View
2
Download
0
Category
Preview:
Citation preview
To view the webinar, click here
Sponsored byhttp://www.catalent.com/
Developing site - Specifically Modified ADCs:
Using a Chemoenzymatic Approach
A special article collection companion to the webinar broadcast 23rd March 2016.
Table of Contents
IntroductionDeveloping Site-Specifically Modified ADCs Using a Chemoenzymatic Approach
pages 1-2Antibody-drug conjugates: current status and future directionsDrug Discovery Today, Volume 19, Number 7, July 2014, pages 869-881Heidi L. Perez, Pina M. Cardarelli, Shrikant Deshpande, Sanjeev Gangwar, Gretchen M. Schroeder, Gregory D. Vite, Robert M. Borzilleri pages 3-15
An emerging playbook for antibody–drug conjugates: lessons from the laboratory and clinic suggest a strategy for improving efficacy and safetyCurrent Opinion in Chemical Biology, Volume 28, October 2015, Pages 174-180Penelope M Drake, David Rabuka pages 16-22
Exploring the effects of linker composition on site-specifically modified antibody-drug conjugatesEuropean Journal of Medicinal Chemistry, Volume 88, 17 December 2014, Pages 3-9Aaron E. Albers, Albert W. Garofalo, Penelope M. Drake, Romas Kudirka, Gregory W. de Hart, Robyn M. Barfield, Jeanne Baker, Stefanie Banas, David Rabuka pages 24-30
An FDA oncology analysis of antibody-drug conjugatesRegulatory Toxicology and Pharmacology, Volume 71, Issue 3, April 2015, Pages 444-452Haleh Saber, John K. Leighton pages 31-39
Introduction
There are relatively few occasions in science that define a generation. Of course, the elucidation of the 3 dimensional structure of DNA was one. For my generation, however, I would suggest that a similarly powerful one was the discovery of monoclonal antibodies by Cesar Milstein at Mill Hill in London. Tools became available for the unambiguous estimation and examination of proteins. It also became clear that monoclonal antibodies could represent a potential class of therapeutic agents. Initially, despite the promise, the translation of monoclonal antibodies into useful medicines was somewhat underwhelming. There were many reasons for the difficulty of translation, some would be related to the poor choice of target, pharmacoki-netic and pharmacodynamics issues and issues related to immunogenicity. More recently, advances in protein engineering have addressed and, to a great extent, solved these issues. The number of monoclonal antibody therapies on the market and in devel-opment bears testament to the concept of using such molecules as therapeutics and the ingenuity of biologists and protein engineers.
One of the hottest topics in monoclonal antibody-based therapeutics at the moment is that of antibody-drug conjugates (ADCs). This approach is particularly attractive to the oncology therapeutic area and relies upon tethering a cytotoxic drug (generally) to a monoclonal antibody that targets the drug to the tumour cell and, upon internalisation, releases its toxic payload to the cell. The advantage of the approach is that there is a highly-efficient delivery of a lethal level of drug to a tumour. Such concentrations may be impossible to achieve in the patient because of cost and toxicity of systemic administration. In this eBook we have included articles to highlight some of the most up-to-date developments in ADC technology.
The first article in this supplement is by Haleh Saber and John K. Leighton of the US Food and Drug Admin-istration, Center for Drug Evaluation and Research, Office of Hematology and Oncology Products, 10903 New Hampshire Ave, Silver Spring, MD 20903,United States, entitled “An FDA oncology analysis of anti-body-drug conjugates”. The article deals with approaches that have been adopted as preclinical surrogates for first-in-human dose estimation and as part of a preclinical safety package. Analysis of the various approaches has allowed the authors to establish appropriate rules for the estimation of appropriate first-in-human doses and dose escalation studies to aid with Phase 1 clinical trial design.
The second article, from Penelope M Drake and David Rabuka of Catalent Pharma Solutions, Emeryville, CA 94608, USA, entitled “An emerging playbook for antibody–drug conjugates: lessons from the laboratory and clinic suggest a strategy for improving efficacy and safety”. The authors highlight the prominent posi-tion occupied by ADCs in the oncology drug portfolios of Pharmaceutical companies. They suggest that, the field has become somewhat bifurcated in that most of the developing strategies for development rely upon existing technologies. Another, minority, approach is concerned with developments in conjugation chemistry and linker optimisation. Such an approach indicates that there may be some advantages of such approaches over the more conventional methods. The authors suggest that the success of two pioneer ADCs, KadcylaTM and AdcetrisTM, may be as a result of somewhat unique target antigens and development of new therapeutic agents may require the use of novel linker techniques.
In the third of our four articles, the authors present some concepts which, to some degree, support the conclusions of Drake and Rabuka in that they suggest that the most effective ADCs will require careful
1
antigen selection and conjugation optimisation. The article: “Antibody–drug conjugates: current status and future directions” by Heidi L. Perez, Pina M. Cardarelli, Shrikant Deshpande, Sanjeev Gangwar, Gretchen M. Schroeder, Gregory D. Vite and Robert M. Borzilleri all from Bristol Myers Squibb, is a tour de force of the history, current status, mode of action, optimisation and future directions for this exciting class of biophar-maceuticals.
Last, but by no means least, is the article from Aaron E. Albers, Albert W. Garofalo, Penelope M. Drake, Romas Kudirka, Gregory W. de Hart, Robyn M. Barfield, Jeanne Baker, Stefanie Banas, David Rabuka entitled “Exploring the effects of linker composition on site-specifically modified antibody-drug conjugates”. As the title suggests, this article deals with how modification of the antibody-drug linker may improve potency or reduce the propensity to interact with multidrug transporter proteins. Optimising these parameters will lead to a drug with a much improved profile of an ADC and can be used to engineer improved properties.
Steve Carney was born in Liverpool, England and studied Biochemistry at Liverpool University, obtaining a BSc.(Hons) and then read for a PhD on the Biochemistry and Pathology of Connective Tissue Diseases in Manchester University, in the Departments of Medical Biochemistry and Histopathology. On completion of his PhD he moved to the Kennedy Institute of Rheumatology, London, where he worked with Professor Helen Muir FRS and Professor Tim Hardingham, on the biochemistry of experimental Osteoarthritis. He joined Eli Lilly and Co. and held a number of positions in Biology R&D, initially in the Connective Tissue Department, but latterly in the Neuroscience Department. He left Lilly to take up his present position as Managing Editor, Drug Discovery Today, at Elsevier. Currently, he also holds an honorary lectureship in Drug Discovery at the University of Surrey, UK. He has authored over 50 articles in peer-reviewed journals, written several book chapters and has held a number of patents.
2
Reviews�KEYNOTEREVIEW
Drug Discovery Today � Volume 19, Number 7 � July 2014 REVIEWS
Recent advances in the discovery and development of antibody–drug conjugateshave led to FDA approvals and a rich clinical pipeline of promising
new cancer therapies.
Antibody–drug conjugates: currentstatus and future directions
Heidi L. Perez1, Pina M. Cardarelli2, Shrikant Deshpande2,Sanjeev Gangwar2, Gretchen M. Schroeder1, Gregory D. Vite1
and Robert M. Borzilleri1
1Bristol-Myers Squibb Research & Development, Princeton, NJ 08543, USA2Bristol-Myers Squibb Research & Development, Redwood City, CA 94063, USA
Antibody–drug conjugates (ADCs) aim to take advantage of the specificity of
monoclonal antibodies (mAbs) to deliver potent cytotoxic drugs selectively
to antigen-expressing tumor cells. Despite the simple concept, various
parameters must be considered when designing optimal ADCs, such as
selection of the appropriate antigen target and conjugation method. Each
component of the ADC (the antibody, linker and drug) must also be
optimized to fully realize the goal of a targeted therapy with improved
efficacy and tolerability. Advancements over the past several decades have
led to a new generation of ADCs comprising non-immunogenic mAbs,
linkers with balanced stability and highly potent cytotoxic agents.
Although challenges remain, recent clinical success has generated intense
interest in this therapeutic class.
IntroductionThe past decade has seen significant advances in new cancer treatments through the development
of highly selective small molecules that target a specific genetic abnormality responsible for the
disease [1,2]. Although this approach has seen great success in application to malignancies with a
single, well-defined oncolytic driver, resistance is commonly observed in more complex cancer
settings [3,4]. Traditional cytotoxic agents are another approach to treating cancer; however,
unlike target-specific approaches, they suffer from adverse effects stemming from nonspecific
killing of both healthy and cancer cells. A strategy that combines the powerful cell-killing ability
of potent cytotoxic agents with target specificity would represent a potentially new paradigm in
cancer treatment. ADCs are such an approach, wherein the antibody component provides
specificity for a tumor target antigen and the drug confers the cytotoxicity. Here, we present
key considerations for the development of effective ADCs and discuss recent progress in ADC
technology for application to the next wave of cancer therapeutics. Advances in other modalities
of antibody-mediated targeting, such as immunotoxins, immunoliposomes and radionuclide
conjugates, have been extensively reviewed elsewhere [5,6].
Corresponding author:. Borzilleri, R.M. (robert.borzilleri@bms.com)
1359-6446/06/$ - see front matter � 2013 Published by Elsevier Ltd. http://dx.doi.org/10.1016/j.drudis.2013.11.004 www.drugdiscoverytoday.com 869 3
Review
s�K
EYNOTEREVIEW
REVIEWS Drug Discovery Today � Volume 19, Number 7 � July 2014
Historical perspectiveThe origin of ADCs can be traced back over a century to the
German physician and scientist Paul Ehrlich, who proposed the
concept of selectively delivering a cytotoxic drug to a tumor via a
targeting agent (Fig. 1) [7,8]. Ehrlich coined the term ‘magic bullet’
to describe his vision, similar to the descriptors ‘warhead’ or
‘payload’ commonly used for the drug component of current
ADCs. Nearly 50 years later, Ehrlich’s concept of targeted therapy
was first exemplified when methotrexate (MTX) was linked to an
antibody targeting leukemia cells [9]. Early research relied on
available targeting agents, such as polyclonal antibodies, to enable
preclinical efficacy studies in animal models with both noncova-
lent-linked ADCs and later covalently linked ADCs [10–12]. In
1975, the landmark development of mouse mAbs using hybri-
doma technology by Kohler and Milstein greatly advanced the
field of ADCs [13]. The first human clinical trial followed less than
a decade later, with the antimitotic vinca alkaloid vindesine as the
cytotoxic payload [14]. Further advances in antibody engineering
enabled the production of humanized mAbs with reduced immu-
nogenicity in humans compared with the murine mAbs used for
early ADCs [15].
First-generation ADCs typically used clinically approved drugs
with well-established mechanisms of action (MOAs), such as anti-
metabolites (MTX and 5-fluorouracil), DNA crosslinkers (mitomy-
cin) and antimicrotubule agents (vinblastine) [16]. In addition to
1913
19501910 19401930 19601920
1958 19
Paul Ehrlich describedthe concept of a‘magic bullet’ anddrug targeting (i.e. a‘heptophore’ candeliver a ‘toxophore’selectively to a tumor)
MTX linked to anantibody directed
toward leukemia cells
ADCs proposedimmunoradioactiv
agent disclose
Nonlin
tested
NH2H2N
CO2H
CO2H
N
N
O
HN
N
N
N
FIGURE 1
Antibody–drug conjugate (ADC) timeline. Abbreviations: mAbs, monoclonal antibo
870 www.drugdiscoverytoday.com
the immunogenicity issues observed with murine mAbs, these
early attempts were met with limited success for several reasons,
including low drug potency, high antigen expression on normal
cells and instability of the linker that attached the drug to the mAb
[17]. Lessons learned from these initial failures led to a new
generation of ADCs, several of which entered and later failed
human clinical trials. For example, doxorubicin conjugate 1
(BR96-DOX) was designed using a bifunctional linker, wherein
the drug was appended via a hydrazone, and a maleimide enabled
conjugation to the BR96 antibody via cysteine residues (Fig. 2)
[18]. Although curative efficacy was observed in human tumor
xenograft models, the relatively low potency of doxorubicin
necessitated high drug:antibody ratios (DARs, eight per antibody)
and high doses of the ADC to achieve preclinical activity. In
clinical trials, significant toxicity was observed due to nonspecific
cleavage of the relatively labile hydrazone linker and expression of
the antigen target in normal tissue [19].
Further advancements, including higher drug potency and care-
fully selected targets, ultimately led to the first ADC to gain US Food
and Drug Administration (FDA) approval in 2000 (Mylotarg1,
gemtuzumab ozogamicin, 2) [20,21]. Despite initially encouraging
clinical results, Mylotarg1 was withdrawn from the market a
decade later owing to a lack of improvement in overall survival.
In 2011, following an accelerated approval process, a second ADC
(Adcetris1, brentuximab vedotin, 3) gained marketing approval
2000199019801970 2010
67
1972
1975
1983
1991
;ed
covalentked ADCin animal
models
Production ofmAbs usinghybridoma-based technology
Covalent linkedADCs tested inanimal models
Clinical trialsw/ ADC
vindesine-αCEA
Immunogenicityof mouse mAbs aserious limitationin developmentof ADCs
ADC w/ highlypotent
cytotoxincalicheamicin
First FDAapproved ADC
(Mylotarg®)
Mylotarg®
clinical failuresdue to MDR
Adcetris®
approved
Mylotarg®
withdrawnfrom market
Curative efficacy inhuman tumor
xenograft models w/BR96-DOX ADC
Kadcyla®
approved
2013
2001 2011
1993
1988
HumanizedmAbs reported
Drug Discovery Today
dies; MDR, multidrug resistance; MTX, methotrexate.
4
Drug Discovery Today � Volume 19, Number 7 � July 2014 REVIEWS
OMe
O
O
OH
OH
NOH
O
O
OHNH2
OH
3Adcetris®
(anti-CD30 auristatin conjugate)
OO
HN
NNH
O
SS
OHO
H
OONHO O
OO
S
OH
O
I
OOMe
OMe
NMeO
OOHO
MeO
O
OH
HN
O
N
O
O
S
2Mylotarg®
(anti-CD33 calicheamicin conjugate)
NHCO2Me
O N
OHN
ON
O
OMe O
N
OMe
HN
O
OH
NH
OHN
NH
NH2O
ONH
O
N OO
S
1BR96-DOX
(anti-BR96 doxorubicin conjugate)
H
O
NH
O
MeO OH
MeO NCl O O
O
O
N
N
O
OO
HNO
S
4Kadcyla®
(anti-Her2 maytansine conjugate)
Drug Discovery Today
FIGURE 2
Structures of antibody–drug conjugates (ADCs).
Reviews�KEYNOTEREVIEW
from the FDA for the treatment of Hodgkin’s (HL) and anaplastic
large-cell lymphomas (ALCL) [22,23]. Most recently, Kadcyla1 (ado-
trastuzumab emtansine, T-DM1, 4), which combines the huma-
nized antibody trastuzumab with a potent antimicrotubule cyto-
toxic agent using a highly stable linker, was approved for the
treatment of patients with Her2-positive breast cancer [24,25]. With
nearly 30 additional ADCs currently in clinical development, the
potential of this new therapeutic class might finally be coming to
fruition [26].
ADC designAlthough simple in concept, the success of a given ADC depends
on careful optimization of each ADC building block: antibody,
drug and linker (Fig. 3) [27]. The chosen antibody should target a
well-characterized antigen with high expression at the tumor site
and low expression on normal tissue to maximize the efficacy of
the ADC while limiting toxicity. Bifunctional linkers with attach-
ment sites for both the antibody and drug are used to join the two
components. With respect to the mAb, existing linker attachment
strategies typically rely on the modification of solvent-accessible
cysteine or lysine residues on the antibody, resulting in hetero-
geneous ADC populations with variable DARs. Given that low drug
loading reduces potency and high drug loading can negatively
impact pharmacokinetics (PK), DARs can have a significant impact
on ADC efficacy. In addition, the linker must remain stable in
systemic circulation to minimize adverse effects, yet rapidly cleave
after the ADC finds its intended target antigen. Upon antigen
recognition and binding, the resulting ADC receptor complex is
internalized through receptor-mediated endocytosis [28]. Once
inside the cell, the drug is released through one of several mechan-
isms, such as hydrolysis or enzymatic cleavage of the linker or via
degradation of the antibody. Typically, the unconjugated drug
should demonstrate high potency, ideally in the picomolar range,
to enable efficient cell killing upon release from the ADC.
Target antigens and antibody selectionAlthough the basic premise that a successful ADC should target a
well-internalized antigen with low normal tissue expression and
high expression on tumors remains true, the field is evolving to
refine these parameters. For example, antigen expression on nor-
mal tissues can be tolerated if expression on vital organs is minimal
or absent. The FDA approval of Kadcyla1 for Her2-positive breast
cancer highlights this point since Her2/neu, a member of the
epidermal growth factor receptor (EGFR) family, is not only
expressed in breast tissue, but also in the skin, heart and on
epithelial cells in the gastrointestinal, respiratory, reproductive
and urinary tracts [29]. In addition, prostate-specific membrane
antigen (PSMA) is an ADC target expressed both on prostate cancer
cells as well as normal prostate and endothelial tissue [30]. Given
that most patients with prostate cancer undergo surgery to remove
their prostate, selective expression relative to normal prostate cells
might not be crucial in this setting. Furthermore, apical expression
of PSMA on the kidney and gastrointestinal tract might prevent
the ADC from accessing these tissues. Other possible exceptions
include hematological malignancies in which normal target tis-
sues are able to regenerate, supported by the case of rituximab
where depletion of normal B cells was not a major safety issue in
patients [31]. Accessibility of the ADC to the target antigen is also
www.drugdiscoverytoday.com 871 5
REVIEWS Drug Discovery Today � Volume 19, Number 7 � July 2014
Attachment site Linker Cytotoxin
Antibody
- Targets a well-characterizedantigen with high tumorexpression and limited normal
tissue expression- Maintains binding, stability,
internalization, PK, etc. when conjugated to a cytotoxin- Minimal nonspecific binding
- Typically through nonspecificmodification of cysteine or
lysine residues on the antibody- Mixture of conjugates with
variable drug:antibody ratios- Site-selective conjugation
technologies can produce more homogeneous ADCs
- Cleavable or noncleavable- Stable in circulation- Selective intracellular
release of drug (e.g. viaenzymatic cleavage or
antibody degradation)
- Highly potent- Non-immunogenic- Amenable to modifications for linker attachment- Defined mechanism of action
Drug Discovery Today
FIGURE 3
Key components of an antibody–drug conjugate (ADC). Abbreviation: PK, pharmacokinetics.
TABLE 1
Target antigens for ADCs in preclinical and clinical development
Indication Targets
NHL CD19, CD20, CD21, CD22, CD37, CD70, CD72,
CD79a/b and CD180
HL CD30
AML CD33
MM CD56, CD74, CD138 and endothelin B receptor
Lung CD56, CD326, CRIPTO, FAP, mesothelin, GD2,
5T4 and alpha v beta6
CRC CD74, CD174, CD227 (MUC-1), CD326 (Epcam),
CRIPTO, FAP and ED-B
Pancreatic CD74, CD227 (MUC-1), nectin-4 (ASG-22ME)and alpha v beta6
Breast CD174, GPNMB, CRIPTO, nectin-4 (ASG-22ME)
and LIV1A
Ovarian MUC16 (CA125), TIM-1 (CDX-014) and mesothelin
Melanoma GD2, GPNMB, ED-B, PMEL 17 and endothelinB receptor
Prostate PSMA, STEAP-1 and TENB2
Renal CAIX and TIM-1 (CDX-014)
Mesothelioma Mesothelin
Abbreviations: AML, acute myeloid leukemia; MM, multiple myeloma; CRC, colorectal
cancer.
Review
s�K
EYNOTEREVIEW
an important consideration. In addition to high interstitial pres-
sure in the tumor, endothelial, stromal and epithelial barriers can
limit ADC uptake, resulting in only a small percentage of the
injected dose reaching the intended tumor target [32]. From a
biology perspective, the design of an effective ADC relies on
selection of an appropriate target antigen, taking into account
tumor expression levels, rates of antigen internalization and anti-
body Fc format.
Tumor types
Table 1 highlights the broad range of hematologic and solid tumor
indications targeted by ADCs currently in preclinical or clinical
development. Several of these tumor-associated antigens exhibit
remarkable specificity, such as CD30 for HL and MUC16 for
ovarian cancer. Other antigens, such as CD74, are expressed in
multiple tumor types.
Antigen expression
In general, optimal ADC targets are homogeneously and selec-
tively expressed at high density on the surface of tumor cells.
Homogenous tumor expression, although preferred, is likely not
an absolute requirement owing to the ability of some ADCs to
induce bystander killing. Under these circumstances, a membrane-
permeable free drug liberated after intracellular cleavage of the
linker can efflux from the cell and enter neighboring cells to
facilitate cell death [33]. Most advanced ADCs in the clinic target
hematological indications, in part due to the largely homogeneous
expression of antigen in liquid tumors, despite frequently low
receptor densities. Although the treatment of solid tumors with
heterogeneous antigen expression might benefit from bystander
killing, the potential to harm normal cells could contribute to
systemic toxicity.
Current experimental evidence generally suggests that tumor
antigen density (expression level) does not directly correlate
with ADC efficacy [34]. When patient samples are accessible,
the number of receptors per cell can be quantified using flow
872 www.drugdiscoverytoday.com
cytometry, immunohistochemistry (IHC) or radiolabeled satura-
tion-binding studies to assess the relation between target expres-
sion and efficacy [35]. In non-Hodgkin’s lymphoma (NHL) cell
lines, high CD79b expression was found to be a prerequisite for in
vitro response to an anti-CD79b auristatin conjugate (RG-7596,
Roche-Genentech); however, a wide range of sensitivities were
observed, indicating that a minimal expression threshold exists
[36]. Likewise, melanoma cells lines with receptor densities vary-
6
Drug Discovery Today � Volume 19, Number 7 � July 2014 REVIEWS
Reviews�KEYNOTEREVIEW
ing from 20,000 to 280,000 binding sites per cell were sensitive to
an anti-p97-auristatin conjugate [37]. This threshold level varies
among different targets based on the unique factors of the anti-
gen, such as rate of internalization and binding affinity for the
ADC. For example, approximately 5,000–10,000 copies of CD33,
the antigen target for Mylotarg1, are expressed per cell [38]. As
with Mylotarg1, no significant correlation was observed between
the activity of a preclinical anti-CD33 pyrrolobenzodiazepine
conjugate (SG-CD33A, Seattle Genetics) and CD33 levels in a
panel of acute myeloid leukemia (AML) cell lines [39]. An anti-
PSMA auristatin conjugate (PSMA ADC, Progenics/Seattle Genet-
ics) demonstrated potent in vitro cytotoxicity versus cells expres-
sing >105 PSMA molecules per cell, with 104 receptors per cell
serving as a threshold level [40]. For some tumor antigens, how-
ever, a relatively proportional relation between efficacy and
receptor expression level has been observed. In the case of an
anti-endothelin B receptor (EDNBR) auristatin conjugate,
improved efficacy against human melanoma cells lines and xeno-
graft tumor models generally correlated with increasing EDNBR
expression (1,500–30,000 copies per cell) [41].
Antigen internalization
Ideally, once an ADC binds to a tumor-associated target, the ADC–
antigen complex is internalized in a rapid and efficient manner.
Although poorly understood, various factors are likely to influence
the rate of internalization, such as the epitope on the chosen target
antigen bound by the ADC, the affinity of the ADC–antigen
interaction and the intracellular trafficking pattern of the ADC
complex [42–44]. For example, anti-Her2 antibodies that bind
distinct epitopes on Her2 have been shown to impact downstream
trafficking and lysosomal accumulation differentially, despite
binding to the same cell surface receptor [45]. Several ADCs,
including Adcetris1, have been shown to internalize with rates
similar to or greater than the corresponding unconjugated anti-
bodies [46–48,37]. Certain antigens mediate exceptionally rapid
accumulation of ADCs inside cells. When bound to ligand-acti-
vated EGFR, Her2 monomer is internalized at a rate up to 100-fold
greater than carcinoembryonic antigen (CEA) [49,50]. Likewise,
the catabolic rate of antibodies targeting CD74 is approximately
100 times faster than other antibodies that are considered to
rapidly internalize, such as anti-CD19 and anti-CD22 [51]. The
preclinical data for milatuzumab-DOX (Immu-110), an anti-CD74
doxorubicin conjugate in early clinical trials, suggest this agent is
equipotent to ADCs comprising more potent drug payloads that
target slower internalizing antigens [52].
Alternative approaches have been explored in which antigen
internalization is not required for efficient cell killing. The extra-
domain B (ED-B) of fibronectin is a marker of angiogenesis
undetectable in healthy tissue, but highly expressed around
tumor blood vessels [53]. Anti-ED-B antibodies have been shown
to localize to the subendothelial extracellular matrix of tumor
vasculature. Conjugation of these antibodies with a photosen-
sitizer has led to agents that selectively disrupt tumor blood
vessels upon irradiation, resulting in curative efficacy in mouse
models [54].
Impact of format
The biological activity of an antibody can depend on the interac-
tion of its Fc portion with cells that express Fc receptors (FcRs).
Therefore, selection of the appropriate antibody format for an
ADC is an important consideration. Broad understanding of the
relation between antibody Fc format and ADC function is lacking
since species differences in immune systems complicate preclini-
cal studies. In one study, McDonagh et al. conjugated anti-CD70
antibody immunoglobin G (IgG) variants (IgG1, IgG2 and IgG4) to
an auristatin (ADC toxin monomethyl auristatin F; MMAF) to
determine the effect of format on ADC function [55]. In addition,
the Fc regions of IgG1 and IgG4 were mutated (IgG1v1 and
IgG4v3) to examine the influence of IgG receptor (FcgR) binding.
Although all the ADCs demonstrated potent in vitro cytotoxicity
and were well tolerated in mice, the engineered IgGv1-MMAF
conjugate displayed improved antitumor activity and increased
exposure, which correlated with a superior therapeutic index
compared to the parent IgG1 conjugate.
