Design, synthesis, and evaluation of linker-duocarmycin payloads: Toward selection of HER2-targeting antibody-drug conjugate SYD985

Article abstract of DOI:10.1021/mp500781a

Antibody-drug conjugates (ADCs) that are currently on the market or in clinical trials are predominantly based on two drug classes: auristatins and maytansinoids. Both are tubulin binders and block the cell in its progression through mitosis. We set out to develop a new class of linker-drugs based on duocarmycins, potent DNA-alkylating agents that are composed of a DNA-alkylating and a DNA-binding moiety and that bind into the minor groove of DNA. Linker-drugs were evaluated as ADCs by conjugation to the anti-HER2 antibody trastuzumab via reduced interchain disulfides. Duocarmycin 3b, bearing an imidazo[1,2-a]pyridine-based DNA-binding unit, was selected as the drug moiety, notably because of its rapid degradation in plasma. The drug was incorporated into the linker-drugs in its inactive prodrug form, seco-duocarmycin 3a. Linker attachment to the hydroxyl group in the DNA-alkylating moiety was favored over linking to the DNA-binding moiety, as the first approach gave more consistent results for in vitro cytotoxicity and generated ADCs with excellent human plasma stability. Linker-drug 2 was eventually selected based on the properties of the corresponding trastuzumab conjugate, SYD983, which had an average drug-to-antibody ratio (DAR) of about 2. SYD983 showed subnanomolar potencies against multiple human cancer cell lines, was highly efficacious in a BT-474 xenograft model, and had a long half-life in cynomolgus monkeys, in line with high stability in monkey and human plasma. Studies comparing ADCs with a different average DAR showed that a higher average DAR leads to increased efficacy but also to somewhat less favorable physicochemical and toxicological properties. Fractionation of SYD983 with hydrophobic interaction chromatography resulted in SYD985, consisting of about 95% DAR2 and DAR4 species in an approximate 2:1 ratio and having an average DAR of about 2.8. SYD985 combines several favorable properties from the unfractionated ADCs with an improved homogeneity. It was selected for further development and recently entered clinical Phase I evaluation.

Full text of DOI:10.1021/mp500781a

Subscriber access provided by GAZI UNIV
Article
Design, Synthesis, and Evaluation of Linker-Duocarmycin Payloads:
Towards Selection of HER2-Targeting Antibody–Drug Conjugate SYD985
Ronald Elgersma, Ruud Coumans, Tijl Huijbregts, Wiro Menge, John Joosten, Henri Spijker,
Vincent de Groot, Miranda van der Lee, Ruud Ubink, Diels van den Dobbelsteen, David
Egging, Wim Dokter, Gijs Verheijden, Jacques Lemmens, Marco Timmers, and Patrick Beusker
Mol. Pharmaceutics, Just Accepted Manuscript • DOI: 10.1021/mp500781a • Publication Date (Web): 30 Jan 2015
Downloaded from http://pubs.acs.org on February 2, 2015
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Molecular Pharmaceutics is published by the American Chemical Society. 1155
Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 70
Molecular Pharmaceutics
1
2
3
4
5
6
7
8
9
Design, Synthesis, and Evaluation of Linkerꢀ
Duocarmycin Payloads: Towards Selection of
HER2ꢀTargeting Antibody–Drug Conjugate
SYD985
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
#
#
#
#
Ronald C. Elgersma, Ruud G. E. Coumans, Tijl Huijbregts, Wiro M. P. B. Menge, John A. F.
#
#
#
†
Joosten, Henri J. Spijker, Franciscus M. H. de Groot, Miranda M. C. van der Lee, Ruud
†
†
†
†
Ubink, Diels J. van den Dobbelsteen, David F. Egging, Wim H. A. Dokter, Gijs F. M.
‡
‡
‡
,#
Verheijden, Jacques M. Lemmens, C. Marco Timmers, Patrick H. Beusker*
#
†
‡
Departments of Medicinal & Protein Chemistry, Preclinical, and New Molecular Entities,
Synthon Biopharmaceuticals BV, Microweg 22, 6545 CM Nijmegen, the Netherlands
KEYWORDS: duocarmycin, linkerꢀduocarmycin, antibody–drug conjugate, ADC, structure–
activity relationship, drug targeting, design, secoꢀDUBA, DUBA, SYD983, SYD985, HER2
ACS Paragon Plus Environment
1

Molecular Pharmaceutics
Page 2 of 70
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
ABSTRACT
Antibody–drug conjugates (ADCs) that are currently on the market or in clinical trials are
predominantly based on two drug classes: auristatins and maytansinoids. Both are tubulin
binders and block the cell in its progression through mitosis. We set out to develop a new class
of linkerꢀdrugs based on duocarmycins, potent DNAꢀalkylating agents that are composed of a
DNAꢀalkylating and a DNAꢀbinding moiety and that bind into the minor groove of DNA.
Linkerꢀdrugs were evaluated as ADCs by conjugation to the antiꢀHER2 antibody trastuzumab via
reduced interchain disulfides. Duocarmycin 3b, bearing an imidazo[1,2ꢀa]pyridineꢀbased DNAꢀ
binding unit, was selected as the drug moiety, notably because of its rapid degradation in plasma.
The drug was incorporated into the linkerꢀdrugs in its inactive prodrug form, secoꢀduocarmycin
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
3a. Linker attachment to the hydroxyl group in the DNAꢀalkylating moiety was favored over
linking to the DNAꢀbinding moiety, as the first approach gave more consistent results for in vitro
cytotoxicity and generated ADCs with excellent human plasma stability. Linkerꢀdrug 2 was
eventually selected based on the properties of the corresponding trastuzumab conjugate,
SYD983, which had an average drugꢀtoꢀantibody ratio (DAR) of about 2. SYD983 showed
subnanomolar potencies against multiple human cancer cell lines, was highly efficacious in a
BTꢀ474 xenograft model, and had a long halfꢀlife in cynomolgus monkeys, in line with high
stability in monkey and human plasma. Studies comparing ADCs with a different average DAR
showed that a higher average DAR leads to increased efficacy, but also to somewhat less
favorable physicochemical and toxicological properties. Fractionation of SYD983 with
hydrophobic interaction chromatography resulted in SYD985, consisting of about 95% DAR2
and DAR4 species in an approximate 2:1 ratio and having an average DAR of about 2.8.
SYD985 combines several favorable properties from the unfractionated ADCs with an improved
ACS Paragon Plus Environment
2

Page 3 of 70
Molecular Pharmaceutics
1
2
3
4
5
6
7
8
9
homogeneity. It was selected for further development and recently entered clinical Phase I
evaluation.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ACS Paragon Plus Environment
3

Molecular Pharmaceutics
Page 4 of 70
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
INTRODUCTION
Antibody–drug conjugates (ADCs) are more and more considered to be valuable therapeutics
for cancer therapy, and a still increasing number of companies are actively developing ADCs.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
The recent approvals of Adcetris (brentuximab vedotin) and Kadcyla (adoꢀtrastuzumab
2
emtansine) for the treatment of CD30ꢀ and HER2ꢀpositive cancers, respectively, have
significantly contributed the excitement to the use of ADC therapeutics for cancer treatment.
Over 30 ADCs are currently in clinical trials and more are in preclinical development. Most of
these ADCs, including brentuximab vedotin and adoꢀtrastuzumab emtansine, use auristatins or
maytansinoids as payloads. Both are tubulin polymerization inhibitors that bind to the vinca
domain in tubulin and that cause cell cycle arrest in the G2ꢀM phase, eventually leading to
apoptotic cell death. There are only a small number of ADCs in clinical development that contain
different drug classes, including calicheamycins, camptothecins, anthracyclines, and
3
pyrrolobenzodiazepine dimers, which all target DNA or DNAꢀrelated proteins. Other drug
4
4
5
classes, such as aminitins, tubulysins, and monofunctional indolinobenzodiazepine dimers, are
being explored preclinically and may enter the clinic in the years to come.
Although ADCs generally show an enhanced selectivity towards tumor cells compared to nonꢀ
targeted chemotherapeutics, still a considerable number of offꢀtarget toxicities and side effects
are noted in a significant number of patients. This is most likely due to the fact that ADCs, like
traditional chemotherapeutics, are dosed at the maximum tolerated dose, and consequently,
adverse effects are notable. For brentuximab vedotin, given to patients at a dose level of 2.4
mg/kg every three weeks, peripheral sensory neuropathy and neutropenia are commonly
6
,7,8
observed grade ≥3 adverse events.
Similarly, thrombocytopenia and neutropenia are
frequently observed for adoꢀtrastuzumab emtansine, which is administered at a dose level of 3.6
ACS Paragon Plus Environment
4

