Small molecule inhibitors of histone arginine methyltransferases: Homology modeling, molecular docking, binding mode analysis, and biological evaluations

Article abstract of DOI:10.1021/jm061213n

The screening of the inhibition capabilities of dye-like small molecules from a focused library against both human PRMT1 and Aspergillus nidulans RmtA is reported as well as molecular modeling studies (homology modeling, molecular docking, and 3-D QSAR) o

Full text of DOI:10.1021/jm061213n

J. Med. Chem. 2007, 50, 1241-1253
1241
Small Molecule Inhibitors of Histone Arginine Methyltransferases: Homology Modeling,
Molecular Docking, Binding Mode Analysis, and Biological Evaluations
Rino Ragno,*,† Silvia Simeoni,† Sabrina Castellano,‡ Caterina Vicidomini,‡ Antonello Mai,† Antonella Caroli,†
Anna Tramontano,§ Claudia Bonaccini,§,+ Patrick Trojer,| Ingo Bauer, Gerald Brosch,*, and Gianluca Sbardella*,‡
Istituto Pasteur - Fondazione Cenci Bolognetti, Dipartimento di Studi Farmaceutici, UniVersita` degli Studi di Roma “La Sapienza”, P.le Aldo
Moro 5, I-00185 Roma, Dipartimento di Scienze Farmaceutiche, UniVersita` degli Studi di Salerno, Via Ponte Don Melillo, I-84084 Fisciano
(SA), Dipartimento di Scienze Biochimiche, A. Rossi Fanelli, UniVersita` degli Studi di Roma “La Sapienza”, P.le Aldo Moro 5, 00185 Roma,
Howard Hughes Medical Institute, Department of Biochemistry, DiVision of Nucleic Acids Enzymology, Robert Wood Johnson Medical School,
UniVersity of Medicine and Dentistry of New Jersey, 683 Hoes Lane, Piscataway, New Jersey 08854, and DiVision of Molecular Biology,
Biocenter-Innsbruck Medical UniVersity, Fritz-Preglstrasse 3, 6020 Innsbruck, Austria
ReceiVed October 17, 2006
The screening of the inhibition capabilities of dye-like small molecules from a focused library against both
human PRMT1 and Aspergillus nidulans RmtA is reported as well as molecular modeling studies (homology
modeling, molecular docking, and 3-D QSAR) of the catalytic domain of the PRMT1 fungal homologue
RmtA. The good correlation between computational and biological results makes RmtA a reliable tool for
screening arginine methyltransferase inhibitors. In addition, the binding mode analyses of tested derivatives
reveal the crucial role of two regions, the pocket formed by Ile12, His13, Met16, and Thr49 and the SAM
cisteinic binding site subsite. These regions should be taken into account in the design of novel PRMT
inhibitors.
Introduction
highly conserved catalytic domain. Seven of them (including
the recently discovered FBXO11/ PRMT9) catalyze the transfer
of a methyl group from S-adenosylmethionine (SAM or
AdoMet, 1, Chart 1) to guanidino nitrogen atoms of arginine
residues,21 resulting in S-adenosylhomocysteine (2, Chart 1) and
mono- or dimethylated (symmetric or asymmetric) arginine. The
fact that PRMTs are known coactivators for nuclear receptors
makes them likely candidates to be overexpressed in prostate
and breast cancers. The inhibition of both PRMT1 and CARM1
(coactivator-associated arginine methyltransferase-1) can sup-
press estrogen and androgen receptor-mediated transcriptional
activation.22 In addition, the ability of PRMT5, when overex-
pressed, to promote anchorage-independent cell growth also
points to this enzyme as a candidate for deregulation in
transformed cellular states. Actually, two types of compounds
are used for the inhibition of PRMTs: inhibitors of S-
adenosylhomocysteine catabolism (like 2′,3′-acycloadenosine-
2′,3′-dialdehyde 3, Chart 1)23 and SAM analogues (like meth-
ylthioadenosine, 4, and sinefungin, 5, Chart 1),24,25 but, though
effective, they both have displayed limited specificity, indis-
criminately targeting all SAM-dependent enzymes. For this
reason the development of novel small molecule selective
inhibitors of PRMTs is highly desirable.
Similar to other posttranslational covalent modifications, such
as phosphorylation, acetylation, and ubiquitination, histone
methylation regulates a broad range of DNA and chromatin-
templated nuclear events, including transcription.1,2 Histones can
be methylated on lysine as well as arginine residues, preferen-
tially on the amino-terminal tails of histones H3 and H4, and
this methylation is a stable epigenetic mark that does not alter
the overall charge of the histone tails. Nevertheless, the recent
identification and structural characterization of “histone
demethylases”3-12 led us to consider that histone methylation
might also play a dynamic role in the epigenetic code.3-5
However, with increasing methylation13 comes an increase in
basicity, hydrophobicity, and an influence on the affinity for
anionic molecules such as DNA.14,15 As a matter of fact,
similarly to histone acetylation, histone methylation can modu-
late histone interaction with DNA and chromatin associated
proteins, resulting in an alteration of nucleosomal structures and
functions and ultimately contributing to different biological
processes.16 Histone methyltransferases display remarkable
specificity in the level of methylation they catalyze, and the
latest findings suggest that this could have functional signifi-
cance in transcription.17,18 The nine mammalian protein arginine
methyltransferases (PRMTs) thus far identified19,20 share a
Recently a few compounds selected through a random
screening were described as reversible arginine methyltrans-
ferase inhibitors (AMIs) by Bedford and co-workers.22 Being
interested in small molecule modulators of epigenetic targets
and, particularly, of histone-modifying enzymes,26-28 we focused
our attention on AMI structures and noticed that all of them
were dyes or dye-like derivatives. Particularly, two scaffolds
(A and B, Chart 2) emerged as privileged ones.
* To whom correspondence should be addressed. R.R., Tel.: +39-06-
4991-3937; Fax: +39-06-491491; e-mail: rino.ragno@uniroma1.it;
G.B., Tel.: +43-512-9003-70211; Fax: +43-512-9003-73100; e-mail:
gerald.brosch@i-med.ac.at; G.S., Tel.: +39-089-96-9770; Fax:+39-089-
96-9602; e-mail: gsbardella@unisa.it.
† Istituto Pasteur - Fondazione Cenci Bolognetti, Dipartimento di Studi
Farmaceutici, Universita` degli Studi di Roma “La Sapienza”.
‡ Universita` degli Studi di Salerno.
This finding, together with the aim to develop a rational
approach toward arginine methyltransferase inhibitors, prompted
us to explore the affinity for the target binding site of selected
AMIs as well as of a focused library (Chart 3) of dye-like small
molecule compounds (mostly containing privileged structures
scaffolds A and B).
§ Dipartimento di Scienze Biochimiche, Universita` degli Studi di Roma
“La Sapienza”.
| University of Medicine and Dentistry of New Jersey.
Biocenter-Innsbruck Medical University.
+ Current address: Dipartimento di Scienze Farmaceutiche, Polo Sci-
entifico di Sesto Fiorentino, Universita` degli Studi di Firenze, Via U. Schiff
6, I-50019 Sesto Fiorentino (FI).
10.1021/jm061213n CCC: $37.00 © 2007 American Chemical Society
Published on Web 02/27/2007