In the absence of definitive guidelines for selecting an optimal
antibody format, all human IgG isotypes, except for IgG3, are
currently used for ADCs in clinical trials. IgG1, the most com-
monly used format, can potentially engage secondary immune
functions, such as antibody-dependent cellular cytotoxicity
(ADCC) or complement-dependent cytotoxicity (CDC). These
inherent effector functions could prove beneficial by providing
additional antitumor activity, as in the case of Kadcyla1, which
was shown to activate ADCC in preclinical models [56]. Adcetris1,
however, demonstrated minimal ADCC and no detectable CDC
despite its IgG1 format [57]. The absence of effector functions is
potentially advantageous as binding of an ADC to effector cells
could reduce tumor localization, hinder internalization and lead
to off-target toxicity [55]. Unlike IgG1, IgG2 and IgG4 typically
lack Fc-mediated effector functions. Mylotarg1 and inotuzumab
ozogamicin (CMC-544) exhibited no ADCC or CDC in preclinical
studies, consistent with their IgG4 format [58]. Overall, the con-
tribution of IgG effector functions to the efficacy, selectivity and
toxicity of ADCs is not yet well understood.
In addition to effector functions, ADCs often retain other
biological properties associated with their parent mAbs, such
as immunogenicity potential. Limited therapeutic efficacy of
early ADCs comprising murine mAbs prompted the develop-
ment of chimeric and humanized antibodies, which minimize
human immune response. Conversion of murine mAbs to
human IgGs also results in longer retention in systemic circula-
tion due to recognition by the human neonatal Fc receptor
(FcRn) and a greater ability to elicit ADCC [59]. Technologies
for the generation of fully human mAbs include the use of either
phage display or transgenic mouse platforms, in which a mouse
strain is engineered to produce human rather than mouse anti-
bodies [60].
Linker technology and stabilityThe identity and stability of a linker that covalently tethers the
antibody to the cytotoxic drug is crucial to the success of an ADC.
Sufficient linker stability is necessary to enable the conjugate to
circulate in the bloodstream for an extended period of time before
reaching the tumor site without prematurely releasing the free
drug and potentially damaging normal tissue. Once the ADC is
internalized within the tumor, the linker should be labile enough
to efficiently release the active free drug. Linker stability also
influences overall toxicity, PK properties and the therapeutic index
of an ADC. The lack of adequate therapeutic index for earlier
www.drugdiscoverytoday.com 873 7
REVIEWS Drug Discovery Today � Volume 19, Number 7 � July 2014
TABLE 2
Examples of ADC drug linkers
Cleavable linkers Noncleavable linkers
Valine-Citrulline (protease sensitive)
CytotoxinN-Maleimidomethylcyclohexane-1-carboxylate (MCC)
Cytotoxin
Hydrazone (acid-sensitive)
Cytotoxin
Maleimidocaproyl
Cytotoxin
Disulfide (glutathione-sensitive)
Cytotoxin Mercaptoacetamidocaproyl
Cytotoxin
Review
s�K
EYNOTEREVIEW
ADCs, such as BR96-DOX and Mylotarg1, has been attributed to
poor linker stability (Fig. 1) [19,61].
The two main classes of ADC drug linkers currently being
explored take advantage of different mechanisms for release of
the drug payload from the antibody (Table 2). The first is a
cleavable linker strategy, with three different types of release
mechanism within this class.
(i) Lysosomal protease sensitive linkers. This strategy utilizes
lysosomal proteases, such as cathepsin B (catB), that
recognize and cleave a dipeptide bond to release the free
drug from the conjugate [62]. Many ADCs in the clinic use a
valine–citrulline dipeptide linker, which was designed to
display an optimal balance between plasma stability and
intracellular protease cleavage [63]. This linker strategy was
successfully utilized by Seattle Genetics/Millennium in the
case of Adcetris1 [64].
(ii) Acid sensitive linkers. This class of linkers takes advantage of
the low pH in the lysosomal compartment to trigger
hydrolysis of an acid labile group within the linker, such
as a hydrazone, and release the drug payload. In preclinical
studies, hydrazone linker-based conjugates have shown
stability (t1/2) ranges from 2 to 3 days in mouse and human
plasma, which may not be optimal for an ADC [65].
Hydrazone linkers were used in Mylotarg1 (anti-CD33
calicheamicin conjugate) and recently in inotuzumab
ozogamicin (anti-CD22 calicheamicin conjugate) [66,67].
The withdrawal of Mylotarg1 from the market was attributed
to toxicities related to hydrazone linker instability, which
resulted in increased fatalities in patients treated with
Mylotarg1 plus chemotherapy as opposed to chemotherapy
alone [65]. Similarly, inotuzumab ozogamicin was recently
withdrawn from a phase III clinical trial owing to a lack of
improvement in overall survival.
874 www.drugdiscoverytoday.com
(iii) Glutathione sensitive linkers. This strategy exploits the
higher concentration of thiols, such as glutathione, inside
the cell relative to the bloodstream. Disulfide bonds within
the linker are relatively stable in circulation yet are reduced
by intracellular glutathione to release the free drug. To
further increase plasma stability, the disulfide bond can be
flanked with methyl groups that sterically hinder premature
cleavage in the bloodstream [68]. This class of linker has been
used in several clinical candidates, such as SAR3419 (anti-
CD19 maytansine conjugate), IMGN901 (anti-CD56 may-
tansine conjugate) and AVE9633 (anti-CD33 maytansine
conjugate) developed by ImmunoGen and its partners [67].
The second strategy is one that uses noncleavable linkers. This
approach depends on complete degradation of the antibody after
internalization of the ADC, resulting in release of the free drug
with the linker attached to an amino acid residue from the mAb. As
such, noncleavable linker strategies are best applied to payloads
that are capable of exerting their antitumor effect despite being
chemically modified. This type of strategy has been used success-
fully by Genentech/Immunogen with Kadcyla1 (trastuzumab-
MCC-DM1). The released modified payload (lysine-MCC-DM1)
demonstrated similar potency compared with DM1 alone,
although the charged lysine residue is likely to impair cell perme-
ability and hence abate the bystander killing observed with the
free drug [69]. One potential advantage of noncleavable linkers is
their greater stability in circulation compared with cleavable
linkers. However, no significant difference in terminal half-life
(t1/2) values was observed in the clinic between Kadcyla1 [24],
which contains a noncleavable linker, and Adcetris1, which
employs a cleavable linker [22].
Preclinically, linker strategies continue to evolve [70,71].
Additional tumor-associated proteases, such as legumain, have
been identified that release the ADC payload in nonlysosomal
8
Drug Discovery Today � Volume 19, Number 7 � July 2014 REVIEWS
Reviews�KEYNOTEREVIEW
compartments (i.e. the endosome) [72]. Other nonprotease
enzymes have recently been exploited for the selective cleavage
of b-glucuronidase and b-galactosidase sensitive linkers in the
lysosome [73,74]. Demonstrating expanded utility, these
approaches enable drug linkage via a phenol functional group
in addition to a more traditional basic amine residue.
Cytotoxic agentsPayload classes and MOAs
There are two main classes of ADC payloads undergoing clinical
evaluation. The first class comprises drugs that disrupt microtu-
bule assembly and play an important role in mitosis. This class
includes cytotoxics, such as dolastatin 10-based auristatin analogs
(3, Adcetris1) [64] and maytansinoids (4, Kadcyla1) [75]. The
second class of payloads consists of compounds that target DNA
structure and includes calicheamicin analogs, such as Mylotarg1
(2), that bind the minor groove of DNA causing DNA double-
strand cleavage [76]. Duocarmycin analogs (MDX-1203, 5) [77]
participate in a sequence-selective alkylation of adenine-N3 in the
minor groove of DNA to induce apoptotic cell death (Fig. 4).
One common feature among these cytotoxic agents is that they
demonstrate at least 100–1000-fold greater potency in in vitro
NH
NONN
OHN
HN
NH
HN
O
O
HN
NH2O
OO
5Duocarmycin analog
Cl
NHO
HN
ONHO
N
O
O
S
7α-Amanitin
NHO
HN
O
HN
O
NH
O
O
HN
HNO
HN
O
NH
H
OH2N
HOH
O
NH
SO OH
O
OO
O
S
O
FIGURE 4
Representative antibody–drug conjugate (ADC) payload structures.
proliferation assays against a broad range of tumor cell lines
compared with conventional chemotherapeutic agents, such as
paclitaxel and doxorubicin [78,79]. The high potency of these
alternative payloads is crucial since only an estimated 1–2% of the
administered ADC dose will ultimately reach the tumor site,
resulting in low intracellular drug concentrations [80]. Unlike
earlier ADCs that failed to make a meaningful impact in the clinic
owing to low drug potency and suboptimal delivery, newer, more
potent cytotoxic compounds are now the focus of preclinical
research. For example, pyrrolobenzodiazepine (PBD) dimers 6
covalently bind the minor groove of DNA, resulting in a lethal
interaction due to cross-linking of opposing strands of DNA [81].
a-Amanitin 7, a cyclic octapeptide found in several species of the
Amanita genus of mushrooms, strongly inhibits RNA polymerase
II, leading to inhibition of DNA transcription and cell death [82].
Tubulysins 8, similar to auristatins and maytansine, inhibit tubu-
lin polymerization to induce apoptosis [83–85].
Addressing drug resistance
In addition to potency, the sensitivity of cytotoxic agents to
multidrug resistance (MDR) mechanisms is a factor to consider
in selecting the optimal payload for an ADC. Cancer cells have the
ability to become resistant to multiple drugs via increased efflux of
NH
ON
O
O
S
O
OMe
N
N
O
H
MeO N
N
O
HO
OMe
N
HN
ON
O O
S
N
HN
O
CO2H
O
8Tubulysin analog
6PBD
ON
O
Drug Discovery Today
www.drugdiscoverytoday.com 875 9
REVIEWS Drug Discovery Today � Volume 19, Number 7 � July 2014
Cysteine specific
Unnatural aminoacid insertion
Enzyme-assistedligation
(a) (b)
Transglutaminase SortaseFormylglycine-generating enzyme(chemoenzymatic)
Lysine
Cysteine,S-S reduction
Drug Discovery Today
FIGURE 5
Random and site-specific conjugation strategies. Antibody–drug conjugate (ADC) products of random conjugation comprise chemically heterogeneous species
(a), whereas site-specific conjugation methods produce fairly homogeneous product profiles (b).
Review
s�K
EYNOTEREVIEW
the drug by either P-glycoprotein (Pgp) or other multidrug-resis-
tance proteins (e.g. MRP1 and MRP3) [86]. The sensitivity of
cytotoxic drugs to MDR mechanisms can be measured in vitro.
In the case of Mylotarg1, in vitro cytotoxicity assays in AML cell
lines indicated that Pgp expression altered the potency of the
calicheamicin payload and that drug potency could be restored
by adding known efflux transporter antagonists to inhibit Pgp and
MRP-1 proteins [87]. These results were relevant for patients with
AML as levels of Pgp expression in the clinic were found to
correlate directly with responders and nonresponders [88,89].
Another interesting example related to MDR mechanisms
involves AVE9633, which comprises an anti-CD33 antibody linked
through a disulfide bond to the maytansine analog DM4. In vitro
data clearly demonstrated that the cytotoxicity of AVE9633 and
the DM4 free drug were highly dependent on the expression level
of Pgp protein in myeloid cell lines [90]. As with the calicheamicin
payload of Mylotarg1, the potency of DM4 could be restored in
Pgp-overexpressing cell lines by adding known inhibitors of Pgp.
However, Pgp activity was not found to be a major mechanism of
resistance for the AVE9633 conjugate in cells from patients with
AML. Reasons for the lack of correlation are unclear; other
mechanisms such as microtubule alteration were proposed for
chemoresistance to AVE9633.
Conjugation strategiesFor most ADCs in clinical development, conjugation of the drug
payload to the antibody involves a controlled chemical reaction
with specific amino acid residues exposed on the surface of the
mAb. Given that this process results in a mixture of ADC species
with variable DARs and linkage sites, alternative conjugation
strategies aimed at minimizing heterogeneity have been devel-
oped. In the overall design of an ADC, selection of the appropriate
876 www.drugdiscoverytoday.com
drug-conjugation strategy significantly impacts efficacy, PK and
tolerability. As such, careful consideration of the various conjuga-
tion technologies for ADC generation is warranted (Fig. 5).
Chemical conjugation
In one type of chemical conjugation, a reactive moiety pendant to
the drug–linker is covalently joined to the antibody via an amino
acid residue side chain, commonly the e-amine of lysine. As
demonstrated with Mylotarg1, direct conjugation of lysine resi-
dues on gemtuzumab was achieved using an N-hydroxysuccini-
mide (NHS) ester appended to the drug–linker to form stable amide
bonds [91]. A two-step process can also be used in which surface
lysines on the antibody are first modified to introduce a reactive
group, such as a maleimide, and then conjugated to the drug–
linker containing an appropriate reactive handle (e.g. a thiol) [92].
Such a strategy was utilized in the case of Kadcyla1. Alternatively,
controlled reduction of existing disulfide bonds can liberate free
cysteine residues on the antibody, which then react with a mal-
eimide attached to the drug–linker. This approach, used in the
preparation of Adcetris1, takes advantage of the reducible disul-
fide bonds of IgG antibodies in which controlled conditions
enable reduction of only interchain disulfide bonds while intra-
chain disulfides remain unaffected, thus minimizing major struc-
tural disruptions to the antibody [19].
The random conjugation processes described above produce
heterogeneous mixtures of conjugated species with variable
DARs. Adding to the complexity, the site of conjugation could
be different for each ADC species containing even only one drug.
When lysines are used for conjugation, heterogeneity in overall
charge can impact solubility, stability and PK [93]. Therefore, the
clinical success of an ADC produced by random conjugation
depends on robust manufacturing processes that provide
the ability to monitor, control and purify the heterogeneous
10
Drug Discovery Today � Volume 19, Number 7 � July 2014 REVIEWS
Reviews�KEYNOTEREVIEW
products. Several organizations have developed expertise in this
area to overcome the process development and manufacturing
challenges associated with ADC commercialization [94].
Site-specific conjugation
Despite the success of Adcetris1 and Kadcyla1, considerable
enthusiasm for the next generation of ADCs has focused on the
development of homogeneous products derived via site-specific
conjugation. Currently, three strategies are at the forefront: inser-
tion of cysteine residues in the antibody sequence by mutation or
insertion, insertion of an unnatural amino acid with a bio-ortho-
gonal reactive handle, and enzymatic conjugation.
Building on early studies that explored the introduction of
surface cysteines on recombinant antibodies [95], several cysteine
engineered antibodies have been produced and tested for use in
site-specific attachment of cytotoxic drugs to yield homogeneous
ADCs [96]. Junutula et al. reported a class of THIOMAB-drug
conjugates (TDCs) prepared by taking advantage of: (i) phage
display techniques to identify ideal sites for mutation and pro-
duce antibodies with minimal aggregation issues, and (ii) meth-
ods to reduce and re-oxidize the antibody under mild conditions
to present only thiols of mutated cysteines for conjugation
[92,97]. Compared with a conventional, randomly conjugated
ADC, the analogous TDC displayed minimal heterogeneity with
similar in vivo activity, improved PK and a superior therapeutic
index. Moreover, McDonagh et al. engineered antibodies in
which interchain cysteines were replaced with serines to reduce
the number of potential conjugation sites, yielding ADCs with
defined DARs (two or four drugs per antibody) and attachment
sites [98]. Broad application of this approach to future ADCs will
depend on further studies to evaluate the effect of these mutations
on the overall stability and biological function of the engineered
antibody.
Encouraged by studies with cysteine engineered antibodies,
several investigators reasoned that the site and stoichiometry of
conjugation could be controlled by inserting unnatural amino
acids with orthogonal reactivity relative to the 20 natural amino
acids. Axup et al. genetically engineered an orthogonal amber
suppressor tRNA/aminoacyl-tRNA synthetase pair to insert site-
specifically p-acetylphenylalanine (pAcPhe) in recombinantly
expressed antibodies [99]. As a test case, pAcPhe was introduced
at one of several positions in the constant region of trastuzumab
(anti-Her2). These mutants were then conjugated to an alkoxy-
amine auristatin derivative via formation of a stable oxime bond.
The resulting chemically homogeneous ADCs demonstrated
improved PK compared with nonspecifically conjugated ADCs
and were highly efficacious in a Her2-positive human tumor
xenograft model. In addition to pAcPhe, other unnatural amino
acids are being explored through the use of appropriate tRNA–
aminoacyl-tRNA synthetase pairs [100]. Recently, in vitro transcrip-
tion and translation processes have also been developed and
optimized to insert unnatural amino acids in antibodies for site-
specific conjugation [101].
In addition to inserting unnatural amino acids into mAb
sequences, chemoenzymatic approaches have been explored to
generate bio-orthogonal reactive groups for selective conjugation.
Bertozzi and co-workers utilized formylglycine-generating enzyme
(FGE), which recognizes a CXPXR sequence and converts a
cysteine residue to formylglycine to produce antibodies with
aldehyde tags [102,103]. The reactive aldehyde functionality
can then undergo conjugation to the drug–linker via oxime
chemistry or a Pictet–Spengler reaction [104].
Harnessing enzymatic post-translational modification processes
for site-specific labeling of proteins is a recently reviewed approach
for the preparation of homogenous ADCs [105]. Bacterial trans-
glutaminase (BTG) catalyzes the ligation of glutamine side chains
with the primary e-amine of lysine residues, resulting in a stable
isopeptide bond. Jegar et al. exploited BTG to load four chelates on
a deglycosylated antibody with an N297Q mutation in a site-
specific manner [106]. Recently, Strop et al. conducted BTG-
assisted conjugations by inserting LLQG sequences at different
sites on an antibody [107]. These studies clearly demonstrated that
the site of conjugation has a significant impact on the stability and
PK of the ADC. Another enzyme, sortase A (SrtA), catalyzes hydro-
lysis of the threonine–glycine bond in a LPXTG motif to form a
new peptide bond between the exposed C-terminus of threonine
and an N-terminal glycine motif [108].
Next-generation ADCsKey clinical assetsThe nearly 30 ADCs currently in clinical development have been
reviewed in detail elsewhere [109], and representative examples of
the most advanced agents are summarized in Table 3. In addition
to the FDA-approved ADCs discussed in preceding sections, several
compounds are in late-stage clinical testing for both hematological
and solid tumor indications. Despite the withdrawal of Mylotarg1
from the market in 2010, promising results from ongoing clinical
studies have shown that when combined with chemotherapy
Mylotarg1 increased overall survival in patients with newly diag-
nosed AML compared to those treated with chemotherapy alone
[110]. Inotuzumab ozogamicin, which uses the same calicheami-
cin payload and cleavable hydrazone linker found in Mylotarg1,
recently failed to demonstrate improved survival in a phase III
study for patients with refractory aggressive NHL (Pfizer Inc. press
release; May 20, 2013). No unexpected safety concerns were iden-
tified, however, and phase III studies continue for acute lympho-
blastic leukemia (ALL) patients.
The vast majority of remaining ADCs in clinical development
use either auristatin [monomethyl auristatin E (MMAE) or MMAF]
or maytansinoid (DM1 or DM4) payloads, both potent inhibitors
of tubulin polymerization. Several MMAE conjugates with clea-
vable linkers are currently under evaluation in phase II studies for
various indications based on the target antigen. In general, these
agents were well tolerated in phase I trials with toxicities consis-
tent with the known mechanism of action for the auristatins (e.g.
neutropenia or neuropathy) [22,111]. SAR3419, an anti-CD19
DM4 conjugate with a cleavable disulfide linker, demonstrated a
dose-limiting toxicity (DLT) of reversible severely blurred vision in
a phase I study for refractory B cell NHL, but was well tolerated on a
modified dosing schedule [112]. Recently advanced to phase II
studies for colorectal cancer (CRC), labetuzumab-SN-38 employs a
cathepsin B-cleavable dipeptide linker and SN-38, the active meta-
bolite of the clinically used anticancer agent irinotecan, as a
payload. Initial phase I data indicated that labetuzumab-SN-38
was generally safe and well tolerated at effective clinical doses
[113]. Lorvotuzumab mertansine utilizes a maytansinoid payload
(DM1) and a disulfide linker to target CD56. No serious DLTs or
www.drugdiscoverytoday.com 877 11
REVIEWS Drug Discovery Today � Volume 19, Number 7 � July 2014
TABLE 3
Representative ADCs undergoing clinical evaluationa
Agent Sponsor (licensee) Status Indication Antigen Cytotoxin Linker
AdcetrisW (brentuximab
vedotin, SGN-35)
Seattle Genetics
(Millennium)
Launched HL, ALCL CD30 MMAE Cleavable, Val-Cit
KadcylaW (ado-trastuzumab
emtansine, T-DM1)
Roche-Genentech
(ImmunoGen)
Launched Her2+ metastatic
breast cancer
HER2 DM1 Non-cleavable, thioether
MylotargW (gemtuzumabozogamicin)
Pfizer (UCB) Withdrawn AML CD33 Calicheamicin Cleavable, hydrazone(Ac-But acid)
Inotuzumab ozogamicin
(CMC-544)
Pfizer (UCB) Ph III ALL, NHL CD22 Calicheamicin Cleavable, hydrazone
(Ac-But acid)
RG-7596 Roche-Genentech Ph II DLBCL, NHL CD79b MMAE Cleavable, Val-Cit
Glembatumumabvedotin CDX-011)
Celldex (SeattleGenetics)
Ph II Advanced breastcancer, melanoma
GPNMB MMAE Cleavable, Val-Cit
PSMA-ADC Progenics (Seattle
Genetics)
Ph II HRPC PSMA MMAE Cleavable, Val-Cit
SAR3419 Sanofi (ImmunoGen) Ph II Hematologic tumors CD19 DM4 Cleavable, disulfide
Labetuzumab-SN-38(IMUU-130)
Immunomedics Ph II Metastatic CRC CEACAM5 SN-38 Cleavable, Phe-Lys
Lorvotuzumab mertansine
(IMGN901)
ImmunoGen Ph I/II MM, solid tumors CD56 DM1 Cleavable, disulfide
Milatuzumab-DOX
(IMMU-110)
Immunomedics Ph I/II MM CD74 Doxorubicin Cleavable, hydrazone
BT-062 Biotest AG(ImmunoGen)
Ph I MM CD138 DM4 Cleavable, disulfide
BAY-94-9343 Bayer Schering
(ImmunoGen)
Ph I Solid tumors Mesothelin DM4 Cleavable, disulfide
ASG-5ME Astellas (SeattleGenetics)
Ph I Solid tumors AGS-5 MMAE Cleavable, Val-Cit
SGN-75 Seattle Genetics Ph I NHL, RCC CD70 MMAF Non-cleavable, MC
IMGN529 ImmunoGen Ph I Hematologic tumors CD37 DM1 Non-cleavable, thioether
SAR-566658 Sanofi (ImmunoGen) Ph I Solid tumors DS6 DM4 Cleavable, disulfide
a Abbreviations: CEACAM5, carcinoembryonic antigen cell adhesion molecule 5; HRPC: hormone refractory prostate cancer; MC: maleimidocaproyl; RCC: renal cell carcinoma; SN-38, 7-
ethyl-10-hydroxycamptothecin.
Review
s�K
EYNOTEREVIEW
drug-related adverse events were reported in early-phase multiple
myeloma (MM) studies [114].
Although the modest potency of doxorubicin payloads limited
the efficacy of early ADCs (BR96-DOX), milatuzumab-DOX targets
CD74, an antigen with unique internalization and surface re-
expression, and is currently in phase I/II trials based on encoura-
ging preclinical efficacy in hematopoietic cancer xenograft models
[52]. Select agents in phase I trials include ADCs containing DM1
or DM4 cytotoxic drugs under evaluation by ImmunoGen, and
several ADCs with MMAE or MMAF developed by Seattle Genetics,
each targeting a different antigen across a variety of tumor indica-
tions. Available data for these and other phase I agents generally
provide initial evidence of efficacy and tolerability. Similar to
SAR3419 (anti-CD19 DM4 conjugate), the DM4-based anti-
mesothelin conjugate BAY-94-9343 has also been reported to
induce Grade 2 and 4 ocular toxicity [115].
ADC PKADCs typically retain the PK properties of the antibody compo-
nent, as opposed to the appended drug, and thus exhibit relatively
low clearance and long half lives. Compared with the unconju-
gated antibody, ADCs can exhibit somewhat higher clearance due
878 www.drugdiscoverytoday.com
to introduction of an additional metabolic pathway (i.e. cleavage
of the drug from the antibody). In addition, ADCs with higher
DARs tend to clear faster than those with lower DARs [116].
Variable DARs and attachment sites, a consequence of current
random conjugation methods, result in heterogeneous ADCs with
PK parameters that can vary substantially compared to the uncon-
jugated antibody [117]. Each ADC component, along with their
respective metabolites, can potentially impact efficacy, safety and
tolerability [118]. Both the type of linker used and the site of
conjugation can influence the extent to which the drug is prema-
turely released from the antibody. Deconjugation of the payload
from the antibody can result in ADCs with lower DARs, reduced
efficacy and potentially increased toxicity owing to release of a
highly potent cytotoxic drug in systemic circulation.
The PK parameters of Adcetris1 and Kadcyla1 were evaluated in
mouse, rat and monkey in preclinical toxicity studies. Overall, these
ADCs demonstrated similar PK properties, albeit with a few differ-
ences in mouse and monkey. The t1/2 of Adcetris1 in mouse, rat and
monkey was 14, 10 and 2 days, respectively. The rapid clearance of
Adcetris1 in monkeys as compared with mouse or rat was hypothe-
sized to result from nontherapeutic antibodies, target-mediated
disposition and other factors [119]. In the case of Kadcyla1, the
12
Drug Discovery Today � Volume 19, Number 7 � July 2014 REVIEWS
Reviews�KEYNOTEREVIEW
t1/2 in mouse, rat and monkey was 5.7, 8.3 and 11.6 days, respec-
tively [120,121]. In humans, the PK characteristics of these two
conjugates were similar, with a t1/2 ranging from 3.5 to 5 days
[22,24]. Of note, the t1/2 of an ADC is often significantly shorter
in humans compared with other species. The optimal t1/2 for an ADC
remains to be determined, but the clinical success of Adcetris1 and
Kadcyla1 indicate that the range of 3–4 days is appropriate.