Page 5 of 70
Molecular Pharmaceutics
1
2
3
4
5
6
7
8
9
9,10,11,12
mg/kg every three weeks.
For these ADCs, as well as for other ADCs based on the same
drug classes that are still in clinical development, many of the reported toxicities appear to be
identical and targetꢀindependent, suggesting that many toxicities are related to (systemic)
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
13
exposure to the drug.
The firstly discovered member of the naturally occurring duocarmycins was CCꢀ1065 (1,
1
4
Figure 1), isolated from Streptomyces bacteria in the late 1970s. Duocarmycins act by binding
to the minor groove of DNA in AꢀT rich regions and subsequently alkylating adenine residues at
1
5
the N3 position through the cyclopropyl group in the DNAꢀalkylating moiety. Although CCꢀ
16
1
065 was shown to be very potent in vitro, it demonstrated only moderate in vivo activity and
1
7
delayed lethality accompanied by irreversible hepatic toxicity in animal models. Many groups
have been working on the development of synthetic duocarmycin analogs with the aim to
18,19,20,21
improve the biological profile of this class of compounds.
Halogenꢀcontaining seco derivatives of the duocarmycins are generally believed to require a
ring closure to their cyclopropylꢀcontaining spiro analogs (Figure 1) to become cytotoxic,
although seco derivatives that cannot rearrange to their spiro counterparts have been shown to
2
2,23
have DNAꢀalkylating capabilities as well, albeit much more attenuated.
Thus, by trapping the
duocarmycin in its seco form through conjugation of the crucial hydroxyl group to a promoiety,
the duocarmycin drug can be (largely) inactivated until the promoiety is removed and ring
closure to the spiro form can occur. This concept has for example been applied to duocarmycin
prodrugs carzelesin and KWꢀ2189, which both progressed into clinical development.
2
4
Unfortunately, like nonꢀprodrug adozelesin, which was also evaluated in the clinic, both
compounds lacked sufficient therapeutic window, related to hematological toxicity, and clinical
25,26
development was halted.
ACS Paragon Plus Environment
5

Molecular Pharmaceutics
Page 6 of 70
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
In an effort to develop a class of linkerꢀdrugs that would allow for ADCs with an improved
therapeutic index and with a mechanism of action distinct from that of ADCs currently on the
market or in clinical development, we have explored the duocarmycin class of compounds as
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
27,28,29,30
ADC drugs. Others
have also studied this drug class for use in ADCs, but only one
3
1
32
duocarmycinꢀbased ADC, MDXꢀ1203 (now BMSꢀ936561 ), was progressed into phase I
clinical trials. However, to the best of our knowledge, at the time of writing this paper, no
clinical trials are currently ongoing with this ADC.
Here, we describe a new duocarmycin derivative with the favorable in vitro and in vivo
properties to use it as an ADC. This duocarmycin was incorporated into novel linkerꢀ
duocarmycin constructs using linkers optimized to provide a perfect synergy with the drug.
ADCs based on these linkerꢀduocarmycins and the antiꢀHER2 antibody trastuzumab were
extensively evaluated, eventually culminating in selection of linkerꢀduocarmycin 2 as the lead
linkerꢀdrug for our antiꢀHER2 ADC program and subsequent development of HER2ꢀtargeted
ADC SYD985, which recently entered clinical development.
RESULTS
Duocarmycins. The hydrophobic nature of duocarmycins conveys a strong binding into the
minor groove of DNA, but at the same time it limits the water solubility of this class of
2
8
compounds. Furthermore, it was shown by Jeffrey and coworkers to adversely affect
physicochemical properties of duocarmycinꢀbased ADCs, as relatively high amounts of high
molecular weight (HMW) species were observed when a linker was used that had successfully
been employed for other drug classes. By reducing linker hydrophobicity, the amount of HMW
species could be lowered. We envisioned that hydrophobicity could potentially also be reduced
ACS Paragon Plus Environment
6

Page 7 of 70
Molecular Pharmaceutics
1
2
3
4
5
6
7
8
9
by introduction of heteroatoms at selected positions within the core ring structure of the
duocarmycin itself. An extensive structure–activity relationship study focusing on the effects of
structural modifications on CMC and preclinical properties of the duocarmycins and their
corresponding ADCs led to identification of the imidazo[1,2ꢀa]pyridine series as an interesting
class of duocarmycins. Analog 3a (secoꢀDUBA, Figure 1) was selected from this class based on
its favorable properties.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
secoꢀDUBA was prepared from the corresponding DNAꢀalkylating and DNAꢀbinding
moieties. The DNAꢀalkylating moiety, which is based on the 1,2,9,9aꢀ
33,34
tetrahydrocyclopropa[c]benzo[e]indoleꢀ4ꢀone framework,
was prepared according to the
route depicted in Scheme 1.
Starting from oꢀtolualdehyde (4) and dimethyl succinate (5), a 9:1 mixture of acids 6a and 6b
35
was obtained through a Stobbe condensation. Ring closure of the mixture of acids was
accomplished with trifluoroacetic anhydride and gave alcohol 7, which was then protected with
benzyl chloride to afford benzyl ether 8. Ensuing hydrolysis of the methyl ester group in 8 to
afford carboxylic acid 9 was followed by a Curtius rearrangement in a mixture of toluene and
tertꢀbutyl alcohol to provide carbamate 10. Bromination with Nꢀbromosuccinimide to give
bromide 11 proceeded smoothly. Besides chromatographic resolution and diastereomeric
derivatization, asymmetric synthetic approaches, such as enzymatic desymmetrization
3
6
37
37
reactions, asymmetric hydroboration, and Jacobsen epoxidation, have been used to obtain
the alkylating moiety with high enantiomeric excess. We decided to alkylate bromide 11 with
3
8,39,40
(S)ꢀglycidyl nosylate
to be able to introduce the fiveꢀmembered ring. Alkylation proceeded
well in the presence of potassium tertꢀbutoxide and gave epoxide 12 in good yield. Ring closure
proved unsuccessful with many organometallic reagents, most likely due to steric congestion at
ACS Paragon Plus Environment
7