1242 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 6
Ragno et al.
Chart 1. Nonselective SAM-Dependent Methyltransferases
Inhibitors
(compounds 16-19) of the core structure of compounds 6a-f
yielded inactive or scarcely active derivatives (Table 1).
As far as the molecules of the B set (Chart 3) are concerned,
we observed a slightly different activity ranking against RmtA
(10c < 10b < 10a < 11 <9) and PRMT1 (10c < 10a < 11 <
10b < 9). Yet, in both cases the symmetrical ureidic derivative
9 was the most active. The replacement of one of the two
hydroxynaphthalene sulfonic substituents with the shorter and
nonpolar phenyl group resulted in a significantly reduced
inhibitory potency (compare IC50 values of 9 and 10b), this
decrease being even more marked if the oxygen of the urea
moiety was replaced with a sulfur atom (compound 10c).
Therefore, it is not surprising that suramin (derivative 15), a
polar polysulfonated symmetrical urea was significantly more
active than 9 in the inhibition of both tested enzymes (compare
IC50 values of compounds 9 and 15). On the contrary, it was
unexpected that derivative 12, with a diazo group replacing the
ureidic moiety and only one of the two naphthyl substituents
functionalized with polar groups, was comparable to 9 in
inhibiting both RmtA and PRMT1 (compare IC50 values of
compounds 9 and 12). Finally, derivatives 13, 14, and 20 showed
a very low (if any) activity.
Chart 2. Privileged Scaffolds
Homology Modeling. Computational approaches to the
prediction of protein structures are valuable tools for struc-
ture-based research.34 Indeed, in many cases homology models
have been shown to be valuable research tools for identifi-
cation, validation, and optimization of compounds,35 even
though the success in using such models is very dependent on
their quality, which is in turn correlated to the sequence
similarity between the target protein and the template structures
used for modeling.36
Our previous studies on Aspergillus nidulans genome29
revealed the existence of three distinct PRMTs exhibiting in
vitro histone methyltransferase (HMT) activity as recombinant
proteins. One of these proteins, termed RmtA, showed signifi-
cant sequence similarity (Figure 1) and similar biochemical
properties compared to human PRMT1.29 As we had access to
a large amount of the PRMT1 orthologue RmtA, similarly to
our approach to design new HDAC inhibitors,26,27,37-44 we
decided to verify if RmtA could serve as workhorse for our
enzymatic assays and if the RmtA three-dimensional model
could be used for computer-aided ligand design studies.
The PDB database45,46 contains the crystal structures of three
representative PRMTs from two different families: the rat
PRMT1 (rPRMT1)47 (residues 41-353, pdb entry code 1OR8,
R ) 2.35 Å) and its yeast homologue RMT1/Hmt148 (residues
30-348, pdb entry code 1G6Q, R ) 2.90 Å), and the rat
PRMT3 (rPRMT3) catalytic core49 (residues 208-528, pdb
entry code 1F3L, R ) 2.03 Å). All these structures present a
conserved core of about 310 residues formed by two domains:
the N-terminal catalytic domain, which contains the SAM-
binding site, and the C-terminal barrel-like domain. The all-
beta structure of the latter is interrupted by an arm-like motif
consisting of two alpha helices connected by a short loop, which
seems to be implicated in PRMT dimerization and protein
structure stabilization by hydrophobic interaction with the outer
surface of the catalytic domain of a neighboring molecule.
Among the available structures, all with the exception of the
yeast RMT1/Hmt1 have been determined in complex with the
demethylated cofactor S-adenosyl-homocysteine (AdoHcy). Rat
PRMT1 has also been complexed with peptides mimicking the
protein substrate.
In this paper we report structure-based (SB) and ligand-based
(LB) molecular modeling studies (homology modeling, molec-
ular docking, and 3-D QSAR studies) to investigate the binding
mode of several small molecule AMIs (6-20) in the catalytic
domain of the PRMT1 fungal homologue RmtA.29 Along with
computational studies, enzymatic assays on both RmtA and
PRMT1 using SAM and histones as substrates are also reported.
Chemistry. Derivatives 6e, 6f,30 9,31 10b32, 10c, and 1733
were synthesized as described in the Experimental Section.
All other derivatives were purchased from Lancaster Syn-
thesis, Milan (Italy), or Sigma-Aldrich, Milan (Italy).
Results and Discussion
Biological Results. Compounds 6-20 were tested for their
inhibition capabilities against human PRMT1 as well as
Aspergillus nidulans RmtA (Table 1).
Interestingly and as expected, as a general rule, derivatives
6-20 exhibited similar inhibitory potency against hPRMT1 and
RmtA. In fact, IC50 data for the two enzymes correlate quite
well, displaying a correlation coefficient of 0.91 (Supporting
Information, Figure A).
Derivatives 6c, 6f, and 9 confirmed their activity against both
tested enzymes. Among the molecules of the A set (Chart 3)
we found that derivatives 6a-f exhibited the same activity rank
(6d < 6a < 6b < 6f < 6e < 6c) against both hPRMT1 and
RmtA. In particular, we observed that bromine substituents in
the positions 2, 4, 5, and 7 of the xanthenic moiety greatly
enhanced inhibitory potency (compare IC50 of compounds 6b
and 6c). On the contrary, the introduction of an amino group in
the position 5 of the phenyl ring was detrimental for the activity
(compare derivative 6d with 6b) unless counteracted by the
effects of bromine substituents (compound 6e) or of the
N-dichloro-triazino moiety (derivative 6f). The formal opening
of the xanthenic moiety resulting in the triarylmethane scaffold
(compounds 7a-d and 8) significantly decreased inhibition
capability even when the carboxylic group was replaced with a
sulfonic group or with the introduction of bulky alkyl substi-
tuents in the positions 2, 4, 5, and 7 (compare IC50 value of
compound 6b with those of compounds 7a, 7c, and 7b,
respectively). Similarly, other simplifications or modifications
Optimal superposition of the three structures yielded a root-
mean-square deviation (rmsd) over the main chain atoms of less

Inhibitors of Histone Arginine Methyltransferases
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 6 1243
Chart 3. Structures of Derivatives 6-20
than 1 Å. The sequence identity ranges between 41% and 47%
in the conserved core and increases to more than 60% in the
N-terminal catalytic domain (Figure 1).50 Interestingly, almost
any residue within the target sequence is represented in at least
one structure, with few exceptions principally located in exposed
regions. In particular all the residues observed as forming
interactions with both the AdoHcy cofactor and Arg substrate
are very well conserved, with only minor differences in
hydrophobic side chain dimensions.
The crystal structures of rat PRMT1 (pdb code 1OR8, 1ORI,
1ORH) were obtained at a nonphysiological pH value (pH 4.7;
maximum enzymatic activity at pH 8.5) and therefore are not
suitable as target structures for virtual screening. Additionally,
an important helical section near to the binding pocket was not
resolved in the PRMT1 X-ray structures.47 For better results
we therefore decided to apply a multiple template approach51
and to make use of all the structures (from yeast RMT1, rat
PRMT1, and rat PRMT3) to build our RmtA homology model
by the SWISS-MODEL server.52 The obtained model was then
subjected to energy minimization in the presence of both SAM
and Arg substrates in order to optimize side-chain positions
(Figure 2).
Similarly as reported,47 SAM and Arg pockets lie close to
each other to allow the monomethylation or the dimethyla-
tion of the guanidine nitrogen atoms of arginine side chain
(Figure 2).
Docking Studies and Binding Mode Analysis of Com-
pound 6c, 6f, and 9. The binding modes of tested compounds
6a-20 into the modeled RmtA were analyzed by the means of
the program Autodock,53 and the X-Score54 external scoring
function was applied to select the binding conformations. To
assess the reliability of the Autodock method either the SAM

1244 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 6
Ragno et al.
Table 1. Inhibition Activities (IC50) of 6-20 against hPRMT1 and RmtA
IC50 (µM)
cmpd
Y
R1
R2
R3
R4
R5
R6
R7
RmtA
PRMT1
6a
6b
6c
6d
6e
6f
H
H
H
H
H
NH2
NH2
H
H
H
H
Br
H
OH
H
Br
H
Br
H
H
H
Br
H
Br
H
H
H
Br
H
Br
H
OH
H
Br
H
Br
H
160
123
0.18
710
22.3
126
75
1.4
519
4.8
18.9
COOH
COOH
COOH
COOH
COOH
H
H
H
Br
H
73
7a
7b
7c
7d
8
COOH
COOH
SO3H
H
i-Pr
H
H
H
H
H
Me
H
H
i-Pr
H
H
H
H
H
Me
H
H
-
-
296
>217
>263
>34
515
885
>217
1050
>34
280
SO3H
i-Pr
CH2N(CH2COOH)2
Me
i-Pr
CH2N(CH2COOH)2
Me
-
9
88
92
10a
10b
10c
11
12
13
14
15
16
17
18
19
20
H
590
928
529
990
540
96
1220
440
18.1
1070
476
1160
>164
>510
CONHPh
CSNHPh
1080
1360
220
103
490
2100
5.9
1100
406
1220
>164
>510
or the Arg substrates rPRMT1 experimental bound conformation
were effectively reproduced and successfully selected by the
X-Score scoring function (see Supporting Information, Figure
B). The DOCK55 program was also tried as alternative and faster
docking routine. Although DOCK performed rather well in
reproducing the SAM binding mode (not shown), it was only
partially able to correctly pose the Arg substrate (not shown).
In their report, Cheng et al.22 identified three compounds,
AMI-1, AMI-5, and AMI-6 (Cheng’s paper numbers), as
methyltransferase inhibitors, AMI-1 and AMI-6 being selective
for PRMT1 while AMI-5 showed to be active against both lysine
and arginine methyltransferases. Therefore, initial binding mode
analyses were focused on these three compounds, herein
reported as 6c (AMI-5), 6f (AMI-6), and 9 (AMI-1).
Interestingly, the Autodock conformations selected by X-
Score for these three compounds were found docked in differ-
ent binding sites (Figure 3). In particular, 6c binds exclusively
to the SAM binding pocket whereas 9 and especially 6f lie
between SAM and Arg binding sites. These results are consistent
with the outcomes of kinetics experiments performed by Cheng
et al.22
A deeper inspection revealed that 6c binds mainly to the
SAM-adenosine binding pocket formed by Val45, Gly46, Cys47,
Gly48, Val67, Asp68, Met69, Ser70, Gly94, Lys95, Met96,
Glu97, Glu112, Met123, and Thr126 residues and makes several
hydrophobic (Gly46, Met69, Ser70, Met123, and Thr126) and
electrostatic (Gly46, Val67, Asp68, Ser70, and Thr126) interac-
tions and a hydrogen bond between the 6c carboxylate group
and the amidic NH of Met69 (2.6 Å, Figure 3c).
As regards to 9, about half of its structure is buried exclusively
in the Arg site (Glu15, Met16, Glu112, Met114, Tyr116,
Glu121, His261, and Trp262, Figures 3a and 3b) with a strong
hydrogen bond interaction occurring between one of the sulfonic
groups and the amidic NH of His261 (Figure 3b, SO39‚‚‚
NHHis261 ) 2.95 Å) and a second one between the ureidic NH
and the arginine anchoring Glu112 side-chain (HN9‚‚‚OOCGlu112
) 2.80 Å, Figure 3b). The other half of the structure of 9
stretches out in the SAM binding site, slightly overlapping with
the cofactor methyl donor group and making at least three
hydrogen bonds involving the second sulfonic group and Arg22
guanidine side chain (3.08 Å), Gly48 amidic NH (2.75 Å), and
Thr49 hydroxyl group (3.22 Å), respectively. Several hydro-
phobic (Glu15 and Met16), electrostatic (Glu15, Met16, Gly48,
Thr49, Glu112, Met114, and His261), and π-π (Try116,
His261, and Trp262) interactions are also detectable between 9
and RmtA.
Similarly and to a higher extent than that of 9, the 6f binding
conformation partially occupies both SAM and Arg sites (Figure
3d). In particular, part of the fused tricyclic moiety is super-
imposed with the SAM methionine residue, making at least two
hydrogen bonds (phenolic-OH6f‚‚‚OOCAsp44 ) 3.20 Å; endocy-
clic-O‚‚‚N-guanidineArg22 ) 2.74 Å), while the dichlorotriazine
portion fills the guanidine binding site of the Arg substrate,
making positive contacts with Trp113, Met114, Gly115, Glu121,
Ser122, Met123, and Trp262. A further hydrogen bond interac-
tion is also made between carboxylic group of 6f and the Gly48
amidic NH (2.89 Å). Several hydrophobic (His14, Met16, Thr49,
Glu112, Trp113, Try116, Ser122, and Met123) and electrostatic
(Asp44, Gly46, Gly48, Thr49, Glu112, Met114, and Glu121)
interactions are also observed between 6f and RmtA.
Docking Studies and Binding Mode Analysis of Other
Compounds. On the basis of biological results, we analyzed
the binding mode of compounds showing a measurable
inhibitory activity (20 molecules).56 We divided the selected
docked conformations into three groups: (a) molecules docked
in the Arg pocket (DAP: 6a, 12, 13, 16b,57 and 17, Figure 4a);
(b) molecules docked in the SAM pocket (DSP: 6b, 6c, 6d,
10a, and 18, Figure 4b), and (c) molecules partially over-
lapping with both sites (docked in both pockets, DBP: 6e, 6f,
7a, 8, 9, 10b, 10c, 14, and 16a, Figure 4c). Derivative 11
was found to dock only outside either the SAM or Arg binding
sites.

