Discovery of cytotoxic dolastatin 10 analogues with N-terminal modifications

Article abstract of DOI:10.1021/jm501649k

Auristatins, synthetic analogues of the antineoplastic natural product Dolastatin 10, are ultrapotent cytotoxic microtubule inhibitors that are clinically used as payloads in antibody-drug conjugates (ADCs). The design and synthesis of several new auristatin analogues with N-terminal modifications that include amino acids with α,α-disubstituted carbon atoms are described, including the discovery of our lead auristatin, PF-06380101. This modification of the peptide structure is unprecedented and led to analogues with excellent potencies in tumor cell proliferation assays and differential ADME properties when compared to other synthetic auristatin analogues that are used in the preparation of ADCs. In addition, auristatin cocrystal structures with tubulin are being presented that allow for the detailed examination of their binding modes. A surprising finding is that all analyzed analogues have a cis-configuration at the Val-Dil amide bond in their functionally relevant tubulin bound state, whereas in solution this bond is exclusively in the trans-configuration. This remarkable observation shines light onto the preferred binding mode of auristatins and serves as a valuable tool for structure-based drug design.

Full text of DOI:10.1021/jm501649k

Subscriber access provided by University Libraries, University of Memphis
Article
Discovery of cytotoxic Dolastatin 10 analogs with N-terminal modifications
Andreas Maderna, Mattew Doroski, Chakrapani Subramanyam, Alexander Porte, Carolyn Leverett,
Beth C Vetelino, Zecheng Cheng, Hud Risley, Kevin Parris, Jayvardhan Pandit, Alison H. Varghese,
Suman Shanker, Cynthia Song, Sai Chetan K Sukuru, Kathleen A. Farley, Melissa M Wagenaar,
Michael J. Shapiro, Sylvia Musto, My-Hanh Lam, Frank Loganzo, and Christopher J. O'Donnell
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/jm501649k • Publication Date (Web): 28 Nov 2014
Downloaded from http://pubs.acs.org on December 3, 2014
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.
Journal of Medicinal Chemistry 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 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Discovery of cytotoxic Dolastatin 10 analogs with
N-terminal modifications
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
Andreas Maderna a*, Matthew Doroski a, Chakrapani Subramanyam a, Alexander Porte a,
Carolyn Leverett a, Beth C. Vetelino a, Zecheng Chen a, Hud Risley a , Kevin Parris a,
Jayvardhan Pandit a, Alison Varghese a, Suman Shanker a, Xi Song a, Sai Chetan K. Sukuru a,
Kathleen A. Farley a, Melissa M. Wagenaar a, Michael J. Shapiro a, Sylvia Musto b, My-Hanh
Lam b, Frank Loganzo b, Christopher J. O’Donnell a
a Pfizer Worldwide Research and Development, Worldwide Medicinal Chemistry, Oncology
Eastern Point Road, Groton, CT 06340 (USA)
b Pfizer Worldwide Research and Development, Oncology Research Unit, 401 N. Middletown
Road, Pearl River, NY 10965 (USA)
KEYWORDS (Dolastatin 10 analogs, N-terminal modifications, cytotoxicity, tubulin-bound co-
crystal structures, auristatin binding mode, antibody drug conjugates)
ACS Paragon Plus Environment
1

Journal of Medicinal Chemistry
Page 2 of 66
1
2
3
4
ABSTRACT
5
6
7
8
9
Auristatins, synthetic analogs of the antineoplastic natural product Dolastatin 10, are ultra-potent
cytotoxic microtubule inhibitors that are clinically used as payloads in antibody-drug conjugates
(ADCs). The design and synthesis of several new auristatin analogs with N-terminal
modifications that include amino acids with α,α-disubstituted carbon atoms are described,
including the discovery of our lead auristatin, PF-06380101. This modification of the peptide
structure is unprecedented and led to analogs with excellent potencies in tumor cell proliferation
assays and differential ADME properties when compared to other synthetic auristatin analogs
that are used in the preparation of ADCs. In addition, auristatin co-crystal structures with tubulin
are being presented that allow for the detailed examination of their binding modes. A surprising
finding is that all analyzed analogs have a cis-configuration at the Val-Dil amide bond in their
functionally relevant tubulin bound state, whereas in solution this bond is exclusively in the
trans-configuration. This remarkable observation shines light onto the preferred binding mode of
auristatins and serves as a valuable tool for structure based drug design.
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
INTRODUCTION
In 1987 Pettit et al. described the discovery of Dolastatin 10 (1), a highly cytotoxic peptide that
was isolated from the sea hare Dolabella auricularia in the Indian Ocean.1 This remarkable
discovery was the result of fifteen years of research in which a number of cytotoxic peptides
were isolated from this sea creature with Dolastatin 10 (1) being the most potent analog in tumor
cell proliferation assays. It was soon recognized that this compound was an extremely potent
mitotic spindle poison and one of several compounds produced by cyanobacteria in the intestines
ACS Paragon Plus Environment
2

