent

732

increases from 6 to 10 months of age, and reached a plateau at 10 and 13 months of age. mAb X

733

binding to target was first detectable at 6 months and considerably increased with age. Therefore, 6

734

months was selected as the optimal starting age for toxicology studies.

735

Pharmacology studies administered mAb X via IP administration, and were suggestive of anti-drug

736

antibody (ADA) development. To better understand potential for ADA formation, a pharmacokinetics

737

(PK) study was conducted in 6-month-old Tg mice, in which 6 mice/sex/group were dosed on Days 1

738

and 8 at 60 mg/kg IV or 100 mg/kg SC injection. There were no sex differences in PK parameters,

739

there was moderate SC bioavailability (approximately 60%), half-life was approximately 3 days, and

740

there were no indications for significant ADA development. Serum concentrations of mAb X were

741

maintained for up to one week after dosing. In addition, the PK study demonstrated feasibility for

742

sparse sampling to minimize animal numbers.

AC C

EP

TE D

M AN U

SC

RI PT

718

Page 28 of 46

ACCEPTED MANUSCRIPT An important consideration in the decision to evaluate potential toxicity in animal models of disease is

744

the lack of historical background knowledge of the model. This may be especially true for a Tg

745

animal model in which the transgene and its associated protein may be expressed very differently

746

from the normal protein and from the human disease of interest. For example, the animal model may

747

have alterations in behavior, clinical pathology parameters, or tissue histology. With that in mind,

748

toxicology studies in animal models are best conducted as hazard identification studies, with focus on

749

particular tissues or organ systems, rather than a comprehensive toxicological evaluation. In this case,

750

however, a full toxicological evaluation was intended and therefore, preliminary characterization of

751

background histopathological findings was conducted. To reduce and refine animal usage, tissues

752

were collected from Tg mice in a pharmacology study in which Tg mice were dosed with mAb X, one

753

of three alternate antibodies, a positive control or a negative control. Mice received 0.5 mg antibody

754

per week via IP administration for up to 18 weeks, starting at 6 months of age. Selected tissues (heart,

755

lung, kidney, liver, GI tract, and hindlimb joint and muscle) were evaluated from available animals in

756

the negative control group (N = 8), the mAb X group (N = 10), and two of the alternate antibody

757

groups (N = 7 and N = 11). There were no statistical differences in survival or cause of death (via

758

neoplasia or other causes), no changes in body weight, and no differences in types or incidence of

759

neoplasia between different experimental groups. Serum concentrations of mAb X were decreased by

760

6 weeks, suggestive of ADA development. Histopathological changes in kidney and degenerative

761

joint disease were common and attributable to aging, whereas skeletal muscle degeneration and

762

regeneration was severe in some animals and may have been a background change in the Tg model.

763

The preliminary PK and histopathologic studies informed study design for a GLP-compliant repeated

764

dose toxicology study. Tg mice were administered mAb X at 0 (vehicle control), 60 and 200

765

mg/kg/week IV (low and high doses), and 200 mg/kg/week SC (mid dose) for 4 weeks. Animal

766

numbers were increased to 18/sex/group to account for expected mortality up to 10%. ADA

767

development remained an unknown factor; therefore, the GLP study included two additional groups

768

dosed at 0 and 200 mg/kg/week IV for 13 weeks to evaluate PK and ADA, to inform feasibility for a

769

study longer than 4 weeks. In the 4- and 13-week GLP study, there was anticipated mortality (15%)

AC C

EP

TE D

M AN U

SC

RI PT

743

Page 29 of 46

ACCEPTED MANUSCRIPT during the study but also unanticipated mortality on the first day of dosing. The unanticipated deaths

771

were attributed to technical difficulty with IV injections into tail veins of dark-colored mice.

772

(Toxicology studies are often conducted in CD-1 mice, which are light-colored and tail veins are

773

easily observed on pink tails.) There was no relationship to dose and no clinical signs or

774

histopathologic changes suggestive of test article-related death; therefore the deaths on study were

775

attributed to procedure, aging and/or disease state. The mAb X half-life was < 2 days (IV), and 62%

776

of test article-dosed animals had detectable ADA. Only 1 (of 15) animals had detectable mAb X

777

concentrations after the 13th weekly dose of 200 mg/kg, confirming that a study longer than 4 weeks

778

would not be feasible. Histopathologic evaluation revealed anticipated effects (chronic nephropathy,

779

skeletal muscle degeneration/inflammation) that were common in control and test article-treated

780

animals and therefore considered age- and Tg-related but not mAb X-related.