The long t1/2 typical of ADCs and mAbs results from FcRn
recycling [122]. In this process, antigen-independent internaliza-
tion by endothelial cells is followed by FcRn binding and then
FcRn-mediated return to the bloodstream. FcRn recycling essen-
tially protects ADCs from catabolism; however, diversion of FcRn-
bound ADCs to the lysosome can increase the risk of off-target
toxicities. Although the factors that influence this process are
poorly understood, the drug, linker, antibody and antigen can
each affect FcRn-mediated ADC trafficking [123]. Another
mechanism of off-target toxicity involves soluble cell-surface man-
nose receptors (MRs), which interact with agalactosylated glycans
on the antibody Fc domain [124]. Cell-surface MRs can internalize,
effectively delivering the ADC to the endosome and lysosome
compartments where the potent cytotoxic drug is released. Impor-
tantly, locations of off-target ADC activities reportedly coincide
with cell-surface MR locations. The shedding of antigen from the
tumor cell surface into circulation may also increase the risk of
toxicity. Binding of an ADC to shed antigen can, in some cases,
lead to higher ADC clearance and impaired tumor localization as
well as immune complex formation and accumulation in the
kidney [125].
To determine the effect of linker stability on PK and efficacy, the
noncleavable thioether linker of Kadcyla1 was compared to the
cleavable disulfide linker of a T-SPP-DM1 conjugate [121]. The
nonreducible thioether-linked Kadcyla1 demonstrated superior
PK with greater plasma exposure (area under the curve) and
increased maytansinoid tumor concentration. The disulfide-
linked ADC demonstrated higher plasma clearance owing to the
presence of the metabolically labile linker. Despite the difference
in PK, both conjugates had similar in vivo efficacy. It was hypothe-
sized that the drug released from the disulfide-linked T-SPP-DM1
conjugate would benefit from the bystander killing effect, whereas
Kadcyla1 ultimately liberates a maytansinoid appended to a
charged lysine residue, which limits diffusion to neighboring
tumor cells. Taken together, these results illustrate how minor
structural changes can profoundly impact ADC PK and efficacy.
In addition to the type of linker used to join the drug and
antibody, the conjugation site on the antibody has been shown to
influence stability and, therefore, PK. A recent study examined the
stability of MMAE conjugated to Her2 via a maleimide at various
site-specifically engineered cysteines [126]. Highly solvent acces-
sible conjugation sites were found to be labile, undergoing mal-
eimide exchange with reactive thiols in the plasma, such as
glutathione, albumin or free cysteine. At less accessible sites,
the succinimide ring of the linker underwent hydrolysis, which
served to protect the linker from maleimide exchange and resulted
in enhanced stability and efficacy. In a separate study, the stability
of monomethyl auristatin D (MMAD) conjugated to an anti-M1S1
antibody was examined using BTG to introduce the drug payload
site specifically at either the heavy or light chain [107]. The
conjugation site was found to influence stability and PK, with
ADCs appended to the heavy chain demonstrating a higher rate of
drug loss in rats via proteolysis of the valine–citrulline linker.
Interestingly, these results were species specific since both con-
jugates demonstrated comparable stability in mice, which also
serves to highlight the potential pitfall of performing safety and
efficacy studies in different species.
Concluding remarks and future directionsDespite complexities in designing ADCs, the promise of this
therapeutic class has generated intense interest for decades. A
robust clinical pipeline and the recent FDA approvals of Adcetris1
and Kadcyla1 suggest that the potential benefit of ADCs may
finally be realized. Evolving clinical data will continue to drive
technological advancements in the field. Current methods for
preclinical lead selection typically rely on systematic in vitro
evaluation of a matrix of various mAbs, linkers and cytotoxic
payloads. Whether in vitro models are sufficient to predict response
remains to be seen; until further understanding of ADCs is rea-
lized, early in vivo studies might be crucial. Progress in site-specific
conjugation modalities, optimization of linkers with balanced
stability and identification of novel, potent cytotoxic agents
should pave the way for greater insight into the contribution of
these various factors to ADC efficacy, PK and safety. Challenges in
target tumor selection will be addressed as the roles of antigen
expression, heterogeneity and internalization rate are further
elucidated. Guiding principles for the selection of an ideal anti-
body Fc format are, as of yet, lacking and prompt validation of
current assumptions regarding antibody-dependent properties,
such as specificity and immune effector functions. Ongoing efforts
to address these issues will continue to broaden the impact of
ADCs as targeted therapeutics for the treatment of cancer and
potentially other diseases.
References
1 McDermott, U. and Settleman, J. (2009) Personalized cancer therapy with selective
kinase inhibitors: an emerging paradigm in medical oncology. J. Clin. Oncol. 27,
5650–5659
2 Weinstein, I.B. (2005) Addiction to oncogenes – the Achilles heal of cancer. Science
297, 63–64
3 Giroux, S. (2013) Overcoming acquired resistance to kinase inhibition: the case of
EGFR, ALK, and BRAF. Bioorg. Med. Chem. Lett. 23, 394–401
4 Rosenzweig, S.A. (2012) Acquired resistance to drugs targeting receptor tyrosine
kinases. Biochem. Pharmacol. 83, 1041–1048
5 Janthur, W.-D. et al. (2012) Drug conjugates such as antibody drug conjugates
(ADCs), immuotoxins and immunoliposomes challenge daily clinical practice. Int.
J. Mol. Sci. 13, 16020–16045
6 Steiner, M. and Neri, D. (2011) Antibody–radionuclide conjugates for cancer
therapy: historical considerations and new trends. Clin. Cancer Res. 17, 6406–6416
7 Ehrlich, P. (1906) The relationship existing between chemical constitution,
distribution and pharmacological action. In Collected Studies on Immunity. pp. 441–
450, Wiley & Sons
8 Ehrlich, P. (1913) Chemotherapy. Proceedings of 17th international congress of
medicine. In Collected Papers of Paul Ehrlich (Himmelwiet, F., ed.), pp. 505–518,
Pergamon Press
9 Mathe, G. et al. (1958) Effet sur la leucemie L1210 de la souris d’une combinaison
par diazotation d’A-methopterine et de gamma-globulines de hamsters porteur de
cette leucemie par heterogreffe. C. R. Acad. Sci. 246, 1626–1628
10 Ghose, T. et al. (1967) Immunoradioactive agent against cancer. Br. Med. J. 1, 90–93
www.drugdiscoverytoday.com 879 13
REVIEWS Drug Discovery Today � Volume 19, Number 7 � July 2014
Review
s�K
EYNOTEREVIEW
11 Ghose, T. et al. (1972) Antibody as carrier of chlorambucil. Cancer 29, 1398–1400
12 Rowland, G.F. et al. (1975) Suppression of tumor growth in mice by a drug-
antibody conjugate using a novel approach to linkage. Nature 255, 487–488
13 Kohler, G. and Milstein, C. (1975) Continuous cultures of fused cells secreting
antibody of predefined specificity. Nature 256, 495–497
14 Ford, C.H.J. et al. (1983) Localization and toxicity study of a vindesine-anti-CEA
conjugate in patients with advanced cancer. Br. J. Cancer 47, 35–42
15 Riechmann, L. et al. (1988) Reshaping human antibodies for therapy. Nature 332,
323–332
16 Pietersz, G.A. and Krauer, K. (1994) Antibody-targeted drugs for the therapy of
cancer. J. Drug Targeting 2, 183–215
17 Petersen, B.H. et al. (1991) The human immune response to KS1/4-
desacetylvinblastine (LY256787) and KS1/4-desacetylvinblastine hydrazide
(LY203728) in single and multiple dose clinical studies. Cancer Res. 51, 2286–2290
18 Trail, P.A. et al. (1993) Cure of xenografted human carcinomas by BR96–
doxorubicin immunoconjugates. Science 261, 212–215
19 Senter, P.D. (2009) Potent antibody drug conjugates for cancer therapy. Curr. Opin.
Chem. Biol. 13, 235–244
20 Sievers, E.L. et al. (2001) Efficacy and safety of Mylotarg (gemtuzumab ozogamicin)
with CD33-postive acute myeloid leukemia in first relapse. J. Clin. Oncol. 19, 3244–
3254
21 Linenberger, M.L. et al. (2001) Multidrug-resistance phenotype and clinical
responses to gemtuzumab ozogamicin. Blood 98, 988–994
22 Younes, A. et al. (2010) Brentuximab vedotin (SGN-35) for relapsed CD30-positive
lymphomas. N. Engl. J. Med. 363, 1812–1821
23 Senter, P. and Sievers, E.L. (2012) The discovery and development of brentuximab
vedotin for use in relapsed Hodgkin lymphoma and systemic anaplastic large cell
lymphoma. Nat. Biotechnol. 30, 631–637
24 LoRusso, P.M. et al. (2011) Trastuzumab emtansine: a unique antibody–drug
conjugate in development for human epidermal growth factor receptor 2-positive
cancer. Clin. Cancer Res. 17, 6437–6447
25 Verma, S. et al. (2012) Trastuzumab emtansine for HER2-positive advanced breast
cancer. N. Engl. J. Med. 367, 1783–1791
26 Mullard, A. (2013) Maturing antibody–drug conjugate pipeline hits 30. Nat. Rev.
Drug Discov. 12, 329–332
27 Sievers, E.L. and Senter, P.D. (2013) Antibody–drug conjugates in cancer therapy.
Annu. Rev. Med. 64, 15–29
28 Ritchie, M. et al. (2013) Implications of receptor-mediated endocytosis and
intracellular trafficking dynamics in the development of antibody drug
conjugates. mAbs 5, 13–21
29 Press, M.F. et al. (1990) Expression of the HER-2/neu proto-oncogene in normal
human adult and fetal tissues. Oncogene 5, 953–962
30 Silver, D.A. et al. (1997) Prostate-specific membrane antigen expression in normal
and malignant human tissues. Clin. Cancer Res. 3, 81–85
31 Solal-Celigny, P. (2006) Safety of rituximab maintenance therapy in follicular
lymphomas. Leuk. Res. 30S1, S16–S21
32 Christiansen, J. and Rajasekaran, A.K. (2004) Biological impediments to
monoclonal antibody-based cancer immunotherapy. Mol. Cancer Ther. 3, 1493–
1501
33 Kovtun, Y.V. et al. (2006) Antibody–drug conjugates designed to eradicate tumors
with homogeneous and heterogeneous expression of the target antigen. Cancer
Res. 66, 3214–3221
34 Polson, A.G. et al. (2011) Investigational antibody–drug conjugates for
hematological malignancies. Expert Opin. Investig. Drugs 20, 75–85
35 Clark, T. et al. (2013) Insights into antibody–drug conjugates: bioanalysis and
biomeasures in discovery. Bioanalysis 5, 985–987
36 Dornan, D. et al. (2009) Therapeutic potential of an anti-CD79b antibody–drug
conjugate, anti-CD79b-vc-MMAE, for the treatment of non-Hodgkin lymphoma.
Blood 114, 2721–2729
37 Smith, L.M. et al. (2006) Potent cytotoxicity of an auristatin-containing antibody–
drug conjugate targeting melanoma cells expressing melanotransferrin/p97. Mol.
Cancer Ther. 5, 1474–1482
38 Tanimoto, M. et al. (1989) Restricted expression of an early myeloid and
monocytic cell surface antigen defined by monoclonal antibody M195. Leukemia
3, 339–348
39 Sutherland, M.S.K. et al. (2013) SGN-CD33A: a novel CD33-targeting antibody–
drug conjugate utilizing a pyrrolobenzodiazepine dimer is active in models of
drug-resistant AML. Blood 122, 1455–1463
40 Wang, X. et al. (2011) In vitro and in vivo responses of advanced prostate tumors to
PSMA ADC, an auristatin-conjugated antibody to prostate-specific membrane
antigen. Mol. Cancer Ther. 10, 1728–1739
41 Asudi, J. et al. (2011) An antibody–drug conjugate targeting the endothelin B
receptor for the treatment of melanoma. Clin. Cancer Res. 17, 965–975
880 www.drugdiscoverytoday.com
42 Yoshikawa, M. et al. (2013) Robo4 is an effective tumor endothelial marker for
antibody–drug conjugates based on the rapid isolation of the anti-Robo4 cell-
internalizing antibody. Blood 121, 2804–2813
43 Ackerman, M.E. et al. (2008) Effect of antigen turnover rate and expression
level on antibody penetration into tumor spheroids. Mol. Cancer Ther. 7,
2233–2240
44 Carter, P. (2006) Potent antibody therapeutics by design. Nat. Rev. Immunol. 6,
343–357
45 Owen, S.C. et al. (2013) Targeting HER21 breast cancer cells: lysosomal
accumulation of anti-HER2 antibodies is influenced by antibody binding site and
conjugation to polymeric nanoparticles. J. Control. Release 172, 395–404
46 Sutherland, M.S.K. et al. (2006) Lysosomal trafficking and cysteine protease
metabolism confer target-specific cytotoxicity by peptide-linked anti-CD30-
auristatin conjugates. J. Biol. Chem. 281, 10540–10547
47 Hamann, P.R. et al. (2005) A calicheamicin conjugate with a fully humanized anti-
MUC1 antibody shows potent antitumor effects in breast and ovarian tumor
xenografts. Bioconjug. Chem. 16, 354–360
48 Law, C.L. et al. (2004) Efficient elimination of B-lineage lymphomas by anti-CD20-
auristatin conjugates. Clin. Cancer Res. 10, 7842–7851
49 Schmidt, M.M. et al. (2008) Kinetics of anti-carcinoembryonic antigen antibody
internalization: effects of affinity, bivalency, and stability. Cancer Immunol.
Immunother. 57, 1879–1890
50 Hendriks, B.S. et al. (2003) Quantitative analysis of HER2-mediated effects on
HER2 and epidermal growth factor receptor endocytosis: distribution of
homo- and heterodimers depends on relative HER2 levels. J. Biol. Chem. 278,
23343–23351
51 Hansen, H.J. et al. (1996) Internalization and catabolism of radiolabelled
antibodies to the MHC class-II invariant chain by B-cell lymphomas. Biochem. J.
320, 293–300
52 Sapra, P. et al. (2005) Anti-CD74 antibody-doxorubicin conjugate, IMMU-110,
in a human multiple myeloma xenograft and in monkeys. Clin. Cancer Res. 11,
5257–5264
53 Neri, D. and Bicknell, R. (2005) Tumour vascular targeting. Nat. Rev. Cancer 5, 436–
446
54 Palumbo, A. et al. (2011) A chemically modified antibody mediates complete
eradication of tumours by selective disruption of tumour blood vessels. Br. J.
Cancer 104, 1106–1115
55 McDonagh, C.F. et al. (2008) Engineered anti-CD70 antibody–drug conjugate with
increased therapeutic index. Mol. Cancer Ther. 7, 2913–2923
56 Junttila, T.T. et al. (2011) Trastuzumab-DM1 (T–DM1) retains all the mechanisms
of action of trastuzumab and efficiently inhibits growth of lapatinib insensitive
breast cancer. Breast Cancer Res. Treat. 128, 347–356
57 European Medicines Agency (2012) Adcetris CHMP Assessment Report, EMA/
702390/2012.
58 DiJoseph, J.F. et al. (2006) Antitumor efficacy of a combination of CMC-544
(inotuzumab ozogamicin), a CD22-targeted cytotoxic immunoconjugate of
calicheamicin, and rituximab against non-Hodgkin’s B-cell lymphoma. Clin.
Cancer Res. 12, 242–249
59 Goldmacher, V.S. and Kovtun, Y.V. (2011) Antibody–drug conjugates: using
monoclonal antibodies for delivery of cytotoxic payloads to cancer cells. Ther.
Deliv. 2, 397–416
60 Lonberg, N. (2008) Fully human antibodies from transgenic mouse and phage
display platforms. Curr. Opin. Immunol. 20, 450–459
61 Boghaert, E.R. et al. (2008) Determination of pharmacokinetic values of
calicheamicin-antibody conjugates in mice by plasmon resonance analysis
of small (5 micron) blood samples. Cancer Chemother. Pharmacol. 61,
1027–1035
62 Dubowchik, G.M. and Firestone, R.A. (1998) Cathepsin B-sensitive dipeptide
prodrugs. 1. A model study of structural requirements for efficient release of
doxorubicin. Bioorg. Med. Chem. Lett. 8, 3341–3346
63 Dubowchik, G.M. et al. (2002) Cathepsin B-labile dipeptide linkers for lysosomal
release of doxorubicin from internalizing immunoconjugates: model studies of
enzymatic drug release and antigen-specific in vitro anticancer activity. Bioconjug.
Chem. 13, 855–869
64 Doronina, S.O. et al. (2003) Development of potent monoclonal antibody
auristatin conjugates for cancer therapy. Nat. Biotechnol. 21, 778–784
65 Ducry, L. and Stump, B. (2010) Antibody–drug conjugates: linking cytotoxic
payloads to monoclonal antibodies. Bioconjug. Chem. 21, 5–13
66 Hamann, P.R. et al. (2002) Gemtuzumab ozogamicin, a potent and selective anti-
CD33 antibody–calicheamicin conjugate for the treatment of acute myeloid
leukemia. Bioconjug. Chem. 13, 47–58
67 Sapra, P. et al. (2011) Investigational antibody drug conjugates for solid tumors.
Expert Opin. Investig. Drugs 20, 1131–1149
14
Drug Discovery Today � Volume 19, Number 7 � July 2014 REVIEWS
Reviews�KEYNOTEREVIEW
68 Sato, G. et al. (2003) Drug delivery strategy utilizing conjugation via reversible
disulfide linkages: role and site of cellular reducing activities. Adv. Drug Deliv. Rev.
55, 199–215
69 Erickson, H.K. (2010) Tumor delivery and in vivo processing of disulfide-linked and
thioether-linked antibody–maytansinoid conjugates. Bioconjug. Chem. 21, 84–92
70 Yurkovetskiy, A.V. Mersana Therapeutics, Inc. Protein–polymer–drug conjugates.
U.S. Patent Appl. 20130101546.
71 Williams, J.C. et al. Meditope Biosciences, Inc. Meditopes and meditope-binding
antibodies and uses thereof. U.S. Patent Appl. 20120301400.
72 Bajjuri, K.M. et al. (2011) The legumain protease-activated auristatin prodrugs
suppress tumor growth and metastasis without toxicity. ChemMedChem 6, 54–59
73 Jeffrey, S.C. et al. (2010) Expanded utility of the b-glucuronide linker: ADCs that
deliver phenolic cytotoxic agents. ACS Med. Chem. Lett. 1, 277–280
74 Legigan, T. et al. (2012) The first generation of b-galactosidase-responsive prodrugs
designed for the selective treatment of solid tumors in prodrug monotherapy.
Angew. Chem. Int. Ed. 51, 11606–11610
75 Chari, R.V. et al. (1992) Immunoconjugates containing novel maytansinoids:
promising anticancer drugs. Cancer Res. 52, 127–131
76 Hinmann, L.M. et al. (1993) Preparation and characterization of monoclonal
antibody conjugates of the calicheamicins: a novel and potent family of antitumor
antibiotics. Cancer Res. 53, 3336–3342
77 Coccia, M.A. et al. Medarex, Inc. Human antibodies that bind CD70 and uses
thereof. U.S. Patent Appl. 20100150950.
78 Pietersz, G.A. et al. (1994) Chemoimmunoconjugates for the treatment of cancer.
Adv. Immunol. 56, 301–387
79 Tolcher, A.W. et al. (1999) Randomized phase II study of BR96-doxorubicin
conjugate in patients with metastatic breast cancer. J. Clin. Oncol. 17, 478–484
80 Teicher, B.A. and Chari, R.V.J. (2011) Antibody conjugate therapeutics: challenges
and potential. Clin. Cancer Res. 17, 6389–6397
81 Hartley, J.A. (2011) The development of pyrrolobenzodiazepines as antitumor
agents. Expert Opin. Investig. Drugs 20, 733–744
82 Moldenhauer, G. et al. (2012) Therapeutic potential of amanitin-conjugated anti-
epithelial cell adhesion molecule monoclonal antibody against pancreatic
carcinoma. J. Natl. Cancer Inst. 104, 1–13
83 Jackson, D.Y. and Edward, H.A. Igenica, Inc. Antibody–drug conjugates and
related compounds, compositions, and methods. U.S. Patent Appl. 20130224228.
84 Low, P.S. and Ku-Laratne, S.A. Purdue Research Foundation. PSMA binding ligand–
linker conjugates and methods for using. U.S. Patent Appl. 20100324008.
85 Cheng, H. et al. Medarex, Inc. Antiproliferative compounds, conjugates thereof,
methods therefor, and uses thereof. U.S. Patent 8,394,922 B2.
86 Benderra, Z. et al. (2005) MRP3, BCRP, and P-glycoprotein activities are prognostic
factors in adult acute myeloid leukemia. Clin. Cancer Res. 11, 7764–7772
87 Naito, K. et al. (2000) Calicheamicin-conjugated humanized anti-CD33
monoclonal antibody (gemtuzumab zogamicin, CMA-676) shows cytocidal effect
on CD33-positive leukemia cell lines, but is inactive on P-glycoprotein-expressing
sublines. Leukemia 14, 1436–1443
88 Walter, R.B. et al. (2003) Multidrug resistance protein attenuates gemtuzumab
ozogamicin-induced cytotoxicity in acute myeloid leukemia cells. Blood 102,
1466–1473
89 Linenberger, M.L. (2005) CD33-directed therapy with gemtuzumab ozogamicin in
acute myeloid leukemia: progress in understanding cytotoxicity and potential
mechanisms of drug resistance. Leukemia 19, 176–182
90 Tang, R. et al. (2009) P-gP activity is a critical resistance factor against AVE9633 and
DM4 cytotoxicity in leukaemia cell lines, but not a major mechanism of
chemoresistance in cells from acute leukaemia patients. BMC Cancer 9, 1–10
91 Bross, P.F. et al. (2001) Approval summary: gemtuzumab ozogamicin in relapsed
acute myeloid leukemia. Clin. Cancer Res. 17, 1490–1496
92 Junutula, J.R. et al. (2008) Site-specific conjugation of a cytotoxic drug to an
antibody improves the therapeutic index. Nat. Biotechnol. 26, 925–932
93 Boylan, N.J. et al. (2013) Conjugation site heterogeneity causes variable
electrostatic properties in Fc conjugates. Bioconjug. Chem. 24, 1008–1016
94 Rohrer, T. (2012) Considerations for the safe and effective manufacturing of
antibody drug conjugates. Chem. Oggi/Chem. Today 30, 76–79
95 Lyons, A. et al. (1990) Site-specific attachment to recombinant antibodies via
introduced surface cysteine residues. Protein Eng. 3, 703–708
96 Stimmel, J.B. et al. (2000) Site-specific conjugation on serine ! cysteine variant
monoclonal antibodies. J. Biol. Chem. 39, 30445–30450
97 Junutula, J.R. et al. (2008) Rapid identification of reactive cysteine residues for site-
specific labeling of antibody-Fabs. J. Immunol. Methods 332, 41–52
98 McDonagh, C.F. et al. (2006) Engineered antibody–drug conjugates with defined
sites and stoichiometries of drug attachment. Protein Eng. Des. Sel. 19, 299–307
99 Axup, J.Y. et al. (2012) Synthesis of site-specific antibody–drug conjugates using
unnatural amino acids. Proc. Natl. Acad. Sci. U. S. A. 109, 16101–16106
100 Young, T.S. and Schultz, P.G. (2010) Beyond the canonical 20 amino acids:
expanding the genetic lexicon. J. Biol. Chem. 285, 11039–11044
101 Yin, G. et al. (2012) Aglycosylated antibodies and antibody fragments produced in
a scalable in vitro transcription–translation system. mAbs 4, 217–225
102 Carrico, I.S. et al. (2007) Introducing genetically encoded aldehydes into proteins.
Nat. Chem. Biol. 3, 321–322
103 Rabuka, D. et al. (2012) Site-specific chemical protein conjugation using
genetically encoded aldehyde tags. Nat. Protoc. 7, 1052–1067
104 Agarwal, P. et al. (2013) A Pictet-Spengler ligation for protein chemical
modification. Proc. Natl. Acad. Sci. U. S. A. 110, 46–51
105 Sunbul, M. and Yin, J. (2009) Site specific protein labeling by enzymatic
posttranslational modification. Org. Biomol. Chem. 7, 3361–3371
106 Jeger, S. et al. (2010) Site-specific and stiochiometric modification of antibodies by
bacterial transglutaminase. Angew. Chem. Int. Ed. 49, 9995–9997
107 Strop, P. et al. (2013) Location matters: site of conjugation modulates stability and
pharmacokinetics of antibody drug conjugates. Chem. Biol. 20, 161–167
108 Madej, N.P. et al. (2012) Engineering of an anti-epidermal growth factor receptor
antibody to single chain format and labeling by sortase A-mediated protein
ligation. Biotechol. Bioeng. 109, 1461–1470
109 Lambert, J.M. (2012) Drug-conjugated antibodies for the treatment of cancer. Br. J.
Clin. Pharmacol. 76, 248–262
110 Burnett, A.K. et al. (2011) The addition of gemtuzumab ozogamicin to intensive
chemotherapy in older patients with AML produces a significant improvement in
overall survival: results of the UK NCRI AML16 randomized trial. Blood 118, 582
111 Dumontet, C. and Sikic, B.I. (1999) Mechanisms of action of and resistance to
antitubulin agents: microtubule dynamics, drug transport, and cell death. J. Clin.
Oncol. 17, 1061–1070
112 Younes, A. et al. (2012) Phase I multidose-escalation study of the anti-CD19
maytansinoid immunoconjugate SAR3419 administered by intravenous infusion
every 3 weeks to patients with relapsed/refractory B-cell lymphoma. J. Clin. Oncol.