Molecular Pharmaceutics
Page 8 of 70
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
the reaction site. The reaction with nꢀbutyllithium gave the best results and provided a mixture of
desired compound 13 and debrominated, rearranged derivative 14. The ratio between 13 and 14
was strongly dependent on the reaction conditions that were used. Yields for 13 proved to be
highest when tetrahydrofuran was used as the solvent and the reaction temperature was kept
between –25 °C and –20 °C. Under these conditions, 13 and 14 were obtained in an approximate
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1:1 ratio. Separation of the two compounds proved difficult, but workꢀup with pꢀtoluenesulfonic
acid resulted in conversion of 14 to 10, which made separation possible. Mesylation of the
hydroxyl group in 13 followed by chloride substitution using lithium chloride gave key
intermediate 15.
The DNAꢀbinding moiety of secoꢀDUBA was prepared according to the route depicted in
4
1
Scheme 2. A Chichibabin cyclization reaction between commercially available ethyl
bromopyruvate (16) and 5ꢀnitropyridinꢀ2ꢀamine (17) was used to obtain nitro compound 18.
Reduction of the nitro group with zinc under acidic conditions gave amine 19. Next, coupling
with methoxymethyl (MOM)ꢀprotected 4ꢀhydroxybenzoic acid (20), prepared from methyl 4ꢀ
hydroxybenzoate through reaction with chloromethyl methyl ether followed by ester
4
2
hydrolysis, gave ethyl ester 21, which was hydrolyzed with sodium hydroxide in aqueous 1,4ꢀ
dioxane to provide acid 22.
secoꢀDUBA was synthesized from DNAꢀalkylating unit 15 and DNAꢀbinding moiety 22 by
removal of the tertꢀbutoxycarbonyl (Boc) protective group from 15 under acidic conditions
followed by EDCꢀmediated coupling of amine 23 to acid 22. Compound 24 was then fully
deprotected in two consecutive steps to afford secoꢀDUBA as its HCl salt.
secoꢀDUBA was evaluated in three human cancer cell lines, selected based on their HER2
status: breast carcinoma cell line SKꢀBRꢀ3 (HER2 3+), ovarian carcinoma cell line SKꢀOVꢀ3
ACS Paragon Plus Environment
8

Page 9 of 70
Molecular Pharmaceutics
1
2
3
4
5
6
7
8
9
43
(
HER2 2+), and colon carcinoma cell line SW620 (HER2ꢀnegative). All cell lines were highly
sensitive to secoꢀDUBA with IC values ranging from 90 to 430 pM (Table 1).
50
The presence of the second nitrogen atom in the ring structure of the DNAꢀbinding moiety of
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
secoꢀDUBA proved to have a positive effect on logP. secoꢀDUBA and its analog 26 (Chart 1)
4
4
were calculated to have logP values of 4.32/4.14 and 4.64/4.68, respectively, indicating a
decreased hydrophobicity for secoꢀDUBA with respect to its analog 26. Calculated values for
secoꢀDUBA were in line with the experimentally determined logD value of 3.7 at pH 7.4,
especially if one considers that the DNAꢀbinding moiety in secoꢀDUBA is expected to be partly
protonated at physiological pH, since the pKa of the imidazo[1,2ꢀa]pyridine group is estimated
4
5,46
to be close to 6.
DUBA (3b), formed through spontaneous spirocyclization of secoꢀDUBA (Figure 1), was
shown to be equally potent and efficacious as secoꢀDUBA in SKꢀBRꢀ3 cells, which suggests that
spirocyclization is not rateꢀlimiting. DUBA was found to be relatively unstable in buffer at
physiological pH (data not shown) and in plasma from Balb/c mice, Wistar rats, cynomolgus
monkeys, and man when incubated at a concentration of 5 nM. Halfꢀlives were determined to be
5.3, 0.6, 1.6, and 1.0 h in mouse, rat, monkey, and human plasma, respectively. Percentages of
DUBA remaining after 180 min of incubation were 68.2% in mouse plasma, 2.6% in rat plasma,
44.5% in monkey plasma, and 27.0% in human plasma. Plasma stability of DUBA was about
threefold lower than that of the spirocyclized analog of 26, indicating that the presence of the
additional heteroatom in the DNAꢀbinding moiety had a substantial effect on plasma stability. A
main degradation pathway of DUBA was shown to comprise hydrolysis of the amide bond
between the DNAꢀalkylating moiety and DNAꢀbinding moiety, resulting in formation of DNAꢀ
alkylating unit 27 and DNAꢀbinding unit 28 (Chart 1).
ACS Paragon Plus Environment
9

Molecular Pharmaceutics
Page 10 of 70
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Both 27 and 28 showed over 50,000ꢀfold reduced potency against SKꢀBRꢀ3 cells (Figure 2A,
Table 1), indicating that hydrolysis is an important detoxification route of the drug. When ring
closure of secoꢀDUBA to its spiro analog was prevented by coupling of a linker to the hydroxyl
group in the alkylating moiety, hydrolysis was not detected, suggesting that the spiro form is
required for hydrolysis to occur.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Although low plasma stability for DUBA may seem deleterious for biological activity, we
believed that it could positively affect the profile of ADCs based on secoꢀDUBA, as rapid
degradation of DUBA in plasma would translate to low systemic exposure to the drug and
consequently a lower probability for offꢀtarget toxicity.
To corroborate the low stability of DUBA in vivo, a pharmacokinetic study with secoꢀDUBA
and DUBA was conducted in Wistar rats. This study indicated once more that secoꢀDUBA is
likely converted to DUBA almost instantaneously, as DUBA pharmacokinetics were very similar
after administration of DUBA and secoꢀDUBA at an equimolar dose (Figure 2B). No secoꢀ
DUBA could be detected in plasma, and DUBA was rapidly cleared from the systemic
circulation, showing a very high clearance of 17 L/(h·kg) and a terminal halfꢀlife of 1.1 h. The
instability observed in plasma in vitro is likely to contribute to the high clearance of DUBA from
the systemic circulation.
Linker Optimization. secoꢀDUBA contains two hydroxyl groups, which can each be used for
coupling to an antibody via a linker. If the linker is coupled to the hydroxyl group in the
alkylating moiety, complete elimination of the linker after linker cleavage at the tumor site
would be necessary to liberate the hydroxyl group and allow ring closure to the active spiro
compound. If linker attachment occurs through the DNAꢀbinding moiety, both nonꢀcleavable and
30
cleavable linkers might be acceptable, as was shown for some duocarmycin derivatives similar
ACS Paragon Plus Environment
10

Page 11 of 70
Molecular Pharmaceutics
1
2
3
4
5
6
7
8
9
to secoꢀDUBA. When the linker is coupled to the DNAꢀbinding moiety, the hydroxyl group in
the alkylating moiety needs to be protected as well to prevent premature ring closure to the spiro
compound and subsequent degradation of the antibodyꢀbound duocarmycin to 27 and the
antibodyꢀbound DNAꢀbinding moiety. The additional requirement for a protective group when
the linker is coupled to the DNAꢀbinding moiety thus implies that in this approach, formation of
the spiro drug requires two activating steps at the tumor site.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Linkerꢀduocarmycins 29 and 30 were prepared to evaluate the differences between linkage to
the DNAꢀalkylating unit and linkage to the DNAꢀbinding moiety. Both linkerꢀdrugs contain a
maleimide group for conjugation to a cysteine thiol group on the antibody, the dipeptide valineꢀ
4
7
citrulline, which is a substrate for cathepsin B, and two selfꢀelimination spacers in between the
dipeptide and secoꢀDUBA. The pꢀaminobenzyl alcohol (PABA) selfꢀelimination spacer is a
commonly used selfꢀelimination spacer, which is for example present in the linker of
brentuximab vedotin and in all other ADCs in the clinic bearing the monomethylauristatin E
(MMAE) drug. The bisamine cyclization spacer was incorporated to allow for a carbamate
linkage between linker and secoꢀDUBA, as such a linkage is generally more stable than a
carbonate linkage, which would be present if only a PABA spacer were used. The N,N'ꢀdimethyl
cyclization spacer was initially selected, as it had been successfully used in prodrugs
4
8,49,50
before.
Linkerꢀdrug 30 furthermore contains a 4ꢀmethylpiperazinꢀ1ꢀylcarbonyl promoiety,
in analogy with MDXꢀ1203, designed to stabilize the secoꢀduocarmycin as long as it is still
attached to the antibody and to release the drug through an esteraseꢀmediated cleavage inside the
5
1,52
cell.
Linkerꢀdrug 29 was prepared according to the route depicted in Scheme 3. The linker building
block was synthesized by starting with a condensation reaction between commercially available
ACS Paragon Plus Environment
11