Inhibitors of Histone Arginine Methyltransferases
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 6 1245
Figure 1. Sequence alignment of the target A. nidulans (A.N.) RmtA and human PRMT1to the structural alignment of yeast RMT1, rat PRMT1,
and rat PRMT3. Cylinders and arrows indicate the common secondary structural elements (helices and strands, respectively) in the template structures.
Red boxes point out the catalytic domain regions. Residue numbering is indicated on the left of each sequence row. Small letters indicate residues
which are not structurally aligned. Identical residues between target and templates are bold. Residues highlighted in yellow are identical in all the
sequences. Red asterisks indicate residues having a role in cofactor binding and/or catalytic mechanism; blue asterisks indicate residues important
for the dimerization process.
tricycle portion is lying between Met16 and His261 side chains
(Supporting Information, Figure C).
On the other hand, derivative 12, the most active of the DAP
group, makes important hydrogen bonding interactions between
its sulfonic group and the guanidine moiety of Arg295. In
addition the unsubstituted naphthyl ring stacks between the
Trp262 indole and the Tyr116 phenol rings making π-π
interactions. (Supporting Information, Figure D). As regards to
the less active derivatives 13, 16b, and 17, although a few
relevant interactions (i.e., hydrogen bond with Glu15) are
present, as in the case of 6a and 12, major hydrophobic and
electrostatic interactions are lacking (Supporting Information,
Figure E). Additional docking studies performed in the presence
of the SAM cofactor revealed similar results with differences
only for compounds 12 and 17. (See Supporting Information,
Figure F): derivative 12 displays a different binding pose, while
rPRMT1 X-rays structure (green). The SAM binding site, the Arg
Figure 2. Superimposition of RmtA homology model (blue) and
17 is shifted toward SAM and the negatively charged carboxy-
late seems to make electrostatic interaction with the positively
charged SAM methyl group. No relevant interactions are made
between 6a, 13, 12, and 16b with SAM.
binding site, and the backbone of histone H4 are highlighted in yellow,
cyan, and purple, respectively.
As illustrated in Figure 4a, the molecules of the DAP group
occupy the Arg pocket plus a portion of the area surrounding
the entry site of the substrate side chain (not shown). The
formation of a hydrogen bond with the Glu15 side-chain
carboxyl is the most common interaction in this group, notice-
able for three compounds (6a, 16b, and 17) out of five. Among
them, 6a, the second most active molecule of the set, also makes
positive interactions with the hydroxyl of Tyr116 and amidic
nitrogen of His261. Moreover, the phenyl group seems to make
a π-π interaction with the indole ring of Trp262, while the
As regards to compounds of DSP group, it is noteworthy that
the two similar derivatives 6b and 6d display a different binding
scenario from the one above-described for 6c, as they are shifted
toward the Arg site and while losing the interactions with amidic
nitrogen Asp68, they gain some interactions with key residues
at the interface between the SAM and Arg sites (Glu112 or
Asp68 carboxylate for 6b or 6d, respectively; see Supporting
Information, Figure G). Compounds 10a and 18 occupy the
same space but, though making a few important interactions
such as an hydrogen bond with the Thr49 hydroxyl, most

1246 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 6
Ragno et al.
Figure 3. Autodock/X-Score selected binding conformations of compounds 6c, 6f, and 9 into RmtA catalytic site. (a) Superimposed conformations
of 6c (purple), 6f (green), and 9 (orange); (b) binding mode of 9 (in stick, carbon atoms in gray); (c) binding mode of 6c (in stick); (d) binding
mode of 6f (in stick). The volumes occupied by Arg and SAM are represented in yellow and cyan, respectively. The RmtA residues within 4.0 Å
from the docked inhibitor are shown in wire. For the sake of clarity, hydrogen atoms are not displayed.
Figure 4. Binding conformations of tested molecules, divided according to the binding modes of 6c, 6f, and 9 (see text). In yellow and cyan are
represented the Arg and SAM sites, respectively. (a) 6a (aquamarine), 12 (yellow), 13 (orange), 16b (red) and 17 (cyan); (b) 6b (orchid), 6c (red),
6d (light gray), 10a (yellow), and 18 (sky blue); (c) 6e (magenta), 6f (light gray), 7a (green), 8 (cornflower blue), 9 (sky blue), 10b (purple), 10c
(orange), and 16a (red).
hydrophobic interactions are lacking due to their smaller size
(Supporting Information, Figure H).
porting Information, Figure I). Among these residues, Arg22
(6f, 10b and 14), Gly46 (6f, 7a, 10, and 14), Glu112 (6f, 7a, 8,
9, 10b, 10c, 14, and 16a), Met114 (6f, 7a, 10b, and 10c), and
His261 (9, 10c, and 16a) are the most involved in ligand
binding. The most active compound 6e shows peculiar hydrogen
bonding interactions: a strong one with the hydroxyl group of
Tyr116 (3.07 Å) and a weak one with His13 (3.73 Å) (not
shown). However, the strong point of 6e seems to be the number
of positive contacts that the six bromine atoms make in mostly
The docking area of the third group of molecules (DBP),
partially overlapping both Arg and SAM sites, indicates a
different interaction profile. Taken together, the molecules of
this group make contacts within 3 Å that define a binding site
comprising at least 18 residues (Ile12, His13, Glu15, Leu17,
Arg22, Gly46, Cys47, Gly48, Thr49, Leu52, Asp68, Glu112,
Met114, Gly115, Tyr116, Glu121, His261, and Trp262, Sup-