781

In conclusion, the preliminary studies conducted to evaluate PK parameters after different routes of

782

administration and histopathological changes in the Tg mouse model of interest were instrumental to

783

the design of a 4-week GLP repeated-dose toxicology study and to the interpretation of study results.

784

Furthermore, the addition of dose groups administered control or mAb X for 13 weeks confirmed

785

development of ADA and associated decreased test article concentration, thereby providing evidence

786

that a study longer than 4 weeks would not be feasible.

787

Flavio Crameri (Roche) presented a combined surrogate and in vitro safety approach in support of

788

first-in-human dosing for CEA-IL2v (INN: cergutuzumab amunaleukin), a Roche proprietary tumor

789

antigen targeted immunocytokine currently in phase 1 clinical trials. CEA-IL2v is composed of a

790

monomeric human IL-2 with mutations (IL2v) to abolish binding to the high affinity IL-2 receptor

791

(IL-2Rαβγ). IL2v is fused to the c-terminal end of a bivalent anti-carcinoembryonic antigen (CEA)

792

human IgG1 with silenced effector functions. The anti-CEA component of the clinical candidate lacks

793

cross-reactivity with any toxicology species, while the IL2v component has been shown to be cross-

794

reactive in the cynomolgus monkey and the mouse. Given the lack of a fully cross-reactive species for

795

safety assessment it was decided to start the development of a cynomolgus monkey reactive anti

796

CEA-IL2v (cyCEA-IL2v) and at the same time to discuss and obtain feedback from a European health

AC C

EP

TE D

M AN U

SC

RI PT

770

Page 30 of 46

ACCEPTED MANUSCRIPT authority (HA) regarding the proposed preclinical safety strategy. The HA considered a surrogate

798

approach with cyCEA-IL2v appropriate for qualitative safety assessment once functional equivalence

799

of cyCEA-IL2v to CEA-IL2v had been shown. The surrogate approach however was to be combined

800

with an extensive quantitative in vitro safety assessment using CEA-IL2v in human cell based assays

801

to derive a minimum anticipated biological effect level (MABEL). The HA further commented that

802

CEA-IL2v may not be considered high risk based on the extensive previous clinical experience with

803

anti-CEA mAbs for imaging and with rhuIL-2 (Proleukin®) as well as the tumor targeted approach to

804

be used. In addition, the HA acknowledged that a starting dose based on a MABEL may be

805

unacceptably low and indicated that it was open to accept a combination of MABEL with other

806

approaches to determine a safe starting dose for CEA-IL2v. Based on the HA feedback received,

807

development of the surrogate cyCEA-IL2v was continued. Binding and functional equivalence of cy

808

CEA-IL2v when compared to CEA-IL2v was shown in several comparative in vitro assays; i.e.

809

affinity to recombinant CEA and IL-2Rβγ and IL2v-mediated functional activity using human and

810

cyno cells (cell signaling, cell activation and proliferation as well as cytokine release). The surrogate

811

cyCEA-IL2v was then tested in pilot and GLP studies in the cynomolgus monkey. The safety profile

812

obtained with cyCEA-IL2v in the cynomolgus monkey was consistent with the known safety profile

813

of wild type IL-2 (Proleukin®) thus validating the surrogate-IL2v for qualitative safety assessment.

814

No vascular leak syndrome, a major toxicity of wild type IL-2, was seen and there was no evidence

815

for any potential CEA-targeting related toxicity in tissues known to express CEA. The safe starting

816

dose for the FiH study was then selected based on a combination of the MABEL (IC20) obtained with

817

CEA-IL2v and published clinical data from competitor IL-2 immunocytokines for which safety and

818

exposure data were available. The resulting starting dose used for fist-in-human was approximately 3-

819

fold higher than the MABEL and was shown to be safe in the ongoing phase 1 clinical trial. This

820

approach helped to balance safety and number of patients without potential clinical benefit.

821

The second case study from Roche discussed a FiH that was supported exclusively by in vitro safety

822

data and a safe starting dose selection based on a MABEL only.

AC C

EP

TE D

M AN U

SC

RI PT

797

Page 31 of 46

ACCEPTED MANUSCRIPT Katharine Bray-French (Roche) presented a case study for CEA TCB which is a novel, Roche

824

proprietary T cell bispecific (TCB) antibody targeting carcinoembryonic antigen (CEA) expressed on

825

tumor cells and CD3 epsilon chain (CD3e) present on T cells that is currently in Phase 1 clinical trials

826

(NCT02324257) for the treatment of CEA positive solid tumors. Because the human CEA (hCEA)

827

binder of CEA TCB does not cross-react with cynomolgus monkey and CEA is absent in rodents,

828

alternative nonclinical safety evaluation approaches were considered. These included the development

829

of a cynomolgus monkey cross-reactive homologous (surrogate) antibody (cyCEA TCB) for

830

evaluation in cynomolgus monkey and the development of double transgenic mice, expressing hCEA

831

and human CD3e (hCEA/hCD3e Tg), as a potential alternative species for nonclinical safety studies.