30, 2776–2782
113 Segal, N.H. et al. (2013) A phase I study of IMMU-130 (labetuzumab-SN38) anti-
CEACAM5 antibody–drug conjugate (ADC) in patients with metastatic colorectal
cancer (mCRC) [abstract. In Proceedings of the 104th Annual Meeting of the American
Association for Cancer Research. Abstract nr LB-159, AACR
114 Berdeja, J.G. et al. (2010) Phase I study of lorvotuzumab mertansine (IMGN901) in
combination with lenalidomide and dexamethasone in patients with CD56-
positive relapsed or relapsed/refractory multiple myeloma – a preliminary safety
and efficacy analysis of the combination. Blood 116, 1934
115 Bendell, J. et al. (2013) First-in-human phase I dose escalation study of a novel anti-
mesothelin antibody drug conjugate (ADC), BAY 94-9343, in patients with
advanced solid tumors [abstract]. In Proceedings of the 104th Annual Meeting of the
American Association for Cancer Research. Abstract nr LB-291, AACR
116 Hamblett, K.J. et al. (2004) Effects of drug loading on the antitumor activity of a
monoclonal antibody drug conjugate. Clin. Cancer Res. 10, 7063–7070
117 Lin, K. and Tibbitts, J. (2012) Pharmacokinetic considerations for antibody drug
conjugates. Pharm. Res. 29, 2354–2366
118 Lu, D. et al. (2013) Strategies to address drug interaction potential for antibody–
drug conjugates in clinical development. Bioanalysis 5, 1115–1130
119 Ferrante, K. and Fox, I. (2011) Brentuximab vedotin (SGN-35). Physician’s Brochure
2, 15–18
120 Bender, B. et al. (2012) Multicompartmental population PK model elucidating the
complex disposition of trastuzumab emtansine (T-DM1): an antibody–drug
conjugate for the treatment of HER2-positive cancer [abstract]. Abstracts of the
Annual Meeting Population Approach Group in Europe. Abstract nr 2607, PAGE pp. 21
121 Erikson, H.K. et al. (2012) The effect of different linkers on target cell catabolism
and pharmacokinetics/pharmacodynamics of trastuzumab maytansinoid
conjugates. Mol. Cancer Ther. 11, 1133–1142
122 Roopenian, D.C. and Akilesh, S. (2007) FcRn: the neonatal Fc receptor comes of
age. Nat. Rev. Immunol. 7, 715–725
123 Gerber, H.-P. et al. (2013) The antibody–drug conjugate: an enabling modality for
natural-product based cancer therapeutics. Nat. Prod. Res. 625–639
124 Gorovits, B. and Krinos-Fiorotti, C. (2013) Proposed mechanism of off-target
toxicity for antibody–drug conjugates driven by mannose receptor uptake. Cancer
Immunol. Immunother. 62, 217–223
125 Carter, P. and Senter, P. (2008) Antibody–drug conjugates for cancer therapy.
Cancer J. 14, 154–168
126 Shen, B.-Q. et al. (2012) Conjugation site modulates the in vivo stability and
therapeutic activity of antibody–drug conjugates. Nat. Biotechnol. 30, 184–189
www.drugdiscoverytoday.com 881 15
An emerging playbook for antibody–drug conjugates:lessons from the laboratory and clinic suggest astrategy for improving efficacy and safetyPenelope M Drake and David Rabuka
Available online at www.sciencedirect.com
ScienceDirect
Antibody–drug conjugates (ADCs) have become de rigueur for
pharmaceutical oncology drug development pipelines. There
are more than 40 ADCs undergoing clinical trials and many more
in preclinical development. The field has rushed to follow the initial
successes of KadcylaTM and AdcetrisTM, and moved forward with
new targets without much pause for optimization. In some
respects, the ADC space has become divided into the clinical
realm — where the proven technologies continue to represent the
bulk of clinical candidates with a few exceptions — and the
research realm — where innovations in conjugation chemistry
and linker technologies have suggested that there is much room
for improvement in the conventional methods. Now, two and four
years after the approvals of KadcylaTM and AdcetrisTM,
respectively, consensus may at last be building that these two
drugs rely on rather unique target antigens that enable their
success. It is becoming increasingly clear that future target
antigens will require additional innovative approaches. Next-
generation ADCs have begun to move out of the lab and into the
clinic, where there is a pressing need for continued innovation to
overcome the twin challenges of safety and efficacy.
Address
Catalent Pharma Solutions, Emeryville, CA 94608, USA
Corresponding author: Rabuka, David (david.rabuka@catalent.com)
Current Opinion in Chemical Biology 2015, 28:174–180
This review comes from a themed issue on Synthetic biomolecules
Edited by Christian Hackenberger and Peng Chen
http://dx.doi.org/10.1016/j.cbpa.2015.08.005
1367-5931/# 2015 Elsevier Ltd. All rights reserved.
IntroductionAntibody–drug conjugates (ADCs) are a class of drugs
that have long been heralded as the answer to nonspecific
toxicity induced by systemically-dosed chemotherapies.
As their name implies, ADCs comprise a small molecule
cytotoxic drug conjugated through a linker to an antibody
that directs the ADC to tumor cells expressing the cog-
nate antigen (Figure 1a). Upon internalization by the cell,
the ADC is degraded and the cytotoxic payload is re-
leased and diffuses into the cytosol or nucleus where it
Current Opinion in Chemical Biology 2015, 28:174–180
binds to its targets and induces cell death (Figure 1b). In
theory, the targeting feature should endow ADCs with an
impressive therapeutic index — the difference between
the therapeutic and the toxic dose — and thus should
enable effective tumor eradication with minimal side
effects. This promise is now being tested by more than
40 molecules currently in clinical trials [1]. Two ADCs,
KadcylaTM and AdcetrisTM, are approved by the FDA for
treatment of HER2+ metastatic breast cancer and Hodg-
kin lymphoma or anaplastic large-cell lymphoma, respec-
tively. A third ADC, MylotargTM, approved to treat acute
myeloid leukemia, was voluntarily pulled from the mar-
ket after failure of the required post-approval study [2]. As
of early 2015, the majority of ADCs in clinical trials are in
Phase I studies [3]. The most significant hurdle for the
field will be to demonstrate that the ADC approach can be
broadly applied to generate drugs that are safe at effective
dose levels. Although promising results have been
reported for isolated therapies — such as the recently
released Phase I results from ImmunoGen’s anti-folate
receptor alpha-DM4 conjugate for the treatment of ovari-
an cancer [4] — the overall picture has yet to emerge. For
a summary of the clinical ADC pipeline, we refer readers
to the excellent review by Wang and Watanabe [1]. What
we do know is that the current iteration of ADCs has
limitations, including the targeting efficiency. It is esti-
mated that a very low percentage of the injected conju-
gate reaches its tumor target [5]. The remaining ADC is
eliminated predominantly by the hepatobiliary system
[6], inducing toxic side effects in the process. By boosting
the percentage of dosed ADC that reaches the intended
target, efficacy and safety can be increased in parallel.
These challenges are beginning to be addressed on a
number of fronts (Figure 2).
Current ADC technologies and associatedchallengesDue to poor targeting efficiency, an ADC’s cytotoxic
payload must be sufficiently potent to induce target cell
death at low doses; subnanomolar activity is desired.
Only a handful of cytotoxic payloads are represented
on ADCs in the clinic, including the microtubule-bind-
ing auristatins and maytansinoids, and the DNA-target-
ing natural products pyrrolobenzodiazepine (PBD) and
calicheamicin [3,7]. Amanatin, an RNA polymerase II
inhibitor, is being evaluated in preclinical studies [8��].The linker joining the payload to the antibody may be
either cleavable — designed to be stable in the circulation
www.sciencedirect.com
16
An emerging playbook for antibody–drug conjugates Drake and Rabuka 175
Figure 1
(a)
(b)
Current Opinion in Chemical Biology
An idealized ADC acts as a ‘magic bullet’ directing the cytotoxic
payload to tumors. (a) The ADC binds to cognate antigen on the
tumor cell surface and is internalized. (b) The ADC traffics through the
endosomes to the lysosome, where it is degraded into component
pieces. The cytotoxic payload (star) diffuses through cell membranes
to reach its targets in the cytoplasm or nucleus.
but readily degraded after internalization by the target
cell — or noncleavable — designed to persist through an-
tibody degradation in the lysosome [9,10]. The current
generation of ADCs, which constitute nearly all of the
Figure 2
Conjugate structural features Effects on
Conjugationchemistry
Linkerchemistry
Cytotoxicpayload
Antibodyselectivityfor tumor
Tunable ADC structural characte
106
105
104
103
ng/m
L
Total ATotal A
0 100
0.0001 0.01 1 Drug (n
100
80
60
40
20
0
viab
ility
(%
)
ADC structural features that directly impact function can be compared throu
configuration for safety and efficacy. (left panel) Tunable features include lin
modifications that limit target-mediated or FcgR-mediated drug disposition
impact critical pharmacokinetic (PK) and pharmacodynamic (PD) attributes
PD largely governs treatment outcomes such as safety and efficacy. The ho
thereby offer patients excellent anti-tumor efficacy with minimal side effects
www.sciencedirect.com
compounds in the clinic, are characterized by heteroge-
neous payload conjugation to lysine or cysteine residues.
The random nature of both conjugation approaches yields a
mixture of ADC species with drug-to-antibody ratios
(DARs) ranging from 0 to 8. The desired average DAR
is usually �3.5; this mixture includes some unconjugated
(and thus, inactive) antibody and some highly conjugated
antibody. The more highly-conjugated species appear to
be cleared faster by the liver [11,12�], likely due to their
increased hydrophobicity; in turn, this faster clearance
reduces efficacy and probably increases toxicity. Accord-
ingly, the heterogeneous mixtures found in conventional
ADCs include both inactive and more toxic isoforms, both
of which reduce the therapeutic index of the resulting
construct.
Clinical studies indicate that the drug-linker,not the target antigen, is driving toxicityA fascinating recent report by the FDA analyzed the
relationship between nonclinical study results and Phase
I study outcomes by looking at data from investigational
new drug applications filed for 20 ADCs from December
2012 to August 2013 [13��]. The authors found that ADCs
bearing the same drug-linker but with distinct target
antigens showed similar toxicity results, including the
PK/PD Treatment outcomes
Safety and efficacycharacteristics
ristics change functional readouts
Minimumefficacious dose
Therapeuticindex
Toxic dose
Dru
g do
se
ntibodyDC
200 300 400 500Hours post-dose
100 10000
M)
Dru
g do
se
Current Opinion in Chemical Biology
gh structure-activity relationship mapping to determine the optimal
ker and conjugation chemistries, cytotoxic payload selection, and
in non-tumor tissues. (middle panel) The structural features can directly
such as ADC distribution, half-life, and potency. (right panel) The PK/
pe for ADCs is that they will have a wide therapeutic index and
.
Current Opinion in Chemical Biology 2015, 28:174–180
17
176 Synthetic biomolecules
affected organs and the maximum tolerated dose (MTD)
levels. The target antigen, even when the epitope was
present (e.g., in non-human primate studies) had no effect
on either the target organs affected or on the MTD. In fact,
the FDA concluded that the main driver of toxicity was the
drug-linker component. These findings strongly support
the observations that only a low percentage of an adminis-
tered antibody finds its way to the tumor (or indeed to any
target-expressing tissues), and that toxicity is primarily
driven by non-target mediated disposition. The vast major-
ity (�80%) of the injected dose is eliminated within the first
7 days through excretion in the bile and feces [6] — leading
to predictable and dose-limiting toxicities. Finding ways to
mitigate this non-specific elimination will be key to im-
proving the safety and efficacy of next-generation ADCs.
The relationship between linker chemistry,ADC stability, and conjugate efficacyIt may seem that improving the in vivo stability of the
conjugate would increase the proportion of payload that
reaches its target — and thus, improve efficacy. This
trend generally holds true, but there are a number of
factors that enter the equation [3,9,14]. First, there are
target-dependent requirements for linker-payloads,
which are likely related to cell surface expression levels
as well as internalization and recycling kinetics. Polson
et al. performed a systematic study comparing the
effects of a cleavable (SPP-DM1) and a noncleavable
(SMCC-DM1) linker carrying the maytansinoid pay-
load, DM1 [15]. Antibodies against a panel of seven
potential non-Hodgkin lymphoma antigens were con-
jugated to these linkers and tested for in vivo efficacy.
All of the ADCs with cleavable linkers showed efficacy,
while ADCs with noncleavable linkers were active
against only two of the seven target antigens. In addition
to revealing a relationship between target antigen, link-
er design, and efficacy, these data also highlight the
general importance of target antigen selection during
ADC development, a topic worthy of its own review
article [16��,17,18].
Second, there is a positive correlation between ADC
exposure and efficacy, such that ADCs with longer circu-
lating half-lives tend to show better efficacy if other
parameters (e.g., target antigen and cytotoxic payload)
are held equal. ADC half-lives are determined by at least
two components: linker stability and total antibody half-
life [19]. Conjugated antibodies often display shorter half-
lives as compared to their unconjugated counterparts
[12�,20,21]; evidence suggests that hydrophobicity is
the major driver of this increased clearance [11,22�]. In
cases where the total antibody half-life is relatively short,
Alley et al. showed that improving linker stability only
minimally improves the overall ADC exposure, and thus
has little effect on efficacy [23]. However, when the total
antibody half-life is relatively long, improving linker
stability can lead to better efficacy. Examples of this
Current Opinion in Chemical Biology 2015, 28:174–180
effect came from both Pfizer and Seattle Genetics, who
took different approaches to efficiently hydrolyze malei-
mide-containing linkers for improved ADC stability; both
groups found that the new technologies led to longer
circulating ADC half-lives, longer overall exposure to the
conjugate, and improved efficacy against xenograft mod-
els [19,24].
Third, linker design can impact both active and passive
cell permeability of the ADC’s active metabolite. With
respect to active permeability, linker-payload suscepti-
bility to efflux by multidrug transporters (e.g., P-glyco-
protein/MDR1) relates to efficacy against multidrug
resistant cells. Namely, if other variables are held con-
stant, then linker-cytotoxins that are targets for MDR1-
mediated efflux have worse efficacy relative to those that
are not. Linker design can be exploited to reduce binding
of the cytotoxic payload to MDR1, and thus rescue ADC
efficacy in multidrug resistant cells. In general, linkers
that impart hydrophilicity, charge, or steric bulk show
reduced MDR1 binding, as the transporter prefers hydro-
phobic, weakly cationic, planar substrates [9,25,26]. With
respect to passive cell permeability, in vivo potency
against solid tumors is governed partially by the bystander
killing effect, where non-targeted (perhaps target anti-
gen-negative) cells are killed by cytotoxic payloads that
cross plasma membranes. Kellogg et al. observed that a
linker with intermediate plasma stability gave the best invivo efficacy results, in spite of having lower overall
exposure as compared to a noncleavable ADC bearing
the same payload [27]. The authors postulated that by-
stander killing, a trait that the noncleavable linker lacked,
contributed additional potency in vivo. Notably, the most
labile linker, which also could mediate bystander killing,
was cleared rapidly from the circulation and showed
relatively poor efficacy. Therefore, the linker that pro-
vided a balance of in vivo potency and exposure yielded
the best overall effect.
Finally, a number of groups have noted that the in vivostability of a given linker can vary depending upon the
model species [28–30], often with worse stability in
rodents and improved stability in primates. This likely
reflects increased enzymatic (e.g., protease or esterase)
activity in rodent plasma, leading to linker degradation.
The disparity in stability among species can make it
challenging to develop meaningful efficacy studies, which
are usually performed in immunocompromised rodent
models. In summary, there is not one best way to build
a linker for efficacy; rather the linker should be consid-
ered a tunable ADC feature that can be optimized for a
given target and payload.
The connection between linker stability, ADCpharmacokinetics, and conjugate tolerabilityThe relationship between ADC stability and safety is
similarly nuanced. Toxicity can be mediated both by
www.sciencedirect.com
18
An emerging playbook for antibody–drug conjugates Drake and Rabuka 177
deconjugated cytotoxic payloads and by intact ADCs
that — through off-target effects — damage healthy
cells. The majority of deconjugated small molecules
are metabolically inactivated in the liver and cleared
through the biliary/fecal route. ADCs that are over-
conjugated with a hydrophobic payload can also be
cleared by the liver. Conventional, heterogeneous con-
jugates suffer from this issue, which can be alleviated by
using site-specific conjugation (discussed below) or by
increasing the hydrophilicity of the linker-payload
[11,22�]. A less often discussed avenue for toxicity is
through ADC interactions with Fcg receptors (FcgRs)
that are expressed on a variety of hematopoietic cells. A
recent study by Uppal et al. assessed the role of FcgIIRa
as a potential mechanism for thrombocytopenia follow-
ing treatment with T-DM1 (KadcylaTM) [31��]. The
authors found that the ADC inhibited a crucial step in
platelet development by exhibiting an FcgIIRa-depen-
dent cytotoxic effect against differentiating megakar-
yocytes. The data are particularly compelling because
thrombocytopenia is the dose-limiting toxicity associ-
ated with the clinical use of KadcylaTM. The recent
FDA analysis concluded that antibody isotype had no
impact on an ADC’s MTD or safety profile, suggesting
that immune-mediated effector functions (e.g., anti-
body-dependent cell-mediated cytotoxicity) were not
a major factor. However, the study did not rule out
FcgR-dependent uptake into hematopoietic cells as a
toxicity mechanism. Interestingly, overly conjugated/
hydrophobic ADCs may be at particular risk for this
clearance route, as low-affinity FcgRs (which bind to all
IgG isotypes) can only bind to aggregated or opsonized
antibody [32].
In animal models, if ADC hydrophobicity is held con-
stant, then ADCs with more stable linkers tend to show
improved tolerability. For example, in both the previ-
ously discussed linker studies [15,24], the more stable
linkers were less toxic when administered to rodents.
Interestingly, in humans the picture is more complex;
clinical data from a number of maytansinoid-based
ADCs made with different linkers indicates that toler-
ability is inversely related to stability [9]. Namely, the
most labile linker (the disulfide-based SPP-DM1)
has an MTD of 6.0–8.0 mg/kg every 3 wk. By contrast,
the intermediately labile and most stable linkers (the
hindered disulfide-based SPDB-DM4 and the nonclea-
vable SMCC-DM1, respectively) have MTDs of 3.5–7.0 and 3.6 mg/kg every 3 wk. The dose-limiting toxi-
cities are different for each of these linker-payload
combinations; the most labile linker is limited by liver
toxicity, while the most stable is limited by thrombo-
cytopenia. These observations bolster the conclusion
that a variety of factors contribute to toxicity, including
FcgR-mediated disposition to the hematopoietic com-
partment and deconjugated payload detoxification by
the liver.
www.sciencedirect.com
Site-specifically conjugated ADCs can widenthe therapeutic windowA number of studies have shown that reducing the
heterogeneity of an ADC through the use of site-specific
conjugation approaches can simultaneously improve effi-
cacy, pharmacokinetics, and safety [33,34]. Furthermore,
the analytics are greatly simplified, leading to easier
quality control and batch-to-batch consistency. For all
of these reasons, the field is steadily moving towards site-
specific conjugation as a critical feature of the next-
generation ADC. Three site-specific ADCs are already
in the clinic, including two from Seattle Genetics bearing
PBD payloads [35,36], and one from Rinat-Pfizer carrying
a microtubule inhibitor [37]. A variety of approaches have
been developed to enable site-specific conjugation [10],
including the introduction of engineered cysteine resi-
dues (Genentech/Roche, Seattle Genetics, Medimmune)
[38�], the incorporation of non-natural amino acids
(Ambrx, Sutro) [39,40], and chemoenzymatic approaches
such as the SMARTagTM method (Catalent), which uses
an enzyme to install a bioorthogonal chemical group for
subsequent conjugation [41,42], and the transglutaminase
method (Rinat-Pfizer) [28], which uses an enzyme to
perform the ligation between payload and protein. The
relative merits of these approaches will be sorted out in
the coming months through in vivo efficacy and safety
outcomes, and through manufacturing feasibility studies.
It is likely that several, if not all, of them will eventually
be tested in humans.
In a satisfying example of convergence, the research
groups studying site-specifically conjugated ADCs made
using different approaches are coming to similar conclu-
sions. For example, while a higher DAR (more cytotoxic
payload per antibody) generally leads to increased poten-
cy in vitro, in vivo the best efficacy can sometimes be
obtained by using a lower DAR species, particularly when
comparing a site-specifically conjugated lower DAR to a
heterogeneously-conjugated higher DAR construct
[41,43,44]. This outcome also leads to an increased ther-
apeutic index and improved safety, as the same or better
efficacy can be achieved with less cytotoxic payload
[41,43,44]. Furthermore, conjugation site does not affect
ADC potency in vitro, but greatly impacts efficacy in vivo[28,41,45]. Evidence suggests that these observations are
likely explained by the improved pharmacokinetics of
site-specific relative to conventional conjugates, and the
fact that some site placements yield longer ADC half-
lives than others [46]. The latter point highlights a
particular strength of site placement: the ability to per-
form structure activity relationship (SAR) studies, where
the effect of payload placement on biophysical (e.g.,
aggregation) and functional parameters (e.g., pharmaco-
kinetics and efficacy) can be determined. While a number
of groups have demonstrated that site placement matters,
consistent rules for payload placement have yet to be
developed. It is likely that optimal site placement may
Current Opinion in Chemical Biology 2015, 28:174–180 19
178 Synthetic biomolecules
vary according to the particular conjugation chemistries
and linker-payloads employed.
A multipronged strategy may be needed tominimize ADC targeting of healthy tissuesWhile the approaches outlined to improve ADC stability
and efficacy can improve the therapeutic index, at a
certain point this treatment window will be limited by
the tumor specificity of the target antigen. A recent
bioinformatics study surveyed healthy tissue mRNA
expression of target antigens for 27 ADCs that are
currently in the clinic [16��]. Most target antigens
showed poor differential expression between tumors
and normal tissues and were also widely expressed
across organs. This study highlights the fact that addi-
tional innovations are needed in order to improve ADC
tumor targeting. Target antigen selection is becoming
more sophisticated [7], and is beginning to acknowledge
the roles played by biological function, absolute — as
Figure 3
(4) Antigen-specifictumor targeting
(3) Site-specificconjugation
(2) Payloadand DAR
(1) Linker/conjugation chemistryCurrent Opinion in Chemical Biology
Design considerations for building better ADCs. (1) The choice of
conjugation and linker chemistry will govern many of the biophysical and
functional characteristics of the resulting ADC. Linker design can alter
stability, membrane permeability, and resistance to P-glycoprotein-
mediated efflux — all of which can impact efficacy. Furthermore, both
the linker and the conjugation chemistry can affect the overall
hydrophobicity of the construct, which directly relates to in vivo stability,
PK, and tolerability. (2) The choice of cytotoxic payload and the DAR
affect both efficacy and safety. The latter is at least partially dependent
on the conjugation and linker chemistry for success, that is, higher DAR
ADCs tend to need more solubilizing linkers to prolong PK and improve
efficacy. (3) The use of site-specific conjugation facilitates analytics and
offers the ability to perform structure-activity relationship mapping to
identify optimal ADC configurations for the best efficacy and safety. (4)
Modifications, perhaps to the CDR or Fc regions, should be introduced
to improve the specificity of ADC targeting to the tumor in order to
reduce off-target toxicity in healthy tissues. Whereas points 1–3 have
been explored to some extent by the community and a consensus
opinion has begun to form around how to approach these
considerations (as indicated by the solid arrows), point 4 represents a
new direction for the field and relatively little has been done as yet in
this vein (as indicated by the dashed arrow). It will remain to be seen
how this next challenge is best addressed.
Current Opinion in Chemical Biology 2015, 28:174–180
well as relative — expression levels, internalization pa-
rameters, and biomarker availability [16��,17]; judicious
choices may lead to more specific drug delivery. Another
approach is to reduce the visibility of healthy tissue anti-
gens to ADCs by either predosing with unconjugated
antibody to saturate binding sites [47], or by using anti-
bodies with masked CDRs that are unmasked by tumor-
specific proteases to allow localized target binding [48].
ConclusionsThe emerging strategies for improving ADC performance
focus on tunable ADC features that affect ADC stability,
potency, and targeting efficiency — which together de-
termine the efficacy and safety profiles (Figure 3). These
in turn define the therapeutic index and govern whether
an efficacious dose can be safely administered. Increasing
the therapeutic index has been the motivation driving
ADC development; we are beginning to understand the
rules of the game and can start moving the ball down the
field towards that goal.
AcknowledgementsThe authors would like to thank Patrick Holder for fruitful discussions andRaghuveer Parthasarathy for Figure 1.
References and recommended readingPapers of particular interest, published within the period of review,have been highlighted as:
� of special interest�� of outstanding interest
1. Wang J, Watanabe JH: AAPS Advances in the PharmaceuticalSciences Series. Springer International Publishing; 2015.
2. Bross PF, Beitz J, Chen G, Chen XH, Duffy E, Kieffer L, Roy S,Sridhara R, Rahman A, Williams G et al.: Approval summary:gemtuzumab ozogamicin in relapsed acute myeloid leukemia.Clin Cancer Res 2001, 7:1490-1496.
3. Chari RVJ, Miller ML, Widdison WC: Antibody–drug conjugates:an emerging concept in cancer therapy. Angew Chem Int Ed2014, 53:3796-3827.
4. Moore K, Martin L, Seward S, Bauer T, O’Malley D, Perez R, Oza A,Jeong W, Kirby M, Zhou Y et al.: Preliminary single agent activityof IMGN853, a folate receptor alpha (FRa)-targeting antibody–drug conjugate (ADC), in platinum-resistant epithelial ovariancancer (EOC) patients (pts): Phase I trial. Abstract 5518, ASCO,Annual Meeting; Chicago, IL: 2015.