Molecular Pharmaceutics
Page 12 of 70
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
3
1 and 2ꢀ(2ꢀaminoethoxy)ethanol (32) to give alcohol 33, which was then converted to reactive
carbonate 34 through reaction with 4ꢀnitrophenyl chloroformate. Coupling of 34 to HꢀValꢀCitꢀ
5
3
PABA (35), prepared according to literature procedures, resulted in formation of linker 36,
which was treated with bis(4ꢀnitrophenyl) carbonate to give activated linker 37.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
The other building block for linkerꢀdrug 29, cyclization spacerꢀduocarmycin construct 40, was
prepared from MOMꢀprotected duocarmycin 25 in two steps. Consecutive treatment of 25 with
4
ꢀnitrophenyl
chloroformate
and
commercially
available
tertꢀbutyl
methyl(2ꢀ
(methylamino)ethyl)carbamate (38) gave construct 39. Removal of the Boc and MOM protective
groups in 39 with trifluoroacetic acid (TFA) provided 40 as its TFA salt. Linkerꢀdrug 29 was
synthesized through reaction of activated linker 37 with cyclization spacerꢀduocarmycin
construct 40 under slightly basic conditions. Under these conditions, selfꢀelimination of the
cyclization spacer and resulting formation of 3a was suppressed.
Linkerꢀdrug 30 was synthesized in seven steps from linker 37 and alcohol 25 using a reaction
sequence largely analogous to that of linkerꢀdrug 29. Compound 25 was treated consecutively
with 4ꢀnitrophenyl chloroformate and 1ꢀmethylpiperazine to give carbamate 41. Ensuing
removal of the MOM protective group was followed by reaction of alcohol 42 with 4ꢀnitrophenyl
chloroformate and cyclization spacer 38 to give 43. Removal of the Boc protective group and
subsequent coupling of 44 with linker 37 finally afforded linkerꢀduocarmycin 30.
Linkerꢀduocarmycins 29 and 30 were conjugated to trastuzumab following partial reduction of
the interchain disulfide bonds of trastuzumab using tris(2ꢀcarboxyethyl)phosphine (TCEP) as the
reducing agent in such an amount that on average one disulfide bond per antibody was reduced.
The resulting ADCs, Tmabꢀ29 and Tmabꢀ30, were obtained with average drugꢀtoꢀantibody ratios
(DARs) of approximately 2, as determined by hydrophobic interaction chromatography (HIC).
ACS Paragon Plus Environment
12

Page 13 of 70
Molecular Pharmaceutics
1
2
3
4
5
6
7
8
9
Tmabꢀ29 and Tmabꢀ30 were evaluated for in vitro cytotoxicity in the SKꢀBRꢀ3, SKꢀOVꢀ3, and
SW620 cell lines. Both ADCs demonstrated cytotoxic activity that was dependent on HER2
expression (Table 1 and Figure 3A–C), as no significant activity was observed for both ADCs
against the HER2ꢀnegative SW620 cell line. While Tmabꢀ29 showed subnanomolar activity
against the SKꢀBRꢀ3 and SKꢀOVꢀ3 cell lines with 90–100% efficacy, Tmabꢀ30 was only equally
potent and efficacious against the SKꢀOVꢀ3 cell line and considerably less potent against the SKꢀ
BRꢀ3 cell line.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Both ADCs were also incubated with SW620 cells in the presence of 1% human or mouse
plasma in order to assess plasma stability. If the linker is unstable in plasma, this could lead to
release of the active drug and, consequently, increased cytotoxicity against SW620 cells in the
presence of plasma. This assay would not detect degradation pathways that do not lead to release
of the active drug.
In the presence of 1% mouse plasma, the linkerꢀdrug in Tmabꢀ29 proved less stable than that
in Tmabꢀ30. On the basis of the IC values of 0.46 nM for Tmabꢀ29 (drug equivalents) in the
5
0
presence of mouse plasma, ±100 nM for Tmabꢀ29 in the absence of mouse plasma, and 0.09 nM
for secoꢀDUBA, a significant amount of conjugated drug was released in the sixꢀday assay. In
the presence of 1% human plasma, both ADCs did not seem to release any considerable amount
of drug within the sixꢀday assay as IC values were at least 450ꢀfold higher than that of secoꢀ
5
0
DUBA.
In a separate assay, linkerꢀdrugs 29 and 30 were incubated in human plasma and release of
degradation products was followed over time. When linkerꢀdrugs 29 and 30 were added to
human plasma, almost instantaneous conjugation to human serum albumin (HSA) took place –
complete conversion was observed within 5 minutes – and therefore stability of HSAꢀconjugated
ACS Paragon Plus Environment
13

Molecular Pharmaceutics
Page 14 of 70
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
2
9 and 30 was in fact assessed. Quantities of DUBA and DNAꢀalkylating unit 27 were measured
over time as a measure of plasma stability. Results clearly showed that HSAꢀ29 was highly
stable in human plasma, in line with the data from the cellular assay, whereas HSAꢀ30 was
shown to degrade significantly (Figure 3D). Only formation of DNAꢀalkylation unit 27 was
detected. As it was shown separately that the total measurable amount of DUBA and 27 slowly
decreased over time after incubation of DUBA in plasma, absolute quantities of DUBA and 27
that were formed after incubation of linkerꢀdrugs 29 and 30 could not be determined accurately.
The plasma stability and cytotoxicity assays made clear that linkerꢀdrug 30, being relatively
unstable in human plasma and, at least in SKꢀBRꢀ3 cells, not sufficiently cytotoxic, is
outperformed by linkerꢀdrug 29 and that linking through the DNAꢀbinding moiety, like in 30, is
thus disfavored over linking to the DNAꢀalkylating moiety, like in 29. As no other promoieties
showing substantially improved properties could be identified and it was considered an
advantage that only one enzymatic cleavage step would be required to release the drug, we
decided to move forward with the concept of linking to the DNAꢀalkylating moiety.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
The relatively low stability of Tmabꢀ29 in mouse plasma was found to be caused by rodentꢀ
4
3
specific carboxylesterase 1c (CES1c). Although this enzyme is not expressed in man or
cynomolgus monkey, optimization of mouse plasma stability was considered useful to enable a
better translation of observed efficacy in mouse models. As studies indicated that CES1c cleaves
the carbamate linkage between duocarmycin and linker, a large series of linkerꢀduocarmycins
with different substituents on the nitrogen atom of the linking carbamate group was synthesized
and evaluated. From this series, linkerꢀduocarmycins 2 (Figure 1) and 45 (Chart 1) were selected
for more extensive evaluation.
ACS Paragon Plus Environment
14