832

However, a battery of nonclinical in vitro/ex vivo experiments demonstrated that neither of the above

833

approaches provided a suitable and pharmacologically relevant model to assess the safety of CEA

834

TCB. Therefore, an alternative approach, a MABEL based on an in vitro tumor lysis assay was used

835

to determine a safe starting dose for the first-in-human clinical study.

836

Benno Rattel (Amgen) presented three different nonclinical testing strategies for bispecific T-cell

837

engagers, commonly referred to as BiTE® antibody constructs. These molecules are comprised of two

838

different flexibly-linked single-chain antibodies, one directed against a tumor antigen and the other

839

targeting CD3. BITE® antibody constructs are designed to link transiently tumor cells with resting

840

polyclonal T-cells for induction of a surface target antigen-dependent re-directed lysis of tumor cells.

841

First-generation BiTE® antibodies cross-reacted only with respective antigens from chimpanzees

842

therefore, to facilitate in vivo safety testing, surrogate BiTE® antibodies were generated that are

843

cross-reactive with murine antigens. Firstly, the nonclinical safety package of Blincyto®, a CD19-

844

targeting BiTE® antibody construct, was presented (Nagorsen et. al, 2012; Topp et. al, 2015). It

845

included up to 3-month repeat-dose toxicity studies and an embryofetal toxicity study with the mouse

846

surrogate, which mimicked the parent compound`s pharmacological characteristics and successfully

847

supported the molecule`s recent approval both in the US and the EU (FDA, 2014; EMA, 2015) The

848

second case study described the determination of the clinical starting dose for a CEA-targeting BiTE®

849

antibody construct. For this tumor target, a surrogate molecule could not be developed that was

AC C

EP

TE D

M AN U

SC

RI PT

823

Page 32 of 46

ACCEPTED MANUSCRIPT comparable to the parent molecule, since both the mouse and the cynomolgus surrogates evaluated

851

caused unspecific T cell activation. Therefore, the FIH dose was based on in vitro MABEL data, only.

852

Finally, the current nonclinical safety assessment paradigm for defining a safe clinical starting dose

853

was presented for the new generation of BiTE® antibody constructs that cross-react with cynomolgus

854

monkeys (Friedrich et al, 2012).

855

“Hot Topic” - Breakout Sessions

856

Breakout Session Question 1:

857

advantages over combinations ? For this breakout session Lolke de Haan (MedImmune) first

858

gave a brief presentation outlining the potential challenges in the nonclinical and clinical development

859

of bispecific antibodies and antibody combinations. Subsequent discussions focused on whether or

860

not bispecifics would provide drug molecules with additional, mechanistically differential, mode of

861

action. Or, alternatively, if bispecifics simply represent a more convenient way for the delivery of 2

862

pharmacodynamically active modalities, and an opportunity for replacing 2 separate injections with

863

monoclonal antibodies with 1. The consensus appeared to be that convenience was the main driver for

864

developing bispecifics. It was recognised that the ratio of the 2 bioactive components in a bispecific is

865

fixed, and that this could be a potential disadvantage. On the other hand, for combinations flexibility

866

could become a disadvantage, as with the number of variables available (antibody ratio and dosing

867

frequency and level) the number of combinations that could be tested are endless, and this could

868

complicate or hamper clinical development. With respect to targets suitable for bispecifics, there was

869

consensus that soluble targets are preferable, however, there was no expectation that both arms of the

870

bispecifics would engage 2 different soluble targets simultaneously. For membrane-bound targets,

871

binding 2 targets per bispecific molecule was generally not deemed likely, however, it would still be

872

possible to modulate 2 membrane-bound targets through monovalent interactions. Finally, the

873

bioanalytical challenges associated with bispecifics in the nonclinical and clinical setting were

874

discussed. For bispecifics, it was agreed that anti-idiotype antibodies against each bispecific arm to

875

determine the amount of free bispecific would be required. In addition, assays demonstrating stability

RI PT

850

AC C

EP

TE D

M AN U

SC

“Do Bispecifics offer nonclinical and clinical development

Page 33 of 46

ACCEPTED MANUSCRIPT 876

of the bispecific, e.g. capture with an anti-idiotype combined with an anti-Fc detection, were

877

considered desirable.

878

Breakout Session Question 2: “Is the paradigm “SC more immunogenic than IV” correct ?”

879

chaired by Benno Rattel (Amgen). Biotherapeutics have to be administered via a parenteral route,

880

because approaches to oral administration have failed so far.

