Development and evaluation of 4-(pyrrolidin-3-yl)benzonitrile derivatives as inhibitors of lysine specific demethylase 1

Article abstract of DOI:10.1016/j.bmcl.2017.08.052

As part of our ongoing efforts to develop reversible inhibitors of LSD1, we identified a series of 4-(pyrrolidin-3-yl)benzonitrile derivatives that act as successful scaffold-hops of the literature inhibitor GSK-690. The most active compound, 21g, demonstrated a K_d value of 22 nM and a biochemical IC₅₀ of 57 nM. In addition, this compound displayed improved selectivity over the hERG ion channel compared to GSK-690, and no activity against the related enzymes MAO-A and B. In human THP-1 acute myeloid leukaemia cells, 21g was found to increase the expression of the surrogate cellular biomarker CD86. This work further demonstrates the versatility of scaffold-hopping as a method to develop structurally diverse, potent inhibitors of LSD1.

Full text of DOI:10.1016/j.bmcl.2017.08.052

Accepted Manuscript
Development and evaluation of 4-(pyrrolidin-3-yl)benzonitrile derivatives as
inhibitors of lysine specific demethylase 1
Daniel P. Mould, Ulf Bremberg, Allan M. Jordan, Matthis Geitmann, Alison E.
McGonagle, Tim C.P. Somervaille, Gary J. Spencer, Donald J. Ogilvie
PII:
S0960-894X(17)30853-3
DOI:
http://dx.doi.org/10.1016/j.bmcl.2017.08.052
BMCL 25246
Reference:
Bioorganic & Medicinal Chemistry Letters
To appear in:
Received Date:
Revised Date:
Accepted Date:
24 July 2017
21 August 2017
23 August 2017
Please cite this article as: Mould, D.P., Bremberg, U., Jordan, A.M., Geitmann, M., McGonagle, A.E., Somervaille,
T.C.P., Spencer, G.J., Ogilvie, D.J., Development and evaluation of 4-(pyrrolidin-3-yl)benzonitrile derivatives as
inhibitors of lysine specific demethylase 1, Bioorganic & Medicinal Chemistry Letters (2017), doi: http://dx.doi.org/
10.1016/j.bmcl.2017.08.052
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Development and evaluation of 4-(pyrrolidin-3-yl)benzonitrile derivatives as
inhibitors of lysine specific demethylase 1
Daniel P. Moulda, Ulf Brembergc, Allan M. Jordana, Matthis Geitmannc, Alison E. McGonaglea, Tim C.
P. Somervailleb, Gary J. Spencerb, and Donald J. Ogilviea
aDrug Discovery Unit, Cancer Research UK Manchester Institute, University of Manchester, Wilmslow
Road, Manchester, M20 4BX, UK
bLeukaemia Biology Laboratory, Cancer Research UK Manchester Institute, University of Manchester,
Wilmslow Road, Manchester, M20 4BX, UK
cBeactica AB, Uppsala Business Park, Virdings allé 2, 75450, Uppsala, SE
Abstract
As part of our ongoing efforts to develop reversible inhibitors of LSD1, we identified a series of 4-
(pyrrolidin-3-yl)benzonitrile derivatives that act as successful scaffold-hops of the literature inhibitor
GSK-690. The most active compound, 21g, demonstrated a Kd value of 22 nM and a biochemical IC50 of
57 nM. In addition, this compound displayed improved selectivity over the hERG ion channel compared
to GSK-690, and no activity against the related enzymes MAO-A and B. In human THP-1 acute myeloid
leukaemia cells, 21g was found to increase the expression of the surrogate cellular biomarker CD86. This
work further demonstrates the versatility of scaffold-hopping as a method to develop structurally diverse,
potent inhibitors of LSD1.
Keywords
epigenetics; LSD1; KDM1A; reversible inhibitor; stem cell differentiation; cancer therapy; epigenetic
therapy; acute myeloid leukaemia