Page 3 of 66
Journal of Medicinal Chemistry
1
2
of Dolabella auricularia that were postulated to act as defense agents against predators.2,3,4 It
was demonstrated that in addition to inhibiting tubulin polymerization 1 has a strong inhibitory
effect on tubulin dependent GTP hydrolysis.2 In order to obtain sufficient quantities for
structure identification and biological testing Pettit et. al. collected 1000 kg of wet animal
material whose combined ethanol / 2-propanol extract yielded 28.7 mg of pure Dolastatin 10
after extensive purification (Figure 1).1
3
4
5
6
7
8
9
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
Figure 1. Structures of Dolastatin 10 (1) and key amino acid components.
Dolastatin 10 (1) is comprised of the four amino acids dolavaline (Dov), valine (Val),
dolaisoleuine (Dil), dolaproine (Dap) and the C-terminal amine dolaphenine (Doe).1,5 The
binding site for 1 is located adjacent to the vinca binding site at the α,β-tubulin heterodimer
interface (Figure 5).6 Dolastatin 10 (1) was advanced into clinical trials, but despite encouraging
preclinical efficacy data, no appreciable therapeutic index could be attained due to its significant
toxicity at the maximum tolerated dose (MTD).7-13 Shortly after the discovery of 1 a large
number of synthetic analogs, termed auristatins, were prepared by Pettit et al., Miyazaki et al.
and other groups with the goal to investigate the influence of structural changes on cell based
14-16
potencies in tumor cell proliferation assays (Figure 2).
reviewed.17
These efforts have been extensively
ACS Paragon Plus Environment
3

Journal of Medicinal Chemistry
Page 4 of 66
1
2
3
4
5
6
7
8
9
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
Figure 2. Structures of key auristatin analogs with three different C-termini. Top: Dolastatin 10
(1) and monomethyl auristatin-D (MMAD, 2). Middle: auristatin-E (3) and monomethyl
auristatin-E (MMAE, 4). Bottom: auristatin-PHE (5) and monomethyl auristatin-F (MMAF, 6).
Color coding of C-termini: dolaphenine, (orange), (1S,2R)-(+)-norephedrine (yellow) and
phenylalanine (green). Red arrow: removal of one N-Me group does not lead to significant
potency loss.
Figure 2 shows key auristatin analogs containing the three different C-termini groups
dolaphenine, (1S,2R)-(+)-norephedrine and phenylalanine. All three analogs were first described
by Pettit. et al. as their N,N-dimethyl valine (dolavaline) N-termini constructs. 1,18,19 Miyazaki et
al. discovered that removal of one methyl group from the N-terminal amine of the natural
product Dolastatin 10 (1) gave mono-methyl auristatin-D (MMAD, 2) and did not compromise in
14
vivo antitumor activity.
This finding was later recognized by Senter et al., who made the
ACS Paragon Plus Environment
4

Page 5 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
remarkable discovery that these auristatins having a secondary amine at their N-terminus could
be attached to a linker and subsequently conjugated to monoclonal antibodies. This clever idea
led to generation of highly potent and efficacious antibody-drug conjugates (ADC’s).20-22
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
Figure 3. Hotspots for structural modifications of auristatin analogs.
Figure 3 displays the auristatin core structure and indicates the locations where the majority of
chemical variations have been published. Notably, after the original publications of Pettit et al.
and Miyazaki et al. most of the subsequent auristatin research has focused on analogs with C-
terminal modifications and N-terminal extensions without modifications of the peptide structure
(N-Me)Val-Val-Dil-Dap with only few exceptions.14,23-28 Of particular note in this regard is the
naturally occurring analog Symplostatin 1 that differs from Dolastatin 10 (1) by the presence of
a terminal N,N-dimethylisoleucine instead of a terminal dolavaline residue. 3,29 The cytotoxicity
of Symplostatin 1 is reported to be only slightly less when compared to 1. To evaluate potential
opportunities for structural modification, analysis of the detailed binding mode of the auristatins
to tubulin is important. At the start of our studies, the only reported auristatin co-crystal structure
using the [(Tc)2R] tubulin construct was obtained with the analog soblidotin (TZT-1027) (7,
Figure 4)15 by Cormier, Knossow et al..6 The structure was disordered in some regions of the co-
crystal complex but allowed for the basic assessment of the orientation and binding mode in the
nucleotide exchange binding cleft of the tubulin construct used in this study. However, due to the
ACS Paragon Plus Environment
5