881

administration, SC administration has the advantage of improved convenience and patient compliance

882

as well as prolonged half-life (Richter and Jacobsen , 2014). However, the SC route is perceived to be

883

problematic due to a greater potential for undesired immunogenicity (Braun et al, 1997; Ponce et al,

884

2009; FDA 2014). As there is generally a poor concordance between immunogenicity in animals and

885

humans, and, with few exceptions, human biologics are generally more immunogenic in animals than

886

in humans (Bugelski and Treacy, 2004; Ponce et al, 2009), ideally the validity of the above paradigm

887

should be verified with actual clinical experience of head-to-head comparisons of SC and IV

888

administrations of biotherapeutics.

889

There are only few published clinical examples directly comparing the two administration routes:

890

Rituximab (Bittner et al, 2014; Davies et al, 2014), Trastuzumab (Jakisch et al, 2015), Tocilizumab

891

(Ogata et al, 2014; Burmester et al, 2014), Belimumab (Shida et al, 2014), ATR-107 (Hua et al, 2014)

892

and Abatacept (Schiff, 2013). These clinical data as well as unpublished internal company data were

893

discussed, and the group agreed that these data do not support the above paradigm. The group argued

894

that a number of patient- and product-specific factors such as origin, post-translational modification,

895

aggregates, degradation products, impurities, formulation, container closure and, especially,

896

immunomodulatory properties (Parenky et al, 2014) are at least equally important for an unwanted

897

immune response as the administration route. Therefore, the group concluded that the generalization

898

that the SC administration bears a higher risk of immunogenicity compared to the IV route cannot be

899

universally upheld.

900

Breakout Session Question 3: For 3Rs considerations of NHP use “Why do we need a 6 month

901

toxicology study for a mAb when we already have a 13 week toxicology study?” chaired by Katja

AC C

EP

TE D

M AN U

SC

RI PT

Compared to the IV route of

Page 34 of 46

ACCEPTED MANUSCRIPT Hempel (AbbVie) and Peter Ulrich (Novartis). Before the meeting a survey was send out by Peter

903

Ulrich to BioSafe members with the goal to collect the experience of the pharmaceutical industry with

904

toxicity studies supporting clinical development of monoclonal antibodies in order to evaluate

905

repeated dose toxicity strategy. This survey was limited to results generated with monoclonal

906

antibodies given via the intravenous or subcutaneous route at similar dose levels between the toxicity

907

studies and selected according to toxicology standard practice. Pure oncology mAb were not included

908

in this survey, since 26 week studies are not required for those mAbs. The findings asked for in the

909

survey were divided into pharmacodynamic (PD), target organ toxicity, and safety pharmacology

910

findings leading to a NOAEL.

911

Results: 26 case examples were provided for the survey by five BioSafe member companies. In none

912

of the 26 case examples were differences seen in the toxicological profile between 3 and 6 months. In

913

addition, one company informed verbally that they had 4 mAbs with again no difference in

914

toxicological profile. During the plenary session two additional interesting examples were mentioned

915

where toxicological effects were only seen in studies of longer duration. Taking into consideration the

916

results of the long study, and reevaluating the short term study, there was a hint of this toxicological

917

finding. With regard to the challenge as to when a 3 month stand-alone study might be sufficient,

918

there was agreement in panel that this needs to be a case by case decision also taking the target into

919

account. In addition, the use of sexually mature animals was recommended. A working group will be

920

formed to distribute the survey to companies that have not yet received the survey to include as many

921

case examples as possible, including especially the ones with a different outcome other than

922

mentioned in the plenum discussion. Additionally, it was suggested to collect data from marketed

923

products to update the list from former publications (Clarke et al, 2008). Goal is to publish the

924

current pharmaceutical expertise to leverage experience for mAbs with potential future extension to

925

more complex ADCs or DVDs, if possible.

926

Breakout Session Question 4:

927

Preclinically and does it Translate Clinically” chaired by Rajni Fagg (GSK). The breakout

928

session opened with an example from the literature (Heyden et. al, 2014) in which an investigative

AC C

EP

TE D

M AN U

SC

RI PT

902

“How Do we Investigate Immune Complex Formation

Page 35 of 46

ACCEPTED MANUSCRIPT 929

study was conducted to characterise mAb induced immune complex (IC) formation in cynomolgus

930

monkeys. This preclinical investigative study was also designed to identify potential biomarkers that

931

could be indicative of immune complex formation and determine if the induced immune complex

932

formation was reversible. The discussions in the breakout session concluded: 1. Most likely, as antigen/mAb therapeutic ratios approach molar equivalence, the ICs are larger

934

and when clearance mechanisms are blocked or saturated, large ICs can deposit in tissue,

935

activate complement, and cause tissue damage.

