Fragment‐based drug design facilitated by dynamic combinatorial chemistry and NMR spectroscopy
Peter Kroon, s1687735 18/03/2015
Abstract
Drug development is a challenging process, which often fails. To ameliorate this, we attempted to develop an efficient methodology for fragment‐based drug design based on dynamic combinatorial chemistry and NMR spectroscopy. To this end, two libraries were designed based on crystallographic data from the group of Klebe for the model protein endothiapepsin using computer aided modelling.
The designed compounds all feature an acylhdyrazone linker and at least one fluorine atom. The first library was discarded due to problems with the synthesis. The second library was analysed using quantitative 19F‐NMR and 1H‐STD‐NMR spectroscopy, using trifluoroacetic acid as an internal reference for the fluorine NMR spectra. No binding compounds could be identified, even though one of the compounds was a known binder. It is hypothesised this is due to the trifluoroacetic acid since the 1H‐STD‐NMR spectrometry also produced no results. However, it could be concluded that
19F‐NMR spectrometry provides better peak resolution with little loss of sensitivity compared to
1H‐NMR spectrometry when analysing mixtures; and that STD‐NMR requires far less protein than regular NMR spectroscopy for analysing templated DCLs.
Table of Contents
Abstract ... 1
Table of Contents ... 2
Introduction ... 3
Structure‐based drug design ... 3
Fragment‐based drug design ... 4
Dynamic combinatorial chemistry... 5
NMR techniques ... 7
Quantitative NMR... 7
Saturation‐transfer difference NMR ... 7
Endothiapepsin ... 9
Library design ... 10
Modelling ... 14
Discussion ... 16
Quantitative 19F‐NMR ... 16
1H‐STD‐NMR ... 16
19F‐STD‐NMR ... 16
Conclusions ... 17
Acknowledgements ... 17
Bibliography ... 17 Supporting Information ...S1 Library A — Compounds 1 to 8 ...S1 Modelling methods ...S1 Modelling results ...S2 Synthesis ... S11 Library B — Compounds 10 to 18... S27 Modelling results ... S27 NMR experiments ... S32 Preparation of NMR samples ... S32 Quantitative 19F‐NMR ... S32
1H‐STD‐NMR ... S32 Quantitative 19F‐NMR results ... S33
1H‐STD‐NMR spectrum ... S36
Introd
Drug de number and ana dynamic required
Structu
Drug dev enzymes these en selected dimensio consider and opti
Figure 1: T
Medicina the enzy started.
heavily synthesi enzyme, crystal s inhibitor found.[1, intricate consumi
duction
evelopment of compoun alyse each o c combinato d in synthesis
ure‐based
velopment is s involved in nzymes kills o
an inhibitor onal structu red at this st
misation) an
The process fro
al chemistry yme is obta
By using co on the intu sed, and its , it is attemp structure is t r that do no
2] This mea e) compound
ing.
nowadays is nds that nee of these com
rial chemist s, and nuclea
d drug de
s a long and n the disease
only the pat r can be desi ure of the age.[1] If a po nd finally sub
m disease to ne
y is involved ained (usual mputer‐aide uition and s s efficacy is pted to obta then used in t bind efficie ns that bef ds have to
s plagued b ed to be con mpounds.[1,2 try based on ar magnetic r
esign
costly proce e need to be
hogen (Targ gned based enzyme. C otent inhibit bmitted for c
ew drug. Figure
mainly in lea ly via crysta ed modelling skill of the
tested in v ain a co‐cryst n silico to ide
ently. The p ore a suffic be synthesi
y long deve nsidered, and
2] We hope n already kn resonance (N
ess, progress e identified a et identifica
on known in Commonly, a tor is found, clinical trials
e taken from Jia
ad optimisat allography), g techniques medicinal c vitro. If it is tal structure entify empty process is rep ciently poten ised, purifie
elopment tim d the amoun
to ameliora nown inhibit NMR) spectr
sing through and it has to tion and vali nhibitors, en
around 100 this lead is f (Figure 1).[1,2
ang et al.[2]
tion.[1] Once an iterative s, a potentia chemist invo
confirmed e of the enzy
y pockets in peated until nt inhibitor d and analy
mes, stemm nt of time re
ate this situ tors to grea oscopy to fa
h several disc o be confirm
idation). Onc zyme functio 0,000 separa urther optim
2]
a three‐dim e cycle, dep l inhibitor is olved.[3] This that the com yme with thi
the enzyme a sufficient is found a ysed. This is
ing mainly f equired to sy uation by e atly reduce cilitate analy
crete steps.
ed that knoc ce a target e on, or ideally ate compou mised (Lead d
mensional str picted in Fig s designed; t s compound mpound inh is inhibitor.
e, or moietie tly potent in
multitude o s usually ve
from the ynthesise mploying the time ysis.
First, the cking out enzyme is y a three‐
unds are discovery
ucture of ure 2, is this relies d is then hibits the This new es of the hibitor is of (often ery time‐
Figure 2: The process of structure‐based drug design.
Fragment‐based drug design
Fragment‐based drug design is a methodology that can be used to facilitate structure‐based drug design, and in particular the lead‐identification and optimisation steps. Classically, many large compounds are screened against a target, in the hope one or more will bind to the protein (high‐throughput screening, HTS). However, because the compounds screened are usually rather large, and enzymes bind compounds very selectively, many compounds often need to be screened to find a successful candidate.[1] Two alternative approaches can be used when applying fragment‐
based drug design: fragment linking, and fragment growing.