Inhibitors of Histone Arginine Methyltransferases
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 6 1247
Table 2. Statistical Parameters of the Three 3-D QSAR Models
model
group
probe
no.a
variablesb
PCsc
r2
q2
SDEPLOO
LOO
1
2
3
DAP
DSP
DBP
OH2-C3
O
OH2-C3
5
5
9
1240
2641
1063
2
2
2
0.99
0.99
0.99
0.88
0.91
0.93
0.12
0.40
0.17
a Number of molecules in the group. b Number of GRID variable after the FFD selections. c Number of principal components.
Figure 5. DAP group grid plots of the PLS coefficients for the OH2
(a) and C3 (b) probes. Compounds 6a (yellow), 12 (magenta), 13 (red),
16b (blue), and 17 (green) are also displayed. The cyan contours
represent negative coefficients under -0.0003 energy value while the
yellow contours represent the positive coefficients over 0.0003.
Figure 7. Recalculated and predicted pIC50 from model 3. LOO: cross-
validation with the leave one out method; L5O: cross-validation with
the leave some out method using five groups; L2O: crossvalidation
with the leave some out method using two groups.
complex was geometry optimized we applied the Autodock/X-
SCORE protocol to inspect the binding site of derivatives 6-20
(data not shown). As expected the binding modes of the
compounds in the h-PRMT1 were similar to those described
for RmtA with only marginal differences in the structure
conformations, consistent with the aforementioned good cor-
relation between pIC50 values.
3-D QSAR Studies. As learned from the foregoing docking
studies, the three different binding modes adopted by molecules
6-20 could represent structure-based aligned training sets to
build 3-D QSAR models useful as validation tools for the
docking studies themselves as well as tools to design new PRMT
inhibitors. With this aim, three 3-D QSAR models were derived
using the GRID/GOLPE procedure.58 Obviously, it has to be
stressed that due to the limited number of molecules in each
group the 3-D QSAR models are of limited validity and were
only built to gain qualitative support to the above-described
binding modes and to get some insights to design new PRMT
inhibitors.
In building the 3-D QSAR models, different GRID probes
alone (C3, N1, O, OH, OH2, and DRY) and combinations of
them (C3+N1, DRY+N1, DRY+O, O+C3, O+N1, OH2+C3,
OH2+DRY, OH+C3, OH+DRY, and OH+N1) were tried, and
only the preliminary GOLPE/GRID models showing the higher
q2 values were refined by sequential fractional factorial design
(FFD) selections. The combination of the water (OH2) and the
methyl (C3) probes gave the best preliminary models for the
DAP and DBP groups while only the carbonyl oxygen (O) probe
supplied the best model for the DSP group alignment. After
having imported the molecular interaction fields (MIF) into
GOLPE, a series of fractional factorial design (FFD) selections
allowed the definition of the final 3-D QSAR models whose
statistical parameters are listed in Table 2. As can be seen, in
Figure 6. (a) DSP group grid plot of the PLS coefficients. Compounds
6b (magenta), 6c (red), 6d (green), 10a (blue) and 18 (purple) are also
displayed. The cyan contours represent negative coefficients under
-0.0005 energy value while the yellow contours represent the positive
coefficients over 0.0015. (b) Activity contribution plot relative to
compound 6c.
filling the Arg binding area and part of SAM binding area
(Supporting Information, Figure J). The less active derivatives
7a, 8, 10b, 10c, 14, and 16a though showing docking positions
similar to that of 6e are smaller in volume (data not shown)
and make less positive contacts.
In addition to the above-mentioned docking studies, we also
generated the human PRMT1 (hPRMT1) using the same
protocol described for RmtA homology modeling. Subsequently,
we performed a 3-D comparison of the RmtA and hPRMT1
models and we found that only five (13.5%) of the 37 residues
that are in contact with the SAM and arginine substrates had
different side-chains, and more interestingly, four out of five
showed some similarities (Cys47RmtA f Ser39hPRMT1; Val67RmtA
f Ile59hPRMT1; Asp68RmtA f Val60hPRMT1; Met69RmtA
f
Cys61hPRMT1; Met96RmtA f Val88hPRMT1; see Supporting In-
formation, Figure K). Once the hPRMT1/SAM/H4 ternary

1248 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 6
Ragno et al.
Figure 8. DBP group grid plots of the PLS coefficients for the OH2 (a) and C3 (b) probes. The cyan contours represent negative coefficients under
-0.0006 energy value while the yellow contours represent the positive coefficients over 0.0006. Aminoacids Ile12 (green), His13 (purple), Thr49
(magenta), and Tyr116 (red) are also displayed. Hydrogen atom are not displayed.
Figure 9. Derivative 6e activity contribution plots for the OH2 (a) and C3 (b) probes. The 6e docked conformation is displayed as atom type
colors. The cyan contours represent negative activity contribution areas while the yellow contours represent the positive activity contribution areas.
Amino acids Ile12 (green), His13 (purple), Thr49 (magenta), and Tyr116 (red) are also displayed. Hydrogen atoms are not displayed.
term of cross-validation data (q2 and SDEP values), all
three models, although developed with a limited number of
compounds, were statistically robust with q2 and SDEP values
ranging from 0.88 to 0.93 and from 0.12 to 0.40, respectively.
Although in models 1 and 2 the molecules seem not to be
optimally aligned, the 3-D QSAR models seem to support
the above-described binding modes obtained by docking
studies.
As already outlined in the binding mode analysis section, only
derivative 12, the most active out of the DAP group, is fully
buried in the arginine substrate site, which is correctly recog-
nized by the PLS analysis of the OH2 and C3 GRID fields. In
particular in the OH2 PLS coefficient plot (Figure 5a) a large
positive yellow polyhedron runs over the 12 diaza group which
is close to the guanidine group of the arginine substrate.
Although in the above binding mode analysis the 12 diaza group
did not seem to make any significant interaction, the PLS
analysis of both the actual field and activity contribution plots
(Supporting Information, Figure L) indicates that a π-rich group
should be retained in designing new compounds competitive
for the arginine site. This consideration is also supported by
the observation that less active molecules such as 13 and 16b
possessed hydrogen bonding acceptor groups in the same 12
diaza moiety area. To prevent the model from being guided by
the misalignment of 12, we also built a 3-D QSAR model with
only four compounds, and, taking into account the statistical
limits of such a model, the final 3-D QSAR model was charac-
terized with r2 and q2 values of 0.99 and 0.84, respectively.
Regarding the 3D-QSAR model 2, it was built using only
five molecules which were docked exclusively in the SAM
binding site; derivative 6c, the most active among all tested
molecules, belongs to this group. As in the case of compound
12 in model 1, 6c seems misaligned (Figures 4b and 6), yet a
3-D QSAR model endowed with high r2 (0.99) and q2 (0.68)
values without compound 6c was also obtained, confirming that
model 2 reflects a structure-activity correlation belonging to
all five molecules in the set.
In Figure 6a is reported the PLS coefficients plot of the 3-D
QSAR model for the DSP group in which one large yellow
polyhedron (positive values) and a few small cyan polyhedrons
(negative values) are present. To properly understand the role
of the yellow polyhedron, the activity contribution plot for
compound 6c is also reported (Figure 6b). As shown, the most
important contribution to the activity seems due to the benzoate
group while no polyhedron can be associated to the bromine
atoms, the importance of which still emerged from biological
assays. This disagreement could be attributed to the fact that
the SAM binding site is not fully represented by the aligned
molecules, and thus the 3-D QSAR model here only supports
the hypothesis that there is some correlation between the
inhibitory activity with both the bound conformations and their
spatial disposition (alignment).