Page 15 of 70
Molecular Pharmaceutics
1
2
3
4
5
6
7
8
9
54
Both linkerꢀdrugs were prepared analogously to 29. The most notable difference was the
requirement for 1ꢀhydroxybenzotriazole as a catalyst in the coupling reaction between the Bocꢀ
protected cyclization spacer and the 4ꢀnitrophenyl carbonate of 25, likely due to the lower
reactivity of the amine as a consequence of steric congestion. Furthermore, the presence of the
amino acid group in the cyclization spacer of linkerꢀdrug 45 required a slightly different
protective group strategy. The synthesis of 45 was in general somewhat more laborious than that
of linkerꢀdrug 2, and the final steps of the synthesis of linkerꢀdrug 45 proceeded with somewhat
lower yields.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Linkerꢀduocarmycins 2 and 45 were conjugated through their maleimide group to Nꢀ
acetylcysteine, serving as an antibody surrogate, to give compounds 46a and 46b, respectively
(Scheme 5). Both 46a and 46b were efficiently cleaved by cathepsin B at pH 5 at a similar rate,
as monitored by formation of cyclization spacerꢀduocarmycin compounds 47a and 47b (data not
shown). Selfꢀelimination of the cyclization spacers in 47a and 47b at pH 7.4 and 25 °C occurred
readily with halfꢀlives of 27 and 37 min, respectively. Halfꢀlives at pH 7.4 and 37 °C were even
considerably shorter. As expected, halfꢀlives at pH 5 and 37 °C were longer, being
approximately 10 h for both cyclization spacers.
Linkerꢀdrugs 2 and 45 were conjugated to trastuzumab following the same procedure as for
Tmabꢀ29. Tmabꢀ2 (SYD983) and Tmabꢀ45 (SYD981) were obtained with average DARs of
approximately 2, as confirmed by HIC. Percentages of HMW species, as determined by size
exclusion chromatography (SEC), were 1.8 and 3.0% for SYD983 and SYD981, respectively,
slightly lower than the percentage of HMW species for Tmabꢀ29 (4.1%), as was anticipated
based on the less lipophilic nature of the substituents in the cyclization spacer.
ACS Paragon Plus Environment
15

Molecular Pharmaceutics
Page 16 of 70
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Both ADCs were evaluated for cytotoxicity in the same three cell lines as Tmabꢀ29 (Table 1).
SYD983 was slightly more potent than SYD981 against the HER2ꢀpositive cell lines SKꢀBRꢀ3
(
0.22 nM vs 0.31 nM) and SKꢀOVꢀ3 (0.44 nM vs 1.20 nM). Both were significantly less active
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
against the HER2ꢀnegative SW620 cell line.
In vitro plasma stability of SYD983 and SYD981 was assessed by incubating the ADCs in
mouse, rat, monkey, and human plasma. SYD981 demonstrated halfꢀlives ranging from 39 to
2
91 h, while SYD983 showed halfꢀlives between 6 and 252 h (Table 2). Halfꢀlives of conjugated
antibody were determined with a sandwich ELISA assay in which ADCs were captured with an
antiꢀdrug antibody and detected with an antiꢀidiotype antibody. All conjugates that contained at
least one drug were detected. Conjugated antibody halfꢀlives in plasma were mainly driven by
linkerꢀduocarmycin degradation instead of degradation of the antibody component, as a
sandwich ELISA assay to detect total antibody, that is, conjugated and unconjugated antibody,
using antiꢀidiotype antibodies for both capture and detection, showed that 75–92% of antibody
was remaining after seven days of incubation in mouse, rat, monkey, or human plasma.
Thus, especially in mouse and rat plasma, SYD981 proved significantly more stable than
SYD983. These results were confirmed in a cellular assay, which showed that cytotoxic activity
in the HER2ꢀnegative SW620 cell line in the presence of 1% mouse plasma was considerably
higher for SYD983 (IC = 0.93 nM) than for SYD981 (IC = 3.95 nM), while no significant
5
0
50
activity was observed for both ADCs in the presence of 1% human plasma (Table 1).
In vivo PK data for SYD981 and SYD983 were generated in healthy female Balb/c mice
(
Figure 4A), Sprague Dawley rats (Figure 4B), and cynomolgus monkeys (Figure 4C,E). The in
vivo PK results were well in line with the in vitro plasma stability data. Exposure to conjugated
antibody, expressed in SYD983 equivalents, was about three times lower for SYD983 than for
ACS Paragon Plus Environment
16

Page 17 of 70
Molecular Pharmaceutics
1
2
3
4
5
6
7
8
9
SYD981 in female Balb/c mice. For both ADCs, exposure to total antibody was considerably
higher, indicating substantial degradation of both ADCs. PK profiles in female Sprague Dawley
rats were similar to those in Balb/c mice, although the ratio between total antibody exposure and
conjugated antibody exposure was somewhat smaller for both ADCs. In cynomolgus monkeys,
both ADCs demonstrated a very similar PK profile and conjugated antibody exposure was very
close to total antibody exposure, indicating that linkerꢀdrug degradation was very limited and
clearance of intact ADC was mainly responsible for the gradual decrease of plasma
concentrations of conjugated antibody and total antibody over time.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
In line with the slightly higher stability of SYD981 in mouse and rat plasma in vitro and in
vivo, the tolerability and toxicity profile of SYD981 was slightly better than that of SYD983 in
these species (data not shown). However, the limited plasma stability of either linkerꢀdrug in
rodents indicates that these species are of limited value in the quantitative risk assessment for
man, where high stability is expected based on human plasma stability data.
In vivo efficacy of the ADCs was assessed in Balb/c nu/nu mice bearing breast carcinoma BTꢀ
474 xenografts. In vitro, SYD981 and SYD983 demonstrated IC50 values of 0.15 and 0.06
ꢁg/mL, respectively, against the BTꢀ474c cell line, the same subclone as used for the xenograft
studies. In the xenograft study, both ADCs were dosed intravenously either once or three times
(days 0, 7, and 14) at 0.2, 1, and 5 mg/kg (Figure 5A,C). Both ADCs showed clear dose–
response relationships, especially for the multiꢀdose regimens. Durable remissions were observed
for both SYD981 and SYD983 at 5 mg/kg (qw × 3), while tumor growth delay was observed for
both 1 mg/kg (qw × 3) groups. Mice treated with 0.2 mg/kg (qw × 3) did not show a significant
effect on tumor growth compared to vehicle. From the groups that received a single dose of
ADC, only both 5 mg/kg groups were significantly different from the vehicle group and showed
ACS Paragon Plus Environment
17

Molecular Pharmaceutics
Page 18 of 70
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
nearly complete tumor regression followed by tumor regrowth. Thus, antiꢀtumor activities of
SYD983 and SYD981 were, at least in this xenograft model, very comparable, which could be
explained by the lower ADC exposure for SYD983, counteracted by the somewhat higher
potency of SYD983.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
On the basis of the preclinical profiles of SYD981 and SYD983, the more straightforward
synthesis of linkerꢀdrug 2, and the lower tendency of SYD983 to form HMW species, linkerꢀ
drug 2 was selected as the lead linkerꢀdrug.
Average DAR. We next set out to determine the optimum DAR. Conjugates of trastuzumab
and 2 with average DARs of about 1 (SYD982) and 3 (SYD984) were prepared analogously to
SYD983 using different amounts of TCEP. The ADCs were then compared to SYD983 for
physicochemical and preclinical properties.
Percentages of HMW species in SYD982, SYD983, and SYD984 were 1.3, 2.4, and 6.0%,
respectively. As a consequence of the random conjugation approach, percentages of
unconjugated antibody in the ADCs decreased from approximately 60% for SYD982 to 10% for
SYD984, while the percentages for the total of DAR6 and DAR8 species increased from
approximately 2% for SYD982 to 10% for SYD984. Despite the presence of different amounts
of HMW species and the different percentages for the individual DAR species, a shortꢀterm
stability study, in which ADCs were stored for four weeks at temperatures up to 40 °C, could not
differentiate between the ADCs. No significant differences were observed when the ADCs were
analyzed for changes in solubilized content, amount of HMW species, average DAR, individual
DAR species, onset temperature for aggregation, melting temperature, charged species, and
HER2 binding (data not shown).
ACS Paragon Plus Environment
18