937

2. Other factors that influence IC formation include high

antigen

and

antibody

SC

936

RI PT

933

concentration/density as well as antigen and mAb charge.

3. Although in the published example high anti-mAb antibody response in the monkeys

939

preceded the clinical pathology findings of thrombocytopenia, complement activation and

940

neutrophilia by several weeks, this was not considered to be always the case.

942

4. No clear correlation of preclinical findings of mAb therapeutic induced IC formation has been identified in the clinical setting.

TE D

941

M AN U

938

The fifth and final “hot topic” during the breakout session was a group discussion of “Adversity in

944

Nonclinical Safety Studies: STP and ESTP opinion, your opinion needed !”, chaired by John

945

Burkhardt (AbbVie) and Paul Germann (AbbVie). As an introduction John presented the STP (see

946

Figure 1, Kerlin et al, 2016; flow chart was developed for NCEs/small molecules but concept is also

947

relevant for biologics), and Paul the ESTP point of view (see Figure 2, Palazzi et al, 2015). Finally

948

several case studies from different companies were presented followed by a lively discussion on

949

whether a specific finding will be assessed as adverse and will therefore impact the NOAEL setting.

950

In essence, the tiered approach in histopathological adversity assessments, as depicted in Figure 2,

951

should guide younger colleagues towards a correct assessment of adversity. What still remains an

952

area of concern is the clear disconnect between the humoral reaction against biological products like

953

ADA formation and the histopathological findings which do not correlate in most of the cases.

AC C

EP

943

954

Page 36 of 46

ACCEPTED MANUSCRIPT Figure 1: Schematic diagram delineating the hierarchy by which “adverse” test-article-related effects

956

are communicated (see Figure 2 from Kerlin et al, 2016)

M AN U

SC

RI PT

955

AC C

EP

TE D

957

Page 37 of 46

ACCEPTED MANUSCRIPT Figure 2: Tiered approach for evaluating adversity (see figure 1 from Palazzi et al, 2015)

959

TE D

M AN U

SC

RI PT

958

Conclusions

961

The opportunities for the application of biotherapeutics for the treatment or prevention of human

962

diseases continue to grow, and new molecular formats, including bi- or multi-specific molecules,

963

antibody fragments, and ADCs, are likely to reach the market in the near future.

964

complexity and unique properties of these novel modalities, standardized approaches for nonclinical

965

safety testing are becoming increasingly obsolete, and each molecule needs to be evaluated on case-

966

by-case basis, tailoring the nonclinical safety program and strategy to the molecule.

967

This meeting brought experts together from the nonclinical safety (toxicology, pathology, safety

968

pharmacology), pharmacokinetics and bioanalytical field to exchange knowledge and experience. In

969

addition to the presentations, podium discussions and breakout sessions, the meeting offered ample

AC C

EP

960

Page 38 of 46

Given the

ACCEPTED MANUSCRIPT 970

opportunity for participants to share ideas, development strategies and case studies. This is becoming

971

increasingly important for the biopharmaceutical industry, as this often represent the only way in

972

which to share data on failed development programs, studies and strategies.

973

exchange of knowledge promotes the optimization of future biotherapeutics non-clinical development

974

programs, which will not only contribute to a responsible and justified use of animals, but hopefully

975

also to improved human risk assessment of innovative biotherapeutics.

976

Conflict of interest

977

Guenter Blaich, Rob Caldwell, and Edit Tarcsa are employees of AbbVie. Andreas Baumann is an

978

employee of Bayer Pharma AG; Sven Kronenberg, Wolfgang Richter and Flavio Crameri are

979

employees of Roche. Lolke de Haan is an employee of MedImmune; Peter Ulrich is an employee of

980

Novartis Pharma; Jay Tibbitts is an employee of UCB Celltech and Simon Chivers is an employee of

981

ADC Therapeutics. AbbVie and affiliates participated in the preparation of the workshop report,

982

interpretation of data, review, and approval of the manuscript.

983

Acknowledgements

984

The authors would like to thank all speakers for their contributions and members of the BioSafe

985

Leadership committee for valuable comments during the review of this manuscript.

986

References

987

Baumann A, Tuerck D, Prabhu S, Dickmann L and Sims J. (2014) Pharmacokinetics, metabolism and

988

distribution of PEGs and PEGylated proteins: quo vadis? Drug Discov Today 19(10):1623-1631.

