Andrew S T Lim1, Isabel M Vincent2, Michael P Barrett2,3, Ian H Gilbert1. 1. Drug Discovery Unit, Wellcome Centre for Anti-Infectives Research, Division of Biological Chemistry and Drug Discovery, University of Dundee, Dundee DD1 5EH, U.K. 2. Glasgow Polyomics, University of Glasgow, Wolfson Wohl Cancer Research Centre, Garscube Campus, Bearsden G61 1QH, U.K. 3. Wellcome Centre for Molecular Parasitology, Institute of Infection, Immunity and Inflammation, University of Glasgow, Glasgow G12 8TA, U.K.
Abstract
The global prevalence of antibacterial resistance requires new antibacterial drugs with novel chemical scaffolds and modes of action. It is also vital to design compounds with optimal physicochemical properties to permeate the bacterial cell envelope. We described an approach of combining and integrating whole cell screening and metabolomics into early antibacterial drug discovery using a library of small polar compounds. Whole cell screening of a diverse library of small polar compounds against Staphylococcus aureus gave compound 2. Hit expansion was carried out to determine structure-activity relationships. A selection of compounds from this series, together with other screened active compounds, was subjected to an initial metabolomics study to provide a metabolic fingerprint of the mode of action. It was found that compound 2 and its analogues have a different mode of action from some of the known antibacterial compounds tested. This early study highlighted the potential of whole cell screening and metabolomics in early antibacterial drug discovery. Future works will require improving potency and performing orthogonal studies to confirm the modes of action.
The global prevalence of antibacterial resistance requires new antibacterial drugs with novel chemical scaffolds and modes of action. It is also vital to design compounds with optimal physicochemical properties to permeate the bacterial cell envelope. We described an approach of combining and integrating whole cell screening and metabolomics into early antibacterial drug discovery using a library of small polar compounds. Whole cell screening of a diverse library of small polar compounds against Staphylococcus aureus gave compound 2. Hit expansion was carried out to determine structure-activity relationships. A selection of compounds from this series, together with other screened active compounds, was subjected to an initial metabolomics study to provide a metabolic fingerprint of the mode of action. It was found that compound 2 and its analogues have a different mode of action from some of the known antibacterial compounds tested. This early study highlighted the potential of whole cell screening and metabolomics in early antibacterial drug discovery. Future works will require improving potency and performing orthogonal studies to confirm the modes of action.
Antibacterial resistance,
an issue under the umbrella term of antimicrobial
resistance (AMR), is becoming a major global public health issue,
affecting both developed and developing countries alike, although
the impact in low and middle income countries is disproportionately
high.[1−3] It has been estimated that the global death toll
associated with AMR could rise to 10 million and the cumulative economic
cost to U.S. $100 trillion by the year 2050 if AMR is not tackled
urgently.[3]Maintaining a healthy
pipeline of new antibiotics is a key strategy
in combating AMR as bacterial resistance against antibiotics is inevitable.
An evaluation of the antibiotic pipeline by the Pew Charitable Trust[4] in September 2017 found that there were 48 antibiotics
in various phases in clinical development. Despite these seemingly
healthy numbers, there are several caveats to this observation. First,
there is the risk of attrition in each phase of the clinical development
of antibiotics.[5] Second, it was observed
that most of the antibiotics in the pipeline were iterations of some
of the major classes of antibiotics such as β-lactams, quinolones,
tetracyclines, and oxazolidinones. This strategy, while maintaining
a stream of new antibiotics, may suffer from the impact of cross-resistance,
in which resistance against one antibiotic confers cross-resistance
to other antibiotics of similar class, either through mutation of
the molecular target or some common means of efflux or drug inactivation
(e.g., β-lactamases).As such, it is vital to maintain
a healthy pipeline of antibiotics
with either new modes of action or new chemical scaffolds. There have
been a very few new classes of antibiotic uncovered since the golden
age of antibiotic discovery (in the 1950s and 1960s),[6] and the introduction of new classes of antibiotics post-2000
(such as oxazolidinones, lipopeptides, and mutilins) can be traced
to their discovery before 1990.[7]A key issue in the development of new antibiotics is obtaining
sufficient compound levels in cells, due to restricted permeation
of the bacterial cell envelope and ejection of compounds through efflux
pumps. The physicochemical properties of compounds play a role in
compound uptake. The wealth of discussions[8−12] on the physicochemical properties of antibiotics
and the comparison to other nonantibiotics highlights the challenges
of designing an antibiotic with favorable physicochemical properties.