In fragment linking and growing, smaller compounds are screened instead. These will usually bind the target enzyme only weakly, but chemical space can be sampled far more efficiently. If multiple, small compounds bind the protein in different pockets, these can be thought of as fragments of the ultimate inhibitor. In this case, if the binding modes of these fragments can be elucidated, it may be possible to link the fragments together in order to obtain a highly potent inhibitor in an efficient manner. However, great care must be exercised: if there is too much strain in the linker between the fragments, or if it slightly too long or too short, this will prevent the inhibitor from binding like the fragments, leading to a loss of potency. Because of this strict geometrical requirement on the linker, fragment linking is usually considered hard, but there are examples in which this methodology has been successfully applied, resulting in very potent inhibitors.[4–6]
If instead there is only binding information available for one fragment, this fragment may be "grown"
in order to make more beneficial interactions with the protein. This process is generally more elaborate than fragment linking, but it is easier. There are still pitfalls however. It is often very easy to identify empty hydrophobic pockets in a three‐dimensional model of an enzyme. Growing the inhibitor in this direction would probably result in an insoluble compound however. This
Compound synthesis and
purification
Evaluation
Crystallography Modelling and
design
methodo Figure 3.
In practi not unco modellin
Figure 3: F large com small orga manner. B pocket in o
Dynam
Dynamic compou template accordin of chang affinity;* noted th template change i
*
ology has a .
ice, the disti ommon to g ng is used to
Fragment‐based pounds are scr anic molecules Bottom panel:
order to form a
mic combi
c combinato nds is form e, and will in ng to Le Chât ge in conce
* this degree hat the amp
e, and the to n concentra
lso been us
inction betw grow one fra facilitate the
d drug design. T eened against t bind the targe If only one sm a potent inhibit
inatorial c
orial chemist med.[12–14] Ne n this way be telier's princi ntration of e of change
lification fac otal concentr tion, a lot of
ed successf
ween fragme agment in o e process.
Top panel: the the target enzy et protein thes all organic mol or. Figure taken
chemistry
try is an app ext, if a tem e depleted fr
iple, making individual li
in concentr ctor is a func
ration of bui f template m
ully, resultin
nt linking an rder to mim
classical appro yme, but only a se may be link lecule binds th n from Erlanson
y
proach in w mplate is ad rom their eq more of the brary memb ation is calle ction of: the lding blocks.
may be requir
ng in comm
nd fragment mic a second
oach to drug de few — if any — ked together to e protein, this n.[1]
which a libra dded, some quilibrium. Th e compounds bers can be
ed an ampli e affinity of . This means red.
ercially avai
growing is .[11] In all th
esign is high‐thr
— will bind. Mid o form a poten fragment may
ary of dynam library mem he library wi s that bind (F
used as a fication fact the binder, , that in orde
ilable drugs.
not so clear hese cases, c
roughput scree ddle panel: if tw nt inhibitor in y be grown into
mically inter mbers can ill then re‐eq Figure 4). Th measure for tor.[12–14] It s
the concent er to see a s
.[7–10] See
‐cut. It is computer
ning; many wo or more an elegant o an empty
rchanging bind this quilibrate he degree r binding hould be tration of ignificant
DCC has building efficient significa member be desig (few) dif target pr Because there are
Figure 4: A protein be
Since th connecti forming Diels‐Ald conditio aqueous stopped To these can rev equilibra nucleoph of condit
s the advant blocks can b screening o ntly accelera r has to be sy gned around
fferent ways rotein.
of this, dyn e numerous
A schematic re est, it is selected
e only requi ions can be reactions c der reaction ns which ca s environme
at will by ch e ends we h versibly com
ation occurs hilic catalyst tions (Schem
tage that rel be synthesise of large libra
ates both the ynthesised, d one or mo the system
amic combin examples in
presentation o d from the libra
irement for used betwe an be used, , etc. The o an be tolera nt that is sli hanging the r ave opted fo mbine to fo at an acidi t such as ani me 1).
latively few ed in paralle ries with rel e synthesis a
purified and re fragment will show wh
natorial chem which this t
of dynamic com ary. Figure take
DCC is that t een the bui , such as im only prerequ
ted by the ghtly acidic reaction cond or acylhydra orm an acy
c pH. Addit line,[20] allow
building blo el, or may ev atively little and analysis
analysed se ts. For exam hich linker ha
mistry is inc technique ha
mbinatorial che en from Mamid
the library c lding blocks mine format uisite is that
template.[12 or basic. La ditions, since zones as lin ylhydrazone tionally, acyl wing this rea
ocks need to ven be comm
effort,[12,14]
of potential eparately; pa mple, if two f as the best o
reasingly be as been used
mistry. Since th yala and Finn.[1
constituents s. In principl ion, transac the reactio
,14,19] In case stly, it is des e this facilita king group.
e. Acylhydra hydrazone e action to also
o be synthes mercially avai which mean inhibitors, si rticularly if a fragments ca orientation a
ing used in d d successfully
he compound A
13]
can intercha e, all revers ylation, disu on runs at a
e of a prote sirable that t ates analysis.
Here an alde azones are
exchange ca o be used un
sised, and th ilable. This a ns that this t
ince not eve a dynamic lib
an link toge and propertie
drug design, y.[15–18]
A2‐B3 binds th
ange, many sible, covale ulfide forma n acceptabl ein this is u the equilibri .[12,14]
ehyde and h rather sta an be cataly nder a broad
hat these allows for echnique ry library brary can ether in a es for the
[12,14] and
e template
different nt bond‐
tion, the e rate in sually an ia can be
hydrazide ble, and sed by a der range
Scheme 1:
Classical (LC‐MS).
be stopp and a lo some of
NMR te
Quantit NMR spe
1H‐NMR sensitive structura abundan Secondly result in An alter addition sensitive from –20 solvents referenc In order so, two of releva spectra.
required Saturat An altern measure (STD‐NM which is been sat the nucl saturate
: The formation
ly, DCLs are . These tech ped, separat ot of protein these drawb
echnique
tative NMR ectroscopy i spectroscop e nucleus, an al informatio nt, many solv y, 1H‐NMR si
overlapping rnative to 1H
, its gyroma e as 1H‐NMR
00 to 20 ppm do not cont ce might be d to analyse a spectra are ant peaks a
Since the ch d.
tion‐transfe native to me e the bindin MR) was dev s not of inte turated, this ear Oberhau ed, and no lo
n of an acylhydr
e analysed u niques have tion may be is required backs by usin
es
s a common py. There ar nd 1H atoms on. Unfortun vents also co ignals span o g signals.