RI PT

SC

M AN U

TE D

EP

AC C

989

Furthermore, this

990

Bessa J, Boeckle S, Beck H, Buckel T, Schlicht S, Ebeling M, Kiialainen A, Koulov A, Boll B,

991

Weiser T, Singer T, Rolink AG, Iglesias A (2015). The immunogenicity of antibody aggregates in a

992

novel transgenic mouse model. Pharm Res 32:2344-2359.

993 994

Bittner, B., et al. (2014). "Non-clinical pharmacokinetic/pharmacodynamic and early

Page 39 of 46

ACCEPTED MANUSCRIPT 995

clinical studies supporting development of a novel subcutaneous formulation for the

996

monoclonal antibody rituximab." Drug Res (Stuttgart) 64(11): 569-575.

997 Bohrmann B, Baumann K, Benz J, Gerber F, Huber W, Knoflach F, Messer J, Oroszlan K,

999

Rauchenberger R, Richter WF, Rothe C, Urban M, Bardroff M, Winter M, Nordstedt C, Loetscher H

1000

(2012). Gantenerumab: a novel human anti-Aβ antibody demonstrates sustained cerebral amyloid-β

1001

binding and elicits cell-mediated removal of human amyloid-β. J Alzheimers Dis. 28:49-69.

RI PT

998

1002

Braun, A., et al. (1997). "Protein aggregates seem to play a key role among the

1004

Parameters influencing the antigenicity of interferon alpha (IFN-alpha) in normal and

1005

transgenic mice." Pharm Res 14(10): 1472-1478.

M AN U

SC

1003

1006

Bugelski, P. J. and G. Treacy (2004). "Predictive power of preclinical studies in

1008

animals for the immunogenicity of recombinant therapeutic proteins in humans." Curr

1009

Opin Mol Ther 6(1): 10-16.

TE D

1007

1010

Burmester, G. R., et al. (2014). "A randomised, double-blind, parallel-group study of

1012

the safety and efficacy of subcutaneous tocilizumab versus intravenous tocilizumab

1013

in combination with traditional disease-modifying antirheumatic drugs in patients with

1014

moderate to severe rheumatoid arthritis (SUMMACTA study)." Ann Rheum Dis 73(1):

1015

69-74.

AC C

1016

EP

1011

1017

Chen X, Hickling TP, Vicini P (2014). A mechanistic, multiscale mathematical model of

1018

immunogenicity for therapeutic proteins: part 1-theoretical model. CPT Pharmacometrics Syst

1019

Pharmacol 3:e134.

1020 1021

Clarke, J., et al. (2008). Duration of chronic toxicity studies for biotechnology-derived

1022

pharmaceuticals: Is 6 months still appropriate? Regul Toxicol Pharmacol, 50(1): 2-22. Page 40 of 46

ACCEPTED MANUSCRIPT 1023 1024

Davies, A., et al. (2014). "Pharmacokinetics and safety of subcutaneous rituximab in

1025

Follicular lymphoma (SABRINA): stage 1 analysis of a randomised phase 3 study."

1026

Lancet Oncol 15(3): 343-352.

RI PT

1027 1028

EMA EPAR on blinatumomab from 7 December 2015.

1029

http://www.ema.europa.eu/ema/index.jsp?curl=pages/medicines/human/medicines/003731/human_me

1030

d_001921.jsp&mid=WC0b01ac058001d124.

SC

1031

FDA Guidance for Industry: Immunogenicity Assessment for Therapeutic Protein

1033

Products (2014).

M AN U

1032

1034 1035

FDA: Approval of Blincyto to treat a rare form of acute lymphoblastic leukemia, 3 December 2014.

1036

http://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm425549.htm.

TE D

1037

Friedrich M et al (2012). Regression of human prostate cancer xenografts in mice by

1039

MT112/BAY2010112, a novel PSMA/CD3-bispecific BiTE® antibody that cross reacts with non-

1040

human primate antigens. Mol Cancer Ther 11(12): 2664-2673.

1041

Gabriel R, von Kalle C, Schmidt Manfred (2015). Mapping the precision of genome editing. Nature

1042

Biotech. 33(2): 150

AC C

1043

EP

1038

1044

Han et al (2013). CYP3A-mediated drug-drug interaction potential and excretion of brentuximab

1045

vedotin, an antibody-drug conjugate, in patients with CD30-positive hematologic malignancies. J Clin

1046

Pharm 53(8), 866-877.

1047

Page 41 of 46

ACCEPTED MANUSCRIPT 1048

Hattori H, Kato M, Tamanaka M, Aoki T, Furuhama K, Manabe S (2007). Development of a novel

1049

mouse anaphylaxis model produced by intermittent intravenous injections of ovalbumin without

1050

adjuvant. J Toxicol Pathol 20:237-244.