Journal of Medicinal Chemistry
Page 6 of 66
1
2
3
4
5
6
7
8
9
low resolution of the structure no detailed information of specific receptor interactions could be
ascertained with good confidence. Nevertheless, when studying the co-crystal structure of 7 with
tubulin, we noted a binding void around the N-terminus which suggested the potential for
additional substitutions on the N-terminal carbon atom (Figure 4). We hypothesized that N-
terminal amino acids with a α,α-dialkyl carbon substitutions could lead to potent new auristatin
analogs. Additionally, we were interested in determining if changes on the N-terminus would
lead to different in vitro ADME properties. Of particular interest was the potential modulation of
in vitro intrinsic clearance profiles because our ultimate goal was to use these molecules as
payloads for the next generation of antibody drug conjugates (ADC’s).22 The design rationale
was that an ADC payload with high plasma clearance would be desirable, as this could reduce
toxicity resulting from premature release of the payload from the antibody.
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
6

Page 7 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
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
Figure 4. Co-crystal structure of soblidotin (TZT-1027) (7) with the (Tc)2R tubulin construct as
described by Cormier, Knossow et al.6 View of the N-terminus with the arrow indicating the N-
terminal secondary carbon atom of interest. The structure of 7 is depicted with the correct
stereochemical assignments as reported for 7 15, but deviates from that reported in ref. 6, see text.
A binding void is apparent in the receptor pocket that indicates the potential for further
substitutions. Protein residues are displayed with their van der Waals radii and color coded based
on atom type (dark blue: nitrogen, red: oxygen, light blue: beta subunit carbon, green: alpha
subunit carbon). For the overall orientation of auristatin analogs at the α,β-tubulin interface see
Figure 5 and ref. 6 .
ACS Paragon Plus Environment
7

Journal of Medicinal Chemistry
Page 8 of 66
1
2
3
4
5
6
7
8
9
Herein, we describe new N-terminal modifications with α,α-disubstituted amino acids that led
to auristatin analogs with excellent potencies in tumor cell proliferation assays. We also present
co-crystal structures of four analogs with tubulin having improved atomic resolution allowing for
the first time the detailed examination of the auristatin binding mode. Furthermore, we
investigate selected in vitro ADME characteristics of the new analogs and describe the
pharmacokinetic profiling of the lead auristatin analog 20a (PF-06380101).
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
CHEMISTRY
Pettit et al. first described the total synthesis of Dolastatin 10 (1) in 1989 and also described
the preparation of the monomer building blocks dolaisoleuine (Dil), dolaproine (Dap) and
dolaphenine (Doe).5 Due to the interest in Dolastatin 10 (1) and related analogs, the preparation
of the monomers has been improved over time. While there are multiple convergent methods of
preparing the linear 1 molecule, Pettit et al. utilized the convergent method of separately
preparing the left hand Dov-Val-Dil tripeptide (8) and right hand Dap-Doe dimer (9). 5,30,31
ACS Paragon Plus Environment
8

Page 9 of 66
Journal of Medicinal Chemistry
1
2
Scheme 1. Convergent synthesis of Dolalstatin 10 (1) based on tripeptide 8 and dimer 9. 5
3
4
5
6
7
8
9
We were interested in designing a flexible route that would allow for both N- and eventually
C-terminal modifications in a modular fashion. For this purpose we sought to modify the
synthetic route to enable introduction of the terminal amino acid at the last reaction step from 18.
Additionally, compound 18 could be generated by coupling of the dimers 15 and 16 (Scheme 2).
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
N
N
O
O
H
H
N
a, b, c
rt
N
FmocHN
O
N
N
FmocHN
S
S
+
N
Boc
N
O
O
O
O
O
O
O
O
16
17
15
N
S
O
H
d
N
N
H2N
N
rt
O
O
O
O
18
a Reagents and conditions: (a) TFA/DCM for 15 (b) HCl dioxane for 16 (c) HATU, i-Pr2NEt,
CH2Cl2, 78 % (d) Et2NH, THF, rt, 87%.
Scheme 2. Synthesis of the tetramer Val-Dil-Dap-Doe (18) by using a modular approach from
dimers 15 and 16. Boc-protected 18 was previously prepared by an alternative method.30,31
The tetramer 18 was then utilized as a building block for the auristatin analogs as shown in
Scheme 3. The amino acids 19 a-l as their corresponding N-Boc or N-Fmoc protected analogs
were coupled to the tetramer 18 using HATU with diisopropylethyl amine, then deprotected
using either trifluoroacetic acid or diethylamine to give the auristatin analogs 20a-l.
ACS Paragon Plus Environment
9

Journal of Medicinal Chemistry
Page 10 of 66
1
2
3
4
5
6
7
8
9
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
Reagents and conditions: (a) HATU, Et3N, DMF; (b) Et2NH, THF (a-e) or TFA, CH2Cl2 (f-l),
(yields: 27-67% over two steps). N-Pg: protected amine.
Scheme 3. Synthesis of new auristatin analogs 20a-l.
Additionally, three specific examples of the above auristatin analogs were prepared with
phenylalanine as the C-terminus instead of the dolaphenine C-terminus (Scheme 4). For this
purpose we applied similar chemistry procedures using the previously described tetramer Val-
Dil-Dap-Phe-OMe (21).32 This tetramer was coupled to protected Aib (19a), N-Me-Aib (19b)
and (S)-2-methylpyrrolidine-2-carboxylic acid (19g) followed by saponification of the methyl
ester and deprotection of the terminal amines.
ACS Paragon Plus Environment
10