H‐NMR may gnetic ratio R. Furthermo m. This mean
tain 19F atom desirable.
a DCL with th recorded, on are normalis
hange in equ
er differenc easuring the g of ligands veloped. The rest, and th s saturation c
user effect ( onger give a
razone from an
using metho e some draw hard to ach since the eq ng NMR spec
n analytical m re good reas
are abunda nately, 1H‐NM
ontain them, only a narrow
be 19F‐NMR is 83% as la ore, 19F‐NM ns that 19F sig ms, no isotop
hese techniq ne with tem
ed against a uilibrium is m
ce NMR change in e s to the tem e template, i
is saturation can transfer (NOE).[21] As signal in NM
n aldehyde and
ods such as wbacks since ieve, all libra quilibrium is ctroscopy.
method in s sons why 1H ant in most o MR has some , demanding w spectral ra
R spectrosco arge as that
R signals sp gnals are oft pically enric
ques, a chan mplate and on
an internal measured dir
quilibrium c mplate. To th
in this case n spreads th to a bound a result, inh MR. This mea
a hydrazide un
liquid chro they are de ary member s measured
ynthetic org H is so popu
organic com e drawbacks g the use of (
ange: roughl
opy. 19F is 1 of 1H. This m pan a much
en well sepa hed solvent
ge in equilib ne without,
reference a rectly, large
aused by ad his end, sat a protein, i hrough the p
inhibitor by hibitors that ans that whe
der influence o
matography estructive, th s need to ha directly. We
ganic chemis ular in NMR
pounds, wh . Firstly, beca (expensive) d y from –2 to
00% natura means that 1
broader spe arated. Also,
is needed;*
brium is mea after which nd compare amounts of
dition of a te uration‐tran s saturated protein. Onc means of sp bind to the en a regular
of catalytic anili
y mass spec he equilibriu ave a differe e hope to cir
stry, and in p since it is t ich results in ause 1H atom deuterated s o 12 ppm, w
lly abundan
19F‐NMR is a ectral range
since many although an
asured direct the signal in ed between
template ar
emplate is to sfer differen in a spectra ce the active pin diffusion e protein wi
1H‐NMR spe
ne.
ctrometry m has to ent mass, rcumvent
particular the most n a lot of ms are so solvents.* hich may
t, and in almost as : roughly common n internal
tly. To do ntensities the two re usually
o directly nce NMR al region, e site has n through ll also be ectrum is
compare decrease Because many lig methods analysed problem
Figure 5: a protein to multiple lig
Unfortun spectros circumve saturate the 1H c explicitly using an capable Of cours least one
ed to one wh e in intensity
of the dyna gand molecu
s. For this to d.[19] Also, ra m.
a schematic dep o ligands that
gand molecule
nately, 1H‐S scopy: a sm ent these hu ed as before, channel, it y transferrin n INEPT‐like
of measurin se, when ana
e 19F atom.
here the prot y; compound
amic binding les, meaning o work howe apid ligand e
piction of STD‐
bind. Due to d s.
TD‐NMR sp all spectral urdles, the te
, and the sa is decouple ng the satura
sequence.[22 g 19F whilst alysing a DCL
tein was satu ds that do no
g and unbind g that this te ever, extensi exchange is
NMR spectrosc dynamic bindin
ectroscopy range and echnique has
turation is t d, and 19F i ation from h
2] This techn saturating 1H L using any 19
urated, the s ot bind the te ding of com echnique req
ive measure required,[21
copy. Saturation ng and unbind
suffers from the require s been modif transferred t is observed hydrogens n nique howev
H.
9F‐NMR tech
signals corre emplate will pounds, one quires far les ement times
] but for we
n is depicted in ing, one prote
m the same ment for is fied to 19F‐ST to the ligand . Additional nuclei in the ver, does req
hnique all lib
sponding to not be affec e protein mo ss protein th
are required eak binders
blue. Saturatio ein molecule ca
e problems otopically e TD‐NMR, in w d. However,
sensitivity e inhibitor to
quire an NM
rary membe
bound inhib cted (Figure 5 olecule may an quantitat d when mixt this is usua
on is transferre an transfer sat
as normal nriched solv which the te
instead of o can be obt o the fluorin MR instrumen
ers need to c
bitors will 5).[21]
saturate tive NMR tures are ally not a
ed from the turation to
1H‐NMR vents. To mplate is observing ained by ne nuclei nt that is
ontain at
Endoth
In order attempt fungus E This clas several s bonds,[23 catalytic These as Some of binding.[
Figure 6:
cartoon. T blue). The
As a mo research complex which is chemistr
hiapepsin
r to demons to design an Endothia par ss of enzyme
serious disea
3–25] and the c dyad: two
spartic acids f the know
[19]
X‐ray crystal The catalytic dya
water molecul
odel enzyme h group of K x with a seri
beneficial fo ry is compat
[19]
n
strate the a n inhibitor fo asitica, and es is widespr ases such as e inhibitors
aspartic aci s are tightly
n inhibitors
structure of e ad (aspartic aci le closely bound
e, endothiap Klebe et al.[2 ies of fragm or acylhydraz tible with en
advantages D or the enzym it is a model
ead, and has s malaria and known toda id residues
bound to a w displace th
ndothiapepsin d residues 32 a d to the catalyt
pepsin is re
28] has previ ents; its me zone exchan ndothiapepsi
DCC can off me endothia
enzyme for s varied biol d AIDS.[23–25]
ay block the (Asp‐32 and water molec his catalytic
(PDB code: 1E and 215) is dep tic dyad is show
eadily availab iously publis echanism is
ge;[23,26] it ha n;[19] and las
fer in fragm pepsin. This aspartic pro ogical functi Its function e active site d Asp‐215) o
cule through water mole
ER8[27]). The pr icted as a stick wn as a red sphe
ble and crys shed co‐crys
fully elucida as been show stly, our grou
ment‐based d protein is o teases.[23]
ions and play n in nature is
.[23,26] This a of which one h hydrogen b
ecule, but o
rotein backbon model (Colour ere.