1051 Hengel et al (2014). Measurements of in vivo drug load distribution of cysteine-linked antibody-drug

1053

conjugates using microscale liquid chromatography mass spectrometry. Anal Chem 86, 3420-3425.

RI PT

1052

1054

Heyen et al (2014). Characterization, Biomarkers, and Reversibility of a Monoclonal Antibody-

1056

induced Immune Complex Disease in Cynomolgus Monkeys (Macaca fascicularis). Toxicologic

1057

Pathology, 1-9.

SC

1055

M AN U

1058 1059

Hua, F., et al. (2014). "Anti-IL21 receptor monoclonal antibody (ATR-107): Safety,

1060

pharmacokinetics, and pharmacodynamic evaluation in healthy volunteers: a phase I,

1061

first-in-human study." J Clin Pharmacol 54(1): 14-22.

TE D

1062

International Conference on Harmonisation; addendum to International Conference on Harmonisation

1064

Guidance on S6 Preclinical Safety Evaluation of Biotechnology-Derived Pharmaceuticals;

1065

availability. Notice. Fed Regist. 2012; 77(97):29665–29666.

EP

1063

1066

Ivens IA, Achanzar W, Baumann A, Brändli-Baiocco A, Cavagnaro J, Dempster M, et al. (2015).

1068

PEGylated Biopharmaceuticals: Current Experience and Considerations for Non-clinical

1069

Development. Toxicol Pathol 43: 959-983.

1070

AC C

1067

1071

Ivens IA, Baumann A, McDonald TA, Humphries TJ, Michaels LA, Mathew P (2013). PEGylated

1072

therapeutic proteins for haemophilia treatment: a review for haemophilia caregivers. Haemophilia.

1073

19(1):1-10.

1074 1075

Jackisch, C., et al. (2015). "Subcutaneous versus intravenous formulation of Page 42 of 46

ACCEPTED MANUSCRIPT 1076

trastuzumab for HER2 positive early breast cancer: updated results from the phase III

1077

HannaH study." Ann Oncol 26(2): 320-325.

1078 Jevsevar S, Kunstelj M, Porekar VG (2010). PEGylation of therapeutic proteins. Biotechnol J.

1080

5(1):113-28.

RI PT

1079

1081

Karle A, Spindeldreher S, Kolbinger F (2016) Secukinumab, a novel anti-IL-17A antibody, shows

1083

low immunogenicity potential in human in vitro assays comparable to other marketed biotherapeutics

1084

with low clinical immunogenicity. MAbs 8:536-550.

1085

SC

1082

Kerlin R, et al. (2016). Scientific and Regulatory Policy Committee: Recommended (“Best”) Practices

1087

for Determining, Communicating, and Using Adverse Effect Data from Nonclinical Studies. Toxicol

1088

Pathol 44(2), 147-162.

M AN U

1086

1089

Lyon RP et al (2014). Self-hydrolyzing maleimides improve the stability and pharmacological

1091

properties of antibody-drug conjugates. Nature Biotech 32, 1059-1062.

TE D

1090

1092

Mehvar R (2000). Modulation of the pharmacokinetics and pharmacodynamics of proteins by

1094

polyethylene glycol conjugation. J Pharm Sci. 3(1):125-136.

EP

1093

AC C

1095 1096

Mihara K, Smit MJ, Krajnc-Franken M, Gossen J, Rooseboom M, Dokter W (2005). Human CXCR2

1097

(hCXCR2) takes over functionalities of its murine homolog in hCXCR2 knockin mice. Eur J Immunol

1098

35(9):2573-2582.

1099 1100

Nagorsen, D et al (2012). Blinatumomab: A historical perspective. Pharmacol Therap. 136(3): 334–

1101

342.

1102

Page 43 of 46

ACCEPTED MANUSCRIPT 1103

Nakaoka R, Tabata Y, Yamaoka T, Ikada Y (1997). Prolongation of the serum half-life period of

1104

superoxide dismutase by poly (ethylene glycol) modification. J Control Release. 46(3):253-61.

1105 Norelli M, Casucci M, Camisa B et al (2015). Early and self-limiting cytokine release syndrome by

1107

CD44v6 CAR-T cells in human hematochimeric NSG mice. Annual Meeting of the European Society

1108

for Gene and Cell Therapy. Helsinki, Finland. Abstract OR003.

1109 1110

Offenberg et al. , Haemophilia (2015), 21 (Suppl. S2),p. 27.

SC

1111

RI PT

1106

Ogata, A., et al. (2014). "Phase III study of the efficacy and safety of subcutaneous versus

1113

intravenous tocilizumab monotherapy in patients with rheumatoid arthritis." Arthritis Care Res

1114

(Hoboken) 66(3): 344-354.