Table of contents graphic
Text:
Lysine specific demethylase 1 (LSD1) has been found to have roles that go far beyond its initially
discovered function of removing methyl marks from histone tails via an oxidative demethylation
mechanism.1-3 The clinical potential of this epigenetic eraser in cancer and associated disease areas has
been unravelled, largely as a result of experiments exploiting the use of irreversible LSD1 inhibitors
developed from tranylcypromine. In a subset of acute myeloid leukaemia (AML), LSD1 has been shown
to be important in the maintenance of leukaemia stem cells and the promotion of a block on
differentiation.4 Additionally, LSD1 inhibition has been shown to reactivate the all-trans retinoic acid
(ATRA) pathway by increasing the expression of myeloid-differentiation genes, suggesting that a
combination of LSD1 inhibitor and ATRA may offer a new therapeutic option for AML subtypes that are
generally insensitive to ATRA therapy alone.5 While tranylcypromine derivatives have facilitated a great
deal of research, there is still an unmet need for a variety of reversible tool compounds that display no off-
target pharmacology.6
GSK-690 (1, Figure 1) is a versatile start-point for the development of reversible LSD1 inhibitors,
however the development of this compound was hindered by a human ether-a-go-go related gene (hERG)
liability.7 Aromatic scaffold replacements are an attractive and logical way to retain potency while

moving into novel intellectual property space.8 To that end, patents that disclose a number of mono- and
bicycle heterocyclic scaffolds that replace the central pyridine ring have recently been disclosed.9-14 We
instead looked to focus on scaffold-hops that featured ring systems that were partly or fully saturated. We
hypothesised that the decreased lipophilicity and increased flexibility of a saturated core structure may be
a valid strategy to try and overcome the persistent hERG liability of 1 and related derivatives.15-16
An increase in sp3 character has the effect of increasing the molecules 3-dimensional (3D) shape and
reducing the aromatic ring count, which have both been noted as important factors in improving the
developability of compounds as potential drugs.17-18 On the other hand, this approach brings the added
complications of potential stereocenters and synthetic challenges of transformations that would be trivial
on an aromatic scaffold but remain complex in the timeframe of hit-to-lead development.19
Herein, we describe our efforts to identify series of reversible inhibitors with scaffolds containing
multiple sp3 centres, and their elaboration and development into potent inhibitors of LSD1 as confirmed
by biochemical, surface plasmon resonance (SPR) and cellular biomarker assays with significantly
reduced activity against the hERG ion channel.
To investigate the potential of saturated scaffold-hops from compound 1, we begin by using Cresset
Spark (Welwyn Garden City, UK) software to identify potential replacements of the pyridine core
incorporating sp3 centres at the attachment points of the two aryl groups. This identified two potential
isosteric replacements that appeared to be synthetically accessible via cycloaddition reactions, namely β-
lactams and pyrrolidines. From there, we refined the spatial overlap between the scaffold-hops and 1
using Cresset Torch to design potential ligands (Figure 1). The cis-β-lactam diastereomers (2) were
predicted to be far more favourable than the trans- (3), as this allows the two aryl groups be oriented in
the same plane as modelled by 1, whereas in the trans-diastereoisomers the aryl groups are forced out of
the plane. Modelling potential pyrrolidine scaffolds on the other hand, the increased flexibility of the fully

saturated ring system means that it is possible for a trans-diastereoisomer (4) to adopt a conformation
where the two hydrogens sit axially, with the aryl rings oriented only slightly out of plane.
Figure 1. Cresset Torch alignment between 1 and new scaffold designs 2 – 4. A similarity score of >0.7
was considered a good score. Absolute stereochemistry indicated.
The β-lactam core can be constructed via a Staudinger [2+2] cycloaddition reaction between an imine and
a ketene which can optionally be formed in situ. We used the method developed by Coulthard and co-
workers,20 which employs propylphosphonic anhydride (T3P) under basic conditions to afford a mixture
of diastereoisomers, with the trans-β-lactam being formed as the major diastereoisomer on similar
substrates to the target molecule. After forming imine 6, T3P-mediated cycloaddition with 2-(p-
tolyl)acetic acid or 2-(4-methoxyphenyl)acetic acid afforded a mixture of diastereometric products in an
almost 1:1 ratio. These were separable by preparative HPLC to afford the cis- and trans- products 7–10 as