Page 19 of 70
Molecular Pharmaceutics
1
2
3
4
5
6
7
8
9
Plasma stability of SYD982 and SYD984 was comparable to that of SYD983. Conjugated
antibody halfꢀlives for SYD982 were between 4 and 278 h, while halfꢀlives for SYD984 ranged
from 10 to over 336 h (Table 2). Halfꢀlives for conjugated antibody were, as expected, rather
short in rodent plasma, but comparable to total antibody halfꢀlives in monkey and human plasma.
Conjugated antibody halfꢀlives increased in the series SYD982<SYD983<SYD984, which was
anticipated, because more drug has to be removed from a higherꢀloaded species before it is no
longer detected as a conjugated antibody. Halfꢀlives of total antibody were similar for SYD982
and SYD983 in plasma from all species. Halfꢀlives of total antibody for SYD984 in mouse and
rat plasma were slightly lower than for SYD983, while they were comparable to those for
SYD983 in monkey and human plasma.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
In vivo PK data for SYD982 and SYD984 were generated following single intravenous bolus
injections of SYD982 and SYD984 in Balb/c mice (Figure 4A), Sprague Dawley rats (Figure
4B), and cynomolgus monkeys (Figure 4D,F), and results were compared to the data for
SYD983. In line with the in vitro plasma stability data, exposure to conjugated antibody in
rodents was highest for SYD984 and lowest for SYD982. In cynomolgus monkeys, time–
concentration curves for conjugated antibody and total antibody were very similar, indicative of
a high in vivo stability of the ADCs. In all species, exposure to total antibody tended to be
somewhat lower for SYD984 than for SYD982 and SYD983, which may be indicative of a
slightly faster clearance of the intact (higher DAR) ADC species in SYD984 from circulation.
SYD982 and SYD984 were also evaluated for in vivo efficacy in the BTꢀ474 xenograft model
(Figure 5B,D). Efficacy clearly depended on drug load. Like SYD983, SYD982 and SYD984
showed a clear dose–response relationship, but at equivalent antibody dose SYD984 was clearly
the most efficacious and SYD982 the least efficacious of the three ADCs. These data were
ACS Paragon Plus Environment
19

Molecular Pharmaceutics
Page 20 of 70
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
confirmed in vitro when cytotoxicity against the BTꢀ474c cell line was assessed. SYD982 and
SYD984 had IC values of 0.13 and 0.04 ꢁg/mL, respectively, which bracket the value found for
50
SYD983 (0.06 ꢁg/mL).
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
SYD982, SYD983, and SYD984 were wellꢀtolerated in female cynomolgus monkeys up to
two dosages of 30 mg/kg 24 days apart. Clinical, clinical pathology, and histopathological effects
4
3
observed were all mild and transient, as previously reported for SYD983, and comparable in
nature for all three ADCs. However, the intensity of the effects in general increased with dose
and increasing average DAR. For example, microscopic examination of skin sections revealed
hyperpigmentation of the extremities as one of the most sensitive and quantitative effects, and
the degree of hyperpigmentation was highest for SYD984. Notably, peripheral neuropathy,
assessed by clinical signs and histopathology, and thrombocytopenia were not observed for any
of the ADCs up to 30 mg/kg for two cycles.
On the basis of the estimated efficacy–safety windows of SYD982, SYD983, and SYD984,
none of the ADCs could be disqualified. From a physicochemical point of view, however,
SYD984 was disfavored because of the higher amount of HMW species. We considered that
fractionation of SYD983 to remove the unconjugated antibody and the higherꢀloaded DAR6 and
DAR8 species could result in an ADC outperforming any of the unfractionated ADCs.
Fractionation of SYD983 was carried out with preparative HIC using a resin with a relatively
low hydrophobicity. Unconjugated antibody did not bind to the resin under the conditions
applied and eluted in the flowꢀthrough fraction. The lower DAR species were then eluted with
elution buffer that did not contain organic coꢀsolvent. The higher DAR species remained largely
bound to the resin until the column was stripped (Figure 6). The fractionated ADC, obtained in
approximately 60% yield, was designated SYD985. It is a mixture consisting of about 95%
ACS Paragon Plus Environment
20

Page 21 of 70
Molecular Pharmaceutics
1
2
3
4
5
6
7
8
9
DAR2 and DAR4 species in an approximate 2:1 ratio and has an average DAR of about 2.8. In
this respect, SYD985 is much more homogeneous than SYD984, while having a similar average
DAR.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
SYD985 was extensively evaluated preclinically, comparing favorably to its unfractionated
4
3
predecessor, SYD983. It was also demonstrated to have several favorable pharmacological
5
5
characteristics with respect to adoꢀtrastuzumab emtansine. Hence, SYD985 was selected for
clinical evaluation, and Phase I clinical evaluation recently started.
DISCUSSION
Most of the ADCs that are currently on the market or in clinical development contain drugs
that belong either to the class of maytansinoids or to the class of auristatins. Both drug classes
inhibit tubulin polymerization, a process that is essential for mitosis, thus inhibiting cell division
and eventually inducing apoptosis. The tubulin binderꢀbased ADCs currently in the clinic have
several toxicities in common, despite the fact that the ADCs bind to different targets. The nature
of the toxicities is in line with the expected adverse effect profile of tubulin inhibitors, that is,
neurotoxicity, neutropenia, lymphopenia, and thrombocytopenia, and this may suggest that these
toxicities are related to the rather high systemic (peak) exposure levels of unconjugated drug that
occur after administration of the ADCs. Alternatively, the similarities in the toxicity profiles of
maytansinoidꢀ and auristatinꢀbased ADCs may be due to common deposition characteristics of
the ADCs, which are independent of target binding and distribution of the target. Either way, offꢀ
target toxicities are often doseꢀlimiting, thereby limiting the therapeutic index of the ADCs.
We developed a linkerꢀdrug technology based on the duocarmycin class of compounds.
Duocarmycins, consisting of a DNAꢀalkylating and a DNAꢀbinding moiety, exert their cell
ACS Paragon Plus Environment
21

Molecular Pharmaceutics
Page 22 of 70
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
killing capabilities by binding to the minor groove in DNA and selectively alkylating the N3 of
adenine residues. We envisioned that this different mechanism of action could enable the
development of ADCs with a different clinical profile, both in terms of indications that could be
treated and doseꢀlimiting side effects that would be observed. Linkerꢀdrugs were evaluated in
ADCs based on the HER2ꢀbinding antibody trastuzumab. Linkerꢀdrug 2, which contains the
inactivated drug secoꢀDUBA (3a), was eventually selected as lead linkerꢀdrug and used to
develop SYD985, an ADC in which linkerꢀdrugs are conjugated to reduced interchain disulfide
bridges and that consists mainly of DAR2 and DAR4 species.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
secoꢀDUBA was identified as an interesting candidate drug from an extensive structure–
activity relationship study. The imidazo[1,2ꢀa]pyridine core structure of the DNAꢀbinding
moiety imposes several favorable properties on secoꢀDUBA, one of which being that it has a
pKa close to 6. This can be exploited during ADC manufacturing and formulation of the ADC by
using (slightly) acidic process and formulation buffers. Protonation will enhance water solubility
of and repulsion between ADC molecules, which in turn may lead to reduced selfꢀassociation, a
56
common phenomenon of ADCs, especially when they bear relatively hydrophobic drugs.
Hydrolysis of the amide in DUBA (3b), resulting from spirocyclization of secoꢀDUBA,
occurred more rapidly in plasma compared to the spirocyclized form of its indole analog 26. This
is likely due to an increased electrophilicity of the carbonyl carbon as a consequence of the
additional inductive effects exerted by the second nitrogen atom. Amide hydrolysis is possibly
catalyzed or facilitated by plasma components, as DUBA was found to be hydrolyzed more
slowly in buffer at physiological pH. On the basis of separate plasma stability studies with
ADCs, it seems plausible that it is the spiro form that is hydrolyzed, and not the seco form, as
amide hydrolysis of the antibodyꢀbound drug, trapped in its inactive seco form through linker
ACS Paragon Plus Environment
22