1115

M AN U

1112

Okeley NM et al (2010). Intracellular activation of SGN-35, a potent anti-CD30 antibody-drug

1117

conjugate. Clin Can Res 16(3), 888-897.

TE D

1116

1118

Palazzi X et al., (2015). Characterizing “Adversity” of Pathology Findings in Nonclinical Toxicity

1120

Studies: Results from the 4th ESTP International Expert Workshop. Toxicol Pathol ??, 1-15.

EP

1119

1121

Parenky, A., et al. (2014). "New FDA draft guidance on immunogenicity." AAPS J 16(3): 499-

1123

503.

1124

AC C

1122

1125

Piccand M, Bessa J, Schick E, Senn C, Bourquin C, Richter WF (2016). Neonatal immune tolerance

1126

induction to allow long-term studies with an immunogenic therapeutic monoclonal antibody in mice.

1127

AAPS J. 18:354-361.

1128 1129

Ponce, R., et al. (2009). "Immunogenicity of biologically-derived therapeutics: assessment

1130

and interpretation of nonclinical safety studies." Regul Toxicol Pharmacol 54(2): 164-182. Page 44 of 46

ACCEPTED MANUSCRIPT 1131 1132

Richter, W. F. and B. Jacobsen (2014). "Subcutaneous absorption of biotherapeutics:

1133

knowns and unknowns." Drug Metab Dispos 42(11): 1881-1889.

1134 Rombach-Riegraf V, Karle AC, Wolf B, Sorde L, Koepke S, et al (2014). Aggregation of human

1136

recombinant monoclonal antibodies influences the capacity of dendritic cells to stimulate adaptive T-

1137

cell responses in vitro. Plos One 9: e86322.

RI PT

1135

1138

Rudmann DG, Alston JT, Hanson JC, Heidel S (2013). High molecular weight polyethylene glycol

1140

cellular distribution and PEG-associated cytoplasmic vacuolation is molecular weight dependent and

1141

does not require conjugation to proteins. Toxicol Pathol. 41(7):970-983.

M AN U

SC

1139

1142

Sakurai T, Takai R, Bürgin H, Shioda A, Sakamoto Y, Amano J, Grimm HP, Richter WF, Higuchi Y,

1144

Chiba S, Kawamura A, Suzuki M, Müller L (2013). The effects of interleukin-6 signal blockade on

1145

immune system, reproductive and skeletal development in juvenile mice. Birth Defects Res B Dev

1146

Reprod Toxicol. 98:170-182.

1147

TE D

1143

Schiff, M. (2013). "Subcutaneous abatacept for the treatment of rheumatoid arthritis."

1149

Rheumatology (Oxford) 52(6): 986-997.

AC C

1150

EP

1148

1151

Shida, Y et al (2014). The pharmacokinetics and safety profiles of belimumab after single

1152

subcutaneous and intravenous doses in healthy Japanese volunteers. J Clin Pharm Ther 39(1): 97-101.

1153 1154

Thway TM, Magana I, Bautista A, Jawa V, Gu W and Ma M (2013) Impact of Anti-Drug Antibodies

1155

in Preclinical Pharmacokinetic Assessment. AAPS Journal 15: 856-863.

1156

Page 45 of 46

ACCEPTED MANUSCRIPT 1157

Topp MS et al (2015). Safety and activity of blinatumomab for adult patients with relapsed or

1158

refractory B-precursor acute lymphoblastic leukaemia: a multicentre, single-arm, phase 2 study.

1159

Lancet Oncol 16: 57-66.

1160 Turecek et al. (2016). PEGylation of Biopharmaceuticals: A Review of Chemistry and Nonclinical

1162

Safety Information of approved drugs. Journal of Pharmaceutical Sciences 105, 460-475.

RI PT

1161

1163

Webster R, Elliott V, Kevin Park B, Walker D, Hankin M, Taupin P (2009). PEG and PEG conjugates

1165

toxicity: towards an understanding of the toxicity of PEG and its relevance to PEGylated biologicals.

1166

In: Veronese FM, editor. PEGylated protein drugs: basic science and clinical applications.

1167

Switzerland: Birkhäuser Basel:127-146.

M AN U

1168

SC

1164

1169

Yamaoka T, Tabata Y, Ikada Y (1995). Fate of water-soluble polymers administered via different

1170

routes. J Pharm Sci. 84(3):349-354.

TE D

1171

Yang BB, Lum PK, Hayashi MM, Roskos LK (2004). Polyethylene glycol modification of filgrastim

1173

results in decreased renal clearance of the protein in rats. J Pharm Sci. 93(5):1367-73.

AC C

1174

EP

1172

Page 46 of 46