stallises eas stal structure ated;[23,25] its wn in the pas
up is current
drug design, originally fou
ys a causativ s to hydroly active site fe
e is protona bonds (Figur others requi
ne is depicted r code: C: green
sily. Addition es of this en s pH optimu
st that acylhy tly also wor
, we will nd in the
ve role in se amide eatures a ated.[23,25]
e 6).[23,25]
ire it for
as a green n, O: red, N:
nally, the nzyme in m is 4.5, ydrazone king with
Based on analysis designed developm
Librar
A library by the g their ori structure order to Fragmen substitut Since it would h acylhydr construc (Scheme Predicte found in
Figure 7: T Colour cod surface is water mol
S4 S2
n the require and contai d a library o
ment can be
ry design
y of compou group of G. K
iginal paper es in Schem o accommod nt F284 was ted with an e is hard to de have to be o razone moie cted by syste e 3).
d binding m the support
The binding m de: O: red, N: b depicted in gre lecule bound to
S1
4
ements that n at least of compoun e greatly acce
n
nds was des Klebe.[28] The r. The bindin e 2. Since th ate an acylh s reduced to ethylamine m esign the ide oriented, a t ety, and fo ematically inc
modes for al ting informat
odes of the tw blue, F: pale cya ey. It can be see o the active site
all potentia one acylhyd ds based on elerated usin
signed based e fragments ng modes o he fragments hydrazone lin o a fluoroph moiety to pre eal linker an total of eigh our featurin cluding meth
l these com tion.
wo fragments c an, Caspartic dyad: en that the frag e upon binding
S2'
al inhibitors s drazone mo n computer ng DCC and S
d on fragmen used will b of these frag
s overlap wh nker, whilst henyl moiety
event interfe d predict ex ht inhibitors ng a "backw
hylene space
mpounds, alo
co‐crystallised w green, CF109: te gments bind in d (translucent re
'
S3 S1
should conta iety to ena
models in o STD‐NMR spe
nts in comple e referred t gments are hen binding, preserving k y, and the h erence with xactly in wha
were desig ward" acylh ers at either
ong with cal
with endothiap eal, CF284: hot p different pocke
d sphere).
ain at least o ble the dyn order to de ectroscopy.
ex with endo o as F109 an depicted in they had to key interactio hydrazide in the acylhydr at way the a gned: four fe hydrazone m
end of the a
culated bind
pepsin (PDB co pink. Endothiap ets, and that fra
one fluorine namic excha
monstrate t
othiapepsin nd F284, in n Figure 7 a
o be made s ons with the fragment F razone excha acylhydrazon
eaturing a "
moiety. The acylhydrazon
ding energie
odes: 3PBZ and pepsin's solvent agment F109 di
atom for ange, we that drug
reported line with and their smaller in e protein.
F109 was ange.
e moiety forward"
ese were ne moiety
es can be
3PMU).[28]
t‐accessible splaces the
Scheme 2 acylhydraz moiety.
Scheme 3
"forward"
of the acyl
Due to p second l The R' endothia and F284 the pos substitue These ni crystal s accelera done fo enabling
2: After obtain zone linker; red
3: The designed (1–4) or "back lhydrazone mo
problems wi ibrary was d fragment w apepsin in u
4, a new libr sition of the
ents were al ine building structures of te the drug r many com g non‐destru
ing binding m ducing F284 to
d inhibitors. Th kward" (5–8) ac iety, and is dep
th the synth esigned, this was replace npublished rary was des e fluorine w so designed
blocks were f endothiape g‐developme mpounds in
ctive analysi
modes for fragm o a fluoropheny
hey were form cylhydrazone m picted in red.
hesis of the s time consis ed for imid work by M.
signed. Since was varied
(Figure 8 an e obtained c epsin in orde
nt process s one experim s of the DCL
ments F109 an yl moiety, and
med from fragm moiety. A methy
R' fragment sting of comm dazole alde
Mondal, N.
e there was a (10–15), a nd Scheme 4,
commercially er to demon since both t ment. The a L.
nd F284, they exchanging th
ments F109 an ylene spacer w
in Scheme 3 mercially ava hyde (9), w
Radeva, A.
ample space nd some co , 16–18).
y and used t nstrate that the synthesi nalysis was
were modifie he hydrazide of
d F284 (Schem was systematica
3 (see Suppo ailable comp which was
Hirsch and G e around F28
ompounds w
to construct the use of D s and biolog
done using
ed to make ro f F109 for an
me 2) by addin ally included at
orting Inform pounds.
co‐crystallis G. Klebe. Ba 84 inside the with trifluo
t a DCL base DCC can sig gical analysi g NMR spect
oom for an ethylamine
ng either a either end
mation) a
sed with ased on 9 e protein, romethyl
ed on co‐
nificantly s can be troscopy,
Figure 8: C cyan, Caspa
grey. It ca site via hy
* Work of
S
Co‐crystal struc
artic dyad: green, n be seen that drogen bonds (
f M. Mondal,
S
S4 S2
ctures of F284 Cimidazol aldehyde: imidazole alde (dashed yellow
N. Radeva, A.
S1'
(PDB code: 3PM : orange, CF284: ehyde requires
lines).