The precise physicochemical properties required for cell entry are
poorly understood. However, analysis suggests that antibacterial compounds
tend to be more polar than other drugs,[10,12] with compounds
acting against Gram-negative bacteria tending to be more polar/charged
than those acting against Gram-positive bacteria. As such, the choice
of libraries for screening bacteria is crucial to enhance the odds
of finding hit compounds with the appropriate antibacterial physicochemical
properties.Current antibiotics target a limited set of enzymes
or RNA structures
within key bacterial cellular processes, such as cell wall biosynthesis,
protein biosynthesis, and DNA biosynthesis. Although these targets
are validated, there should be a search for other targets to treat
bacteria resistant to current antibiotics.There are two main
routes to drug discovery, target-based and whole
cell (phenotypic) approaches. Target-based discovery is hindered by
the challenges of compound uptake/efflux, which means that compounds
active in an enzyme assay are not necessarily active against intact
bacteria. Further there is a risk that compounds that inhibit a single
target can be subject to high rates of resistance. Whole cell approaches
can also suffer from low hit rates due to compound uptake/efflux issues;
however, compounds that are active hit essential targets and are able
to accumulate to sufficient levels at those molecular targets. Whole
cell hit discovery requires some degree of target deconvolution to
ensure that whole cell actives are targeting a novel and/or progressible
target. Knowledge of the binding mode of the lead in the protein is
also valuable for compound optimization and to overcome issues such
as poor pharmacokinetics.Whole cell and target-based drug discovery
processes are complementary
and can be merged into a hybrid model (Figure ). In this hybrid model, a library of compounds
is screened for whole cell activity. Hits from the screen are then
expanded to give various analogues to probe for any indicative structure–activity
relationship (SAR). Target deconvolution can then be used to identify
the molecular target. If the target is suitable for development, target-based
approaches can be used in optimization and scaffold hoping. This approach
has the advantage of starting with compounds that are cell penetrant
and which give rise to the desired whole cell response. The identified
molecular targets possess a higher degree of validation, as it has
been demonstrated that a druglike compound acting against this molecular
target can have a desired whole cell effect on the bacteria.
Figure 1
Steps involved
in the hybrid whole cell/target-based drug discovery.
Steps involved
in the hybrid whole cell/target-based drug discovery.We decided to screen against intact whole bacteria with a
diverse
library of small polar compounds,[10] which
should address the physicochemical properties to some extent and being
low molecular weight should cover a wide range of chemical space as
well.[13,14] Furthermore, small polar fragment-like molecules
would facilitate optimization toward lead-like molecules through the
process of adding complexity.[15]We
decided to apply whole cell metabolomics as a way to compare
modes of action, to be able to identify compounds that have a novel
mechanism of action. Metabolomics can be described as a comprehensive
bioanalytical technique to characterize and quantify metabolites of
a biological system.[16] Metabolites reflect
the biochemical status of the biological system. Since biochemical
changes in a biological system lead directly to changes in its whole
cell response, it follows that metabolomics can facilitate the understanding
of the genotype–phenotype link, as changes in metabolite levels
can be ascribed to changes in protein activity, which is linked to
genes as well as to environmental status, where precursor metabolites
are acquired by the cell. As such, the whole cell response upon exposure
to an antibacterial compound is a function of the change in the metabolites.