a mixture of enantiomers. The characteristic vicinal coupling of the lactam protons by NMR (cis- J = 5.2
Hz; trans- J = 2.3 Hz) allowed for identification of the structure of each product.20 These enantiomeric
mixtures were tested for their inhibitory activity against LSD1 in biochemical time resolved fluorescent
resonance energy transfer (TR-FRET) and biophysical surface plasmon resonance (SPR) assays, as
previously reported.16 While the trans-diastereoisomers (8–10) were inactive in both assay formats (IC50 >
30µM and Kd > 100 µM), we were encouraged to find that the cis-diastereoisomers, which were predicted
to adopt a more favourable conformation to act as an isostere of 1, did show modest activity and
reversibility against LSD1 by SPR (8 Kd = 38 µM, 10 Kd = 53 µM).
Scheme 1. Synthesis of compounds 7 – 10. Reagents and conditions: i) N-Boc-trans-1,4-
cyclohexanediamine, MeOH, reflux, 1 h, 78%. ii) T3P, DIPEA, (p-tolyl)acetic acid (7, 9) or 2-(4-
methoxyphenyl)acetic acid (8, 10), chloroform, reflux, 2 h, then TFA, DCM, 1 h, 24 – 58% (total of both
isomers).
CN
CN
NC
OEt
iv
R1
R1
ii
i
P
OEt
O
R
N
N
R1
12–14
11a–b
R2
R3
O
a
R = Me
15–17 R2 = Me
18–20 R2 = H
b R = CN
21a–h
22a–b
23a–e
iii

Scheme 2. Synthesis of compounds 21–23. Reagents and conditions: i) 12: 11a, 4-formylbenzonitrile,
KOtBu, THF, 0 °C to rt, 16 h, 54%; 13: 11b, 6-methoxynicotinaldehyde, KOtBu, THF, 0 °C to rt, 16 h,
70%; 14: 1-methyl-1H-indazole-5-carbaldehyde, 11b, KOtBu, THF, 0 °C, 1 h, 91%. ii) trimethylamine
oxide, LDA, DMF, 0 °C, 1 h, 13–52%. iii) 1-Chloroethyl chloroformate, THF, reflux, 1 h, 40 °C, then
MeOH, reflux, 1 h, 85–97% iv) Appropriate carboxylic acid, COMU, DIPEA, DMF, 1 h, then 4 M
HCl/dioxane, 10–66%.
Historically 3,4-disubstituted pyrrolidines have been constructed by the reaction of alkenes most
commonly with azomethine ylide sources such as N-(methoxymethyl)-N-
(trimethylsilylmethyl)benzylamine in the presence of trifluoroacetic acid.21 However, we found this
combination of reagents was unable to effect conversion of the stilbene precursors to the cycloaddition
product. The trans-pyrrolidines were synthesised via 1,3-dipolar cycloaddition of trans-stilbene
precursors with trimethylamine oxide in the presence of excess LDA, as developed by Davoren and co-
workers.22 This afforded the N-methyl products, which were deprotected by reaction with 1-chloroethyl
chloroformate followed by refluxing in methanol. Amide coupling, followed by Boc-deprotection,
afforded amines 21a–h, 22a–b and 22a–e.
Pyrrolidines 21a–h were screened against LSD1 by biochemical TR-FRET assay, and we were gratified
to observe that this series transferred the spatial and electrostatic similarity to 1 into promising levels of
potency (Table 1). Optimisation of the basic centre focussed upon varying the length and flexibility of the
linker between the amide and the amine functionalities. Piperidine 21a and trans-cyclohexylamine 21b
showed reasonable levels of activity, with sub-micromolar Kd values. From initial results, introduction of
a methylene unit between the amide and the cyclohexane looked to offer improved affinity (21c–f).
Moving the methylene linker from the amide to the amine led to transexamic acid derivative 21g which
displayed low nanomolar activity by both SPR and biochemical assay, and was only ~2-fold less active
than 1. Methylation of this amine (21h) was poorly tolerated, resulting in a ~4-fold drop-off in potency

The patent literature generally deviates away from the p-tolyl group, and we were keen to explore
replacements that are frequently observed, to see if this would offer an additional route to improve
potency (Table 2). The methoxypyridine replacement (23a–b) showed a disappointing 8–15-fold drop-off
in potency in comparison to matched pairs 21d and 21f. More successful was replacement with a
methylindazole (23a–e), however potency could not be improved further. The matched pair of 21g (23d)
showed only a ~2-fold drop off by SPR. Interestingly, replacing the cyclohexyl motif of 23d with an
aromatic system (23c) shows a ~20-fold decrease in potency, suggesting the importance of a saturated
system in this region.