Hirsch and G
MU)[28] and imi : hot pink. End a water molecu
. Klebe. To be
S2'
dazole aldehyd othiapepsin's s ule (red sphere
e published.
S1
de.* Colour cod solvent‐accessib e) in order to bi
de: O: red, N: b ble surface is d ind to the prot
lue, F: pale depicted in ein's active
Scheme 4:
F284 was hydrazides
: The building b varied, and p s 19–27.
blocks used. The otentially subs
e compounds a stituted for a t
are based on F2 trifluoromethyl
284 and 9, but t l group in ord
he position of t er to yield acy
the fluorine sub ylhydrazones 1
bstituent in 10–18 from
Model
The des inhibitor Modellin acylhydr stationa acylhydr Accordin mediate der Waa 15 are Supporti Each acy suite.[30]
isomeris visually i
Figure 9: T yellow line water mol
S S2
lling
igned librari rs and would ng was perf razone insid ry. Water m razone.
ng to Moloc, d by a wate als interactio shown in F ing Informat ylhydrazone Aspartic aci sm was cons inspected.
The predicted b es. Favourable lecule bound cl
4
ies were firs d bind to the formed usin
e the prote olecules we
, all acylhydr r molecule;
ons were gen igure 9 and tion.
was docked id residue 21 idered. The 3
binding mode o Van der Waa osely to the cat
S1'
st studied in enzyme like g the softw ein, and the
re reset to t
razones sho and occasion nerally very f d 10, respec
d inside the 19 was proto 30 poses wh
of 13. Colour co ls interactions talytic dyad is s
n silico, to s e the fragme ware program
en optimisin their position
uld bind to nally form a favourable. A ctively. Bind
enzyme us onated and hich were mo
ode: O: red, F:
involving fluor shown as a red
S2'
see if the de nts upon wh m Moloc[29]
ng its posit ns according
the catalytic dditional hy As an examp ding modes
ing the Flex key water m ost promising
pale cyan, Cprot
rine atoms are sphere.
esigned com ich they wer by first gen ion whilst k to the cryst
c dyad throu drogen bond ple, the bind
for 10–18
X docking m molecules we g according t
tein: green, C13: depicted as d
S1
mpounds we re based.
nerating the keeping the tal structure
ugh a hydrog ds. Furtherm ding modes o are provide
module in th ere unrestra to the softw
purple, H‐bond dashed, salmon
1
re viable
e desired enzyme for each
gen bond more, Van of 13 and ed in the
he LeadIT ined; E/Z ware were
ds: dashed, n lines. The
Figure 10:
yellow line water mol
Unfortun was bou FlexX re scored b more im To furth poses in yet, it is bound to
S S2
The predicted es. Favourable lecule bound cl
nately, FlexX und to the c sults were d by Hyde[30] d mportantly, th
er analyse t which the li s unknown w
o the catalyt
S4
binding mode Van der Waa osely to the cat
X produced n catalytic dya discarded. So directly; how
he ligand wa this anomaly
igand was bo why the soft tic dyad.
S1'
of 15. Colour c ls interactions talytic dyad is s
no poses for d with a hy ome ligands, wever, the re s not bound y, 9 was doc ound to the tware would
code: O: red, F:
involving fluor shown as a red
any of the a drogen bon , which wer esulting calcu
to the cataly cked using F
catalytic dya d not produ
S2
pale cyan, Cpro
rine atoms are sphere.
cylhydrazon d via a wate e most prom ulated energ
ytic dyad.
FlexX. In this ad as indicat ce poses in
'
tein: green, C15: depicted as d
es in which t er molecule.
mising accor gy of binding
s case, the s ted by the cr which the a
S
purple, H‐bond dashed, salmon
the imidazol . For this re rding to Mol g was very p
software did rystal structu
acylhydrazon
S1
ds: dashed, n lines. The
le moiety eason the loc, were poor, and
produce ure. As of nes were
Discussion
Quantitative
19F‐NMR
Two spectra were recorded, one of a sample containing a high protein concentration, and one without protein (experimental details can be found in the Supporting Information). Peaks were well separated, although less so for trifluoromethylated compounds. The integrals of the peaks from the library members were normalised against the integral of peak from trifluoroacetic acid and compared. Since sum of the concentration of acylhydrazone and the concentration the corresponding hydrazide should be constant, their respective peak areas were summed to 100%. The result is plotted in Figure 11. Unfortunately, upon addition of protein, no significant change in concentrations is observed. This means that none of the compounds bind to the protein according to these data. The raw data and spectra are provided in the Supporting Information.
Figure 11:The difference in peak area for separate library members between a sample containing protein (plus), and one without (minus).
1
H‐STD‐NMR
No signal could be observed from the 1H‐STD‐NMR spectrum, not even one originating from 9 (the spectrum is provided in the Supporting Information). Since 9 is a known inhibitor this could indicate a problem with library, and it is hypothesised that the trifluoroacetic acid interferes with the binding of library members to the protein by either denaturing the protein, blocking the active site, or stopping the acylhydrazone exchange. Therefore, these experiments should be repeated with either a lower concentration of trifluoroacetic acid, a different reference compound, or with a library of known binders in order to validate the methodology.
19
F‐STD‐NMR
Despite our best efforts, no 19F‐STD‐NMR spectrum could be recorded on the NMR spectrometers available. No machine was available with a probe that could be tuned to 19F and 1H simultaneously.
Tuning a probe to 19F and saturating 1H at −1.2 ppm by directly specifying the frequency produced too many artefacts to yield a useful spectrum.
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Hydrazide Acylhydrazone
Conclusions
In conclusion, DCC can play a significant role in both fragment linking and fragment growing, and should be part of every medicinal chemist’s toolkit. However, analysis of the resulting mixtures is often challenging, but here a variety of NMR techniques can help, providing a way to analyse an equilibrating library in a non‐destructive manner. STD‐NMR is particularly interesting in this sense, since it requires very little protein.