5. Sedlacek HH: Antibodies As Carriers of Cytotoxicity. 1992.
6. Erickson HK, Lambert JM: ADME of antibody–maytansinoidconjugates. AAPS J 2012, 14:799-805.
7. Bouchard H, Viskov C, Garcia-Echeverria C: Antibody–drugconjugates — a new wave of cancer drugs. Bioorg Med ChemLett 2014, 24:5357-5363.
8.��
Liu Y, Zhang X, Han C, Wan G, Huang X, Ivan C, Jiang D,Rodriguez-Aguayo C, Lopez-Berestein G, Rao PH et al.: TP53 losscreates therapeutic vulnerability in colorectal cancer. Nature2015 http://dx.doi.org/10.1038/nature14418.
In this highly innovative study, Liu et al. used tumor genomics data toidentify a vulnerable drug target (POLR2A) that was made functionallyhemizygous following genomic deletion of the neighboring TP53 gene.POLR2A is the catalytic subunit of the RNA polymerase II complex and isinhibited by a-amanitin. An ADC composed of an a-amanitin conjugatedto an anti-EpCAM antibody was highly efficacious in mouse models ofhuman colorectal cancer. This information-rich approach to developingADC therapeutics will hopefully be emulated.
www.sciencedirect.com
20
An emerging playbook for antibody–drug conjugates Drake and Rabuka 179
9. Hong EE, Chari R: AAPS Advances in the Pharmaceutical SciencesSeries. Springer International Publishing; 2015.
10. McCombs JR, Owen SC: Antibody drug conjugates: design andselection of linker, payload and conjugation chemistry. AAPS J2015, 17:339-351.
11. Lyon R: Novel ADC Chemistry for Improved Stability andPharmacokinetics. PEGS Boston; 2014.
12.�
Hamblett KJ, Senter PD, Chace DF, Sun MMC, Lenox J,Cerveny CG, Kissler KM, Bernhardt SX, Kopcha AK, Zabinski RFet al.: Effects of drug loading on the antitumor activity of amonoclonal antibody drug conjugate. Clin Cancer Res 2004,10:7063-7070.
Hamblett et al. purify ADC populations with defined DARs and test themfor in vitro potency and in vivo efficacy, pharmacokinetics, and toxicity.They observe that DAR 8 species have faster clearance and a lowertherapeutic index as compared to DAR 2 and DAR 4 species.
13.��
Saber H, Leighton JK: An FDA oncology analysis of antibody–drug conjugates. Regul Toxicol Pharmacol 2015, 71:444-452.
Saber and Leighton analyze data from IND filings for 20 ADCs and assessthe relationships between nonclinical study results and Phase I out-comes. They conclude that the linker-payload, not the target antigen,drives toxicity and limits dose levels.
14. Sukumaran S, Lin K: AAPS Advances in the PharmaceuticalSciences Series. Springer International Publishing; 2015.
15. Polson AG, Calemine-Fenaux J, Chan P, Chang W, Christensen E,Clark S, de Sauvage FJ, Eaton D, Elkins K, Elliott JM et al.:Antibody–drug conjugates for the treatment of non-Hodgkin’slymphoma: target and linker-drug selection. Cancer Res 2009,69:2358-2364.
16.��
Damelin M, Zhong W, Myers J, Sapra P: Evolving strategies fortarget selection for antibody–drug conjugates. Pharm Res 2015http://dx.doi.org/10.1007/s11095-015-1624-3.
Damelin et al. provide a bioinformatics study that reveals the limiteddifferences in tumor versus healthy tissue expression levels for mostADCs in the clinic, highlighting a salient problem facing the field. Theyalso review current and emerging approaches for target antigenselection.
17. Sadekar S, Figueroa I, Tabrizi M: Antibody drug conjugates:application of quantitative pharmacology in modality designand target selection. AAPS J 2015 http://dx.doi.org/10.1208/s12248-015-9766-0.
18. Bander NH: Antibody–drug conjugate target selection: criticalfactors. Methods Mol Biol 2013, 1045:29-40.
19. Tumey LN, Charati M, He T, Sousa E, Ma D, Han X, Clark T,Casavant J, Loganzo F, Barletta F et al.: Mild method forsuccinimide hydrolysis on ADCs: impact on ADC potency,stability, exposure, and efficacy. Bioconjug Chem 2014,25:1871-1880.
20. Boswell CA, Mundo EE, Zhang C, Bumbaca D, Valle NR,Kozak KR, Fourie A, Chuh J, Koppada N, Saad O et al.: Impact ofdrug conjugation on pharmacokinetics and tissue distributionof anti-STEAP1 antibody–drug conjugates in rats. BioconjugChem 2011, 22:1994-2004.
21. Herbertson RA, Tebbutt NC, Lee F-T, MacFarlane DJ, Chappell B,Micallef N, Lee S-T, Saunder T, Hopkins W, Smyth FE et al.: PhaseI biodistribution and pharmacokinetic study of LewisY-targeting immunoconjugate CMD-193 in patientswith advanced epithelial cancers. Clin Cancer Res 2009,15:6709-6715.
22.�
Burke PJ, Hamilton JZ, Jeffrey SC, Hunter JH, Doronina SO,Okeley NM, Anderson ME, Senter PD, Lyon RP: Optimization of aPEGylated glucuronide-auristatin linker for antibody–drugconjugates. Abstract 648, AACR; Philadelphia, PA: 2015.
Burke et al. show that by increasing the hydrophilicity of linker-payloads,they can generate DAR 8 ADCs with half-lives near that of the unconju-gated antibody, and with improved therapeutic indexes relative to morehydrophobic ADCs.
23. Alley SC, Benjamin DR, Jeffrey SC, Okeley NM, Meyer DL,Sanderson RJ, Senter PD: Contribution of linker stability to theactivities of anticancer immunoconjugates. Bioconj Chem2008, 19:759-765.
www.sciencedirect.com
24. Lyon RP, Setter JR, Bovee TD, Doronina SO, Hunter JH,Anderson ME, Balasubramanian CL, Duniho SM, Leiske CI, Li Fet al.: Self-hydrolyzing maleimides improve the stability andpharmacological properties of antibody–drug conjugates. NatBiotechnol 2014, 32:1059-1062.
25. Drake PM, Rabuka D: Antibody–drug conjugates: can couplingcytotoxicity and specificity overcome therapeutic resistance?Resistance to Targeted Anti-Cancer Therapeutics. Springer; 2013::183-200.
26. Kovtun YV, Audette CA, Mayo MF, Jones GE, Doherty H,Maloney EK, Erickson HK, Sun X, Wilhelm S, Ab O et al.: Antibody–maytansinoid conjugates designed to bypass multidrugresistance. Cancer Res 2010, 70:2528-2537.
27. Kellogg BA, Garrett L, Kovtun Y, Lai KC, Leece B, Miller M,Payne G, Steeves R, Whiteman KR, Widdison W et al.: Disulfide-linked antibody–maytansinoid conjugates: optimization of invivo activity by varying the steric hindrance at carbon atomsadjacent to the disulfide linkage. Bioconjug Chem 2011,22:717-727.
28. Strop P, Liu S-H, Dorywalska M, Delaria K, Dushin RG, Tran T-T,Ho W-H, Farias S, Casas MG, Abdiche Y et al.: Location matters:site of conjugation modulates stability and pharmacokineticsof antibody drug conjugates. Chem Biol 2013, 20:161-167.
29. van der Lee MMC, Groothuis PG, Ubink R, van der Vleuten MAJ,van Achterberg TA, Loosveld EM, Damming D, Jacobs DCH,Rouwette M, Egging DF et al.: The preclinical profile of theduocarmycin-based HER2-targeting ADC SYD985 predicts forclinical benefit in low HER2-expressing breast cancers. MolCancer Ther 2015, 14:692-703.
30. Wang H, Rangan VS, Sung M-C, Passmore D, Kempe T, Wang X,Thevanayagam L, Pan C, Rao C, Srinivasan M et al.:Pharmacokinetic characterization of BMS-936561, an anti-CD70 antibody–drug conjugate, in preclinical animal speciesand prediction of its pharmacokinetics in humans. BiopharmDrug Dispos 2015 http://dx.doi.org/10.1002/bdd.1953.
31.��
Uppal H, Doudement E, Mahapatra K, Darbonne WC, Bumbaca D,Shen BQ, Du X, Saad O, Bowles K, Olsen S et al.: Potentialmechanisms for thrombocytopenia development withtrastuzumab emtansine (T-DM1). Clin Cancer Res 2015,21:123-133.
Uppal et al. address the mechanisms underlying thrombocytopeniafollowing T-DM1 dosing by looking at the effects of the ADC and itscomponents on primary human platelets and hematopoietic stem cells invitro. They observe a DM1-induced impairment of megakaryocyte differ-entiation that is mediated by T-DM1 binding to FcgRIIa expressed on thedeveloping megakaryocytes.
32. Bruhns P: Properties of mouse and human IgG receptorsand their contribution to disease models. Blood 2012,119:5640-5649.
33. Panowski S, Bhakta S, Raab H, Polakis P, Junutula JR: Site-specific antibody drug conjugates for cancer therapy. mAbs2013, 6:34-45.
34. Behrens CR, Liu B: Methods for site-specific drug conjugationto antibodies. mAbs 2013, 6:46-53.
35. Jeffrey SC, Burke PJ, Lyon RP, Meyer DW, Sussman D,Anderson M, Hunter JH, Leiske CI, Miyamoto JB, Nicholas NDet al.: A potent anti-CD70 antibody–drug conjugate combininga dimeric pyrrolobenzodiazepine drug with site-specificconjugation technology. Bioconjug Chem 2013, 24:1256-1263.
36. Kung Sutherland MS, Walter RB, Jeffrey SC, Burke PJ, Yu C,Kostner H, Stone I, Ryan MC, Sussman D, Lyon RP et al.: SGN-CD33A: a novel CD33-targeting antibody–drug conjugateusing a pyrrolobenzodiazepine dimer is active in models ofdrug-resistant AML. Blood 2013, 122:1455-1463.
37. Liu S-H, Strop P, Tran T-T, Dorywalsk M, Delaria K, Ho W-H,Dushin R, Pons J, Rajpal A, Shelton D: RN927C, a potent site-specific Trop-2 antibody–drug-conjugate (ADC) for treatmentof solid tumors. Cancer Res 2014, 74 Abstract 2640; AACRAnnual Meeting, April 2014, San Diego, CA.
38.�
Junutula JR, Raab H, Clark S, Bhakta S, Leipold DD, Weir S,Chen Y, Simpson M, Tsai SP, Dennis MS et al.: Site-specific
Current Opinion in Chemical Biology 2015, 28:174–180
21
180 Synthetic biomolecules
conjugation of a cytotoxic drug to an antibody improves thetherapeutic index. Nat Biotechnol 2008, 26:925-932.
This paper by Junutula et al. offered the first example of a site-specificallyconjugated ADC, showing the promise of this approach towardsimproved safety and efficacy.
39. Zimmerman ES, Heibeck TH, Gill A, Li X, Murray CJ,Madlansacay MR, Tran C, Uter NT, Yin G, Rivers PJ et al.:Production of site-specific antibody–drug conjugates usingoptimized non-natural amino acids in a cell-free expressionsystem. Bioconjug Chem 2014, 25:351-361.
40. Axup JY, Bajjuri KM, Ritland M, Hutchins BM, Kim CH, Kazane SA,Halder R, Forsyth JS, Santidrian AF, Stafin K et al.: Synthesis ofsite-specific antibody–drug conjugates using unnatural aminoacids. Proc Natl Acad Sci 2012, 109:16101-16106.
41. Drake PM, Albers AE, Baker J, Banas S, Barfield RM, Bhat AS, deHart GW, Garofalo AW, Holder P, Jones LC et al.: Aldehyde tagcoupled with HIPS chemistry enables the production of ADCsconjugated site-specifically to different antibody regions withdistinct in vivo efficacy and PK outcomes. Bioconjug Chem2014, 25:1331-1341.
42. Albers AE, Garofalo AW, Drake PM, Kudirka R, de Hart GW,Barfield RM, Baker J, Banas S, Rabuka D: Exploring the effectsof linker composition on site-specifically modified antibody–drug conjugates. Eur J Med Chem 2014, 88:3-9.
Current Opinion in Chemical Biology 2015, 28:174–180
43. Jackson D, Atkinson J, Guevara CI, Zhang C, Kery V, Moon S-J,Virata C, Yang P, Lowe C, Pinkstaff J et al.: In vitro and in vivoevaluation of cysteine and site specific conjugated herceptinantibody–drug conjugates. PLoS ONE 2014, 9:e83865.
44. Junutula JR, Flagella KM, Graham RA, Parsons KL, Ha E, Raab H,Bhakta S, Nguyen T, Dugger DL, Li G et al.: Engineered thio-trastuzumab-DM1 conjugate with an improved therapeuticindex to target human epidermal growth factor receptor 2-positive breast cancer. Clin Cancer Res 2010, 16:4769-4778.
45. Shen B-Q, Xu K, Liu L, Raab H, Bhakta S, Kenrick M, Parsons-Reponte KL, Tien J, Yu S-F, Mai E et al.: Conjugation sitemodulates the in vivo stability and therapeutic activity ofantibody–drug conjugates. Nat Biotechnol 2012, 30:184-189.
46. Rabuka D: Distinct in vivo efficacy and PK outcomes of ADCsconjugated site specifically using SMARTagTM technology. 2015.
47. Boswell CA, Mundo EE, Zhang C, Stainton SL, Yu SF, Lacap JA,Mao W, Kozak KR, Fourie A, Polakis P et al.: Differential effects ofpredosing on tumor and tissue uptake of an 111In-labeledanti-TENB2 antibody–drug conjugate. J Nucl Med 2012,53:1454-1461.
48. Sagert J, West J, Wong C, Desnoyers L, Vasiljeva O, Richardson J,Polu K, Lowman H: Transforming Notch ligands into tumor-antigentargets: a probody-drug conjugate (PDC) targeting Jagged 1 andJagged 2.. 2014.
www.sciencedirect.com
22
Every molecule has a challenge. We have a smart biologic solution.Call: + 1 888 SOLUTION (765-8846) Email: solutions@catalent.com Visit: www.catalentbiologics.com
smartag™ protein conjugate technology. Our advanced protein conjugate
technology uses novel, site-
specific protein modification and
linker technologies to create safer,
more stable ADCs with enhanced
potency and optimal efficacy.
˝ 2
016
Cat
alen
t Ph
arm
a So
luti
ons.
All
righ
ts r
eser
ved.
SMA
RTag
is a
tra
dem
ark
of C
atal
ent
Phar
ma
Solu
tion
s.
DEVELOPMENT DELIVERY SUPPLY
SITE SPECIFICProgrammable drug placement enables more uniform attachment of the conjugate at a specific site for improved ADC product consistency and better regulatory compliance.
MODIFIABLEPrecise location control maximizes ADC performance.Placement flexibility results in more homogeneousADCs, optimized PK, and improved efficacy and safety.
ALDEHYDE Combination of novel aldehyde tagging technology and naturally-occurring amino acid sequence (leveraging enzyme modification) delivers enhanced efficacy.
S
M
A
RECOMBINANTUsed with recombinant antibodies, the technology works with any cell line expression system—resulting in simplified analytics and efficient, scalable processing.
R
TAGAbility to incorporate multiple tags enables more potent payloads. Optimization of linker-toxin combinations for specific targets increases efficacy.
T
biologics
lable at ScienceDirect
European Journal of Medicinal Chemistry 88 (2014) 3e9
Contents lists avai
European Journal of Medicinal Chemistry
journal homepage: http: / /www.elsevier .com/locate/ejmech
Original article
Exploring the effects of linker composition on site-specificallymodified antibodyedrug conjugates
Aaron E. Albers, Albert W. Garofalo, Penelope M. Drake, Romas Kudirka,Gregory W. de Hart, Robyn M. Barfield, Jeanne Baker, Stefanie Banas, David Rabuka*
Redwood Bioscience, 5703 Hollis Street, Emeryville, CA 94608, USA
a r t i c l e i n f o
Article history:Received 5 June 2014Received in revised form21 August 2014Accepted 22 August 2014Available online 23 August 2014
Keywords:Antibodyedrug conjugateNoncleavable linkerAldehyde tag
* Corresponding author.E-mail address: drabuka@redwoodbioscience.com
http://dx.doi.org/10.1016/j.ejmech.2014.08.0620223-5234/© 2014 Elsevier Masson SAS. All rights re
a b s t r a c t
In the context of antibodyedrug conjugates (ADCs), noncleavable linkers provide a means to delivercytotoxic small molecules to cell targets while reducing systemic toxicity caused by nontargeted releaseof the free drug. Additionally, noncleavable linkers afford an opportunity to change the chemicalproperties of the small molecule to improve potency or diminish affinity for multidrug transporters,thereby improving efficacy. We employed the aldehyde tag coupled with the hydrazino-iso-Pictet-Spengler (HIPS) ligation to generate a panel of site-specifically conjugated ADCs that varied only in thenoncleavable linker portion. The ADC panel comprised antibodies carrying a maytansine payload ligatedthrough one of five different linkers. Both the linker-maytansine constructs alone and the resulting ADCpanel were characterized in a variety of in vitro and in vivo assays measuring biophysical and functionalproperties. We observed that slight differences in linker design affected these parameters in disparateways, and noted that efficacy could be improved by selecting for particular attributes. These studies serveas a starting point for the exploration of more potent noncleavable linker systems.
© 2014 Elsevier Masson SAS. All rights reserved.
1. Introduction
Antibodyedrug conjugates (ADCs) promise to alter the land-scape of anti-cancer therapeutics by targeting highly cytotoxic drugmolecules directly to cancer cells. The success of currentlyapproved ADCs has inspired a spate of research and developmentefforts in the area; dozens of new ADCs are in pre-clinical or clinicaltrials [1]. ADCs comprise a monoclonal antibody, a cytotoxicpayload, and a linker that joins them together [2]. The monoclonalantibody targets the payload to cells expressing the antigen on theirsurface, and the cytotoxic payload kills the cells upon internaliza-tion of the ADC. The linker is literally the central component of anADC; it contains the reactive group that governs the conjugationchemistry, and serves as a chemical spacer that physically connectsthe drug payload to the antibody. As such, the linker is also themostversatile aspect of the ADC. It can be modified in any number ofways to influence various drug/linker characteristics (e.g., solubil-ity) [3,4] and ADC properties (e.g., potency, pharmacokinetics,therapeutic index, and efficacy in multidrug resistant cells) [5e11].
(D. Rabuka).
served.
There are essentially two broad classes of ADC linkers; thosethat are chemically labile or enzymatically-cleavable, and thosethat are chemically stable or noncleavable [12]. Labile/cleavablelinkers are designed to keep the ADC intact when in circulation butrelease the drug payload upon internalization by the target cell.Some cytotoxic payloadsdfor example, MMAEdrequire a cleavablelinker, as they do not tolerate substitutions [13,14]. By contrast,other cytotoxic payloadsdfor example, maytansinedcan accom-modate substitutions while maintaining potency [15]. Such drugsare good substrates for the development of noncleavable linkers. Bydesign, noncleavable linkers do not contain chemical functional-ities that are readily susceptible to intracellular degradation.Therefore, after an internalized ADC is trafficked to the lysosome,the antibody moiety is proteolytically degraded into amino acidswhile the cytotoxic drug remains attached via the linker to anamino acid residue [16]. The retention of the linker as part of theactive metabolite allows for the modulation of the overall proper-ties of the metabolite (e.g., by altering hydrophobicity, length, andcharge) in order to improve potency.
We previously reported a novel site-specific ligation chemistrythat takes advantage of an aldehyde-tagged protein [17]. Thealdehyde tag is a straightforward means of site-specifically func-tionalizing proteins for chemical modification. The genetically-encoded tag consists of a pentapeptide sequence (CXPXR) that is
24
A.E. Albers et al. / European Journal of Medicinal Chemistry 88 (2014) 3e94
specifically recognized by formylglycine-generating enzyme (FGE)[18-20]. During protein expression in cells, the cysteine residue inthe sequence is recognized by FGE and oxidized co-translationallyto formylglycine. The resulting aldehyde affords a bioorthogonalchemical handle for ligation (Fig. 1). Linkers terminating in a 2-((1,2-dimethylhydrazinyl)methyl)-1H-indole react with the alde-hyde by way of a hydrazino-iso-Pictet-Spengler (HIPS) reaction toform an azacarboline, resulting in a stable CeC bond joining theantibody and payload.
The aldehyde tag platform allows for site-specific conjugationthat yields a highly homogenous product. Accordingly, this tech-nology is well-suited for performing structure activity relationshipstudies in the context of an intact ADC. Here, we isolated linkercomposition as a single variable for optimization while the otherADC componentsdantibody backbone, cytotoxic payload, conju-gation site, drug-to-antibody ratio, and conjugation chemis-trydwere held constant. By characterizing a panel of five drug/linkers and their corresponding conjugates, we explored the impactof small changes in linker design on ADC potency and stability.
2. Materials and methods
2.1. Linker synthesis
Synthetic routes and analytical data are provided in theSupplemental materials.
2.2. Microtubule polymerization assay
We used the Tubulin Polymerization Assay Kit (Cytoskeleton)according to the manufacturer's instructions for the fluorescence-based test. All test articles were used at 3 mM.
2.3. Direct ELISA antigen binding
Maxisorp 96-well plates (Nunc) were coated overnight at 4 �Cwith 1 mg/mL of human HER2-His (Sino Biological) in PBS. The platewas blocked with ELISA blocker blocking buffer (ThermoFisher),and then the aHER2wild-type antibody and ADCswere plated in an8-step series of 2-fold dilutions starting at 100 ng/mL. The platewasincubated, shaking, at room temperature for 2 h. After washing inPBS 0.1% Tween-20, bound analyte was detected with a donkeyanti-human Fc-g-specific horseradish peroxidase (HRP)-conju-gated secondary antibody. Signals were visualized with Ultra TMB(Pierce) and quenched with 2 N H2SO4. Absorbance at 450 nmwasdetermined using a Molecular Devices SpectraMax M5 plate readerand the data were analyzed using GraphPad Prism.
Fig. 1. The aldehyde tag coupled with HIPS ligation yields site-specifically modified aenzyme (FGE) recognition sequence (CXPXR) is site-specifically inserted into the backboneThe aldehyde of formylglycine can then be reacted with nucleophiles to form a stable CeC
2.4. Bioconjugation, purification, and HPLC analytics
Humanized anti-HER2 IgG antibodies (15 mg/mL) bearing thealdehyde tag (LCTPSR) at the C-terminus of the heavy chain wereconjugated to maytansine-containing drug linkers (8 mol equiva-lents drug:antibody) for 72 h at 37 �C in 50 mM sodium citrate,50mMNaCl pH 5.5 containing 0.85% DMA and 0.085% Triton X-100.Free drug was removed using tangential flow filtration. Unconju-gated antibody was removed using preparative-scale hydrophobicinteraction chromatography (HIC; GE Healthcare 17-5195-01) withmobile phase A: 1.0 M ammonium sulfate, 25 mM sodium phos-phate pH 7.0, and mobile phase B: 25% isopropanol, 18.75 mM so-dium phosphate pH 7.0. An isocratic gradient of 33% B was used toelute unconjugated material, followed by a linear gradient of41e95% B to elute mono- and diconjugated species. To determinethe DAR of the final product, ADCs were examined by analytical HIC(Tosoh #14947) with mobile phase A: 1.5 M ammonium sulfate,25 mM sodium phosphate pH 7.0, and mobile phase B: 25% iso-propanol, 18.75 mM sodium phosphate pH 7.0. To determine ag-gregation, samples were analyzed using analytical size exclusionchromatography (SEC; Tosoh #08541) with a mobile phase of300 mM NaCl, 25 mM sodium phosphate pH 6.8.
2.5. In vitro cytotoxicity
The HER2-positive breast carcinoma cell line, NCI-N87, wasobtained from ATCC and maintained in RPMI-1640 medium (Cell-gro) supplemented with 10% fetal bovine serum (Invitrogen) andGlutamax (Invitrogen). 24 h prior to plating, cells were passaged toensure log-phase growth. On the day of plating, 5000 cells/wellwere seeded onto 96-well plates in 90 mL normal growth mediumsupplemented with 10 IU penicillin and 10 mg/mL streptomycin(Cellgro). Cells were treated at various concentrations with 10 mL ofdiluted analytes, and the plates were incubated at 37 �C in an at-mosphere of 5% CO2. After 6 d, 100 mL/well of CellTiter-Glo reagent(Promega) was added, and luminescence was measured using aMolecular Devices SpectraMax M5 plate reader. GraphPad Prismsoftware was used for data analysis, including IC50 calculations.
2.6. In vitro stability
ADCs were spiked into rat plasma at ~1 pmol (payload)/mL. Thesamples were aliquoted and stored at �80 �C until use. Aliquotswere placed at 37 �C under 5% CO2 for the indicated times and thenwere analyzed by ELISA to assess the anti-maytansine and anti-Fabsignals. A freshly thawed aliquot was used as a reference startingvalue for conjugation. All analytes were measured together on oneplate to enable comparisons across time points. First, analytes were
ntibodies. Using standard molecular biology techniques, a formylglycine-generatingof the antibody. FGE co-translationally oxidizes the cysteine residue to formylglycine.bond.
25
A.E. Albers et al. / European Journal of Medicinal Chemistry 88 (2014) 3e9 5
diluted in blocking buffer to 20 ng/mL (within the linear range ofthe assay). Then, analytes were captured on plates coated with ananti-human Fab-specific antibody. Next, the payload was detectedwith an anti-maytansine antibody followed by an HRP-conjugatedsecondary; the total antibody was detected with a directly conju-gated anti-human Fc-specific antibody. Bound secondary antibodywas visualized with TMB substrate. The colorimetric reaction wasstopped with H2SO4, and the absorbance at 450 nm was deter-mined using a Molecular Devices SpectraMaxM5 plate reader. Dataanalysis was performed in Excel. Each sample was analyzed inquadruplicate, and the average values were used. The ratio of anti-maytansine signal to anti-Fab signal was used as a measure ofantibody conjugation.