Page 11 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
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
Reagents and conditions: (a) HATU, Et3N, DMF (b) LiOH, MeOH and (c) TFA, CH2Cl2
(additional step for 22g). Yields: 33-56% over two steps.
Scheme 4. Synthesis of new auristatin analogs 22a,b,g containing a phenylalanine- instead of the
dolaphenine C-terminus present in 20a-l.
RESULTS AND DISCUSSION
Cell potencies of selected reference compounds and Aib-modified auristatins
Based on the crystal structure of 7 with tubulin, our hypothesis was that replacing the N-
terminal amino acid of Dolastatin 10 (1) with an amino acid having two substituents at the α-
position should be tolerated. However, it is well known that amino acids with two alkyl
substituents at the α-position can lead to distortions of the peptidic structure due to induction of
helical turns resulting from the Thorpe-Ingold effect, i.e. as exemplified with the amino acid Aib
(2-aminoisobutyric acid).33 The potential for success of this structural variation was non-obvious
based on work by Miyazaki et al. who replaced the terminal dolavaline with N,N-dimethyl
alanine and N,N-dimethyl glycine. Remarkably, the change from dolavaline to N,N-dimethyl
glycine increased in vivo efficacy in xenograft studies, whereas N,N-dimethyl alanine at this
ACS Paragon Plus Environment
11

Journal of Medicinal Chemistry
Page 12 of 66
1
2
position completely abolished any efficacy.14 This observation appeared to indicate the sensitive
nature of the N-terminus with respect to structural modifications. Consequently, other positions
on the peptide, i.e. the C-terminus, were deemed more favorable for potential structural
modifications.16,17 However, the interpretation of in vivo efficacy results in the absence of any
data regarding cell-based potency, in vitro ADME and pharmacokinetics needs to be done with
great caution. At the onset of our studies we investigated selected reference auristatins in tumor
cell proliferation assays and two new analogs that contain 2-aminoisobutyric acid (Aib) at the N-
terminus. GI50 values of cell viability were used as the primary potency readout taking into
account all compounds in this manuscript showed good inhibition in the cell-free tubulin
polymerization assay (data not shown). However, the tubulin polymerization assay generally
requires significantly higher compound concentrations compared to cell based studies and is not
suited for detailed SAR studies that examine subtle structural changes.34 Of consideration is that
cell potencies are a direct function of target affinity, cell-membrane permeability and potential
active and passive transport processes, amongst other factors. It was determined that the passive
cell permeability for all of the analogs prepared was generally low (RRCK cell line: Papp AB <
0.5 x 10-6 cm/sec),35 and unless active uptake transport or other cell-uptake mechanisms (i.e.
endocytosis) are involved, the cell GI50 values are likely driven by only a fraction of the
extracellular compound concentration that is present inside of the cells. Hence, any change in
observed cell potencies could reflect not just different tubulin binding affinities, but also altered
cell permeability, cell uptake and transporter profiles. Of note is that none of the investigated cell
lines described here had ABCB1 (MDR1) efflux transporters expressed. Having this limitation
regarding cell based assay analysis in mind, BT474 cells (breast carcinoma), N87 cells (liver
metastasis of gastric carcinoma) and MDA-MB-361-DYT2 cells (breast carcinoma) were
3
4
5
6
7
8
9
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
12

Page 13 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
selected for the potency readout (Table 1). As expected, excellent potency was observed for the
reference compounds MMAD (2), MMAE (4) and MMAF-OMe (6-OMe) (Figure 2) that all
share the N-methyl valine N-terminus. To our surprise, the N-methyl glycine and N-methyl
alanine analogs 23 and 24 retained good cell potencies despite the lack of the terminal isopropyl
group. To the best of our knowledge, this is the first time that detailed cellular potencies are
described for MMAD (2), the N-Me-glycine- (23) and N-Me- alanine (24) analogs. Gratifyingly,
the new Aib analogs 20a and 20b having an α,α-geminal dimethyl group at the N-terminus
showed a high degree of potency in all three investigated cell lines. Interestingly, the potency of
these two compounds was not dependent on whether or not the terminal amino group was
methylated. This stands in contrast to the corresponding N-terminal valine matched pair 2
(MMAD) and 25, where the absence of the N-methyl group in 25 led to a significant reduction in
potency.
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
13