In some cases, this can be deconvoluted to reveal affected metabolic
pathways and individual target enzymes. This should provide a fingerprint
to allow comparison of the modes of action of different compounds
and potentially, in some cases, ultimately the protein target. Some
examples illustrate the use of metabolomics in this scenario.[17−21]In this paper, we aim to describe the experience of using
our in-house
small, fragment-like polar compound library as the starting point
in early antibacterial drug discovery. First, we aim to see if low-molecular-weight
“fragment-like” compounds are a viable starting point
for a drug discovery program and second can we use metabolomic fingerprints
at this early stage of the discovery process to compare modes of action.
We will present some early results, alongside thoughts and challenges
that arose during the process.
Results and Discussion
Whole Cell Screening of
In-house Library
An in-house
library of small, fragment-like polar molecules comprising of almost
1300 compounds has been described in detail by Ray et al.[22] The library consisted of a diverse set of small
molecules, with physicochemical properties with low molecular weight
[heavy atom count (HAC) 5–18]; low lipophilicity (log D ≤ 2.5 and log P ≤
2.5), and aromatic rings ≤3 (these represent the upper limits;
the median values are much lower). These physicochemical properties
(low molecular weight and high polarity) should increase the chances
of entry into the bacteria.[8] Further, the
low molecular weight means that it is possible to cover a similar
chemical space to libraries with much larger numbers of leadlike or
druglike compounds. The library was curated to remove known structural
alerts (toxicophores, reactives, and pan-assay interference compounds).[22] The library was mainly composed of commercially
obtained fragments but also contained some proprietary compounds.
The whole cell assay was based on measuring the turbidity of bacterial
culture (as OD600nm) as a proxy of growth. The screening
was performed using 384-well plates to enable high-throughput screening.
As an example of Gram-positive bacteria, Staphylococcus
aureus RN4220, which was used in previous in-house
screenings, was utilized in this whole cell assay. Compounds were
initially screened at a single concentration of 600 μM to reflect
the low molecular weight of these small fragment-like molecules.From the single-point whole cell assay, 23 compounds were returned
as hits, defined as having inhibitory activity at three robust standard
deviations from the median. To remove false positives from the set
of hits, a 10-point dose–response assay, with the top concentration
of 600 μM, was performed at the whole cell level to seek any
dose-dependent inhibitory activity and to estimate the potency of
these hits. Of the 23 hits forwarded into the dose–response
assay, two compounds, 1 and 2 (Figure ), gave clear dose–response
curves. Given their small size, they had a relatively low potency
[pEC50 less than 4, where pEC50 = −log10(EC50)], which is not unexpected. However, the
ligand efficiencies (LE – determined as 1.37 × pEC50/heavy atom count) of the key hits were very promising, 0.30
and 0.35. By analogy with fragment-based drug discovery processes,
the small size allows scope for optimization, while retaining ligand
efficiency.
Figure 2
Dose–response curves of 1 and 2, together with the corresponding structures and potency described
as pEC50.
Dose–response curves of 1 and 2, together with the corresponding structures and potency described
as pEC50.Compound 1 has been described in the literature as
EX-527, a SIRT1 inhibitor.[23,24] Literature investigation
revealed compound 2 to be a ligand of the adenosine A2A receptor discovered using fragment-based drug discovery.[25,26] It was also described as a noncompetitive inhibitor of the human
divalent metal transporter DMT1/SLC1 1A2, with the Ki of approximately 20 μM.[27] Interestingly, the pyrazolylpyrimidine scaffold of 2 was described as having fungicidal activity.[28,29] Additionally, 2 was described to be active against Mycobacterium tuberculosis.[30]Hit expansion of 2 was carried out, as 2 has lower molecular weight and lipophilicity than 1 and a greater chemical tractability, thus allowing greater
complexity
to be added during the hit expansion process.
Resynthesis of 2 and Analogues
Compound 2 is a pyrazolylpyrimidine.