Fig. 2. Inclusion of amino acid residues resulted in highly soluble maytansine-linker constructs with varied chemical composition. Five different maytansine-conjugated linkers were synthesized (as shown in Scheme 1) and characterized,both as free drugs and after conjugation to an a-HER2 antibody.
2.7. Xenograft studies
The animal studies were approved by Charles River LaboratoriesInstitutional Animal Care and Use Committee (IACUC). Female C.B-17 SCIDmicewere inoculated subcutaneously with 1�107 NCI-N87tumor cells in 50% Matrigel. When the tumors reached an averageof 112 mm3, the animals were given a single 5 mg/kg dose of ADC,trastuzumab antibody (untagged), or vehicle alone. The animalswere monitored twice weekly for body weight and tumor size.Tumor volume was calculated using the formula:
Tumor volume�mm3
�¼ w2 � l
2
where w ¼ tumor width and l ¼ tumor length.Tumor doubling times were obtained by averaging the tumor
growth rate curves from four groups of mice. Then, log10 cell killwas estimated using the formula:
log10 cell kill ¼ treated group TTE� control group TTE3:32� tumor doubling time
Treatment over control (T/C) ratios were determined by dividingthe tumor volume of the treatment group by the tumor volume ofthe control group at a designated time point.
3. Results and discussion
3.1. Linker design and synthesis
To examine the effect of linker composition, we tested a varietyof maytansine-linkers that contained functional groups anticipatedto aid in solubility, which improves bioconjugation yields [4].Initially, we used PEGn spacers (with n ¼ 2, 4, or 6), but found thatthe PEG group alone was not sufficiently hydrophilic to overcomethe very hydrophobic contributions from the maytansine and HIPScomponents. The conjugation efficiencies observed with linkerscontaining PEGn spacers alone were poor, e.g., 40% yield with aPEG6-maytansine linker conjugated to a C-terminally-tagged anti-body.We found that a simpleway to incorporate hydrophilicity wasby using amino acid residues as linker components (Fig. 2). In turn,this change resulted in a significant improvement in conjugationefficiency, e.g., 90% yield with a glutamic acid PEG2-maytansinelinker conjugated to a C-terminally-tagged antibody. Here, wetested the effect of using different amino acids as solubilizingagents by evaluating glutamic acid (Linkers 1, 4, and 5), asparagine(Linker 2), and phosphotyrosine (Linker 3). The latter was meant tofunction as a pro-drug, where the phosphorylated form would besoluble, but not membrane permeable. Once inside a cell, the linkerwas intended to be a substrate for phosphorylases, the action ofwhich would yield a more hydrophobic and membrane-permeable
active metabolite. We also incorporated a spacer element into thelinkersdeither PEG2 or n-propyldto improve conjugation effi-ciency and mitigate ADC aggregation. Finally, taking advantage ofthe hydrazino-iso-Pictet-Spengler (HIPS) chemistry, the linkersterminated in either a reactive 2-((1,2-dimethylhydrazinyl)methyl)indole (1, 2, 3, and 5) or 2-((1,2-dimethylhydrazinyl)methyl)pyrrolo[2,3-b]pyridine (4). The latter varied from the former by a singlenitrogen atom (Fig. 2), making it slightly more hydrophilic. Bothreactive groups enabled HIPS ligation of the linker-maytansine toaldehyde-tagged antibodies for ADC production.
A representative synthesis of the linkers is shown in Scheme 1.In the example, a pegylated, protected amino acid, 6, is coupled topentafluorophenyl ester, 7. The product, 8, is then coupled to N-deacetylmaytansine, 9, using HATU followed by hydrolysis of thetert-butyl ester and removal of the Fmoc-protecting group withpiperidine to give the final desired product, 1.
3.2. Linker composition did not alter the payload's ability to inhibitmicrotubule polymerization
As a first step, once the drug/linkers were in hand, we per-formed an in vitro microtubule polymerization assay to confirmthat the incorporated structural variations and elaborations tomaytansine did not impair the drug's ability to inhibit microtubulepolymerization (Fig. 3). As anticipated, due to the known toleranceof maytansine to substitutions at the N-acyl position [21], the panelof drug/linkers resulted in microtubule polymerization inhibitionsimilar to unmodified maytansine. A small spread of values wasnoted, but all werewithin 32% of maytansine itself. As shown in thenext section, these small differences did not appear to impact theIC50 of the drug/linkers when formulated as an ADC.
3.3. Bioconjugation and in vitro assessment of the ADC panel
Conjugation of the drug/linkers to a C-terminally aldehydetagged a-HER2 antibody was carried out by treating the antibody at37 �C with 8e10 equivalents of linker-maytansine in 50 mM so-dium citrate, 50 mMNaCl pH 5.5 containing 0.85% DMA and 0.085%Triton X-100, and the progress of the reaction was tracked byanalytical hydrophobic interaction chromatography (HIC). Uponcompletion, the excess payload was removed by tangential flow
26
Scheme 1. Representative linker synthesis.
0
500
1000
1500
2000
2500
3000
3500
4000
0 5 10 15 20 25 30 35 40 45 50 55 60
Fluo
resc
ence
(Arb
itrar
yU
nits
)
Minutes
Buffer
Maytansine
Paclitaxel
Vinblastine
Linker 1
Linker 2
Linker 3
Linker 4
Linker 5
Fig. 3. Linker selection does not hinder maytansine inhibition of microtubulepolymerization. The drug/linkers were tested in a microtubule polymerization assay.Free maytansine was included as a positive control. Paclitaxel and vinblastine wereincluded as a promoter and inhibitor, respectively, of microtubule polymerization. Thebuffer control indicates the polymerization rate of untreated microtubules. Relative tofree maytansine, the drug/linkers inhibited polymerization by the following amounts:Linker 1, 68%; Linker 2, 92%; Linker 3, 79%; Linker 4, 105%; Linker 5, 73%.
A.E. Albers et al. / European Journal of Medicinal Chemistry 88 (2014) 3e96
filtration, and the unconjugated antibody was removed by pre-parative HIC. These reactions were high yielding, with >90%conjugation efficiency (Table S1). After purification, the ADCscontained an average drug-to-antibody ratio (DAR) of 1.6 asdetermined by hydrophobic interaction chromatography(Figure S2). The drug distribution (ratio of DAR 1e2) was verysimilar among the ADCs made with different linkers (Table S2). Thepreparations were �95% monomeric as assessed by size-exclusionchromatography (data not shown).
With the ADCs in hand, we first askedwhether conjugationwiththe drug/linkers had altered the antibody affinity for the HER2antigen. To test this, we performed a direct-binding ELISA assayusing plates coated with human HER2-His, and compared the EC50of thewild-type (untagged and unconjugated) a-HER2 to the valuesobtained for the panel of a-HER2 ADCs (Fig. 4). Only minimal dif-ferences in affinity were noted, with most of the ADCs appearing tobind with slightly higher affinity than the wild-type antibody.
Next, we tested the in vitro cytotoxicity of the ADCs against theHER2-overexpressing gastric cell line, NCI-N87 (Fig. 5A). As acomparator, we also tested the cytotoxic activity of the corre-sponding free drug/linkers (Fig. 5B). Cell cultures were exposed to
varying concentrations of the analytes for 6 days, and then cellviability was measured by using a CellTiter-Glo assay, whichquantifies ATP levels. All ADCs exhibited picomolar activity, withIC50 values similar to or better than that observed after treatmentwith free maytansine. By contrast to the ADCs, the free drug/linkerswere overall less potent, generally showing IC50 values that were1000- to 2000-fold higher than the corresponding ADCs. Linkers 1and 4, which shared a glutamic acid-PEG2 scaffold, both had freedrug/linker IC50 values above 1 mM, ~10,000-fold higher than theADC versions of those compounds. In addition to the IC50 values, wenoted that the Linker 3 ADC, in spite of its measured picomolaractivity, failed to kill more than 70% of the cells, even at the highestdoses (Fig. 5A). The free version of Linker 3 did not suffer from thissame cytotoxic plateau, reducing cell viability by >93% at thehighest dose (Fig. 5B). We observed the same trends with the un-conjugated and conjugated versions of Linker 3 on a differentantibody and against a different cell line, suggesting that theplateau effect of this linker is translatable across platforms.Although the cytotoxic plateau is commonly observed in thesetypes of assays, the underlying mechanisms involved and the bio-logical significance of the effect is not clear.
As a final in vitro characterization of the ADC panel, we exam-ined the stability of the HIPS-conjugates in plasma for 14 days at37 �C. The assay consisted of an ELISA-basedmethod that comparedthe ratio of anti-payload to anti-Fc signals. As a group, the conju-gates exhibited a high degree of stability, with �85% payloadremaining after 7 days and �74% payload remaining after 14 days(Table 1). The glutamic acid-PEG2-containing scaffolds were themost stable over 14 days, both demonstrating more than 80%retention of payload. The most labile linker, Linker 2, only differedfrom the most stable linker by about 10% over 14 days.
3.4. In vivo efficacy of the ADC panel
To test the in vivo efficacy of the ADC panel, we assessed theconjugates using an NCI-N87 xenograft model in SCID mice. Com-pounds were administered as a single 5 mg/kg dose at the onset ofthe study. All ADCs were well-tolerated with no animal showing>10% weight loss up to 40 days post-treatment (Figure S1). Tumorgrowth was arrested, and some tumors were reduced in size aftertreatment with the a-HER2 ADCs (Fig. 6A), but not after treatmentwith the isotype control ADC (conjugated using Linker 1). Eventu-ally, tumors began to regrow in all animals, sooner in some groupsthan others, depending on the ADC used for treatment. By 60e70days post-dose, there were clear differences in mean tumor vol-umes among groups treated with an a-HER2 ADC; specifically, the
27
0.1 1 10 1000
1
2
3
[Ab ng/ml]
Abs
orba
nce
(A.U
.)Wild-type αHER2Linker 1Linker 2Linker 3Linker 4Linker 5
Fig. 4. a-HER2 ADCs conjugated with the panel of drug/linkers maintain theiraffinity for HER2 antigen. A direct-binding ELISA assay, using immobilized humanHER2-His protein as the capture reagent and an anti-human Fc-g-specific HRP-conjugated antibody as the detection reagent, was used to monitor the affinity of a-HER2 conjugates relative to the wild-type antibody. Standard curves of each analytewere generated and the EC50 value of each curve was calculated using GraphPad Prism.The calculated EC50 values (nM) were: Wild-type antibody, 8.72; Linker 1, 6.20; Linker2, 10.80; Linker 3, 7.10; Linker 4, 5.95; Linker 5, 7.23.
Table 1ADCs made with different linkers show similar stability in plasma at 37 �C.
ADC % Conjugate remaining after 7days
% Conjugate remaining after 14days
aHER2-Linker 1
93 81
aHER2-Linker 2
85 74
aHER2-Linker 3
93 77
aHER2-Linker 4
97 83
aHER2-Linker 5
95 77
A.E. Albers et al. / European Journal of Medicinal Chemistry 88 (2014) 3e9 7
mean tumor volumes ranged from 249 to 487 mm3 at day 60(Fig. 6A). In order to investigate this effect, we looked at thelog10 cell kill for tumors dosed with the various treatments(Table 2). The results indicated that treatment with ADCs conju-gated to Linkers 1 and 4 killed more tumor cells as compared totreatment with the other ADCs. Notably, these two linkers repre-sented the absolute minimum amount of chemical diversity, bothcontained the glutamic acid-PEG2 scaffold and differed from eachother by only a single nitrogen group in the azacarboline that formsduring ligation. This increased potency translated into a survivaladvantage for animals treated with ADCs conjugated to Linkers 1 or4 (Fig. 6B).
The efficacy of Linker 5 was reduced as compared to Linkers 1and 4, with which it shared the glutamic acid moiety. The results ofthis series of linkers suggest that, in this context, inclusion of the n-propyl spacer reduced efficacy as compared to the PEG2 spacer. Theother two linkers, which incorporated different amino acids on thePEG2 scaffold, had varying efficacy. Linker 2 showed an interme-diate log10 cell kill value (reflecting total cells killed throughout thecourse of the study), but was the best performer in the first 10 days
Fig. 5. Small changes in linker composition do not influence the in vitro cytotoxicity ofcytotoxicity in a 6 day assay. Free maytansine (black line) was included as a positive controcontrol to indicate specificity. (A) ADC IC50 values (reflecting the antibody concentrations exLinker 1, 170 pM; Linker 2, 160 pM; Linker 3, 110 pM; Linker 4, 96 pM; Linker 5, 120 pM;measured as follows: free maytansine, 405 pM; Linker 1, 1.58 mM; Linker 2, 342.5 nM; Link
of the study, reducing tumor volume more than any other treat-ment (Fig. 6A). Linker 3 had the poorest in vivo efficacy (p < 0.007,by the log-rank, ManteleCox, test). It is interesting to considerwhether the incomplete in vitro killing of NCI-N87 target cells byADCs conjugated to this linker is related todor perhaps predictiveofdits reduced in vivo efficacy as compared to the other ADCs.
Next, we selected two linkers from the initial panel to take into amultidose efficacy study. We chose Linker 1 on the merits of itsoverall potency, as measured by tumor growth, log10 cell kill, andsurvival. We chose Linker 2 because it showed the fastest initialtumor reduction, and we reasoned that perhaps this quick responsewould translate into increased efficacy in a multidose setting. Themultidose study employed NCI-N87 tumors in SCID mice. Animalswere dosed (10mg/kg) once aweek for four weeks. The experimentemployed two armsdwith dosing beginning when tumors reachedaverage volumes of either 180 or 400 mm3. a-HER2 ADCs madewith both Linkers 1 and 2 were highly active against the smallertumors (Fig. 7A), and resulted in very similar levels of tumor con-trol. By contrast, against the larger tumors, the a-HER2 ADC madewith Linker 2 showed superior efficacy, resulting in a greater levelof tumor inhibition as compared to the ADC made with Linker 1(Fig. 7B). Specifically, the treated/control tumor volumes at day 42were 0.39 and 0.26 for Linkers 1 and 2, respectively.
In conclusion, we developed a panel of C-terminally-conjugateda-HER2 ADCs bearing highly similar linkers, and observed thatrelatively minor structural changes led to dramatic differences inpotency both in vitro and in vivo against the NCI-N87 tumor model.Other biophysical parameters were less impacted. Specifically, weobserved only minor effects of linker architecture on inhibition ofmicrotubule polymerization (at the free drug/linker level), andantibody affinity (at the ADC level). With respect to in vitro cyto-toxicity, as a group, the ADCs were highly efficacious and yieldedvery similar IC50 values, with less than a 2-fold difference
aHER2 ADCs. NCI-N87 cells, which overexpress HER2, were used as targets for in vitrol, and an isotype control ADC (gray line) conjugated to Linker 1 was used as a negativecept in the case of the free drug) were measured as follows: free maytansine, 250 pM;isotype control ADC, could not be determined. (B) Free drug/linker IC50 values wereer 3, 125.8 nM; Linker 4, ~1 mM; Linker 5, 274.9 nM.
28
0
200
400
600
800
1000
1200
1 8 15 22 29 36 47 53 60 67
Mea
ntu
mor
volu
me
(mm
3 )
Days post-dose
A.
50 1000
50
100
Days post-dose
%S
urvi
val
B.
Linker 1
Linker 2
Linker 3
Linker 4
Linker 5
Isotype ADC, Linker 1
Fig. 6. Linker composition affects the in vivo efficacy of aldehyde-tagged a-HER2 ADCs in an NCI-N87 tumor model. CB.17 SCID mice (8/group) were implanted subcutaneouslywith NCI-N87 cells. When the tumors reached ~113 mm3, the animals were given a single 5 mg/kg dose of an a-HER2 conjugated to Linkers 1-5 or of an isotype control antibodyconjugated to Linker 1. (A) Tumor growth was monitored twice weekly. (B) The differences in efficacy among the ADCs tested were reflected in survival curves. Animals wereeuthanized when tumors reached 800 mm3 or on day 112 of the study, whichever occurred first.
Table 2In vivo log10 cell kill of NCI-N87 tumor cells achieved by a single 5 mg/kg ADCdose.
a-HER2 ADC linker composition Log10 cell kill
Linker 1 1.24Linker 2 0.82Linker 3 0.65Linker 4 1.22Linker 5 0.92
A.E. Albers et al. / European Journal of Medicinal Chemistry 88 (2014) 3e98
encompassing the entire panel. However, one ADCdconjugatedwith Linker 3dexhibited a striking viability plateau in vitro, with32% viable cells remaining at the highest doses. By contrast to theADCs, the potency of the free drug/linkers varied morewidely, witha 12-fold difference encompassing the range of IC50 values. Inter-estingly, the rank order potency of the ADC did not directly corre-late with that of the free drug/linker. Furthermore, the
0
200
400
600
800
1000
1200
0 10 20 30 40 50 60 70
Mea
ntu
mor
volu
me
(mm
3 )
Days
VehicleLinker 1 Linker 2
A.
Fig. 7. Multidose xenograft studies reveal differences in efficacy against larger tumors bsubcutaneously with NCI-N87 cells. Tumors were allowed to grow to either ~180 or 400 mmonce a week for four weeks with 10 mg/kg of an a-HER2 ADC conjugated to Linkers 1 or 2. Aeuthanized when tumors reached 800 mm3.
“completeness” of the cell cytotoxicity was not always the samebetween the corresponding ADC and free drug/linker analytes. Forexample, treatment with free Linker 3 abrogated all but 7% of theviable cells. With respect to in vivo cytotoxicity, ADCsmadewith allof the linkers inhibited growth of the HER2-overexpressing NCI-N87 xenograft to some extent. However, the log10 cell kill valuesachieved by the ADCs varied by up to 2-fold, and the median sur-vival time among the groups differed by 17 days, indicating thatlinker structure affected efficacy. Furthermore, we observed a dif-ference in the kinetics of tumor response to ADCs made withdistinct linkers, e.g., Linker 2 vs. Linker 1, whereby Linker 2 wasmore efficacious in the short term (1 wk), but the response wasshort lived. We were able to capitalize on this difference in a followup multidose xenograft study, in which the ADC with faster cyto-toxic kinetics showed superior efficacy against larger tumors.Therefore, we demonstrated that the sensitivity of our system tolinker design affords an opportunity to engineer next-generationADCs with optimized characteristics for improved efficacy.
0
200
400
600
800
1000
1200
0 10 20 30 40 50 60 70
Tum
orvo
lum
e(m
m3 )
Days
VehicleLinker 1 Linker 2
Days
B.
etween ADCs made with Linkers 1 and 2. CB.17 SCID mice (8/group) were implanted3 (Panels A and B, respectively) and then treatment was initiated. Animals were dosedrrows indicate dosing days. Tumor growth was monitored twice weekly. Animals were
29
A.E. Albers et al. / European Journal of Medicinal Chemistry 88 (2014) 3e9 9
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.ejmech.2014.08.062.
References
[1] J.A. Flygare, T.H. Pillow, P. Aristoff, Antibody-drug conjugates for the treat-ment of cancer, Chem. Biol. Drug. Des. 81 (2012) 113e121.
[2] R.V.J. Chari, M.L. Miller, W.C. Widdison, Antibody-drug conjugates: anemerging concept in cancer therapy, Angew. Chem. Int. Ed. Engl. 53 (2014)3796e3827.
[3] R.Y. Zhao, H.K. Erickson, B.A. Leece, E.E. Reid, V.S. Goldmacher, J.M. Lambert,R.V.J. Chari, Synthesis and biological evaluation of antibody conjugates ofphosphate prodrugs of cytotoxic DNA alkylators for the targeted treatment ofcancer, J. Med. Chem. 55 (2012) 766e782.
[4] R.Y. Zhao, S.D. Wilhelm, C. Audette, G. Jones, B.A. Leece, A.C. Lazar,V.S. Goldmacher, R. Singh, Y. Kovtun, W.C. Widdison, J.M. Lambert, R.V.J. Chari,Synthesis and evaluation of hydrophilic linkers for antibodyemaytansinoidconjugates, J. Med. Chem. 54 (2011) 3606e3623.
[5] H.K. Erickson, G.D. Lewis Phillips, D.D. Leipold, C.A. Provenzano, E. Mai,H.A. Johnson, B. Gunter, C.A. Audette, M. Gupta, J. Pinkas, J. Tibbitts, The effectof different linkers on target cell catabolism and pharmacokinetics/pharma-codynamics of trastuzumab maytansinoid conjugates, Mol. Cancer Ther. 11(2012) 1133e1142.
[6] S.C. Jeffrey, J.B. Andreyka, S.X. Bernhardt, K.M. Kissler, T. Kline, J.S. Lenox,R.F. Moser, M.T. Nguyen, N.M. Okeley, I.J. Stone, X. Zhang, P.D. Senter,Development and properties of b-glucuronide linkers for monoclonal anti-bodyedrug conjugates, Bioconjug. Chem. 17 (2006) 831e840.
[7] G.D. Lewis Phillips, G. Li, D.L. Dugger, L.M. Crocker, K.L. Parsons, E. Mai,W.A. Blattler, J.M. Lambert, R.V.J. Chari, R.J. Lutz, W.L.T. Wong, F.S. Jacobson,H. Koeppen, R.H. Schwall, S.R. Kenkare-Mitra, S.D. Spencer, M.X. Sliwkowski,Targeting HER2-positive breast cancer with trastuzumab-DM1, an antibody-cytotoxic drug conjugate, Cancer Res. 68 (2008) 9280e9290.
[8] S.O. Doronina, T.D. Bovee, D.W. Meyer, J.B. Miyamoto, M.E. Anderson,C.A. Morris-Tilden, P.D. Senter, Novel peptide linkers for highly potent anti-bodyeauristatin conjugate, Bioconjug. Chem. 19 (2008) 1960e1963.
[9] S.O. Doronina, B.A. Mendelsohn, T.D. Bovee, C.G. Cerveny, S.C. Alley,D.L. Meyer, E. Oflazoglu, B.E. Toki, R.J. Sanderson, R.F. Zabinski, A.F. Wahl,P.D. Senter, Enhanced activity of monomethylauristatin F through monoclonalantibody delivery: effects of linker technology on efficacy and toxicity, Bio-conjug. Chem. 17 (2006) 114e124.
[10] S.C. Alley, D.R. Benjamin, S.C. Jeffrey, N.M. Okeley, D.L. Meyer, R.J. Sanderson,P.D. Senter, Contribution of linker stability to the activities of anticancerimmunoconjugates, Bioconjug. Chem. 19 (2008) 759e765.
[11] P. Strop, S.-H. Liu, M. Dorywalska, K. Delaria, R.G. Dushin, T.-T. Tran, W.-H. Ho,S. Farias, M.G. Casas, Y. Abdiche, D. Zhou, R. Chanrasekaran, C. Samain, C. Loo,A. Rossi, M. Rickert, S. Krimm, T. Wong, S.M. Chin, J. Yu, J. Dilley, J. Chaparro-Riggers, G.F. Filzen, C.J. O'Donnell, F. Wang, J.S. Myers, J. Pons, D.L. Shelton,A. Rajpal, Location matters: site of conjugation modulates stability andpharmacokinetics of antibody drug conjugates, Chem. Biol. 20 (2013)161e167.
[12] P.M. Drake, D. Rabuka, in: B. Bonavida (Ed.), Resistance to Targeted Anti-Cancer Therapeutics, Springer New York, New York, NY, 2013, pp. 183e200.
[13] P.R. Hamann, R.G. Dushin, AntibodyeDrug Conjugates in Oncology (Chapter7), in: New Frontiers in Chemical Biology: Enabling Drug Discovery, The RoyalSociety of Chemistry, 2011, pp. 224e257. JF e JO e VL e IS e.
[14] S.O. Doronina, B.E. Toki, M.Y. Torgov, B.A. Mendelsohn, C.G. Cerveny,D.F. Chace, R.L. DeBlanc, R.P. Gearing, T.D. Bovee, C.B. Siegall, J.A. Francisco,A.F. Wahl, D.L. Meyer, P.D. Senter, Development of potent monoclonal anti-body auristatin conjugates for cancer therapy, Nat. Biotechnol. 21 (2003)778e784.
[15] B.A. Kellogg, L. Garrett, Y. Kovtun, K.C. Lai, B. Leece, M. Miller, G. Payne,R. Steeves, K.R. Whiteman, W. Widdison, H. Xie, R. Singh, R.V.J. Chari,J.M. Lambert, R.J. Lutz, Disulfide-linked antibodyemaytansinoid conjugates:optimization of in vivo activity by varying the steric hindrance at carbonatoms adjacent to the disulfide linkage, Bioconjug. Chem. 22 (2011) 717e727.
[16] X. Sun, W. Widdison, M. Mayo, S. Wilhelm, B. Leece, R. Chari, R. Singh,H. Erickson, Design of antibodyemaytansinoid conjugates allows for efficientdetoxification via liver metabolism, Bioconjug. Chem. 22 (2011) 728e735.
[17] P. Agarwal, R. Kudirka, A.E. Albers, R.M. Barfield, G.W. de Hart, P.M. Drake,L.C. Jones, D. Rabuka, Hydrazino-pictet-spengler ligation as a biocompatiblemethod for the generation of stable protein conjugates, Bioconjug. Chem. 24(2013) 846e851.
[18] I.S. Carrico, B.L. Carlson, C.R. Bertozzi, Introducing genetically encoded alde-hydes into proteins, Nat. Chem. Biol. Nat. Publ. Group 3 (2007) 321e322.
[19] P. Wu, W. Shui, B.L. Carlson, N. Hu, D. Rabuka, J. Lee, C.R. Bertozzi, Site-specificchemical modification of recombinant proteins produced in mammalian cellsby using the genetically encoded aldehyde tag, Proc. Natl. Acad. Sci. U. S. A.106 (2009) 3000e3005.
[20] D. Rabuka, J.S. Rush, G.W. deHart, P. Wu, C.R. Bertozzi, Site-specific chemicalprotein conjugation using genetically encoded aldehyde tags, Nat. Protoc. 7(2012) 1052e1067.