As demonstrated, 19F‐NMR could extend these techniques further, granting less background signals and excellent peak separation. The only additional requirement is that all relevant compounds must contain at least one 19F atom.
Modelling and the NMR studies are in agreement, and no inhibitor was found in this study. It would be worthwhile to repeat this methodology with a library of known inhibitors, as well as repeating it with a different reference compound to make sure trifluoroacetic acid does not influence the measurements. Once it is conclusively proven that 19F‐STD‐NMR and quantitative 19F‐NMR are suitable for analysing DCLs, it should be explored how little protein is required as a function of the free energy of binding of the library members.
Acknowledgements
I would like to particularly thank Pieter van der Meulen for his extensive help with all the NMR measurements, as well as his limitless patience and perseverance in trying to get 19F‐STD‐NMR to work on a machine that is not designed to do so. Furthermore, I would like to thank Milon Mondal for his daily supervision and help with the syntheses; and I would like to thank Anna Hirsch for supervising this research project in her group.
Bibliography
1. Erlanson, D. A. Introduction to fragment‐based drug discovery. Top. Curr. Chem. 317, 1–32 (2012).
2. Ou‐Yang, S.‐S. et al. Computational drug discovery. Acta Pharmacol. Sin. 33, 1131–1140 (2012).
3. Kutchukian, P. S. et al. Inside the mind of a medicinal chemist: the role of human bias in compound prioritization during drug discovery. PLoS One. 7, e48476 (2012).
4. Hajduk, P. J. et al. NMR‐based modification of matrix metalloproteinase inhibitors with improved bioavailability. J. Med. Chem. 45, 5628–5639 (2002).
5. Park, C. M. et al. Discovery of an orally bioavailable small molecule inhibitor of prosurvival B‐
cell lymphoma 2 proteins. J. Med. Chem. 51, 6902–6915 (2008).
6. Wada, C. K. The evolution of the matrix metalloproteinase inhibitor drug discovery program at abbott laboratories. Curr. Top. Med. Chem. 4, 1255–1267 (2004).
7. Flaherty, K. T. et al. Inhibition of mutated, activated BRAF in metastatic melanoma. N. Engl. J.
Med. 363, 809–819 (2010).
8. Bollag, G. et al. Clinical efficacy of a RAF inhibitor needs broad target blockade in BRAF‐
mutant melanoma. Nature. 467, 596–599 (2010).
9. Wyatt, P. G. et al. Identification of N‐(4‐piperidinyl)‐4‐(2,6‐dichlorobenzoylamino)‐1H‐
pyrazole‐3‐carboxamide (AT7519), a novel cyclin dependent kinase inhibitor using fragment‐
based X‐ray crystallography and structure based drug design. J. Med. Chem. 51, 4986–4999 (2008).
10. Artis, D. R. et al. Scaffold‐based discovery of indeglitazar, a PPAR pan‐active anti‐diabetic agent. Proc. Natl. Acad. Sci. U. S. A. 106, 262–267 (2009).
11. Sandanayaka, V. et al. Discovery of 4‐[(2S)‐2‐{[4‐(4‐Chlorophenoxy)phenoxy]methyl}‐1‐
pyrrolidinyl] butanoic acid (DG‐051) as a novel leukotriene A4 hydrolase inhibitor of leukotriene B4 biosynthesis. J. Med. Chem. 53, 573–585 (2010).
12. Ramström, O. & Lehn, J.‐M. Drug discovery by dynamic combinatorial libraries. Nat. Rev. Drug Discov. 1, 26–36 (2002).
13. Mamidyala, S. K. & Finn, M. G. In situ click chemistry: probing the binding landscapes of biological molecules. Chem. Soc. Rev. 39, 1252–1261 (2010).
14. Mondal, M. & Hirsch, A. K. H. Dynamic combinatorial chemistry: a tool to facilitate the identification of inhibitors for protein targets. Chem. Soc. Rev. 44, 2455–2488 (2015).
15. Erlanson, D. A. et al. In situ assembly of enzyme inhibitors using extended tethering. Nat.
Biotechnol. 21, 308–314 (2003).
16. Maly, D. J., Choong, I. C. & Ellman, J. A. Combinatorial target‐guided ligand assembly:
identification of potent subtype‐selective c‐Src inhibitors. Proc. Natl. Acad. Sci. U. S. A. 97, 2419–2424 (2000).
17. Huc, I. & Lehn, J. M. Virtual combinatorial libraries: dynamic generation of molecular and supramolecular diversity by self‐assembly. Proc. Natl. Acad. Sci. U. S. A. 94, 2106–2110 (1997).
18. Bunyapaiboonsri, T. et al. Dynamic deconvolution of a pre‐equilibrated dynamic combinatorial library of acetylcholinesterase inhibitors. Chembiochem. 2, 438–444 (2001).
19. Mondal, M. et al. Structure‐based design of inhibitors of the aspartic protease endothiapepsin by exploiting dynamic combinatorial chemistry. Angew. Chem. Int. Ed. Engl. 53, 3259–3263 (2014).
20. Bhat, V. T. et al. Nucleophilic catalysis of acylhydrazone equilibration for protein‐directed dynamic covalent chemistry. Nat. Chem. 2, 490–497 (2010).
21. Viegas, A., Manso, J., Nobrega, F. L. & Cabrita, E. J. Saturation‐Transfer Difference (STD) NMR:
A Simple and Fast Method for Ligand Screening and Characterization of Protein Binding. J.
Chem. Educ. 88, 990–994 (2011).
22. Diercks, T. et al. Fluorinated carbohydrates as lectin ligands: versatile sensors in 19F‐detected saturation transfer difference NMR spectroscopy. Chemistry. 15, 5666–5668 (2009).