Journal of Medicinal Chemistry
Page 14 of 66
1
2
3
MDA-MB-
4
5
6
7
ID
BT474 [nM]
361-DYT2
[nM]
N87 [nM]
8
9
0.22
(± 0.15, 9)
0.24
0.49
0.54
4
(MMAE)
(± 0.15, 15)
0.21
(± 0.35, 22)
0.44
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
6-OMe
(MMAF-OMe)
(± 0.18, 24)
(± 0.15, 14)
(± 0.54, 20)
0.12
(± 0.094, 12)
0.49
0.090
(± 0.013, 6)
0.69
0.30
(± 0.22, 13)
1.2
2
(MMAD)
23
(± 0.15, 3)
(± 0.26, 3)
(± 0.39, 3)
0.23
(± 0.014,3)
5.3
0.25
(± 0.081, 4)
9.0
0.42
(± 0.15, 4)
10
24
25
(± 0.91, 3)
0.26
(± 2.4, 2)
0.19
(± 2.2, 2)
0.27
20a
(PF-06380101)
(± 0.37, 16)
0.045
(± 0.073, 19)
0.076
(± 0.16, 18)
0.12
20b
(± 0.012, 3)
(± 0.047, 4)
(± 0.10, 4)
Table 1. Cell potency in tumor cell viability assays expressed in GI50 values with two significant
figures as arithmetic means in nanomolar concentrations. Standard deviations and the count of
individual assay runs (n). Each individual assay run is a duplicate readout. N-terminal amino
acids: 23: N-Me-L-glycine, 24: N-Me-L-alanine, 25: L-valine, 20a: aib, 20b: N-Me-aib. Aib: 2-
aminoisobutyric acid.
ACS Paragon Plus Environment
14

Page 15 of 66
Journal of Medicinal Chemistry
1
2
3
4
Crystal structure analysis of 20a (PF-06380101)
5
6
7
8
9
Encouraged by the excellent cell potency of 20a (PF-06380101) we sought to further analyze
the binding mode by crystallography using the (Tc)2R tubulin construct.6 Numerous tubulin
crystal structures have been published, most notably by elegant work from Gigant, Knossow et
al. 6,36,37 These structures give insight into the mechanism of polymerization of tubulin as well as
the method of inhibition of polymerization via co-structure with various inhibitor classes. For
most of these co-structures, the resolution is between 3.5Å and 4Å. This low resolution limits the
clarity of which the interaction with tubulin can be visualized with these inhibitors. In the case
of the auristatin compound soblidotin (TZT-1027) (7), examination of the structure and data
deposited in the Protein Data Bank (Entry 3E22) show that the last two C-terminal residues are
disordered at 3.8Å and cannot be placed with confidence.6 Recent advances in tubulin
crystallography have provided crystals that diffract to significantly higher resolution prior to
soaking compounds.38 Using this new information, we undertook the structure determination of
auristatin compounds bound to tubulin with the goal of providing a higher resolution co-structure
to aid our structure based design of tubulin inhibitors. This work is exemplified by the co-
structure of tubulin with 20a, which was solved at an improved 3.1Å resolution.
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
15

Journal of Medicinal Chemistry
Page 16 of 66
1
2
3
4
5
6
7
8
9
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
Figure 5. Co-crystal structure of 20a with the (Tc)2R tubulin construct. Overview of the binding
mode with relative orientation of construct components. 20a binds at the α,β-tubulin interface. A
second molecule of 20a is also associated with the exterior β-subunit. PDB code 4X1I.
As previously described by Cormier et al., the binding site of auristatins is in close proximity
to the β-tubulin nucleotide exchange region and the auristatin binding event leads to direct
interference with the GTP/GDP hydrolysis process resulting in compromised tubulin
polymerization and mitotic arrest (Figure 5).6 In the co-structure of 20a, all residues of the
compound can be fit-to-density and as a result the examination of the interaction of this
compound with tubulin was possible with greater clarity than in the 3.8Å structure solved by
Cormier et al..6
ACS Paragon Plus Environment
16

Page 17 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
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
Figure 6. Binding mode of 20a observed in the 3.1Å co-structure with tubulin with selected key
receptor interactions at the N-terminus. Red: β-subunit, blue: α-subunit, GDP in green. For
simplicity the structure of 20a is drawn with trans-configuration of the Val-Dil amide bond
ACS Paragon Plus Environment
17