[21] Y.V. Kovtun, C.A. Audette, M.F. Mayo, G.E. Jones, H. Doherty, E.K. Maloney,H.K. Erickson, X. Sun, S. Wilhelm, O. Ab, K.C. Lai, W.C. Widdison, B. Kellogg,H. Johnson, J. Pinkas, R.J. Lutz, R. Singh, V.S. Goldmacher, R.V.J. Chari, Anti-body-maytansinoid conjugates designed to bypass multidrug resistance,Cancer Res. 70 (2010) 2528e2537.
30
Regulatory Toxicology and Pharmacology 71 (2015) 444–452
Contents lists available at ScienceDirect
Regulatory Toxicology and Pharmacology
journal homepage: www.elsevier .com/locate /yr tph
An FDA oncology analysis of antibody-drug conjugates q
http://dx.doi.org/10.1016/j.yrtph.2015.01.0140273-2300/Published by Elsevier Inc.
Abbreviations: AD, acceptable dose; ADC, antibody-drug conjugate; ADCC,antibody-dependent cellular cytotoxicity; FIH, first-in-human; HNSTD, highestnon-severely toxic dose; ICH, International Conference on Harmonization; LC/MS,liquid chromatography/mass spectrometry; mc, maleimidocaproyl linker; mAb,monoclonal antibody; MTD, maximum tolerated dose; NHP, non-human primate;NOAEL, no observed adverse effect level; PD, pharmacodynamics; PK, pharmacoki-netics; RP2D, recommended Phase 2 dose; SM, small molecule; STD10, severelytoxic dose in 10% of the animals; vc, valine–citrulline linker.
q Disclaimer: This article reflects the views of the authors and should not beconstrued to represent FDA’s views or policies.⇑ Corresponding author.
E-mail address: john.leighton@fda.hhs.gov (J.K. Leighton).
Haleh Saber, John K. Leighton ⇑US Food and Drug Administration, Center for Drug Evaluation and Research, Office of Hematology and Oncology Products, 10903 New Hampshire Ave, Silver Spring, MD 20903,United States
a r t i c l e i n f o
Article history:Received 16 January 2015Available online 7 February 2015
Keywords:Antibody-drug conjugatesFirst-in-human start dose selectionOncology drug development
a b s t r a c t
Antibody-drug conjugates (ADCs) are complex molecules composed of monoclonal antibodies conjugatedto potent cytotoxic agents through chemical linkers. Because of this complexity, sponsors have used dif-ferent approaches for the design of nonclinical studies to support the safety evaluation of ADCs and first-in-human (FIH) dose selection. We analyzed this data with the goal of describing the relationshipbetween nonclinical study results and Phase 1 study outcomes. We summarized the following data frominvestigational new drug applications (INDs) for ADCs: plasma stability, animal study designs and toxi-cities, and algorithms used for FIH dose selection. Our review found that selecting a FIH dose that is1/6th the highest non-severely toxic dose (HNSTD) in cynomolgus monkeys or 1/10th the STD10 inrodents scaled according to body surface area (BSA) generally resulted in the acceptable balance of safetyand efficient dose-escalation in a Phase 1 trial. Other approaches may also be acceptable, e.g. 1/10th theHNSTD in monkeys using BSA or 1/10th the NOAEL in monkeys or rodents using body weight for scaling.While the animal data for the vc-MMAE platform yielded variable range of HNSTDs in cynomolgus mon-keys, MTDs were in a narrow range in patients, suggesting that for ADCs sharing the same small moleculedrug, linker and drug:antibody ratio, prior clinical data can inform the design of a Phase 1 clinical trial.
Published by Elsevier Inc.
1. Introduction
Antibody-drug conjugates (ADCs) are a unique class of drugsconsisting of a monoclonal antibody (mAb) conjugated to a cyto-toxic drug (herein referred to as small molecules; SM) through achemical linker. Most ADCs in development are intended to treatcancer in patients with serious and life-threatening disease. Theantibody component generally directs the SM to a specific epitopeon the cancer cell where the ADC is internalized and releases thecytotoxic SM. The proposed advantage of developing an ADC con-jugate is to direct high concentrations of the SM to the tumor(via tumor-specific or overexpressed antigens) and thereby reduce
systemic exposure to the free SM and associated off-target adverseeffects (Sievers and Senter, 2013; Cancer Drug Discovery andDevelopment; Antibody-Drug Conjugates and Immunotoxins,2013; Roberts et al., 2013). Interest in developing ADCs for thetreatment of solid tumors and hematologic malignancies hasgrown in recent years with the introduction of new technologies,the validation of more tumor targets for anti-cancer therapies,and the approval of two ADCs. The recent FDA approval of twoADCs (brentuximab vedotin and ado-trastuzumab emtansine)and submission of completed Phase 1 study results for investiga-tional new drug applications (INDs) for ADCs, has provided enoughdata for FDA to conduct a preliminary summary analysis of non-clinical development programs for ADCs. FDA reviewed the non-clinical safety data submitted to support 20 separate INDapplications for ADCs with an emphasis on FIH dose selection, plas-ma stability, toxicities in animals, and toxicology study designs.
In the current generation of ADCs, the SMs used are genotoxicagents that target rapidly dividing cells (e.g., crypt cells and bonemarrow) in general toxicology studies by directly interacting withthe DNA or components of DNA synthesis or cell division. The SMsselected for use in ADCs are typically potent and poorly toleratedwhen used as free agents. The SM is covalently attached to theantibody via a linker intended to provide stability in plasma. The
31
H. Saber, J.K. Leighton / Regulatory Toxicology and Pharmacology 71 (2015) 444–452 445
linker generally provides enough stability to allow the ADC toremain intact while in circulation and is labile enough to allowthe release of the small molecule drug after internalization insidethe cell. For protease-resistant noncleavable linkers, the SM isreleased after internalization of the ADC and proteolytic degrada-tion of the antibody moiety. The first generation of linkers includedthe acid-cleavable linker used in Mylotarg (ADC half-life of 1–2 days), but improvements in linker technologies have producedmore stable ADCs with half-lives of approximately one week(Ducry and Stump, 2010). Based on the INDs examined, linkersused were disulfide-based linkers, peptidic linkers such as the pro-tease cleavable valine–citrulline (vc) linker, and protease-resistantlinkers such as the maleimidocaproyl (mc) linker. The SMs con-tained in ADCs used in our analysis were mostly MMAE, MMAF,DM1, and DM4; other small molecules were also included, e.g.MED2460 (a DNA alkylating agent) and calicheamicin.
Estimating a FIH dose that is reasonably safe in preliminary clin-ical studies is an essential goal of a nonclinical development pro-gram. Ideally, the proposed FIH dose for anticancer therapiesshould avoid unacceptable toxicities while minimizing the time ittakes to deliver sufficient exposure to have the intended pharmaco-logical activity. Approaches used to estimate the FIH dose for smallmolecules or biological products may include applying a safety fac-tor to doses identified as tolerable in animal toxicity studies. The ani-mal dose can be converted to the human equivalent dose by scalingaccording to body surface area (BSA), as is typically done for smallmolecules, or according to body weight (BW) as done for some largemolecules, or less frequently by application of a hybrid PK/PD model(FDA, 2005; Haddish-Berhane et al., 2013).
Currently, there is no published guidance regarding method-ology for FIH dose selection for ADCs. Sponsors of IND applicationshave proposed various algorithms such as traditional approachesused for small molecules in oncology, approaches used for biologi-cal products in oncology or non-oncology settings, and scalingeither to BW or BSA. In our review, we reanalyzed animal toxicol-ogy data for 20 ADCs (2 approved and 18 in development) andcompared different methods for estimating the FIH dose basedon either: 1/10th STD10 (using BW or BSA for animal-to-humanconversions), 1/6th the HNSTD (using BW or BSA for conversion),1/10th HNSTD (using BW or BSA for conversion) and 1/10th NOAEL(using BW and BSA for conversion).
2. Methods
2.1. Data collection methodology
The FDA archival database was queried December 2012 throughAugust 2013 to identify INDs for inclusion in the analysis. Anti-body-drug conjugates selected were those that fell within thescope of ICH S9, i.e., drugs to treat advanced malignancies. Theintention was to select INDs that had completed Phase 1 clinicalinvestigations; although as discussed later for a few ADCs, thesponsors decided to inactivate the IND. The nonclinical programsfor these INDs were examined for the toxicology design, toxicityprofile seen in animals, determination of the FIH dose, and plasmastability. The toxicology studies previously reviewed by the FDAwere peer reviewed by one author (HS) and the following informa-tion was collected for each IND, when available: NOAEL, STD10,HNSTD, the isoform and target epitope of the antibody moiety,the type of linker, the structure and mechanism of action of thesmall molecule moiety, the human maximum tolerated dose(MTD) or recommended Phase 2 dose (RP2D). The time to reachthe human MTD/RP2D was identified as the time from the IND sub-mission to the date that the MTD/RP2D was reported. When anMTD or a RP2D was not identified, a human acceptable dose (AD)
was identified (see section below for the definition of AD). FIH doseapproaches were evaluated and the correlation between HNSTDsand human MTD, RP2D, or AD was assessed. Two ADCs in this ana-lysis have been approved; ado-trastuzumab emtansine and bren-tuximab vedotin. None of the ADCs examined had modificationson their antibody for reduced or increased antibody-dependentcellular cytotoxicity activity (ADCC), or site-specific linker attach-ment (e.g. by introducing non-conventional amino acids). Anti-body-drug conjugates that contain bacterial or plant-derivedtoxins and ADCs with radiolabeled SMs were excluded from thisanalysis.
Out of the 20 INDs examined, 16 were analyzed for FIH doseapproaches using data generated in cynomolgus monkeys withthe ADC. Four INDs were excluded due to the following reasons:IND inactivated before an MTD could be established (3 INDs), thetoxicology study design did not support the frequency of adminis-tration in patients (1 IND). For the latter IND, single doses of theADC were administered to monkeys; however, patients receivedtwice weekly doses of the ADC; the first dose in patients was abovethe MTD and doses were subsequently reduced.
Of the 20 INDs examined, 14 were analyzed for FIH doseapproaches using data generated in rodents with the ADC. Six INDswere excluded due to reasons described above or because the ADCwas not used in the rodents (either that a rodent study was notconducted or it was conducted with the small molecule only).
Plasma stability data were reviewed for 8 INDs; several INDsdid not contain in vitro plasma stability.
All 20 INDs were examined for: toxicities in animals, pharmaco-logic relevance of the animal species, and design of toxicologystudies. Of the 20 INDs, 16 were used to obtain the time betweenthe IND submission and reaching an MTD/RP2D/AD.
2.2. Selection of STD10, HNSTD and NOAEL
The IND-enabling studies in rodents and in monkeys were peer-reviewed by one author (HS) to define or confirm the STD10
(rodents), the HNSTD (monkeys) or the NOAEL in both species. Inevery application reviewed, monkeys were the non-rodent specieschosen for toxicity testing. If a dose exceeded the STD10 (e.g. morethan 10% mortality), the next lower dose was defined as the STD10.Similarly, if a dose in the toxicology study in monkeys resulted inunacceptable toxicities, a dose below it was defined as the HNSTD.Estimation of an STD10 or the HNSTD based on modeling as report-ed in some INDs was not done. Eight potential FIH doses were cal-culated from the pivotal toxicology studies using six approaches:1/6th HNSTD (BW), 1/6th HNSTD (BSA), 1/10th HNSTD (BW), 1/10th HNSTD (BSA), 1/10th STD10 (BW), 1/10th STD10 (BSA), 1/10th NOAEL (BW), or 1/10th NOAEL (BSA).
2.3. FIH-dose selection and its proximity to the human MTD
The human MTD or RP2D was obtained from the sponsor’s sub-mission or publicly available information when such informationcould not be obtained from the FDA archive. The date when theMTD/RP2D was first reported and the source of information wererecorded. If the human MTD/RP2D was not reached or the informa-tion was not available, the safety database in FDA’s archival systemwas searched to obtain an ‘‘acceptable dose’’ (AD). An AD is onethat is not clearly identified as being the MTD, but was the nextlowest dose clinically tested. This could be a dose that was expand-ed to further evaluate adverse effects, e.g. a dose above it presentedwith dose-limiting toxicities (DLTs) and hence the sponsor reducedthe dose and expanded the number of patients at the lower dose toevaluate toxicities.
To provide consistent comparison across INDs when assessingthe various approaches to setting the FIH dose, it was necessary
32
446 H. Saber, J.K. Leighton / Regulatory Toxicology and Pharmacology 71 (2015) 444–452
to normalize the number of dose escalations required to reach theMTD/RP2D/AD by using an assumption that all escalations were100% (i.e., dose-doubling). The point of this exercise was to exam-ine various approaches to setting a start dose relative to the humanMTD/RP2D/AD, not to assess appropriate escalation schemes. TheFIH dose approach was considered acceptable if it resulted in adose that was below the MTD/RP2D/AD and took at least one dosedoubling to reach the human MTD. Of note, ‘‘1 dose doubling’’being acceptable indicates that the first two dose cohorts did notshow DLTs. The FIH dose algorithm was considered unsafe if thedose obtained was at or above the human MTD/RP2D/AD or if adose doubling of the FIH dose would be above the human MTD/RP2D/AD. While acceptable, the FIH dose may be too low if 4 ormore artificial dose doublings are required to reach the humanMTD; this is based on the observation that in Phase 1 trials of ADCs,except for the first dose escalation, actual escalations were usuallyless than 100% and 4 artificial escalations in our analysis maytranslate into more than 7 actual escalations in a Phase 1 trial.See an example of how the number of escalations to reach thehuman MTD/RP2D/AD was estimated using toxicity data from ratsand monkeys for ADCs #1 and 2 (Table 1).
3. Results
3.1. Design of IND-enabling GLP toxicology studies
For the INDs examined, general toxicology studies were con-ducted in 2 species, a rodent and a non-rodent. The rodent wasmouse or rat; the non-rodent was cynomolgus monkey. The study
Table 1Examples of possible FIH start dose calculations and estimated escalations to reach a hum
GLP toxicology study FIH dos
BSA con
ADC#1MTD = 3 mg/kg (Q3 W dosing)
Cynomolgus monkeyADC, given IV, Weekly � 4, at 1, 3,6 mg/kg
1/6th HThree d2.5 mg/
HNSTD: 6 mg/kg (72 mg/m2) 1/10thFour dohuman
NOAEL: 3 mg/kg (36 mg/m2) 1/10thFour dohuman
RatADC given IV, Weekly � 4, at 3, 10,30 mg/kgSTD10: 30 mg/kg (180 mg/m2)
1/10thTwo dohuman
NOAEL: 3 mg/kg (18 mg/m2) 1/10thkg)Five dohuman
ADC#2RP2D dose = 3.6 mg/kg (Q3 Wdosing)
Cynomolgus monkeyADC, given IV; single dose at 3, 10,and 30 mg/kg
1/6th HOne do
HNSTD = 30 mg/kg (360 mg/m2) 1/10thOne do
NOAEL = 3 mg/kg (36 mg/m2) 1/10thFive do
Cynomolgus monkeyADC, given IV; every 3 weeks � 4 at0, 3, 10, and 30 mg/kg
1/6th HTwo do1/10th
HNSTD = 10 mg/kg (120 mg/m2) Three dNOAEL: not identified 1/10 NRatADC, IV, at 6, 20, 60 mg/kg
1/10th
STD10 = 20 mg/kg (120 mg/m2) Three d
NOAEL: not identified 1/10th
Q3W: every 3 weeks.Note: Each dose escalation is assumed to be 100% increase from the previous dose.
designs used varied greatly; however, a consistent feature was thatall sponsors used 3 (or more) dose levels of the ADC in the pivotalstudy conducted in cynomolgus monkeys. The ADC was not alwaysused in the rodent. Some sponsors also studied the effects of thesmall molecule alone, the linker-small molecule, or the free anti-body. These studies were conducted as separate studies or armswere added into the study conducted with the ADC.
3.2. Toxicities in animals
If the ADC did not bind to the animal target, toxicities in rodentand cynomolgus monkey in the IND-enabling studies wereobserved mainly in the hematopoietic system, liver, and reproduc-tive organs for all ADCs examined and also in the skin and kidneysfor some ADCs (Table 2). Toxicities to the hematopoietic system,liver, and reproductive organs were directly related to the SM.Toxicities to the kidneys observed for some ADCs may be a directeffect or may have been secondary to an immunogenic response.For the ADCs examined, the antibody did not bind to the targetin rodents but many bound to the intended target in cynomolgusmonkey. When the target was minimally expressed in the monkey(e.g. target was a tumor antigen), toxicities were comparable to theINDs where no binding was reported. When the target was highlyexpressed, DLTs were mainly due to the toxicities in thehematopoietic system and liver toxicity did not occur in the IND-enabling toxicology studies, with the only exception being ado-trastuzumab emtansine. While ado-trastuzumab emtansine recog-nizes the target epitope in the monkey and the target is highlyexpressed in this species, hepatotoxicity was evident in single dose
an MTD.
e calculation
version BW conversion
NSTD: 12 mg/m2 (0.3 mg/kg) 1/6th HNSTD: 1 mg/kgose escalations will result inkg
One dose escalation will result in a humandose of 2 mg/kg
HNSTD: 7.2 mg/m2 (0.19 mg/kg) 1/10th HNSTD = 0.6 mg/kgse escalations will result in adose of 3 mg/kg
Two dose escalation will result in a humandose of 2.4 mg/kg
NOAEL: 3.6 mg/m2 (0.1 mg/kg) 1/10th NOAEL: 0.3 mg/kgse escalations will result in adose of 1.6 mg/kg
Three dose escalations will result in a humandose of 2.4 mg/kg
STD10 = 18 mg/m2 (0.5 mg/kg)se escalations will result in adose of 2 mg/kg
1/10th STD10 = 3 mg/kgThis dose is at the human MTD. No doseescalation could be done.
NOAEL = 1.8 mg/m2 (0.05 mg/ 1/10th NOAEL = 0.3 mg/kg
se escalations will result in adose of 1.6 mg/kg
Three does escalations will result in a humandose of 2.4 mg/kg
NSTD: 60 mg/m2 (1.6 mg/kg) 1/6th HNSTD: 5 mg/kgse escalation to 3.2 mg/kg The dose is above the RP2D
HNSTD: 36 mg/m2 (1 mg/kg) 1/10th HNSTD: 3 mg/kgse escalation to 2 mg/kg Dose escalation will result in a dose above the
RP2DNOAEL: 3.6 mg/m2 (0.1 mg/kg) 1/10th NOAEL: 0.3 mg/kgse escalations to 3.2 mg/kg Three dose escalations to 2.4 mg/kgNSTD: 20 mg/m2 (0.5 mg/kg) 1/6 HNSTD: 1.6 mg/kgse escalations to 2 mg/kg One dose escalation to 3.2 mg/kgHNSTD: 12 mg/m2 (0.3 mg/kg) 1/10 HNSTD: 1 mg/kgose escalations to 2.4 mg/kg One dose escalation to 2 mg/kg
OAEL: cannot be done 1/10 NOAEL: cannot be doneSTD10 = 12 mg/m2 (0.3 mg/kg) 1/10th STD10 = 2 mg/kg
ose escalations to 2.5 mg/kg Dose escalation will result in a dose that isabove 3.6 mg/kg human dose
NOAEL: Could not be done 1/10th NOAEL = could not be done
33
Table 2Examples of INDs with prominent organ toxicities observed in nonclinical studies. Each row represents a separate ADC.
Target binding in cynomolgusmonkeys/Linker-SM
Rodents Cynomolgus monkeys
Yes1/mc-MMAF
Hematopoietic system, kidney, testes, skin, lung, liver, uterus Hematopoietic system, kidney, testes, skin, lung, liver, uterus
Yes1/Peptide-based-MED2460
Hematopoietic system, liver, male and female reproductive organs Hematopoietic system, liver, kidney, male and femalereproductive organs
NI2/vc-MMAE
Hematopoietic system, liver, kidney, skin, male reproductive organs Hematopoietic system
No/disulfide-DM4
Hematopoietic system, liver, GI tract, kidney, male and female reproorgans, nerve cells, skin
Hematopoietic system, liver, GI tract, kidney, male and femalereproductive organs, nerve cells, skin
Yes/vc-MMAE
Hematopoietic system, liver, male reproductive organs Hematopoietic system
Yes/vc-MMAE
Hematopoietic system, liver, male reproductive organs Hematopoietic system
Yes/vc-MMAE
Hematopoietic system, liver, male reproductive organs Hematopoietic system
Yes3/Disulfide-DM4
Hematopoietic system, GI tract liver, eye, skin, neuropathy, male andfemale reproductive organs
Hematopoietic system, GI tract liver, neuropathy, eye, skin
No/Disulfide-DM4
Hematopoietic system, GI tract, liver, male and female reproductiveorgans, skin (lesions)
Hematopoietic system, GI tract, liver, male and femalereproductive organs., skin (lesions)
No/Acid-cleavable-calicheamicin
Hematopoietic system, liver, GI tract, male and female reproductiveorgans, kidney and peripheral neuropathy
Hematopoietic system, liver, GI tract, male and femalereproductive organs
Yes/vc-MMAE
Hematopoietic system, liver, male reproductive organs Hematopoietic system, male reproductive organs
Yesmc-DM1
Hematopoietic system, liver, kidney, male reproductive organs Hematopoietic system, liver, kidney, neuropathy
1 Target is minimally expressed in cynomolgus monkeys.2 NI: no information.3 Minimal binding/mainly surface epithelium cells.
H. Saber, J.K. Leighton / Regulatory Toxicology and Pharmacology 71 (2015) 444–452 447
toxicity studies. Neurotoxicity was observed for several ADCswhere the SM was a microtubule inhibitor, as is expected for thisclass of cytotoxic agents.
3.3. Plasma stability
It is assumed that the higher the stability of the ADCs in plasma,the less early release of the small molecule and potentially the lessoff-target toxicities. Plasma stabilities were therefore examined tosee whether lower plasma stabilities resulted in lower STD10,HNSTD, or human MTDs. For many INDs, in vitro plasma stabilitystudies were not conducted, for some of the INDs this was possiblydue to the development programs preceding the publication of theICH S9 guidance. In vitro plasma stability was measured by incuba-tion of the ADC in human or animal plasma, usually for 96 h. Sta-bility was measured by LC/MS/MS or by an ELISA assay. The lownumber of INDs (n = 8) with plasma stability data and differentapproaches used for determination of plasma stability made it dif-ficult to draw a conclusion on the role of plasma stability on toxi-city (animal STD10, HNSTD and human MTD).
Generally, lower plasma stability was reported when the intactADC was measured (e.g. ELISA was used or LC/MS/MS was usedafter capturing the intact ADC through ELISA to measure the con-jugated SMs) and higher stability reported when the amount offree/unbound small molecule was measured (through LC/MS/MSalone). For methods used to detect ADC and SM see Cancer DrugDiscovery and Development, Antibody-Drug Conjugates andImmunotoxins, 2013.
For an ADC utilizing the vc-MMAE platform, the human plasmastability indicated that the amount of intact ADC was approximate-ly 70% using ELISA to detect the conjugated antibody (intact ADC);�60% using LC/MS/MS to detect the conjugated MMAE post-ELISA(intact ADC), and 99% using LC/MS/MS alone to detect the freeMMAE (Table 3). For this ADC, similar findings were observed inplasma from most species. While the two methods for measuringthe intact ADC versus the free MMAE agreed that at least someof the small molecule will be released in plasma, the percent ofADC remaining intact were different.
3.4. Initiation of the IND to the RP2D/MTD
Based on the INDs examined, the median time to reach theMTD/RP2D/AD was 2 years (range < 1 –5 years). However, someINDs had unknown dates of initial patient treatment after IND sub-mission. Some INDs exceeded the MTD and the slow de-escalationcontributed to the length of time required to reach the MTD. Insome cases, a slow escalation scheme appeared to result in alengthy Phase 1 trial. There are many other variables that couldaffect the length of time required to identify the MTD/R2D, includ-ing slow patient accrual, unanticipated safety signals, manufactur-ing problems, and the FIH dose selected. For one of the INDs, thetiming between the submission of the IND and reporting of theDLTs was 9 months. For this IND, the sponsor used 1/20th STD10
(BSA approach) to select the FIH dose and an accelerated dose esca-lation clinical design. However, the MTD was exceeded and treat-ment-related deaths occurred under this IND, indicating a steepthreshold for toxicity of this drug and the need for careful titrationdesigns above certain dose levels. Many of the INDs used a 3 + 3design that is commonly proposed for dose escalations in Phase1 clinical trials (Simon et al., 1997).
3.5. Predicting human MTDs
For a given small molecule, the human MTD appears to remainin a relatively tight range regardless of the antibody target or iso-type, if the linker, the ratio of small molecule-to-antibody, and thefrequency of administration remain the same. For example, forADCs with vc-MMAE given once every 3 weeks with a SM:Ab ratioof approximately 4, the human MTD ranged from 1.8 mg/kg to2.4 mg/kg (Table 4). The human MTD was independent of the anti-body isotype used (IgG1 or IgG2), indicating a limited role for anti-body-mediated effector functions (e.g., ADCC) in causing toxicities.In addition the MTD was independent of the antibody target.Though the number or examples is small (n = 3), DM4-containingADCs with disulfide linkers had human MTDs of 150–168 mg/m2
(4.1–4.5 mg/kg), when given once every 3 weeks or with a compa-rable schedule (e.g. Days 1 and 8, every 28 days for one of the
34
Table 3Stability data from one ADC program using vc-MMAE platform: Concentration of ADC after 96 h incubation at 37 �C reported as a percentage relative to time zero.
Species source of plasma Intact ADC measured Intact ADC calculated based on free MMAE
ELISA LC/MS/MS1 Free MMAE measured by LC/MS/MS Intact ADC calculated
Human 71.8 60.1 0.3 99.7Monkey 89.1 55.1 0.2 99.8Rat 73.3 56.6 1.6 98.4Mouse 69.3 64.3 25 75Vehicle 90.5 97.9 0.1 99.9
1 Antibodies (ADC and unconjugated mAb) were isolated by ELISA, MMAE was enzymatically separated from the ADC and the MMAE measured by LC/MS/MS. The MMAEreleased represents those that remained conjugated to the ADC.