23. Coates, L. et al. The catalytic mechanism of an aspartic proteinase explored with neutron and X‐ray diffraction. J. Am. Chem. Soc. 130, 7235–7237 (2008).
24. Davies, D. R. The structure and function of the aspartic proteinases. Annu. Rev. Biophys.
Biophys. Chem. 19, 189–215 (1990).
25. Coates, L., Erskine, P. T., Wood, S. P., Myles, D. A. A. & Cooper, J. B. A neutron laue diffraction study of endothiapepsin: Implications for the aspartic proteinase mechanism. Biochemistry.
40, 13149–13157 (2001).
26. Coates, L., Erskine, P. T., Crump, M. P., Wood, S. P. & Cooper, J. B. Five Atomic Resolution Structures of Endothiapepsin Inhibitor Complexes: Implications for the Aspartic Proteinase Mechanism. J. Mol. Biol. 318, 1405–1415 (2002).
27. Pearl, L. & Blundell, T. The active site of aspartic proteinases. FEBS Lett. 174, 96–101 (1984).
28. Köster, H. et al. A small nonrule of 3 compatible fragment library provides high hit rate of endothiapepsin crystal structures with various fragment chemotypes. J. Med. Chem. 54, 7784–7796 (2011).
29. Gerber, P. R. & Müller, K. MAB, a generally applicable molecular force field for structure modelling in medicinal chemistry. J. Comput. Aided. Mol. Des. 9, 251–268 (1995).
30. Augustin, B. G. S. LeadIT 2.1.3. at <http://www.biosolveit.de>
31. Schrödinger LLC. The PyMOL Molecular Graphics System, Version 1.7.2. (2010). at
<http://www.pymol.org>
32. McDonald, C. et al. The N‐iodosuccinimide‐mediated conversion of aldehydes to methyl esters. J. Org. Chem. 54, 1213–1215 (1989).
33. Sharma, A. K., Subramani, A. V. & Gorman, C. B. Efficient synthesis of halo indanones via chlorosulfonic acid mediated Friedel–Crafts cyclization of aryl propionic acids and their use in alkylation reactions. Tetrahedron. 63, 389–395 (2007).
34. Zhou, D. et al. 2‐(Pyrrolidin‐1‐yl)ethyl‐3,4‐dihydroisoquinolin‐1(2H)‐one derivatives as potent and selective histamine‐3 receptor antagonists. J. Med. Chem. 55, 2452–2468 (2012).
35. Surry, D. S. & Buchwald, S. L. Dialkylbiaryl Phosphines in Pd‐Catalyzed Amination: A User’s Guide. Chem. Sci. 2, 27–50 (2011).
Supporting Information
All figures in this work featuring endothiapepin are generated using PyMOL.[31]
Library A — Compounds 1 to 8 Modelling methods
The target compounds were first designed inside the protein using Moloc,[29] and their position was optimised whilst keeping the protein stationary. Each acylhydrazone was docked inside the enzyme using the FlexX docking module in the LeadIT suite.[30] Aspartic acid residue 219 was protonated and key water molecules were unrestrained; E/Z isomerism was considered. 30 poses were kept per ligand.
The resulting poses were visually inspected to ensure they bound to the protein like the original fragments, and were otherwise discarded. The most promising pose was used for Hyde scoring[30] in order to obtain a predicted free energy of binding (ΔGHyde).
Colour code for the figures: O: red, N: blue, F: pale cyan, Cprotein: green, Cligand: orange. The solvent‐
accessible surface of the protein is shown in grey. Hydrogen bonds are shown as dashed, yellow lines.
Modelling results
Table S1 depicts the binding energies as predicted by Hyde.
Compound ΔGHyde (kJ/mol)
1 −30
2 −39
3 −21
4 −30
5 −32
6 −39
7 −19
8 −17
Table S1: The binding energies for compounds 1–8, predicted by Hyde.
1: (E)‐N
ΔGHyde =
N'‐(2‐(2‐Ami
−30 kJ/mol
S4 S2
noethyl)‐5‐
S1'
(diethylami
ino)benzyliddene)‐2‐fluo
S
orobenzohyd
S2'
S
drazide
3 S1
2:(E)‐N'
ΔGHyde =
S
'‐(2‐(2‐Amin
−39 kJ/mol
S4 S2
noethyl)‐5‐(
S1'
(diethylamin
no)benzyliddene)‐2‐(2‐fl
S2
luorophenyl
2'
S3
l)acetohydr
3 S1
razide
3: (E)‐N
ΔGHyde =
S2
N'‐(2‐(2‐(2‐A
−21 kJ/mol
S4
Aminoethyl)
S1'
‐5‐(diethylaamino)phen
nyl)ethyliden
S2'
ne)‐2‐fluoro
S3 S
obenzohydra
S1
azide
4: (E)‐N drazide
ΔGHyde =
S
N'‐(2‐(2‐(2‐A
−30 kJ/mol
S4 S2
Aminoethyl)
S1'
‐5‐(diethyla
amino)phennyl)ethyliden
S2
ne)‐2‐(2‐fluo
2'
S3
orophenyl)a
3 S1
acetohy
5: (E)‐2
ΔGHyde =
‐(2‐Aminoe
−32 kJ/mol
S4 S2
thyl)‐5‐(die
S1'
ethylamino)‐‐N'‐(2‐fluorobenzyliden
S
ne)benzohyd
S2'
S
drazide
S3 S1
6: (E)‐2
ΔGHyde =
‐(2‐Aminoe
−39 kJ/mol
S4 S2
thyl)‐5‐(die
S1'
ethylamino)‐
‐N'‐(2‐(2‐fluuorophenyl)
)ethylidene)
S2'
S
)benzohydra
S3 S1
azide
7: (E)‐2
ΔGHyde =
S
‐(2‐(2‐Amin
−19 kJ/mol
S4 S2
noethyl)‐5‐(
S1'
diethylamin
no)phenyl)‐NN'‐(2‐fluoro
S2
obenzylidene
'
S3
e)acetohydr
S1
razide
8: (E)‐2 drazide
ΔGHyde =
S
‐(2‐(2‐Amin
−17 kJ/mol
S4 S2
noethyl)‐5‐(
S1'
diethylamin
no)phenyl)‐NN'‐(2‐(2‐flu
S2
orophenyl)e
'
S3
ethylidene)a
S1
acetohy
Synthe
The plan obtained and S2).