Page 23 of 70
Molecular Pharmaceutics
1
2
3
4
5
6
7
8
9
attachment to the hydroxyl group in the alkylating moiety, was not detected. Vinylogous amide
5
7
conjugation in spiro compounds with an alkylating moiety similar to that in DUBA was shown
to lengthen the amide C–N bond and allow for amide hydrolysis with lithium hydroxide under
unusually mild conditions. Similarly, DUBA may be more sensitive than its (conjugated) seco
analog towards hydrolysis in buffer at physiological pH and in plasma because of vinylogous
amide conjugation. The in vivo halfꢀlife of DUBA in Wistar rats was found to be only 1.1 h. This
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
5
8
is considerably shorter than halfꢀlives reported for MMAE (5.6 h in rats ) and DM1 (6.3 h at 0.1
5
9
mg/kg in Sprague Dawley rats ), representative drugs from the currently most widely explored
drug classes for ADCs and the drugs in brentuximab vedotin and adoꢀtrastuzumab emtansine,
respectively.
Linkage of secoꢀDUBA to cysteine thiols of partly reduced trastuzumab via a linker connected
to its DNAꢀbinding moiety (Tmabꢀ30) led to lower in vitro cytotoxicity in SKꢀBRꢀ3 cells and
reduced human plasma stability compared to linkage via the DNAꢀalkylating unit (Tmabꢀ29).
Both observations should likely be attributed to the 4ꢀmethylpiperazinꢀ1ꢀylcarbonyl promoiety
that was used to block the hydroxyl group in the alkylating moiety of Tmabꢀ30 to prevent
premature spirocyclization and degradation of the antibodyꢀbound drug. This promoiety has also
5
1
60
been applied in KWꢀ2189, a duocarmycin prodrug that failed in phase II clinical trials, and in
3
1
duocarmycinꢀbased ADC MDXꢀ1203, and was designed to be cleaved by carboxylesterases
5
2
inside the cell. The lower plasma stability of Tmabꢀ30 could be explained by a mechanism in
which (enzymeꢀcatalyzed) removal of the promoiety is followed by spirocyclization and
subsequent hydrolysis of the amide bond in the antibodyꢀbound drug, leading to loss of the
DNAꢀalkylating unit from the ADC. Conversely, inefficient removal of the promoiety in SKꢀBRꢀ
3
cells may be responsible for the reduced cytotoxic activity in SKꢀBRꢀ3 cells. Therefore, as
ACS Paragon Plus Environment
23

Molecular Pharmaceutics
Page 24 of 70
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
intracellular levels of carboxylesterase appear to be dependent on the cell line and
carboxylesterase seems also present in human plasma, the 4ꢀmethylpiperazinꢀ1ꢀylcarbonyl
promoiety seems unsuitable for use in an ADC. As no alternative promoieties could be identified
with substantially improved properties, linking via the DNAꢀbinding moiety was abandoned as a
strategy.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Linking to the DNAꢀalkylating moiety in 3a resulted in excellent stability in monkey plasma
for all ADCs that were evaluated. Furthermore, ADCs SYD983, based on linkerꢀdrug 2, and
SYD981, based on linkerꢀdrug 45, were shown to have conjugated antibody exposures in
cynomolgus monkey that were close to that of total antibody, indicative of high stability of the
intact ADC. Thus, in vitro monkey plasma stability translated well to in vivo PK in monkey. It is
therefore expected that the excellent stability observed in human plasma will translate into high
exposure of intact ADC in patients as well. The slightly faster clearance of conjugated antibody
compared to total antibody after dosage of SYD983 or SYD981 to monkeys is a commonly
observed phenomenon for ADCs and could for example be due to slow linkerꢀdrug degradation
61,62
by plasma enzymes or maleimide exchange.
The latter comprises transfer of the complete
linkerꢀdrug moiety through a retroꢀMichael reaction from the antibody to another molecule with
a thiol group, such as albumin or glutathione.
In mouse and rat plasma, SYD983 and SYD981 were less stable. This has been attributed to
the presence of rodentꢀspecific CES1c, as in plasma from CES1c knockꢀout mice, SYD983 has a
43
stability comparable to that in cynomolgus monkey and human plasma. Increased stability of
5
5
SYD983 in the absence of CES1c has been confirmed in vivo in CES1c knockꢀout mice. As
cyclization spacerꢀduocarmycin intermediates were not observed in plasma stability studies,
ACS Paragon Plus Environment
24

Page 25 of 70
Molecular Pharmaceutics
1
2
3
4
5
6
7
8
9
CES1c likely cleaves the carbamate linkage between cyclization spacer and duocarmycin
payload.
SYD981 was somewhat more stable than SYD983, but in vitro cytotoxicity was slightly better
for SYD983, resulting in in vivo efficacies in a BTꢀ474 xenograft model that were comparable.
As both ADCs contain the same drug, different tumor cell pharmacokinetics could account for
the difference in in vitro cytotoxicity. Furthermore, differences in pharmacokinetics, for example
differences in linker metabolism in circulation, may exist that cannot be detected with the current
ELISA assay to detect conjugated antibody, as this assay specifically detects the duocarmycin
component.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The in vitro cytotoxicity data of linkerꢀdrug 2ꢀbased ADCs SYD982, SYD983, and SYD984,
having average DARs of approximately 1, 2, and 3, respectively, show that potency in BT474c
cells increases with increasing DAR when IC values are expressed in antibody equivalents.
5
0
When expressed in drug equivalents, however, IC values are comparable, suggesting that the
50
total amount of drug that is released inside the BT474c cells correlates with the extracellular
concentration of antibodyꢀbound payload. Similarly, SYD984 was shown to be the most
efficacious ADC in the BTꢀ474 xenograft model on a mg/kg basis, but the three ADCs are more
comparable in terms of efficacy when dosages are expressed in drug equivalents. From an
efficacy point of view, SYD984 may be favored over SYD983 and SYD982 because: (i) a lower
dose of SYD984 is needed to reach a certain level of efficacy; (ii) SYD984 is potentially more
effective in the treatment of tumors that express low copy numbers of HER2 and are only able to
shuttle a limited number of ADC molecules into the cell; (iii) SYD984 contains a lower amount
of unconjugated trastuzumab, which can compete with conjugated trastuzumab for HER2
binding, thereby potentially lowering antiꢀtumor activity.
ACS Paragon Plus Environment
25

Molecular Pharmaceutics
Page 26 of 70
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Results from a doseꢀrange finding toxicity and pharmacokinetic study in cynomolgus monkeys
showed minimal differences between SYD982, SYD983, and SYD984 up to the highest dose
tested (30 mg/kg, q24dx2). Although toxicities were generally somewhat more intense at higher
average DAR, effects were generally mild and reversible. Dependent on the type of tissue
affected and whether or not there is (low) expression of HER2, increased toxicity in the series
SYD982<SYD983<SYD984 may for instance be caused by augmented drug exposure, enhanced
exposure to higherꢀloaded conjugates, or reduced exposure to unconjugated antibody. PK
profiles in cynomolgus monkeys showed that exposure to total antibody was slightly lower for
SYD984 than for SYD982 and SYD983, which could be explained by a faster clearance of intact
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
6
3
higher DAR species from circulation. In contrast, exposure to conjugated antibody was highest
for SYD984. This is in line with expectations, because more time is needed to fully deconjugate
the higherꢀloaded antibody.
SYD985, obtained through HIC fractionation of SYD983, consists of about 95% DAR2 and
DAR4 species in an approximate 2:1 ratio and has an average DAR of about 2.8. It is
comparable to SYD984 in terms of average DAR, but it is considerably more homogeneous and
contains significantly lower amounts of unconjugated antibody and DAR6 and DAR8 species.
Absence of the higher DAR species is considered advantageous as these species are likely to be
6
3
more rapidly cleared, more toxic and more prone to aggregation, while absence of
unconjugated trastuzumab prevents competition at the level of receptor binding and
internalization.
In conclusion, antiꢀHER2 ADC SYD985 was selected as lead clinical candidate following
extensive linkerꢀduocarmycin structure–activity studies, in which both drug and linker
ACS Paragon Plus Environment
26