Journal of Medicinal Chemistry
Page 18 of 66
1
2
3
4
5
6
(indicated with arrow), but in the co-crystal structure the Val-Dil amide bond of 20a bound at the
tubulin 1/2 interface has cis-coniguration, see text. PDB code 4X1I.
7
8
9
In this higher resolution tubulin-20a co-crystal structure (Figures 5,6), the protein chain
tracings are identical to that which has been described previously with the T5 loop flipped out,
exposing Asp 197 to solvent.6 This movement of the T5 loop is required when compounds
bind in the vinca domain peptide region and is regularly observed. Upon insertion of an entity
into the vinca domain region, additional curvature is induced into the protein assembly. For
example, atoms in β-2 subunit move by almost 3Å and the C-terminal residues of the stathmin
domain move by over 5Å. It is presumably this inhibitor-induced curvature that is the basis for
the diffraction sensitivity of the soaked protein crystals. The Aib N-terminus is easily
accommodated in this pocket and slightly larger groups can be tolerated. The N-2 valine interacts
with the β-1 subunit by inserting into a large pocket formed by the side chains of Val177, Tyr210
and Leu227 (not shown). Notable polar interactions include a hydrogen bond network between
the terminal protonated amino group with Asp 197 located on the T5 loop and the amide
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
carbonyl of Phe 351, a bifocal interaction of the N-2 valine with Asn 329 and a critical
hydrogen bond interaction between the backbone amide of Tyr 224 with the terminal carbonyl
groups of Dap and Doe in 20a. This latter hydrogen bond network arranges the aromatic ring Tyr
224 in a π-stacking interaction with the nucleobase of GDP, preventing it’s dissociation from β-
tubulin. The trapping of GDP is believed to be a key mechanism by which auristatins exert their
tubulin-interfering effect. As stated above, the remaining residues including the C-terminus
interact primarily with the β-tubulin subunit and their mode of binding arranges the peptide such
that the carbonyl associated with the final residue is located at the tip of helix H7 aligned with
the helix dipole (Figure 7).
ACS Paragon Plus Environment
18

Page 19 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
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
Figure 7. Binding mode of 20a observed in the 3.1Å co-structure with tubulin, view on the C-
terminus. PDB code 4X1I.
In addition to the molecule of 20a found at the interface of the 1 and 2 subunits, a second
molecule of 20a is bound unproductively to the distal 2 subunit (Figure 5). There is no alpha
subunit associated with the 2 subunit but the compound interacts with the beta subunit in a
similar way as seen at the internal alpha/beta interface.
ACS Paragon Plus Environment
19

Journal of Medicinal Chemistry
Page 20 of 66
1
2
3
4
Configurational analysis
5
6
7
8
9
We do observe a major configurational difference between the two bound molecules 20a, as
the 20a molecule found at the 1/2 tubulin interface has a cis-configuration near the N-
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
terminus between the amino acids Val-Dil while the compound associated with 2 tubulin has
trans-configuration at the Val-Dil amide bond. Superimposing the beta subunits and examining
the fit-to-density for the compounds shows that this difference is real. By adopting the cis-
configuration 20a is able to provide complementary interactions to Asn 329 as seen in Figure 6.
These interactions are not possible if the Val-Dil amide bond has trans-configuration. The lack
of a corresponding  subunit for the exterior copy of 20a explains why the cis-configuration in
the second 20a molecule is not observed.
ACS Paragon Plus Environment
20

Page 21 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Figure 8. The configurational differences between the valine (Val) and dolaisoleuine (Dil)
residues of compound 20a, as observed at the 1/2 tubulin interface (cyan) and observed bound
to 2 tubulin (orange), is highlighted by the red circle. Only in the cis-configuration the carbonyl
of the valine residue is able to make a productive interaction with Asn 329 from the 2 subunit.
PDB code 4X1I.
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
An important observation is that the Val-Dil amide bond of 20a in solution has trans-
configuration, with no observable cis-isomer detected by NMR at room temperature (DMSO),
whereas the amide bond in between Dap and Dil is a mixture of both trans- and cis-rotamers.
However, in the tubulin co-crystal structure the Val-Dil bond amide bond of the functionally
relevant bound form of 20a at the 1/2 tubulin interface is seen to be in cis-configuration. At
first glance this might be surprising since none of this form seems to be present in solution as
detected by NMR, suggesting the energy difference between trans- and cis forms is greater than
2.7 kcal/mol. For tertiary amides the barrier to rotation around the amide bond is often 17-20
kcal/mol.39 At this energy it is still possible to access both trans-and cis-forms kinetically since
the exchange between trans- and cis-forms by rotation around the amide bond is in the msec to
sec range. Thus, tubulin can select the most favorable form giving the lowest overall G of
binding. Apparently, the energy expenditure for the rotation of the Val-Dil amide bond in 20a
from trans (solution state) to cis (tubulin bound) is more than compensated for by the resulting
high binding energy of 20a having picomolar dissociation constants (tubulin binding assay by
fluorescence polarization, data not shown). Interestingly, the other unproductively bound
molecule of 20a appears to not make favorable interactions with tubulin and it remains
associated in its trans form as previously shown. This is in agreement with two studies by Alattia
et al. and Benedetti et al. who investigated the NMR solution structure of 1.40,41 In addition, the
ACS Paragon Plus Environment
21