Inhibitors of Histone Arginine Methyltransferases
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 6 1249
The 3-D QSAR model 3 was developed on molecules
partially occupying both the SAM and Arg binding sites (DBP,
Figure 4c). Differently from the above analyzed models, all nine
molecules of this group were found superimposed to each other,
leading to a robust model (Figure 4c and Supporting Informa-
tion, Figure M), as also confirmed by high q2 values obtained
by cross-validations using the group method (leave some out,
LSO): with five (L5O, q2 ) 0.90) and two (L2O, q2 ) 0.61)
groups the DBP molecules were highly self-predictive
(Figure 7).
inhibition of small molecule inhibitors of histone arginine
methyltransferases. In particular the structure-based 3-D QSAR
high statistical coefficients seem to support the docking experi-
ments. If considered together, the docking and the 3-D QSAR
studies allowed a deep binding mode analysis of tested
molecules. In addition, these analyses hinted that two regions,
the pocket formed by Ile12, His13, Met16, and Thr49 and the
SAM cysteinic portion binding site, should be taken into account
to design novel active inhibitors.
Further molecular modeling, synthetic and biological efforts
are ongoing to develop more robust 3-D QSAR models.
Analysis of the PLS coefficient plots (Figure 8) revealed that
some areas around the structure-based aligned molecules should
be further explored by the inclusion of additional ligands. In
particular, in both the OH2 and the C3 PLS coefficients plot
(Figure 8a and 8b, respectively) at 0.0006 energy level a large
yellow polyhedron is noticeable close to the tricyclic moiety of
6e (magenta circle). This polyhedron is not correlated with any
residue of RmtA but lies in an open space, maybe the entrance
of SAM. A second yellow polyhedron (blue circle) could be
associated to a space close to the dibromoaminobenzoate moiety
of 6e corresponding to a pocket formed by Ile12, His13, Met16,
and Thr49 residues. It is noteworthy that this area is also filled
by 6f and 9, that together with 6e are the most potent compounds
of the DBP group. Two other cyan-colored polyhedrons
(negative PLS coefficients) are visible in Figure 8a (OH2 probe)
and almost absent in Figure 8b (C3 probe): the polyhedrons
highlighted by green circles, mainly associated either with
electrostatic (7a, 8, 10b, and 16a) or hydrogen bonding (6e, 6f,
and 9) interactions with Arg22 (not shown) and the polyhedrons
circled in red, mainly associated either with electrostatic (6f,
10c, and 14, not shown) or hydrogen bonding (7a, 8, 9 10b,
and 10c, not shown) interactions with either Glu112 side-chain
carboxyl (6f, 9, 10b, 10c, and 14, not shown) or Met114
carbonilic oxygen (6f, 7a, 8, 9, 10b, and 10c, not shown). The
negative PLS coefficients generated by a methyl probe (small
cyan polyhedrons in Figure 8b) indicate areas that have to be
filled to maintain or increase the activity. One of these areas
coincides with the binding site of the cysteinic portion of SAM,
a site delimited by Arg22, Asp44, Gly46, Cys47, Ile51, Leu52,
and Glu112 (not shown).
Materials and Methods
Melting points were determined on a Gallenkamp melting point
apparatus and are uncorrected. H NMR spectra were recorded at
1
300 MHz on a Bruker Avance 300 spectrometer; chemical shifts
are reported in δ (ppm) units relative to the internal reference
tetramethylsilane (Me4Si). Electronic impact mass spectrometry (EI-
MS) was performed on a Finnigan LCQ DECA TermoQuest (San
Jose`, CA) mass spectrometer. All compounds were routinely
checked by TLC and 1H NMR. TLC was performed on aluminum-
backed silica gel plates (Merck DC-Alufolien Kieselgel 60 F254
)
with spots visualized by UV light or using a KMnO4 alkaline
solution. All solvents were reagent grade and, when necessary, were
purified and dried by standard methods. Concentration of solutions
after reactions and extractions involved the use of a rotary
evaporator operating at a reduced pressure of ca. 20 Torr. Organic
solutions were dried over anhydrous sodium sulfate. Analytical
results are within (0.40% of the theoretical values. Reagents were
purchased from Lancaster Synthesis, Milan (Italy), or Sigma-
Aldrich, Milan (Italy). As a rule, samples prepared for physical
and biological studies were dried in high vacuum over P2O5 for 20
h at temperatures ranging from 25 to 110 °C, depending on the
sample melting point.
Preparation of 4-Amino-3,5-dibromo-2-(2,4,5,7-tetrabromo-
6-hydroxy-3-oxo-3H-xanthen-9-yl )benzoic Acid (6e). A stirred
mixture of 6-aminofluorescein (0.576 mmol, 0.2 g) in glacial acetic
acid (7 mL) was cooled at 0 °C. A solution of bromine (4.606 mmol,
0.74 g) in acetic acid (3 mL) was then added dropwise, and stirring
was continued at room temperature for additional 15 min. After
completion (TLC monitoring: CHCl3:AcOEt 1:1, on silica gel
plate), the reaction mixture was diluted with water (10 mL) and
filtered by vacuum filtration. The precipitate was washed with water
(3 × 5 mL) and air-dried in a vacuum desiccator to give TLC pure
6e in 91% yield (0.82 g; mp: 291-292 °C), MS (EI, 70 eV) m/z
The aforementioned important role of the interactions of
compounds of the DBP group with the pocket composed of
Ile12, His13, Met16, and Thr49 residues is more clearly
explained in Figure 9, reporting the activity contribution plots
relative to derivative 6e, the most active in this group. In this
figure, yellow polyhedrons, associated with positive activity
contribution areas, are closely related to this pocket. Again the
small cyan polyhedrons, associated with negative activity
contribution areas, highlight regions lacking in positive sterical
interactions (green and red circles in Figure 9b).
1
821; H NMR (DMSO-d6): δ 6.53 (s, 1H), 7.10 (s, 1H), 8.07 (s,
1H), 10.79 (s, 1H).
Preparation of 4-(4,6-Dichloro-1,3,5-triazin-2-ylamino)-2-(6-
hydroxy-3-oxo-3H-xanthen-9-yl)benzoic Acid (6f).30 A solution
of 6-aminofluorescein (0.288 mmol, 0.1 g) in acetone (2 mL) was
added dropwise to a solution of cyanuric chloride (0.2851 mmol,
0.053 g) in acetone (3 mL) magnetically stirred at 5 °C. The
resulting mixture was kept stirring for 3 h at 3-5 °C (TLC
monitoring: AcOEt on silica gel plate) and then filtered to give 6f
as a TLC pure yellow precipitate.(0.13 g, yield, 92%; mp: > 350
°C); MS (EI, 70 eV) m/z 494; 1H NMR (DMSO-d6): δ 6.79-6.88
(m, 4H), 6.95 (s, 2H), 7.73 (s, 1H), 8.18-8.28 (m, 2H), 11.72
(s, 1H).
Preparation of 7,7′-Carbonylbis(azanediyl)bis(4-hydroxynaph-
thalene-2-sulfonic acid) (9).31 J-acid (2.00 g, 3.964 mmol) was
added to 50 mL of water, with dropwise addition of aqueous NaOH
10% to facilitate dissolution; the pH of the final solution was about
7. The solution was then stirred and heated to 60 °C and triphosgene
powder (0.40 g, 1.348 mmol) rapidly added. The resulting mixture
was magnetically stirred for 6 h while keeping temperature (60
°C) and pH (7-8) constant. When the reaction was complete, J-acid
urea 9 was obtained by salting-out. (1.80 g, Yield 90%). MS (EI,
70 ev) m/z: 504; 1H NMR (DMSO-d6): δ 7.04 (m, 1H), 7.51 (m,
1H), 7.55 (m, 1H), 8.00 (m, 1H), 8.02 (m, 1H), 9.27 (s, 1H), 9.61
(s, 1H).
Conclusions
In this report we describe the binding mode analysis of a
focused library of small molecule compounds into the catalytic
domain of the PRMT1 fungal homologue RmtA as well as
enzymatic assays performed on recombinant RmtA and PRMT1
proteins using SAM and histones as substrates. Computational
and biological results were in good agreement. Moreover, the
good correlation between pIC50 values against the two enzymes
(R ) 0.91) supported our hypothesis of using RmtA as a model
to screen for arginine methyltransferase inhibitors.
Moreover, the agreement between molecular modeling studies
(molecular docking and 3-D QSAR) and biological results
allowed us to shed more light on the molecular mechanism of