448 H. Saber, J.K. Leighton / Regulatory Toxicology and Pharmacology 71 (2015) 444–452
INDs). Too few examples were included in our analysis for othersmall molecules (n = 1 or 2) to determine if a narrow MTD orRP2D range holds true for their linker-SM subgroups.
Our analysis suggests that prior human clinical experience forADCs utilizing vc-MMAE given every 3 weeks with SM:Ab ratio of4 may inform the starting dose and escalation plan for subsequentADCs utilizing this platform. The importance of using prior clinicaldata (when available) from other ADCs using the same linker andSM cannot be overstated as human toxicities seen with ADCs canbe severe. For ADCs using the vc-MMAE platform, severe motorand sensory neuropathy as well as bone marrow toxicity, includingtreatment-related deaths, have been seen at doses above 1.8 mg/kggiven on an every 3 week cycle.
For the MMAE and DM4 containing ADCs, the safety profile inanimals was also independent of the antibody isotype or the target
Fig. 1. Data from 8 INDs; for one of the INDs two sets of data are presented.
Fig. 2. Human MTD/RP2D/AD fall within narrow range based on linker-small molecule mX-axis represents an IND.
of the antibody but the HNSTDs did not appear to have a similarlytight range as the human MTDs (Tables 4 and 5). This may be dueto variable frequency of administration used in animals, the largegap between the doses, and lack of pre-specified definition of DLTsfor toxicology studies. There were some consistencies noted how-ever: for vc-MMAE ADCs administered once every 3 weeks for 4doses, the HNSTD ranged from 5 to 6 mg/kg; for single dose studiesusing disulfide linker-DM4 ADCs, the HNSTDs were 10–15 mg/kg.A graphical view of HNSTD (monkeys) and MTD (humans) rangesis presented in Figs. 1 and 2, respectively.
The gap between the doses (low dose versus mid dose versushigh dose) may be too wide in animals to accurately determinethe HNSTD. If there is mortality or severe toxicities at a dose level,the dose below it is defined as the HNSTD, even if this lower doseshows no toxicity. This may occur when there is a large gapbetween the two dose levels. Thus, the HNSTD is the highestnon-severely toxic dose ‘‘tested’’ in that study. Finally, the defini-tion of HNSTD is subjective. There are no set criteria to defineseverity of toxicities in animals, for instance in terms of percentchanges in clinical pathology parameters. In patients, clinical andlaboratory adverse events are monitored utilizing CTCAE criteriaand pre-specified dose-limiting toxicities are defined. The dose isthen adjusted until a true MTD is achieved.
3.6. FIH-dose approach and the number of escalations to reach thehuman MTD
When using the BSA approach for animal-to-human dose con-version, 1/6th the HNSTD and 1/10th HNSTD in cynomolgus mon-keys or 1/10th the STD10 in rodents resulted in a median of 2.5–3dose doublings to reach the human MTD/RP2D/AD (see Table 6).
oiety. Linkers: vc for MMAEs; disulfide for DM4; mc for MMAF. Each number on the
35
Table 4Nonclinical HNSTD and human MTD for vc-MMAE-containing ADCs utilizing an every 3 week administration schedule.
SM: Ab Antibody isotype Animal dosing schedule HNSTD (mg/kg) Human dosing schedule Human MTD/RP2D/AD
3.2–4.8 Human IgG1 Every other week � 3 9 Q3W 2.3 mg/kg (only 1 patient received this dose)�3.5 Human IgG2 Q3 W x 4 6 Q3W *1.8 mg/kg�3.5 Humanized IgG1 Single dose 3 Q3W 2.4 mg/kg�3.5 Humanized IgG1 Q3W � 4 5 Q3W 2.4 mg/kg�4 Chimeric IgG1 Q3W � 11 3 Q3W 1.8 mg/kg�4.5 Human IgG2 Weekly � 4 1 Q3W 1.88 mg/kg�4 Human IgG1 Q3W � 4 5 Q3W 2.3 mg/kg�3.7 Humanized IgG1 Q3W � 4 5 Q3W 2.4 mg/kg
* The MTD is also dependent on the frequency of administration. When the frequency of dosing was increased to weekly administration, the MTD dropped to 1.2 mg/kg forthis ADC.
Table 5Monkey HNSTD and human MTD for disulfide bond-DM4-containing ADCs.
SM:Ab Antibody isotype Animal dosing schedule HNSTD (mg/kg) Human dosing schedule Human MTD/RP2D/AD
3.5 Humanized IgG1 Single dose 10 Days 1/8 every month 150 mg/m2 (4.1 mg/kg)3.9 Humanized IgG1 Single dose 10 Q3W *160 mg/m2 (4.3 mg/kg)3.5 Humanized IgG1 Single dose 15 Q3W 168 mg/m2 (4.5 mg/kg)
* The MTD is also dependent on the frequency of administration. A weekly administration reduced the MTD to 55 mg/m2.
Table 6Median number of escalations and the range for approaches that could be potentially used to select the FIH dose.
*Approach Species Median number for dose doublings to reach human MTD/RP2D/AD Range for dose doubling to reach human MTD/RP2D/AD
1/6th HNSTD (BSA) Monkey 2.5 1–51/10th HNSTD (BSA) Monkey 3 1–51/10th STD10 (BSA) Rodent 3 2–41/10th NOAEL (BSA) Monkey 5 4–71/10th NOAEL (BSA) Rodent 6 5–71/10th NOAEL (BW) Monkey 4 2–51/10th NOAEL (BW) Rodent 3 3–5
* One ADC containing the first generation of linker was excluded. This linker is no longer used in ADCs.
H. Saber, J.K. Leighton / Regulatory Toxicology and Pharmacology 71 (2015) 444–452 449
For one IND, 1/6th HNSTD (BSA approach) was below but close tothe human MTD; this ADC, however, contained the first generationof linker which is unstable and no longer used in any ADCs. ThisIND was excluded from this analysis as the linker is no longer uti-lized. When using the BW approach for conversions, 1/6th theHNSTD (6 INDs) or 1/10th the HNSTD (2 INDs) in monkeys, or 1/10th the STD10 in rodents (10 INDs) resulted in a FIH dose thatwas above the MTD or too close to the human MTD such that adose doubling could not be done. One-tenth the NOAEL incynomolgus monkeys or rodents using the BW approach resultedin a median dose doublings of 3–4 to reach the human MTD/RP2D/AD. One-tenth the NOAEL using the BSA approach resultedin a median of 5–6 dose doublings (range of 4–7) to reach thehuman MTD/RP2D/AD.
For one IND, the sponsor selected the FIH dose of the ADC basedon the doses of the free antibody previously administered topatients under a separate IND. The free antibody was given at dos-es up to 8 mg/kg to patients with acceptable toxicities; however,the dose of ADC which contained that same antibody resulted intwo DLTs at 4 mg/kg and the ADC dose was hence de-escalated.This further re-emphasizes that the conjugated small moleculedrives the human toxicity making the free antibody less informa-tive for FIH dose decisions. Therefore, it is recommended that theFIH dose of the ADC not be based on the doses of the free antibody.
4. Conclusions and discussion
Over the years the improvement in linker technologies hasresulted in ADCs with improved stability. More stable productstogether with the understanding of new targets in sub-populationsand two recent ADC approvals have encouraged sponsors to
develop more ADCs. Our review has identified that there is littleconsistency in the design of toxicology studies and approachestaken to select the FIH dose. Based on accumulating data fromnon-clinical and clinical development, we have generated severalconsiderations for future ADC development.
4.1. Toxicology studies to support FIH trials
A common feature of nonclinical programs to support clinicaltrials was to conduct toxicology studies in the non-human primate(NHP) with 3 (or more) doses of the ADC. While this is done withthe clinical candidate, one sponsor conducted the toxicology studyin the NHP using arms of both the clinical candidate and thecynomolgus surrogate. A toxicology study with the clinical candi-date is considered sufficient at this time as results indicate thatdose limiting toxicities are related to the small molecule, indepen-dent of target binding. Conducting the study with a surrogate mayalso need additional characterization of the surrogate (e.g. epitopebinding, activity and potency, and PK) to find the relevance of theresults and how to use the data for FIH dose selection. The toxicol-ogy study in the NHP with the clinical candidate has providedsufficient information to set the human starting dose and to gainknowledge on organ toxicities for clinical monitoring. For INDsexamined, when the ADC was used in the rodent, the study couldalso be used to select the FIH dose and to define the toxicity profileof the product. In general, the studies conducted in rodents variedgreatly across INDs examined. For several ADCs, the sponsors con-ducted a study with the free small molecule in a rodent species.The advantage of the study with the free small molecule is thatthe sponsor can cross-reference to that IND for another submissionin which the same small molecule is linked to a different antibody,
36
450 H. Saber, J.K. Leighton / Regulatory Toxicology and Pharmacology 71 (2015) 444–452
which simplifies the design of a nonclinical study and also reducesanimal use. Investigating the toxicities associated with the freelinker may not be necessary; toxicology studies conducted withthe ADC is expected to identify potential toxicities associated withthe linker. Moreover, toxicities were comparable and related to thesmall molecule when both arms of the SM and linker-SM wereincluded in the toxicology studies, indicating that linker-relatedtoxicities may be minimal compared to toxicities related to thesmall molecule.
In clinical trials, ADCs were usually dosed weekly, every otherweek, or every 3 weeks. At times, the sponsors conducted a morefrequent dosing in animals (e.g. weekly) if they had uncertaintiesregarding the clinical schedule and to potentially dose patientsmore frequently, hence to support multiple schedules in patients.
In line with ICH S9, some sponsors conducted single dose toxi-cology studies to support every-3-week dosing in patients. Inregard to FIH dose selection, single dose animal toxicology studiessupported a dosing schedule of once every 3 weeks for ADCs whenthe BSA approach was used for scaling (based on 5 single dosestudies in rodents and 5 single dose studies in cynomolgus mon-keys). Since the number of INDs with single-dose administrationis small, more data needs to be collected to ensure that single dosestudies support every-3-week schedule of administration inpatients. For one of the INDs where an old generation acid-cleav-able linker was used, 1/6th HNSTD based on BSA resulted in aFIH dose that was too close to the human MTD. However, this link-er is cleaved by esterases, resulting in a relatively unstable ADCwith a short half-life; this linker is no longer used.
4.2. Selection of FIH dose based on animal data
For small molecules, a common approach, and one discussed inICH S9, has been to use 1/10th the severely toxic dose to 10% of theanimals (STD10), identified in rodents (DeGeorge et al., 1998).When the non-rodent is used to set the clinical start dose, thenthe dose chosen is generally 1/6th the highest non-severely toxicdose (HNSTD). The approach used to select the FIH dose for smallmolecules in oncology is acceptable for ADCs. The approach of 1/10th NOAEL using BW for conversion, (used for certain biologicalproducts) is also acceptable. In addition, 1/10th HNSTD using BSAfor scaling is acceptable as it produced results that were compara-ble to 1/6th HNSTD (BSA approach) and 1/10th STD10 (BSAapproach). One-sixth the HNSTD, 1/10th the HNSTD and 1/10thSTD10 using BW for conversion may not be safe, as this may resultin a FIH dose that is at or above the human MTD, or slightly belowbut too close to the human MTD. One-tenth the NOAEL using theBSA for conversion may be too conservative, resulting in a medianof 5–6 (range of 4–7) dose doublings to reach the human MTD.Most Phase 1 FIH studies utilize a Fibonacci-type algorithm fordose escalation. While the first one or two escalations may utilize100% escalations, additional dose increments are typically less than100%. Therefore, when 1/10th the NOAEL BSA approach is used toset the FIH dose, the number of escalations to reach the humanMTD may be more than 7 steps. Hence this tradeoff in safety (lowerstarting dose) may be offset by a sub-therapeutic dose in the firstfew cohorts as well as an increased time to reach the MTD. Becausethe SM mediates the toxicities of ADCs, the FIH dose should not bebased on doses of the free antibody studied in animals or previous-ly tested in humans.
In a recent review, Ponce (2011) assessed the data of LeTourneau et al. (2010) and concluded that by using the 1/10thSTD10 approach the frequency of unacceptable toxicity in Phase 1trials was low. In the analysis of Le Tourneau, of 81 molecularlytargeted agents studied, the median number of dose levels to reachan MTD or maximum administered dose (MAD) was 5 (range of 1–14 dose levels). For large molecules when no pharmacologically-
relevant animal model is available, or for biologicals with immuneagonist properties, the start dose is usually conservative, e.g.,defined by using an approach based on minimally anticipated bio-logical level (MABEL). For biopharmaceuticals with a pharmaco-logically relevant animal species, when the product is not animmune agonist and drug-related toxicities have been clearlydetected in animals, 1/10th of NOAEL has been a commonly usedapproach to select the FIH dose. Regardless of which method isused, all available scientific data (including pharmacokinetics andpharmacodynamics) are also used to justify the proposed humandose.
4.3. Toxicities observed in patients treated with FDA approved ADCs
There have been three ADCs approved for use in the UnitedStates. The first approved (in 2000), Mylotarg (gemtuzumabozogamicin), is composed of a recombinant humanized IgG4 anti-body to CD33 conjugated to calicheamicin using an acid-cleavablelinker. Calicheamicin binds to DNA creating double-stranded DNAbreaks and CD33 is an antigen expressed primarily on cells of mye-loid lineage. Mylotarg was approved under the acceleratedapproval pathway in 2000 for treatment of acute myeloid leukemia(AML) (Bross et al., 2001). The application was subsequently with-drawn in 2010 because the post-marketing study required foraccelerated approval failed to verify the clinical efficacy of Mylo-targ, particularly in the context of severe toxicities that becamemore apparent in the post-marketing period (Przepiorka et al.,2013).
Upon internalization of the ADC, the calicheamicin is expectedto be released by hydrolysis of the hydrazone in lysosomes of theCD33-positive target cells. However, esterases and carbonyl reduc-tase are also reported to be important in the metabolism/hy-drolysis of gemtuzumab ozogamicin. This may result in a higherthan expected release of the small molecule, resulting in the smallmolecule-related toxicities. The rather short half-life of Mylotarg(1–2 days) compared to ADCs with newer generation linkers fur-ther supports a lower stability compared to contemporary ADCplatforms (Ducry and Stump, 2010). Toxicities seen in human sub-jects receiving Mylotarg are thought to be related to calicheamicinincluding severe and fatal liver toxicity, hepatic veno-occlusive dis-ease, and pulmonary toxicity. Post-marketing data also revealedcases of severe hypersensitivity reactions, including fatalities, indi-cating that the antibody component may also contribute totoxicity.
Adcetris (brentuximab vedotin) is a CD30-directed ADC utiliz-ing a chimeric IgG1 antibody attached to the microtubule inhibitor,monomethyl aurastatin E (MMAE) via a protease cleavable valine–citrulline (vc) linker. Adcetris was granted accelerated approved in2011 for treatment of Hodgkin lymphoma (HL) and systemicanaplastic large cell lymphoma (ALCL) based on high rates of dur-able objective responses in single arm trials (de Claro et al., 2012).Adverse events for this ADC related to MMAE, include neutropeniaand potentially severe cumulative peripheral sensory and motorneuropathy. During post-marketing surveillance, Adcetris-relatedhepatotoxicity was identified as a new safety issue and the FDAlabel was updated to include hepatotoxicity in the warnings andprecautions section. (FDA, 2014).
Kadcyla (ado-trastuzumab emtansine) is an ADC targeting Her2(Poon et al., 2013; Amiri-Kordestani et al., 2014). The antibody moi-ety is a humanized anti-HER2 IgG1, trastuzumab, which is covalent-ly linked to the microtubule inhibitor DM1 (a maytansine derivative)via a protease-resistant, the thioether linker MCC (4-[N-maleimido-methyl] cyclohexane-1-carboxylate). Kadcyla was approved in 2013for treatment of HER2-positive, metastatic breast cancer. The boxedwarning of Kacyla also contains information on hepatotoxicity. Theelimination half-life in humans is 3.5–4 days. Protease-resistance
37
H. Saber, J.K. Leighton / Regulatory Toxicology and Pharmacology 71 (2015) 444–452 451
(aka non-degradable) linkers are expected to increase the half-life ofthe ADC; however, this did not hold for Kadcyla which shows compa-rable or slightly reduced half-life compared to Adcetris.
Overall, there are common toxicities in humans many of whichare related to the small molecule moiety, e.g. myelosuppressionand related effects (neutropenia, sepsis, and hemorrhage), hepato-toxicity, and neurotoxicity. Infusion-related toxicities reported inpatients may be related to the protein component and the largesize of these molecules. While the ADC platforms are designed todeliver these potent cytotoxic agents to the tumor, it is clear thatsystemic toxicities consistent with the small molecule are a risk.The cumulative nature of some adverse events such as neuro-toxicity further complicate optimal dose selection for those treat-ments intended to be administered chronically until diseaseprogression.
4.4. Toxicity profile and MTDs
Based on our review of IND-enabling GLP toxicology studies,when there was no binding to the epitope, toxicities occuredmainly in the liver, hematopoietic system, and reproductiveorgans for all INDs examined, and in the skin (lesions) or inthe kidney for a few INDs. While not common, additional toxici-ties occasionally occurred, e.g. ocular toxicity (Table 2). Whenthe ADC binds to its epitope, as was the case for 5 ADCs testedin monkeys, DLTs observed in monkeys were mainly related totoxicities in the hematopoietic system. The only exceptionappears to be Kadcyla, where liver toxicity was prominent aftera single dose administration in the monkey, even though it doesbind to the monkey Her-2. While this finding is not clearlyunderstood, possible reasons for this exception may include theADC not readily accessing the target and hence high amountsof circulating ADC are taken up by the liver, or the ADC beingrelatively unstable (or a fraction of the ADC being less stable)resulting in some early release of DM1.
For the approved ADCs, many of the toxicities reported, espe-cially those considered dose limiting, are related to the small mole-cule moiety and include bone marrow suppression, hepatotoxicityand neuropathy. As toxicities are related mainly to the small mole-cule and the release of the small molecule is dependent on the link-er used for conjugation, for ADCs using the same linker and smallmolecule, prior human data can be informative. For instance, thedose-finding trials for vc-MMAE containing ADCs have demon-strated similarities with respect to dose-related toxicities. Basedon data from 8 ADC development programs utilizing vc-MMAEplatforms, the recommended Phase 2 human doses are between1.8 mg/kg and 2.4 mg/kg when the ADC is given every 3 weeks.Doses above 2.4 mg/kg have been poorly tolerated, including casesof severe bone marrow toxicity, septic deaths and severe motorneuropathy and doses below 1.0 mg/kg have reported very fewdose-limiting toxicities. With regard to ADC development utilizingvc-MMAE platforms with a SM:Ab ratio of approximately 4, dosesabove 2.4 mg/kg should be explored with extreme caution. Fur-thermore, the peripheral neuropathy seen with MMAE is frequent-ly cumulative; necessitating careful dose modification rules forneuropathy.
4.5. Factors defining the human MTD
Factors that could affect the MTD include the chemistry of thelinker, dose schedule, the small molecule to antibody ratio, andthe small molecule itself. Our review suggests that the antibodyitself does not appear to have a large effect on the MTD. Withrespect to linker, the first generation linkers (acid-cleavable hydra-zone) produced ADCs with a short half-life (e.g. 1–2 days) and lowstability. A linker may increase or decrease the small molecule-re-
lated toxicities by changing the stability of the ADC and result inincreased or reduced release of the small molecule in the plasma.Ducry and Stump (2010) examined the effect of linkers on the plas-ma half-life and concluded that a hydrazone linker (as was used inMylotarg) results in a shorter half-life when compared to two otherlinkers, phenylalanine-lysine and valine–citrulline. While plasmastability data were examined with an to attempt to identify anyrelationship between linker stability and the human (or animal)MTD, no conclusion could be made based on limited data availableand different experimental methods used. It was noted that whenthe intact ADC was measured, plasma stability was reported to beless than when the free small molecule was measured. Further dia-log may be needed with industry regarding the value of varioustechniques employed and to harmonize the approach used formeasuring plasma stability.
Based on human and animal toxicity data, the small moleculeappears to dictate the dose limiting adverse events seen, and there-fore the MTD. The higher the SM: Ab mean ratio, the lower theMTD is anticipated to be. For the INDs examined the ratio wasapproximately 4. Other potential factors may include prior treat-ments; e.g. for ADC containing doxorubicin as the small molecule,prior treatment with doxorubicin may reduce the threshold forcardiac toxicity.
In summary, despite great interest in ADC development, thereremains little uniformity regarding best practices for designinganimal toxicity studies, calculating a FIH starting dose or Phase 1clinical dose escalation strategies. Our review finds that for calcu-lation of human starting dose, 1/6th the HNSTD in cynomolgusmonkeys or 1/10th the STD10 in rodents using the BSA approachfor animal-to-human conversion generally resulted in an accept-able balance of safety and efficiency in dose escalation. Otheracceptable approaches are 1/10th HNSTD in cynomolgus monkeysusing the BSA scaling and 1/10th NOAEL in rodents or monkeysusing the BW scaling. FIH dose selection algorithms that were con-sidered unsafe consist of: 1/6th HNSTD, 1/10th HNSTD, and 1/10thSTD using BW for animal-to-human dose conversions. While theanimal data for the vc-MMAE platform yielded a variable rangeof HNSTDs in cynomolgus monkeys, the human MTD were report-ed with a tighter range and there was a steep dose-dependentthreshold for toxicities in patients. Overall, for all the INDs includ-ed in our analysis, there was a good correlation between HNSTDand human MTD. For ADCs sharing the small molecule drug, thesame linker, and the same SM:Ab ratio, available prior clinical datacan inform the design of a safe yet efficient Phase 1 clinical trialdesign. As the database available for nonclinical and clinical stud-ies using novel ADCs grows, additional analysis will be useful toinform ADC drug development program design andimplementation.
Acknowledgments
The authors would like to thank Paul Kluetz and Anthony Mucciof the Food and Drug Administration for their constructivecomments.
References
Amiri-Kordestani, L., Blumenthal, G.M., Xu, Q.C., Zhang, L., et al., 2014. FDAApproval: ado-trastuzumab emtansine for the treatment of patients with HER2-positive metastatic breast cancer. Clin. Cancer Res. 20, 4436–4441.
Bross, P.F., Beitz, J., Chen, G., Chen, X.H., et al., 2001. Approval summarygemtuzumab ozogamicin in relapsed acute meyloid leukemia. Clin. CancerRes. 7, 1490–1496.
de Claro, A., McGinn, K., Kwitkowski, B., Bullock, J., 2012. U.S. Food and DrugAdministration approval summary: brentuximab vedotin for the treatment ofrelapsed Hodgkin lymphoma or relapsed systemic anaplastic large-celllymphoma. In: Clin. Cancer Res. 18, 5845–5849.
38
452 H. Saber, J.K. Leighton / Regulatory Toxicology and Pharmacology 71 (2015) 444–452
Degeorge, J.J., Ahn, C.-H., Andrews, P.A., Brower, M.E., et al., 1998. Regulatoryconsiderations for preclinical development of anticancer drugs. Cancer.Chemother. Pharmacol. 41, 173–185.
Ducry, L., Stump, B., 2010. Antibody-drug conjugates: linking cytotoxic payloads tomonoclonal antibodies. Bioconjug. Chem. 21, 5–13.
FDA, 2014. Potential signals of serious risks/new safety information identified bythe FDA Adverse Event Reporting System (FAERS) between January – March2014. (<http://www.fda.gov/Drugs/GuidanceComplianceRegulatoryInformation/Surveillance/AdverseDrugEffects/ucm398223.htm>; accessed October 22, 2014).
FDA Guidance for Industry: Estimating the maximum safe starting dose in initialclinical trials for therapeutics in adult healthy volunteers. 2005. Available at:<http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInfor-mation/Guidances/UCM078932.pdf> (accessed October 2, 2014).
Haddish-Berhane, N., Shah, D.K., Ma, D., Leal, M., et al., 2013. On translation ofantibody drug conjugates efficacy from mouse experimental tumors to theclinic: a PK/PD approach. J. Pharmacokinet Pharmacodyn. 40, 557–571.
Le Tourneau, C., Dieras, V., Tresca, P., Cacheux, W., Paoletti, X., 2010. Currentchallenges for the early clinical development of anticancer drugs in the era ofmolecularly targeted agents. Target Oncol. 5, 65–72.
Philips GL (Editor). Cancer Drug Discovery and Development, 2013, Antibody-DrugConjugates and Immunotoxins: From Pre-Clinical Development to TherapeuticApplications.
Ponce, R., 2011. ICH S9: developing anticancer drugs, one year later. Toxicol. Pathol.39, 913–915.
Poon, K.A., Flagella, K., Beyer, J., Tibbitts, J., et al., 2013. Preclinical safety profile oftrastuzumab emtansine (T-DM1): mechanism of action of its cytotoxiccomponent retained with improved tolerability. Toxicol. Appl. Pharmacol.273, 298–313.
Przepiorka, D., Deisseroth, A., Kane, R., Kaminskas, E., et al., 2013. Gemtuzumabozogamicin. J. Clin. Oncol. 31, 1699–1700.
Roberts, S.A., Andrews, P.A., Blanset, D., Flagella, K.M., et al., 2013. Considerations forthe nonclinical safety evaluation of antibody drug conjugates for oncology. Reg.Toxicol. Pharmacol. 67, 382–391.
Sievers, E.L., Senter, P.D., 2013. Antibody-drug conjugates in cancer therapy. Annu.Rev. Med. 64, 15–29.
Simon, R., Freidlin, B., Rubinstein, L., Arbuck, S.G., et al., 1997. Accelerated titrationdesigns for phase I clinical trials in oncology. J. Natl Cancer Inst. 89, 1138–1147.
39
Recommended