Journal of Medicinal Chemistry
Page 22 of 66
1
2
3
4
5
6
7
8
9
crystallographic analysis of the tubulin-unbound isodolastatin 10, an epimer of 1, also only
showed the trans-configuration between Val and Dil.42 It is interesting to speculate what the
binding energy would be in analogs having pre-organized cis-amide bonds at the Val-Dil
juncture.
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
Activities of additional N-modified analogs
Based on the structural information obtained with 20a additional new analogs with N-
terminal modifications were explored (Table 2). In particular, there was an interest to investigate
how various cyclic motifs would be tolerated at the N-terminus and whether the configuration of
the stereocenter present in the chiral amino acids would influence potency. Also, β-amino acids
represented by 20e,f and fluorine containing analogs 20i,j were prepared to determine the impact
of these modifications on activity. Table 2 indicates that all prepared compounds showed good
cell-based potencies comparable with those of MMAE (4) and MMAF-OMe (6-OMe) with only
subtle differences. However, as pointed out previously, the high activity of 20a (PF-06380101) is
notable in that it possesses a primary amine, a trend that is not observed for the natural valine N-
terminus present in MMAD (2), where the N-Me group is necessary for potency (see matched
molecular pair 2 and 25, table 1).
ACS Paragon Plus Environment
22

Page 23 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
ID
BT474 [nM] MDA-MB-361-DYT2 [nM]
N87 [nM]
0.54
0.22
(± 0.15, 9)
0.24
0.49
4
MMAE
(± 0.15, 15)
0.21
(± 0.35, 22)
0.44
6-OMe
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
(MMAF-OMe)
(± 0.18, 24)
(± 0.15, 14)
(± 0.54, 20)
0.12
(± 0.094, 14)
0.26
0.090
(± 0.013, 8)
0.19
0.30
(± 0.22, 13)
0.27
2 (MMAD)
20a
(PF-06380101)
(± 0.37, 16)
(± 0.073, 19)
(± 0.16, 18)
0.045
(± 0.012, 3)
2.9
0.076
(± 0.047, 4)
5.6
0.12
(± 0.10, 4)
6.4
20b
20c
20d
(± 1.1, 5)
0.28
(± 1.6, 2)
0.35
(± 3.8, 4)
0.33
(± 0.11, 10)
0.25
(± 0.20, 8)
0.81
(± 0.14, 10)
1.4
20e
(± 0.10, 3)
(± 0.49, 3)
(± 1.5, 3)
0.25
(± 0.058, 3)
0.14
0.41
(± 0.19, 3)
0.20
1.1
(± 0.53, 3)
0.15
20f
20g
20h
20i
(± 0.019, 3)
0.11
(± 0.19, 4)
0.21
(± 0.036, 3)
0.36
(± 0.040, 3)
0.16
(± 0.11, 3)
0.25
(± 0.15, 3)
0.59
(0.019, 3)
0.29
(± 0.069, 3)
0.60
(± 0.33, 3)
1.1
20j
20k
20l
(± 0.057, 3)
0.16
(± 0.24, 3)
0.20
(± 1.7, 3)
0.25
(± 0.023, 3)
0.35
(± 0.11, 3)
0.40
(± 0.16, 3)
0.52
(± 0.091, 3)
(± 0.011, 3)
(± 0.22, 3)
ACS Paragon Plus Environment
23

Journal of Medicinal Chemistry
Page 24 of 66
1
2
3
4
5
6
7
8
9
Table 2. Cell potency in tumor cell viability assays of additional auristatin analogs expressed in
GI50 values with two significant figures as arithmetic means in nanomolar concentrations.
Standard deviations and the count of individual assay runs (n). Each individual assay run is a
duplicate readout.
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 β-amino acid versions 20e and 20f are only slightly less potent than the
corresponding α-amino acid counterparts 20a and 20b in the DYT2 and N87 cells, once again
showing little influence of the N-methylation status on potency. The cyclopropyl analog 20c is
less active than the cyclopentyl derivative 20d. Of note are the high activities of the pyrrolidine
and piperidine analogs (20g,h and 20k,l) with seemingly no influence of the configuration of the
stereocenter, once more highlighting that structural modifications on the N-terminus are possible
without compromising potency.
Crystal structure analysis of 22a, 22b and 22g
The good activity of the N-terminal pyrrolidine was predicted based on the analysis of the
crystal structure of 20a, which suggested that five and six-membered saturated rings could be
tolerated in the binding pocket that accommodates the auristatin N-terminus. Critical receptor
interactions of 20a were identified to be the hydrogen bond network of the primary amino group
in 20a with the Asp β179 located on the T-5 loop, the carbonyl backbone of Phe α351 as well as
the bifocal interaction with Asn 329. Figure 9 shows the superimposed co-crystal structures of
22a and 22g allowing the direct comparison of the binding mode of the respective N-termini. For
both compounds the C-terminus was chosen to be phenylalanine instead of dolaphenine to
facilitate the crystallization. This change in the C-terminus did not perturb the overall binding
ACS Paragon Plus Environment
24