1250 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 6
Ragno et al.
Preparation of 4-Hydroxy-7-(3-phenylureido)naphthalene-2-
sulfonic Acid (10b).32 A stirred solution of J acid (1.943 mmol,
0.5 g) in 8 mL of alkaline water solution containing NaOH (1.943
mmol, 0.077 g), at room temperature, was treated with a solution
of phenyl isocianate (1.943 mmol, 0.23 g) in tetrahydrofuran (5
mL). The resulting mixture was magnetically stirred at room
temperature for 24 h (TLC monitoring: CHCl3 on silica gel plate),
acidified with a water solution of HCl 2 N (3 mL), evaporated under
reduced pressure, and purified by column chromatography (silica
gel/ethyl acetate:isopropanol: water 12:3:1.). (0.49 g, Yield, 70%;
mM isopropyl â-D-thiogalactopyranoside (IPTG), and the cultures
were grown for 3 h upon induction. Bacterial cells were harvested
by centrifugation (3000g for 20 min), resuspended in sonication
buffer (PBS, 0.01% TritonX100, 0.2 mM PMSF, 1 µg/mL aprotinin,
1µg/mL leupeptin, 1µg/mL pepstatin), and sonicated three times
for 30 s with a sonic dismembrator 550 (Fisher Scientific; amplitude
40%). The sonicated suspension was centrifuged for 20 min at
10000g, and the supernatant was incubated with preequilibrated
glutathione-sepharose-4-fast-flow overnight at 4 °C on a rotating
wheel. The resin was transferred to a gravity-flow column and
washed with five column volumes (CV) of PBS + 0.1% Tri-
tonX100. Bound GST-PRMT1 was released using five times one
CV of elution buffer (50 mM Tris pH 8.0, 10 mM glutathione).
Purified GST-PRMT1 was dialyzed against BC100 buffer (20 mM
Tris pH 7.9, 10% glycerol, 100 mM KCl, 0.2 mM EDTA, 10 mM
2-ME, 0.2 mM PMSF), and protein concentration was estimated
by SDS-PAGE and Bradford assay.
The coding sequence of RmtA was cloned into a pGEX-5X-1
expression vector (Amersham Pharmacia Biotech). RmtA-Protein
was expressed in BL21 cells in LB-medium. Cultures (250 mL)
with an A600 of 0.4 were induced with a final concentration of 1
mM IPTG and grown for 4 h at 37 °C. After centrifugation of cells
at 4000g, the pellet was resuspended in 6 mL of GST-binding buffer
(140 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2-
PO4, pH 7.3) containing one protease inhibitor tablet (Complete,
Roche, Mannheim, Germany) per 50 mL of buffer. For cell lysis,
lysozyme was added at a final concentration of 5 mg/mL binding
buffer, and cells were passed through a french press with a pressure
setting of 1000 psi. The resulting lysate was centrifuged at 20000g
for 10 min at 4 °C. GST fusion protein was purified from soluble
extracts by binding to a GST-HiTrap column (Amersham Pharmacia
Biotech). Proteins were eluted with 50 mM Tris-HCl, 10 mM
reduced glutathione, pH 8.0, and assayed for histone methyltrans-
ferase activity.
Histone Methyltransferase Assay (HMT). For inhibition assays,
affinity purified GST-RmtA and GST-PRMT1 fusion proteins were
used as the enzyme source. HMT activities were assayed as
described29 using chicken erythrocyte core histones as substrate.
A 500 ng amount of GST-RmtA and GST-PRMT1 fusion proteins
was incubated with different concentrations of compounds for 15
min at room temperature, and 20 µg of chicken core histones and
0.55 µCi of [3H]-S-adenosyl-L-methionine (SAM) were added. This
mixture was incubated for 30 min at 30 °C. Reaction was stopped
by TCA precipitation (25% final concentration), and samples were
kept on ice for 20 min. Whole sample volumes were collected onto
a glass fiber filter (Whatman GF/F) preincubated with 25% TCA.
Filters were washed three times with 3 mL of 25% TCA and then
three times with 1 mL of ethanol. After the filters were dried for
10 min at 70 °C, radioactivity was measured by liquid scintillation
spectrophotometry (3 mL of scintillation cocktail).
1
mp: > 350 °C); MS (EI, 70 eV) m/z 358; H NMR (DMSO-d6):
δ 6.92-6.98 (m, 2H), 7.24-7.29 (t, 2H), 7.42-7.48 (m, 4H), 7.85
(s, 1H), 7.95-7.98 (d, 1H), 8.72 (s, 1H), 8.93 (s, 1H), 10.02 (s,
1H).
Preparation of 4-Hydroxy-7-(3-phenylthioureido)naphtha-
lene-2-sulfonic Acid (10c). A stirred solution of J-acid (1.943
mmol, 0.5 g) in 8 mL of alkaline water solution containing NaOH
(1.943 mmol, 0.077 g), at room temperature, was treated with a
solution of phenyl isothiocyanate (1.943 mmol, 0.278 g) in
tetrahydrofuran (5 mL). The resulting mixture was magnetically
stirred at room temperature for 12 h (TLC monitoring: CHCl3 on
silica gel plate), acidified with a water solution of HCl 2 N (3 mL),
evaporated under reduced pressure, and purified by crystallization
(CH3CN) to yield TLC pure 10c (0.55 g, Yield: 75%; mp: >350
°C); MS (EI, 70 eV) m/z 374; 1H NMR (DMSO-d6): δ 7.05-7.10
(m, 2H), 7.31-7.34 (t, 2H), 7.47-7.57 (m, 4H), 7.83-7.84 (s, 1H),
7.97-8.00 (d, 1H), 9.82 (s, 1H), 9.95 (s, 1H), 10.08 (s, 1H).
Preparation of 4-(4,6-Dichloro-1,3,5-triazin-2-ylamino)ben-
zoic Acid (17).33 A solution of 4-aminobenzoic acid (3.656 mmol,
0.5 g) in acetone (10 mL) was added to a solution of cyanuric
chloride (3.61 mmol, 0.67 g) in acetone (12 mL) magnetically
stirred at 0 °C. Aqueous sodium hydrogenocarbonate 0.3 M (12
mL) was then added to the mixture over 30 min while maintaining
a temperature of 0-5 °C. The solution was stirred for 2 h (TLC
monitoring: CHCl3:ethyl acetate 1:1, on silica gel plate), and then
aqueous HCl 12% was added until a precipitate formed which was
filtered and washed with acetone. (0.99 g, Yield: 95%; mp: >350
°C); MS (EI, 70 eV) m/z 284; 1H NMR (DMSO-d6): δ 7.67-7.70
(d, 2H), 7.78-7.95 (d, 2H), 11.37 (s, 1H).
Cloning of Human PRMT1 and Aspergillus nidulans RmtA.
Human PRMT1 (isoform 2) was amplified from a HeLa cDNA
library (Invitrogen) using the following primer set: PRMT1fwd
5′-GGATCCGAGAATTTTGTAGCCACCTTGGCTAA-3′ and
PRMT1rev 5′-CTCGAGTCAGCGCATCCGGTAGTCGGTGGA-
3′. The PCR fragment was subcloned into a pGEM-Teasy vector
(Promega), released by digestion with the restriction endonucleases
BamHI and XhoI (New England Biolabs), and inserted into the
BamHI/XhoI sites of the bacterial expression plasmid pGEX6P1
(Amersham) using T4 Ligase (Roche).
Total Aspergillus RNA was used to prepare cDNA using
Superscript reverse transcriptase (Life Technologies, Inc.) according
to the manufacturer’s instructions employing an oligo-dT17
standard primer. For amplification of fragments of rmtA from the
cDNA of A. nidulans, degenerate oligodeoxynucleotide primers
were used. Forward primer was 5′-GGNATHCAYGARGARATG-
3′ based on the amino acid sequence GIHEEM, and reverse primer
was 5′-ACNCAYTGGAARCARAC-3′ based on THWKQT. A
product of 750 bp was isolated and cloned into pGEM-T vector
(Promega). A full-length genomic copy of rmtA was obtained using
cDNA PCR products to screen a genomic library of Aspergillus
nidulans DNA. For amplification of the 3′ end of rmtA, the 3′ rapid
amplification of cDNA ends (RACE) protocol59 was performed.
The first PCR was performed with forward primer 5′-ATACCGTC-
GAGCTCAAG-3′ and an adapter primer of sequence 5′-GACTC-
GAGTCGACATCGA-3′. For “nested PCR”, primer 5′-GACTA-
GAGTACACGATGGA-3′ (forward) and dT4 adapter primer 5′-
GACTCGAGTCGACATCGATTTT-3′ (reverse) were used. Ampli-
fied products were cloned into pGEM-T vector (Promega).
Preparation of GST-RmtA and GST-PRMT1 Fusion Pro-
teins. Expression of the GST-PRMT1 fusion protein was performed
in BL21 cells. At OD600 nm ) 0.6 expression was induced with 1
Preparation of the RmtA-SAM-H4 Ternary Complex. The
RmtA model from the homology model experiment was merged
with the S-adenosyl-L-homocysteine (AdoHcy) and the 19mer
substrate peptide taken from the experimental PRMT1 structure
(pdb entry code 1or8).47 The AdoHcy was modified to SAM using
an experimental SAM structure extracted from the FTSJ RNA
methyltransferase (pdb entry code 1ej0)60 which showed a SAM
conformation similar to the AdoHcy found in the 1or8 complex.
The experimental 19mer peptide lacked of most of the side-chain
atoms, which were added using the xLeap module of the AMBER
suite.61 The RmtA-SAM-H4 complex was solvated (SOLVA-
TEOCT command) in a box extending 10 Å with 16658 water
molecules (TIP3 model) and neutralized with 10 Na+ ions. The
solvated complex was then refined by minimization using the
SANDER module of AMBER. The parameters for the SAM were
calculated using the antechamber module of AMBER, and the
atomic charges were calculated using the AM1 Hamiltonian.
Molecular Docking. All the tested molecules were built, starting
from ASCII text, using the stand alone version of PRODRG,62 in
conjunction with the GROMACS suite.63 Docking studies were
performed by means of the Autodock 3.0.5 program using a grid