Library d
Scheme S1 with a box night; rt: r
esis
nned and att d starting ma Since not a design).
1: The used syn x; target compo room temperat
tempted syn aterials are s all desired co
nthetic routes f ounds are in b ure; quant.: qu
nthetic route surrounded w ompounds c
or the fluoroph lue; failed reac antitative.
es are depic with a box, a could be obt
henyl compoun ctions are show
cted in the f and target co
tained, a dif
ds. Commercia wn with a red, c
ollowing sch ompounds ar ferent librar
lly obtained co crossed arrow.
hemes. Com re in blue (Sc ry was desig
ompounds are s Abbreviations
mercially cheme S1 gned (see
surrounded : o.n.: over
Scheme S2 obtained c crossed ar
2: The used syn compounds are rrow. Abbreviat
nthetic routes f e surrounded w tions: o.n.: over
for the 2‐(2‐am with a box; tar r night; rt: room
minoethyl)‐5‐(di rget compound m temperature.
ethylamino)be ds are in blue;
.
nzohydrazide c failed reaction
compounds. Co ns are shown w
mmercially with a red,
Methyl 2 N‐Iodosu in dry m stirred i heptane mixture saturate solvent i 7.94 (td, δ= −109.
2‐fluoroben uccinimide (1 methanol (25
n the dark e). Upon com
was extrac ed aqueous
in vacuo yiel , 1H), 7.56 – .50 – −109.7
nzoate (29)[3 1.4 g, 6.22 m 5 mL), and 2 for 2.5 h, a mpletion, wat cted with 50 NaCl solutio ded 29 as a 7.46 (m, 1H 4 (m, 1F).
32]
mmol, 2.4 eq 28 (0.41 mL,
and the rea ter (16 mL) a 0% ether in on, dried ov yellow liquid ), 7.20 (td, 1
) and dried K 0.32 g, 2.5 action was f
and sodium n pentane (4 ver magnesiu d (0.44 g, 2.8 H), 7.14 (ddd
K2CO3 (1 g, 7 6 mmol, 1 e followed by thiosulfate ( 4 ×). The o um sulfate a 85 mmol, 75 d, 1H), 3.93
.24 mmol, 2.
eq) was adde TLC (SiO2; (1.6 g) were
rganic phas and filtered.
5%). 1H‐NMR (s, 3H). 19F‐N
.8 eq) were ed. The mix 20% ethylac added. The se was wash
. Evaporatio R(400 MHz, C NMR(376 MH
dissolved xture was cetate in resulting hed with on of the CDCl3): δ=
Hz, CDCl3)
2‐Fluoro 29 (0.4 g Upon ad the reac (aq) wer and the filtration 2.13 mm 6.84 (td,
obenzohydr g, 2.60 mmo ddition of hyd ction with TL
re added and e resulting m n, and the s mol, 82.5%) a
, 1H), 6.79 –
razide (30)[1 l, 1 eq) was d drazine the s LC (SiO2; 30%
d the metha mixture was olvent was as a bright or
6.69 (m, 1H)
19]
dissolved in solution disc
% ethylacetat nol was rem s sonicated.
removed fro range solid. 1 ), 2.12 (p, 2H
methanol (1 colored. The te in heptan moved in vac Compound om the resu
1H‐NMR(400 H). 19F‐NMR(3
1.7 mL) and N solution was e). Upon co cuo. 1.25 M e ds that did
lting solutio 0MHz, CDCl3) 376 MHz, CD
N2H4∙H2O (1.0 s refluxed fo mpletion a f ethanolic HC
not dissolve n in vacuo y ): δ= 7.39 (td DCl3): δ= −106
05 g, 21 mm or 6 h whilst f few drops of Cl (15 mL) w e were rem
yielding 30 d, 1H), 7.12 (
6.54 (p, 1F).
mol, 8 eq).
following f 2 M HCl as added moved by (329 mg, (tdd, 1H),
Methyl 2 31 (1 g, with a se left to s heptane amount 3H), 3.5 121.56,
2‐(2‐fluorop 6.49 mmol, eptum and a stir overnigh e). Upon com
of pure 32.
8 (s, 2H). 13 121.40, 115.
phenyl)aceta 1 eq) was d acetyl chlorid
t after whic mpletion th
1H‐NMR(40
3C‐NMR(101 .64, 115.43, 5
ate (32)* dissolved in d
de (2.5 mL, ch the conve
e solvent w 0 MHz, CDC MHz, CDCl 52.31, 34.44
dry methano 35.16 mmol ersion was c was evapora Cl3): δ= 7.21
l3): δ= 171.2 4. 19F‐NMR(3
ol (30 mL) in , 5.4 eq) wa checked with ated in vacu
– 7.12 (m, 2 28, 162.41,
76 MHz, CDC
n a round‐bo s added slow h TLC (SiO2; uo which yie
2H), 7.05 – 6 159.96, 131 Cl3): δ= −117
ottom flask e wly. The mix
20% ethyla elded a qua 6.90 (m, 2H) 1.57, 129.22 7.21 – −117.4
equipped xture was cetate in antitative ), 3.61 (s, , 124.26, 46 (m).