Page 25 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
conformation of the auristatin peptide. The inspection of Figure 9 reveals that the overall binding
mode of 22a and 22g is very comparable with good alignment of both bound conformations. Of
particular interest is that in both compounds the N-termini engage in the Asp β179, Phe α351 and
Asn α329 hydrogen bond network. Both analogs retain high tubulin binding affinity at a level
comparable to that of the dolaphenine analogs 20a and 20g (data not shown). Detailed cell-free
tubulin binding affinity studies using a newly developed fluorescence polarization competition
assay will be described in a subsequent publication.
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
Figure 9. Overlayed co-crystal structures of 22a and 22g. Left: structure complexes in the
binding domain superimposed and showing entire peptide structures of 22a and 22g. Right:
close-up view of N-terminus. Visible is the critical hydrogen bond network of both N-termini
with Asp β179, Phe α351 and Asn α329, interactions that are preserved in both molecules and
ACS Paragon Plus Environment
25

Journal of Medicinal Chemistry
Page 26 of 66
1
2
3
4
5
6
believed to be important for binding affinity. The Val-Dil amide bonds are in cis-configuration.
RDB codes: 4X1K (22a), 4X20 (22g).
7
8
9
Figure 10 shows the N-terminal view of the superimposed co-crystal structures of 22b and
22g . The analog 22b differs from 22a by the presence of a secondary N-Me amino group which
does not compromise the hydrogen network of the terminal amino group with Asp β179, Phe
α351 and Asn α329. Of particular note is the difference in the relative orientation of the methyl
group in 22g as compared to the position of both geminal dimethyl groups in 22a and 22b. The
N-methyl group in 22b adjusts the positioning of the N-terminus of the compound relative to
20a. The N-terminal nitrogen moves ~2Å and the N-methyl group is vectored into a channel
between the  and  subunits. This vector is one that can easily accommodate large linear
groups.
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
26

Page 27 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
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
Figure 10. Overlayed co-crystall structures of 22b and 22g, view of the N-terminus. Notable is
the relative orientation of the pyrrolidine Me-group and the two geminal methyl groups in 22b.
In both analogs the Val-Dil amide bonds have cis-configuration and engage with Asn α329 via
bifocal interaction. RDB codes: 4X1Y (22b), 4X20 (22g).
It should be noted that in the deposited structure of soblidotin (TZT-1027) (7), the Val-Dil
amide bond is in the trans-configuration. In contrast, the Val-Dil amide bonds in the functionally
relevant tubulin-bound structures of all the four auristatin analogs (20a, 22a, 22b and 22g)
presented here are in the cis-configuration. In addition to this discrepancy, in the published
crystal structure of 7 three chiral centers (positions 9, 18 and 19a) have different stereochemical
assignments than those originally reported for 7.14,15 In contrast, all compounds described here
ACS Paragon Plus Environment
27

Journal of Medicinal Chemistry
Page 28 of 66
1
2
3
4
5
6
7
8
9
have stereochemical assignments that are consistent with those published for Dolastatin 10 and
other auristatin analogs. The unexpected finding of the cis-configured Val-Dil amide bond in all
crystallized analogs described here, that have a functionally relevant binding mode at the 1/2
tubulin interface, is remarkable and reveals a preferred auristatin binding mode that was not
previously recognized for this compound class. The high potency and conserved binding modes
of the described N-terminal analogs highlight how structural modifications at the N-terminus can
be tolerated without loss of binding affinity. This is also apparent when inspecting the cell-based
potencies of the pyrrolidine epimers 20g and 20h as well as the piperidine epimers 20k and 20l
which indicate that inversion of the pyrrolidine or piperidine stereocenter does not compromise
target affinity or cell based activities.
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 summary, all new compounds show an appreciable amount of activity when compared to
the benchmark auristatins with very little variation in cell-based potencies. This result is
surprising as previous studies suggested that the N-terminal modifications would be less
tolerated. All investigated compounds have generally low passive cell-membrane permeability
with only minor differences (data not shown). Potential explanations for the high cell-potencies,
despite the poor passive cell-membrane permeability, include the possibility for active uptake
mediated by transporters or endocytosis pathways that have been previously described for
peptides.43 It also remains a possibility that the high cell activity is caused by only a fraction of
the drug that penetrates the cell-membrane by passive permeation.
ACS Paragon Plus Environment
28

Page 29 of 66
Journal of Medicinal Chemistry
1
2
3
4
Human liver microsome intrinsic clearance
5
6
7
8
9
Upon profiling of new analogs using an in vitro ADME panel it was observed that all new
compounds with α,α-di-substitution showed significantly elevated human liver microsome
(HLM) apparent intrinsic clearance values (Figure 11).44
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
Figure 11. X-axis: log scale, human liver microsome (HLM) apparent intrinsic clearance values
of selected analogs [μl/min/mg], dose: 1μM. Y-axis: Compound row number, without dimension.
Compounds in the yellow box are compounds with reference N-termini. Significant increase in
ACS Paragon Plus Environment
29