Inhibitors of Histone Arginine Methyltransferases
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 6 1251
spacing of 0.375 Å and 39 × 50 × 56 number of points that
embraced both the SAM and Arg binding sites. The grid was
centered on the mass center of the experimental bound AdoMet
and Arg substrates. The GA-LS method was adopted using the
default setting except for the maximum number of energy evalu-
ations increased from 250000 to 2500000. Autodock generated 100
possible binding conformations for each molecule that were
clustered using a tolerance of 2.0 Å. The AutoDockTool (ADT)
graphical interface64 was used to prepare the enzyme PDBQS file.
The protein atom charges as calculated during the complex
minimization were retained for the docking calculations.
The SAM and N-acetyl,O-methyl-capped Arg PRODRG-gener-
ated conformations were docked into the RmtA to assess the
docking protocol. The SAM and Arg were docked back into their
binding sites, and the Xscore54-selected conformations showed root-
mean-square deviations (rmsd) of 0.60 and 1.47, respectively. The
Autodock scoring function did not select conformations with lower
rmsd values. Trials with the DOCK program did not successfully
dock back the substrates into the RmtA; therefore, we continued
to use Autodock. To analyze the docking results, the ADT was
used and the Chimera 1.2176 program65 was used to produce the
images.
the FFD selection the cross-validation was conducted using five
random groups for 20 times and a maximum of two principal
components.
Cross-Validation. The models were validated using the leave
one out (LOO) method. Molecules were assigned in a random way
to five groups of equal size. Reduced models were built keeping
out one group at a time.
Acknowledgment. Many thanks are due to Prof. Gabriele
Cruciani and Prof. Sergio Clementi (Molecular Discovery and
MIA srl) for the use the GOLPE program in their chemometric
laboratory (University of Perugia, Italy) and for having provided
the GRID program. This work was partially supported by PRIN
2004 (A.M.), AIRC 2005 (A.M.), and Regione Campania 2003,
LR 5/02 (G.S.) grants.
Supporting Information Available: Characterization data for
compounds 6e, 6f, 9, 10b, 10c, and 17, sequence homology, and
alignment and additional binding mode informations. This material
is available free of charge via the Internet at http://pubs.acs.org.
References
(1) Strahl, B. D.; Allis, C. D. The language of covalent histone
modifications. Nature 2000, 403, 41-45.
(2) Jenuwein, T.; Allis, C. D. Translating the histone code. Science 2001,
293, 1074-1080.
Docking studies were not performed on nonactive compounds
(7b, 7c, 7d, and 20) and on compound 15 due to its large
dimensions.
(3) Shi, Y.; Lan, F.; Matson, C.; Mulligan, P.; Whetstine, J. R.; Cole, P.
A.; Casero, R. A.; Shi, Y. Histone demethylation mediated by the
nuclear amine oxidase homolog LSD1. Cell 2004, 119, 941-953.
(4) Forneris, F.; Binda, C.; Vanoni, M. A.; Battaglioli, E.; Mattevi, A.
Human histone demethylase LSD1 reads the histone code. J. Biol.
Chem. 2005, 280, 41360-41365.
(5) Forneris, F.; Binda, C.; Vanoni, M. A.; Mattevi, A.; Battaglioli, E.
Histone demethylation catalysed by LSD1 is a flavin-dependent
oxidative process. FEBS Lett. 2005, 579, 2203-2207.
(6) Chen, Y.; Yang, Y.; Wang, F.; Wan, K.; Yamane, K.; Zhang, Y.;
Lei, M. Crystal structure of human histone lysine-specific demethy-
lase 1 (LSD1). Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 13956-
13961.
(7) Stavropoulos, P.; Blobel, G.; Hoelz, A. Crystal structure and
mechanism of human lysine-specific demethylase-1. Nat. Struct. Mol.
Biol. 2006, 13, 626-632.
(8) Yamane, K.; Toumazou, C.; Tsukada, Y.; Erdjument-Bromage, H.;
Tempst, P.; Wong, J.; Zhang, Y. JHDM2A, a JmjC-containing H3K9
demethylase, facilitates transcription activation by androgen receptor.
Cell 2006, 125, 483-495.
(9) Whetstine, J. R.; Nottke, A.; Lan, F.; Huarte, M.; Smolikov, S.; Chen,
Z.; Spooner, E.; Li, E.; Zhang, G.; Colaiacovo, M.; Shi, Y. Reversal
of histone lysine trimethylation by the JMJD2 family of histone
demethylases. Cell 2006, 125, 467-481.
(10) Tsukada, Y.; Fang, J.; Erdjument-Bromage, H.; Warren, M. E.;
Borchers, C. H.; Tempst, P.; Zhang, Y. Histone demethylation by a
family of JmjC domain-containing proteins. Nature 2006, 439, 811-
816.
(11) Klose, R. J.; Yamane, K.; Bae, Y.; Zhang, D.; Erdjument-Bromage,
H.; Tempst, P.; Wong, J.; Zhang, Y. The transcriptional repressor
JHDM3A demethylates trimethyl histone H3 lysine 9 and lysine 36.
Nature 2006, 442, 312-316.
3-D QSAR Studies. GRID Calculations. The interaction
energies were calculated by using the program GRID66 (version
22) with a grid spacing of 1 Å and the grid dimensions (Å): Xmin
/
Xmax, 58.25/20.25; Ymin/Ymax, 69.25/31.25; Zmin/Zmax, 75.25/37.25.
GOLPE Analyses. PLS models were calculated with GOLPE
4.5.1267 running on a SGI O2 R10000 equipped with the IRIX
operating system 6.5.11. To measure the goodness of the model
the statistical indices r2, q2, and SDEP were employed.
Probe Selection. In this study we started with five probes (Csp3
,
DRY, OH2, NHamide, Osp2) and combinations of them, according to
the nature of the RmtA active site. Preliminary PLS analyses were
run on the DSP, DAP, and DBP groups, and the probe combination
displaying higher values of r2 and q2 was selected to develop the
final model for each group.
Variable Preselection. The resulting probe-target interaction
energies for each compound were unfolded to produce the corre-
sponding one-dimensional vector variables, which were assembled
in the so-called X matrix. This matrix was pretreated first by using
a cutoff of 5 kcal/mol to produce a more symmetrical distribution
of energy values and then zeroing small variable values and
removing variables with small standard deviation, using appropriate
cutoffs. In addition, variables taking only two and three distribution
were also removed.
Smart Region Definition (SRD). A number of seeds (3000)
were selected using a D-optimal design criterion in the weight space.
Structural differences between different molecules in the series will
be reflected in groups of variables, and therefore groups were
generated around each seed in the 3-D space. Variables with a
distance of no more than 1 Å to the seeds were included in the
groups. If two neighboring groups (with a distance smaller than 2
Å) contained the same information the groups were collapsed. The
groups were used in the variable selection procedure replacing the
original variables. The effect of the groups on the predictivity was
evaluated, and groups instead of individual variables were removed
from the data file.
Region Selection. The effect of the grouped variables on the
predictivity was evaluated using a fractional factorial design (FFD)
procedure. A number of reduced models (twice the number of
variables) were built removing some of the variables according to
the FFD design. The effect of dummy variables (20%) on the
predictivity was calculated, and only if a variable had a positive
effect on the predictivity larger than the effect of the average dummy
variable was included in the final model. The FFD selection was
repeated until the r2 and q2 value did not increase significantly. In
(12) Cloos, P. A.; Christensen, J.; Agger, K.; Maiolica, A.; Rappsilber,
J.; Antal, T.; Hansen, K. H.; Helin, K. The putative oncogene GASC1
demethylates tri- and dimethylated lysine 9 on histone H3. Nature
2006, 442, 307-311.
(13) Zhang, Y.; Reinberg, D. Transcription regulation by histone methy-
lation: interplay between different covalent modifications of the core
histone tails. Genes DeV. 2001, 15, 2343-2360.
(14) Byvoet, P.; Shepherd, G. R.; Hardin, J. M.; Noland, B. J. The
distribution and turnover of labeled methyl groups in histone fractions
of cultured mammalian cells. Arch. Biochem. Biophys. 1972, 148,
558-567.
(15) Baxter, C. S.; Byvoet, P. Intercalating agents as probes of the spatial
relationship between chromatin components. Biochem. Biophys. Res.
Commun. 1975, 63, 286-291.
(16) Rice, J. C.; Allis, C. D. Histone methylation versus histone acety-
lation: new insights into epigenetic regulation. Curr. Opin. Cell Biol.
2001, 13, 263-273.
(17) Turner, B. M. Memorable transcription. Nat. Cell Biol. 2003, 5, 390-
393.
(18) Lee, D. Y.; Teyssier, C.; Strahl, B. D.; Stallcup, M. R. Role of protein
methylation in regulation of transcription. Endocr. ReV. 2005, 26,
147-170.