Table 1. Pyrrolidine amide SAR
SPR Kd
(µM)
Compound
1
R
IC50 (µM)
0.041
(0.015)
0.008
0.61
4.2 (0.6)
21a
21b
3.0 (0.6)
0.95 (0)
0.43
0.19
21c
21d
0.57 (0.02) 0.11
0.82 (0.06) 0.12
21e
21f
0.68 (0.05) 0.061
0.057
0.022
(0.003)
21g

0.26 (0.04) 0.098
21h
aIC50 and Kd values for selected compounds against the LSD1 enzyme in biochemical and biophysical
assays. Standard deviation given in parentheses. IC50 determined from 10-point concentration/effect
experiments. Geometric mean of at least two independent experimental determinations given.
Table 2. SAR of p-tolyl replacements and amide elaboration
SPR Kd
(µM)
Compound R1
R2
IC50 (µM)
5.7 (0.2)
0.95
22a
9.0 (0.6)
4.8 (0.2)
1.7
1.4
22b
23a
2.6 (0.4)
4.4 (0.2)
0.21 (0)
0.77
2.0
23b
23c
23d
0.048

1.2 (0.1)
0.54
23e
aIC50 and Kd values for selected compounds against the LSD1 enzyme in biochemical and biophysical
assays. Standard deviation given in parentheses. IC50 determined from 10-point concentration/effect
experiments. Geometric mean of at least two independent experimental determinations given.
Using knowledge of the binding mode of 1 bound to LSD1,7 we used Glide (Schrödinger, New York,
USA),24 to model both enantiomers of 21g into the crystal structure of LSD1 bound to the reversible
ligand tetrahydrofolate (PDB: 4KUM).25 There was no significant difference between the docking scores
of each enantiomer, however we are yet to separate the enantiomers and test this experimentally. The
predicted binding mode of (3R,4R)-21g is given in Figure 2. The compound was predicted to made two
key hydrogen bonding interactions. The nitrile forms a hydrogen bond with Lys661, with the basic centre
being directed towards Asp555 and Asp556. Targeting these acidic residues appears to be a valuable
strategy and has been seen with other structurally distinct series of reversible inhibitors.26-27 The wide-
variety of scaffold hops from 1 appears to be due to the lack of close contacts between the core (in this
case the pyrrolidine ring) and the protein. Thus, this central portion of the molecule could be used as a
versatile handle to improve DMPK. The p-tolyl groups points into a hydrophobic channel formed by
Ile356, Leu677, Leu693, and Trp695.

Figure 2. Predicted binding mode of the (3R,4R)-enantiomer of compounds 21g (in the LSD1 active site
(4KUM). Visualised in PyMOL.
Compound 21g demonstrated excellent selectivity over both MAO-A and MAO-B (IC50 > 25 µM), and
improved selectivity over the hERG ion channel compared to 1 (21g IC50 = 12.1 µM, 1 IC50 = 3.2 µM).28
The increase in expression levels of the cell surface marker CD86 in THP-1 acute myeloid leukaemia
cells as a response to LSD1 inhibition has been used to provide a robust cellular assay format (Figure 3).27
We tested our most active derivative, 21g, alongside 1, and a previously described hydroxypyrazole based
inhibitor of LSD1 (24).16 All three compounds induced the expression of CD86 after 48 h in a dose
dependent manner, as expected, although 21g displayed marginally less activity than 1.
In summary, we have designed a series of 4-(pyrrolidin-3-yl)benzonitrile derivatives that demonstrate
potent activity against LSD1 in biochemical assays and by SPR. The most potent derivative, 21g,
demonstrates comparable levels of cellular activity to other previously described reversible inhibitors of

LSD1, complete selectivity against related monoamine oxidase enzymes and shows improved selectivity
in comparison to 1 against the hERG ion channel. This profile makes 21g a useful cell-active tool
compound to further explore the role of LSD1 in cells. This series represents the first example of a potent
saturated scaffold-hop from 1, and is an excellent start-point for further optimisation.
O
O
H2N
N
N
O
N
N
CN
24
Figure 3. Top: Structure of compound 24. Bottom: CD86 assay results for compounds 21g, 1 and 24. The
upper detection limit of this assay was ~1300 mean fluorescence intensity (MFI) units as determined by
treatment of cells with 250 nM of the irreversible inhibitor OG-86. Error bars display standard deviation
of each mean value.