Page 27 of 70
Molecular Pharmaceutics
1
2
3
4
5
6
7
8
9
components were optimized and the composition of the ADC in terms of average DAR and DAR
species was improved. SYD985 recently entered clinical Phase I evaluation.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
EXPERIMENTAL SECTION
5
3
54
General. The protocols for the synthesis of HꢀValꢀCitꢀPABA and linkerꢀdrugs 2 and 45,
4
3
43
the conjugation of linkerꢀdrugs to trastuzumab, and the HIC purification of SYD983 have
been described elsewhere. The SEC and HIC methods used for characterization of ADCs, the
ELISA methods to quantify total antibody and conjugated antibody, the methods used for data
analysis of in vitro data, and the protocols for measuring in vitro plasma stability and carrying
43
out the in vivo efficacy and safety studies have been described before. Linkerꢀdrugs were
purified by preparative HPLC on a Waters SunFire Prep C18 OBD column (5 µm, 19 × 150 mm)
using acetonitrile and 0.1% TFA in water as the eluents. Purity of compounds was assessed by
HPLC and was ≥95% in all cases. Optical rotation was recorded on a Perkin Elmer 241
polarimeter. Enantiomeric excess was determined by HPLC on a Daicel Chiralcel ODꢀH column
(5 µm, 4.6 × 250 mm; nꢀheptane:isopropyl alcohol, 99:1, v/v; flow rate: 0.8 mL/min). NMR
1
13
spectra were recorded on a Bruker AVANCE400 (400 MHz for H; 100 MHz for C) or Bruker
1
13
AVANCE300 (300 MHz for H; 75 MHz for C) spectrometer. Chemical shifts are reported in
ppm relative to tetramethylsilane as an internal standard. Silica gel for column chromatography
(40–63 ꢁm, 60 Å) was obtained from Screening Devices. All solvents used were reagentꢀgrade
or HPLCꢀgrade.
ACS Paragon Plus Environment
27

Molecular Pharmaceutics
Page 28 of 70
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Methyl 4-Hydroxy-8-methyl-2-naphthoate (7). 2ꢀMethylbenzaldehyde (4) (144 mL, 1.25
mol) and dimethyl succinate (5, 319 g, 2.18 mol) in methanol (400 mL) were treated with
NaOMe (87.8 g, 1.63 mol) at 65–80 °C for 1 h. The mixture was cooled to 20 °C, neutralized
with 3 M hydrochloric acid and diluted with water. Unreacted 5 was removed by extraction with
CH Cl . The mixture was acidified with 3 M hydrochloric acid, and the product was extracted
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
2
into CH Cl . The extract was dried with MgSO , filtered, and concentrated, providing a 9:1
2
2
4
mixture of (E) isomer 6a and (Z) isomer 6b. The crude product was dissolved in THF (1 L), and
the solution was treated with TFAA (165 mL, 1.19 mol) at reflux temperature until complete
conversion to 7. The reaction mixture was neutralized with an aqueous K CO solution. The
2
3
product was extracted with EtOAc. The organic layer was dried with MgSO , filtered, and
4
concentrated, and the residue was crystallized by cooling to –5 °C. The whiteꢀyellow crystals
were filtered, washed with acetonitrile, and dried, providing alcohol 7 (112 g, 42%).
1
H NMR (CDCl , 300 MHz): δ = 2.74 (3H, s, 8ꢀCH ), 3.99 (3H, s, CH O), 6.07 (1H, s, OH),
3
3
3
13
7
.38 (1H, d, H7), 7.48 (1H, dd, H6), 7.53 (1H, s, H3), 8.12 (1H, d, H5), 8.37 (1H, s, H1); C
NMR (DMSOꢀd , 100 MHz): δ = 19.7 (8ꢀCH ), 52.7 (CH O), 107.0 (C3), 117.7 (C1), 120.81
6
3
3
(
C5), 127.3 (C6), 127.5 (C4a, C2), 128.3 (C7), 133.1 (C8a), 135.7 (C8), 154.5 (C4), 167.1
C=O).
(
4-(Benzyloxy)-8-methyl-2-naphthoic Acid (9). Alcohol 7 (83 g, 0.38 mol) was treated with
benzyl chloride (46.4 mL, 403 mmol) in DMF (330 mL) at 80 °C in the presence of K CO (74.3
2
3
g, 537 mmol). When the reaction was complete, the mixture was cooled and diluted with CH Cl2
2
and water. The layers were separated, and the organic layer was concentrated. The residue,
ACS Paragon Plus Environment
28

Page 29 of 70
Molecular Pharmaceutics
1
2
3
4
5
6
7
8
9
comprising methyl 4ꢀ(benzyloxy)ꢀ8ꢀmethylꢀ2ꢀnaphthoate (8), was dissolved in toluene (500 mL)
and methanol (650 mL), and the resultant solution was treated with a 4 M aqueous NaOH
solution (750 mL) under reflux for 2 h. Water (2.9 L) was added in several portions, and
methanol and toluene were removed by distillation. Crude 9 was precipitated by addition of 4 M
hydrochloric acid (330 mL), the suspension was cooled to 10 °C, and the precipitate was filtered
off, washed with cold water, collected, and dried. Crude 9 was crystallized from toluene to give
offꢀwhite crystals. Crystals were filtered, washed with toluene, and dried to give 9 (95.6 g, 82%).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
H NMR (DMSOꢀd , 300 MHz): δ = 2.69 (3H, s, 8ꢀCH ), 5.37 (2H, s, CH Ph), 7.40–7.60 (8H,
6
3
2
13
m, H ), 8.11 (1H, d, H5), 8.29 (1H, s, H1), 13.20 (1H, br s, CO H); C NMR (DMSOꢀd , 100
Ar
2
6
MHz): δ = 19.7 (8ꢀCH ), 70.1 (CH Ph), 104.8 (C3), 119.7 (C1), 120.4 (C5), 127.8 (C4a), 127.9
3
2
(
C2, C2′), 128.4 (C6), 128.5 (C7), 128.6 (C4′), 129.0 (C3′), 132.7 (C8a), 135.9 (C8), 137.4
C1′), 154.8 (C4), 168.1 (C=O).
(
tert-Butyl (4-(Benzyloxy)-8-methylnaphthalen-2-yl)carbamate (10). Acid 9 (100 g, 342
mmol) was reacted with diphenylphosphoryl azide (109 g, 397 mmol), tertꢀbutyl alcohol (65.9 g,
89 mmol), and Et N (52.3 mL, 376 mmol) in toluene (500 mL) at 85 °C for 3 h. After the
8
3
mixture had been cooled to 30 °C, EtOAc and water were added, and the layers were separated.
The organic layer was washed with an aqueous Na CO solution and saturated aqueous NaCl,
2
3
dried with MgSO , filtered, and concentrated. The residue was triturated with isopropyl alcohol.
4
The solid was filtered off, washed with isopropyl alcohol, and dried to provide 10 (110 g, 88%).
1
H NMR (DMSOꢀd , 400 MHz): δ = 1.53 (9H, s, Boc), 2.56 (3H, s, 8ꢀCH ), 5.34 (2H, s,
6
3
CH Ph), 7.23 (1H, m, H6), 7.30–7.39 (3H, m, H1, H7, H4′), 7.42–7.46 (2H, m, H3′), 7.56–7.59
2
1
3
(2H, m, H2′), 7.82 (1H, s, H3), 7.99 (1H, d, H5), 9.53 (1H, s, NH); C NMR (DMSOꢀd , 100
6
ACS Paragon Plus Environment
29