1252 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 6
Ragno et al.
(19) Bedford, M. T.; Richard, S. Arginine methylation an emerging
regulator of protein function. Mol. Cell. 2005, 18, 263-272.
(20) Cook, J. R.; Lee, J. H.; Yang, Z. H.; Krause, C. D.; Herth, N.;
Hoffmann, R.; Pestka, S. FBXO11/PRMT9, a new protein arginine
methyltransferase, symmetrically dimethylates arginine residues.
Biochem. Biophys. Res. Commun. 2006, 342, 472-481.
(21) Gary, J. D.; Clarke, S. RNA and protein interactions modulated by
protein arginine methylation. Prog. Nucleic Acid Res. Mol. Biol. 1998,
61, 65-131.
(40) Ragno, R.; Mai, A.; Massa, S.; Cerbara, I.; Valente, S.; Bottoni, P.;
Scatena, R.; Jesacher, F.; Loidl, P.; Brosch, G. 3-(4-Aroyl-1-methyl-
1H-pyrrol-2-yl)-N-hydroxy-2-propenamides as a new class of syn-
thetic histone deacetylase inhibitors. 3. Discovery of novel lead
compounds through structure-based drug design and docking studies.
J. Med. Chem. 2004, 47, 1351-1359.
(41) Mai, A.; Massa, S.; Rotili, D.; Cerbara, I.; Valente, S.; Pezzi, R.;
Simeoni, S.; Ragno, R. Histone deacetylation in epigenetics: an
attractive target for anticancer therapy. Med. Res. ReV. 2005, 25, 261-
309.
(22) Cheng, D.; Yadav, N.; King, R. W.; Swanson, M. S.; Weinstein, E.
J.; Bedford, M. T. Small molecule regulators of protein arginine
methyltransferases. J. Biol. Chem. 2004, 279, 23892-23899.
(23) a) Bartel, R. L.; Borchardt, R. T. Effects of adenosine dialdehyde
on S-adenosylhomocysteine hydrolase and S-adenosylmethionine-
dependent transmethylations in mouse L929 cells. Mol. Pharmacol.,
1984, 25, 418-424.; b) Schwerk, C.; Schulze-Osthoff, K. Methyl-
transferase inhibition induces p53-dependent apoptosis and a novel
form of cell death. Oncogene 2005, 24, 7002-7011.
(24) Casellas, P.; Jeanteur, P. Protein methylation in animal cells. II.
Inhibition of S-adenosyl-L-methionine:protein( arginine ) N- meth-
yltransferase by analogs of S-adenosyl-L-homocysteine. Biochim.
Biophys. Acta, Nucleic Acids Protein Synth. 1978, 519, 255-268.
(25) Barbes, C.; Sanchez, J.; Yebra, M. J.; Robert-Gero, M.; Hardisson,
C. Effects of sinefungin and S-adenosylhomocysteine on DNA and
protein methyltransferases from Streptomyces and other bacteria.
FEMS Microbiol. Lett. 1990, 69, 239-243.
(26) Massa, S.; Mai, A.; Sbardella, G.; Esposito, M.; Ragno, R.; Loidl,
P.; Brosch, G. 3-(4-aroyl-1H-pyrrol-2-yl)-N-hydroxy-2-propenamides,
a new class of synthetic histone deacetylase inhibitors. J. Med. Chem.
2001, 44, 2069-2072.
(27) Mai, A.; Massa, S.; Ragno, R.; Esposito, M.; Sbardella, G.; Nocca,
G.; Scatena, R.; Jesacher, F.; Loidl, P.; Brosch, G. Binding mode
analysis of 3-(4-benzoyl-1-methyl-1H-2-pyrrolyl)-N-hydroxy-2-pro-
penamide: a new synthetic histone deacetylase inhibitor inducing
histone hyperacetylation, growth inhibition, and terminal cell dif-
ferentiation. J. Med. Chem. 2002, 45, 1778-1784.
(28) Ornaghi, P.; Rotili, D.; Sbardella, G.; Mai, A.; Filetici, P. A novel
Gcn5p inhibitor represses cell growth, gene transcription and histone
acetylation in budding yeast. Biochem. Pharmacol. 2005, 70, 911-
917.
(29) Trojer, P.; Dangl, M.; Bauer, I.; Graessle, S.; Loidl, P.; Brosch, G.
Histone methyltransferases in Aspergillus nidulans: evidence for a
novel enzyme with a unique substrate specificity. Biochemistry 2004,
43, 10834-10843.
(30) Barskii, V. E.; Ivanov, V. B.; Skliar Iu, E.; Mikhailov, G. I.
[Dichlorotriazineylaminofluorescein-a new fluorochrome for cy-
tochemical and histochemical detection of proteins]. IzV. Akad. Nauk.
SSSR Biol. 1968, 5, 744-747.
(31) Peng, X.; Yu, H.; Hang, Y.; Wang, J. N,N′;-phosgenation with
triphosgene in the synthesis of direct dyes containing the ureylene
group. Dyes Pigm. 1996, 32, 193-198.
(42) Mai, A.; Massa, S.; Rotili, D.; Simeoni, S.; Ragno, R.; Botta, G.;
Nebbioso, A.; Miceli, M.; Altucci, L.; Brosch, G. Synthesis and
Biological Properties of Novel, Uracil-Containing Histone Deacety-
lase Inhibitors. J. Med. Chem. 2006, 49, 6046-6056.
(43) Mai, A.; Massa, S.; Valente, S.; Simeoni, S.; Ragno, R.; Bottoni, P.;
Scatena, R.; Brosch, G. Aroyl-Pyrrolyl Hydroxyamides: Influence
of Pyrrole C4-Phenylacetyl Substitution on Histone Deacetylase
Inhibition. ChemMedChem 2006, 1, 225-237.
(44) Ragno, R.; Simeoni, S.; Valente, S.; Massa, S.; Mai, A. 3-D QSAR
studies on histone deacetylase inhibitors. A GOLPE/GRID approach
on different series of compounds. J. Chem. Inf. Model. 2006, 46,
1420-1430.
(45) Kouranov, A.; Xie, L.; de la Cruz, J.; Chen, L.; Westbrook, J.; Bourne,
P. E.; Berman, H. M. The RCSB PDB information portal for
structural genomics. Nucleic Acids Res. 2006, 34, 302-305.
(46) Berman, H. M.; Westbrook, J.; Feng, Z.; Gilliland, G.; Bhat, T. N.;
Weissig, H.; Shindyalov, I. N.; Bourne, P. E. The Protein Data Bank.
Nucleic Acids Res. 2000, 28, 235-242.
(47) Zhang, X.; Cheng, X. Structure of the predominant protein arginine
methyltransferase PRMT1 and analysis of its binding to substrate
peptides. Structure 2003, 11, 509-520.
(48) Weiss, V. H.; McBride, A. E.; Soriano, M. A.; Filman, D. J.; Silver,
P. A.; Hogle, J. M. The structure and oligomerization of the yeast
arginine methyltransferase, Hmt1. Nat. Struct. Biol. 2000, 7, 1165-
1171.
(49) Zhang, X.; Zhou, L.; Cheng, X. Crystal structure of the conserved
core of protein arginine methyltransferase PRMT3. EMBO J. 2000,
19, 3509-3519.
(50) Multiple sequence alignment was accomplished by the use of BioEdit
software v. 7.0.1 and then manually adjusted. For reference see: Hall,
T. A. BioEdit: a user-friendly biological sequence alignment editor
and analysis program for Windows 95/98/NT. Nucl. Acids Symp. Ser.
1999, 41, 95-98.
(51) Venclovas, C. Comparative modeling in CASP5: progress is evident,
but alignment errors remain a significant hindrance. Proteins 2003,
53 Suppl. 6, 380-388.
(52) Schwede, T.; Kopp, J.; Guex, N.; Peitsch, M. C. SWISS-MODEL:
An automated protein homology-modeling server. Nucleic Acids Res.
2003, 31, 3381-3385.
(53) Goodsell, D. S.; Morris, G. M.; Olson, A. J. Automated docking of
flexible ligands: applications of AutoDock. J. Mol. Recognit. 1996,
9, 1-5.
(54) Wang, R.; Lai, L.; Wang, S. Further development and validation of
empirical scoring functions for structure-based binding affinity
prediction. J. Comput.-Aided Mol. Des. 2002, 16, 11-26.
(55) Shoichet, B. K.; Stroud, R. M.; Santi, D. V.; Kuntz, I. D.; Perry, K.
M. Structure-based discovery of inhibitors of thymidylate synthase.
Science 1993, 259, 1445-1450.
(32) Adam, J. M.; Kaser, A. Anionic disazo dyes. EP0290384, 1988.
(33) Birkett, H. E.; Cherryman, J. C.; Chippendale, A. M.; Evans, J. S.
O.; Harris, R. K.; James, M.; King, I. J.; McPherson, G. J. Structural
investigations of three triazines: Solution-state NMR studies of
internal rotation and structural information from solid-state NMR,
plus a full structure determination from powder X-ray diffraction in
one case. Magn. Reson. Chem. 2003, 41, 324-336.
(34) Kuntz, I. D. Structure-based strategies for drug design and discovery.
Science 1992, 257, 1078-1082.
(35) Hillisch, A.; Pineda, L. F.; Hilgenfeld, R. Utility of homology models
in the drug discovery process. Drug DiscoVery Today 2004, 9, 659-
669.
(56) Docking studies were not performed on non-active compounds (7b,
7c, 7d, and 20) and on compound 15 due to its large dimensions.
(57) The two tautomeric forms of compound 16 (keto and enol, indicated
as 16a and 16b, respectively) docked in different sites.
(58) Cruciani, G.; Watson, K. A. Comparative molecular field analysis
using GRID force-field and GOLPE variable selection methods in a
study of inhibitors of glycogen phosphorylase b. J. Med. Chem. 1994,
37, 2589-2601.
(59) Frohman, M. A.; Dush, M. K.; Martin, G. R. Rapid production of
full-length cDNAs from rare transcripts: amplification using a single
gene-specific oligonucleotide primer. Proc. Natl. Acad. Sci. U.S.A.
1988, 85, 8998-9002.
(60) Bu¨gl, H.; Fauman, E.; Staker, B.; Zheng, F.; Kushner, S.; Saper,
M.; Bardwell, J.; Jakob, U. RNA Methylation under Heat Shock
Control. Mol. Cell 2000, 6, 349-360.
(61) Case, D. A.; Cheatham, T. E., 3rd; Darden, T.; Gohlke, H.; Luo, R.;
Merz, K. M., Jr.; Onufriev, A.; Simmerling, C.; Wang, B.; Woods,
R. J. The Amber biomolecular simulation programs. J. Comput.
Chem. 2005, 26, 1668-1688.
(62) Schuttelkopf, A. W.; van Aalten, D. M. PRODRG: a tool for high-
throughput crystallography of protein-ligand complexes. Acta Crys-
tallogr., Sect. D: Biol. Crystallogr. 2004, 60, 1355-1363.
(63) Van Der Spoel, D.; Lindahl, E.; Hess, B.; Groenhof, G.; Mark, A.
E.; Berendsen, H. J. GROMACS: fast, flexible, and free. J. Comput.
Chem. 2005, 26, 1701-1718.
(36) Tramontano, A.; Morea, V. Assessment of homology-based predic-
tions in CASP5. Proteins 2003, 53 Suppl. 6, 352-368.
(37) Mai, A.; Massa, S.; Ragno, R.; Cerbara, I.; Jesacher, F.; Loidl, P.;
Brosch, G. 3-(4-Aroyl-1-methyl-1H-2-pyrrolyl)-N-hydroxy-2-alky-
lamides as a new class of synthetic histone deacetylase inhibitors. 1.
Design, synthesis, biological evaluation, and binding mode studies
performed through three different docking procedures. J. Med. Chem.
2003, 46, 512-524.
(38) Rosato, R. R.; Grant, S. Histone deacetylase inhibitors in cancer
therapy. Cancer Biol. Ther. 2003, 2, 30-37.
(39) Mai, A.; Massa, S.; Cerbara, I.; Valente, S.; Ragno, R.; Bottoni, P.;
Scatena, R.; Loidl, P.; Brosch, G. 3-(4-Aroyl-1-methyl-1H-2-pyrro-
lyl)-N-hydroxy-2-propenamides as a new class of synthetic histone
deacetylase inhibitors. 2. Effect of pyrrole-C2 and/or -C4 substitutions
on biological activity. J. Med. Chem. 2004, 47, 1098-1109.

Inhibitors of Histone Arginine Methyltransferases
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 6 1253
(64) Gillet, A.; Sanner, M.; Stoffler, D.; Olson, A. Tangible interfaces
for structural molecular biology. Structure 2005, 13, 483-491.
(65) Pettersen, E. F.; Goddard, T. D.; Huang, C. C.; Couch, G. S.;
Greenblatt, D. M.; Meng, E. C.; Ferrin, T. E. UCSF Chimera-a
visualization system for exploratory research and analysis. J. Comput.
Chem. 2004, 25, 1605-1612.
cally favorable binding sites on biologically important macromol-
ecules. J. Med. Chem. 1985, 28, 849-857.
(67) GOLPE, Multivariate Infometric Analysis Srl., Viale dei Castagni
16, Perugia, Italy, 1999.
(66) Goodford, P. J., A computational procedure for determining energeti-
JM061213N