Acknowledgments
This work was supported by Cancer Research UK (Grants C5759/A12328, C480/A11411, C5759/A17098
and C5759/A02901). Additional support was provided to DPM by the Society of Chemical Industry
through a Messel Scholarship. We are thankful to Bohdan Waszkowycz for his assistance with Glide. In
vitro pharmacokinetic and hERG data was provided by Cyprotex Discovery (Macclesfield, UK). Images
of protein-ligand docking were captured within the PyMOL Molecular Graphics System, Version 1.7.6.2.
(Schrödinger, LLC, New York).
Notes and References
Tim Somervaille has ongoing research collaborations with Oryzon Genomics and consults for Imago
Biosciences. The other authors declare no competing financial interest.
1.
2.
3.
4.
5.
6.
Shi Y, Lan F, Matson C, et al. Cell. 2004;119(7): 941-953.
Shi Y-J, Matson C, Lan F, Iwase S, Baba T, Shi Y. Mol Cell. 2005;19(6): 857-864.
Amente S, Lania L, Majello B. BBA Gen. Reg Mech. 2013;1829(10): 981-986.
Harris WJ, Huang X, Lynch JT, et al. Cancer Cell. 2012;21(4): 473-487..
Schenk T, Chen WC, Göllner S, et al. Nat Med. 2012;18(4): 605-611
Mould DP, McGonagle AE, Wiseman DH, Williams EL, Jordan AM. Med Res Rev. 2015;35(3):
586-618.
7.
Johnson NW. The Identification of GSK2879552, a Mechanism Based Irreversible Inhibitor of
the Histone Lysine Demethylase LSD1. XXIV EFMC International Symposium on Medicinal Chemistry.
Manchester, UK; 2016.
8.
9.
Sun H, Tawa G, Wallqvist A. Drug Disc Today. 2012;17(7–8): 310-324.
Chen Y, K., Kanouni T, Nie Z, Stafford JA, Veal JM, Sung LM. WO2016004105. Celgene
Quanticel Research Inc.; 2016.

10.
Chen YK, Kanouni T, Kaldor SW, Stafford JA, Veal JM. WO2015089192. Celgene Quanticel
Research Inc.; 2015.
11.
Chen YK, Kanouni T, Nie Z, Stafford JA, Veal JM, Sung LM. WO2016003917. Celgene
Quanticel Research Inc.; 2016.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
Liangxing W, Courter JR, Chunhong H, et al. US20160009720. Incyte Corporation; 2016.
Liangxing W, Konkol LC, Lajkiewicz N, et al. US20160009712. Incyte Corporation; 2016.
Liangxing W, Xiaozhao W, Wenqing Y, Zhang C. US20160009711. 2016.
Moorthy NS, Ramos MJ, Fernandes PA. SAR QSAR Environ Res. 2012;23(5-6): 521-536.
Mould DP, Bremberg U, Jordan AM, et al. Bioorg Med Chem Lett. 2017;27(14): 3190-3195.
Ritchie TJ, Macdonald SJF. Drug Disc Today. 2009;14(21–22): 1011-1020.
Lovering F, Bikker J, Humblet C. J Med Chem. 2009;52(21): 6752-6756.
Ritchie TJ, Macdonald SJF, Young RJ, Pickett SD. Drug Disc Today. 2011;16(3–4): 164-171.
Coulthard G, Unsworth WP, Taylor RJK. Tetrahedron Lett. 2015;56(23): 3113-3116.
Terao Y, Kotaki H, Imai N, Achiwa K. Chem Pharm Bull. 1985;33(7): 2762-2766.
Davoren JE, Gray DL, Harris AR, Nason DM, Xu W. Synlett. 2010;2010(16): 2490-2492.
Friesner RA, Banks JL, Murphy RB, et al. J Med Chem. 2004;47(7): 1739-1749.
Luka Z, Pakhomova S, Loukachevitch LV, Calcutt MW, Newcomer ME, Wagner C. Protein Sci.
2014;23(7): 993-998.
25.
26.
27.
28.
Vianello P, Sartori L, Amigoni F, et al. J Med Chem. 2017;60(5): 1673-1692.
Sartori L, Mercurio C, Amigoni F, et al. J Med Chem. 2017;60(5): 1693-1715.
MAO and hERG inhibition assays were performed by Cyprotex (Macclesfield, U.K.).
Lynch JT, Cockerill MJ, Hitchin JR, Wiseman DH, Somervaille TCP. Anal Biochem.
2013;442(1): 104-106.