The distribution of various tautomeric
forms (pyrimidinone and hydroxypyrimidine) has been discussed[31] using various theoretical methods, albeit in
the gaseous phase. A computational study on complexes of 4-hydroxypyrimidine
with water suggested that the pyrimidinone (4-oxopyrimidine) tautomeric
form predominates over the 4-hydroxypyrimidine tautomeric form at
a 3:1 ratio at room temperature (rt).[32] Nevertheless, the hydroxypyrimidine representation will be used
throughout this paper for consistency purposes and to aid visual comparison
among the analogues synthesized.A key early goal of the project
was to validate the series and to try and increase its activity. The
structure of 2 was amenable to derivatization, to probe
the effect of various substituents at different locations on the scaffold,
to increase potency (Figure ). The choice of substituents for such a study was designed
in a way that a simple change in the substituent can lead to changes
in physicochemical properties such as lipophilicity, hydrogen bond
donor and acceptor count, which will then impact compound activity.
Various synthetic routes were generated to fulfill the various strategies
highlighted in Figure .
Figure 3
Possible strategies in generating analogues of 2.
Possible strategies in generating analogues of 2.First, the pyrazolylpyrimidine compounds were derived
by reacting
the pyrazole of interest with the appropriate pyrimidines. Substituted
2,4-dichloropyrimidines were used in the initial synthesis to generate
the hit compound 2 and other analogues via the 4-chloro
intermediate (for example, 3 and 4) (Scheme ).
Scheme 1
Synthesis of 2 and Analogues from 2,4-Dichloropyrimidine
Derivatives
Synthesis of 2 and Analogues from 2,4-Dichloropyrimidine
Derivatives
Reagent and conditions: (a) 3,5-dimethylpyrazole,
dioxane, microwave (MW) 150 °C, 3 h, 24–35%; (b) 1 M NaOH
(aq), tetrahydrofuran (THF), MW 150 °C, 0.5 h, 68–90%;
(c) appropriate benzyl alcohol, Cs2CO3, MeCN,
reflux, overnight, 36–79%; (d) 1:1 trifluoroacetic acid (TFA)–dichloromethane
(DCM), rt, 1 h, 72%; (e) MeI, K2CO3, dimethylformamide
(DMF), reflux, overnight, 75%.The first step
in Scheme suffered
from regio-isomerism, in which the displacement
by the pyrazole derivative could occur at either the 2-position or
the 4-position of the pyrimidine ring. Using the reaction to generate 3 as an example, it was interesting to note that the formation
of the desired regio-isomer 3 was favored and enhanced
without the presence of a base (Scheme ). However, the isolated yield was compromised by the
difficulty in separating both regio-isomers. The structure of these
regio-isomers was distinguished by performing 1H NMR experiments,
in particular, nuclear Overhauser enhancement spectroscopy (Figures S1–S3).
Scheme 2
Comparison of the
Calculated Regio-isomeric Ratio of the Synthesis
of 3 and 3b by Inclusion of the Base
Calculated ratio was based on
NMR integration.
Comparison of the
Calculated Regio-isomeric Ratio of the Synthesis
of 3 and 3b by Inclusion of the Base
Calculated ratio was based on
NMR integration.Initially, it was envisaged
that the synthesis of the hit compound 2 was carried
out via the 4-methoxybenzyl ether derivative 5 through
a TFA-mediated deprotection reaction. However, it
transpired that direct hydroxylation of the 4-chloro derivatives (3 and 4) was also possible via microwave-assisted
reaction with THF–NaOH solution, thus saving an additional
step in generating analogues of 2. Other benzyl ether
derivatives 6–9 were generated, thus
increasing the diversity of analogues. Additionally, the hit compound 2 was further methylated to give the N-methyl
product 11.This synthetic scheme was extended
to include 2-chloropyrimdine
derivatives as precursors to other analogues with various substitutions
on the 4- and 6-positions of the pyrimidine core (Scheme ). For example, the methoxy
derivatives 14–16 were demethylated
to provide analogues 19–22.
Scheme 3
Synthesis
of Analogues of 2 from 2-Chloropyrimidine
Derivatives
Reagent and conditions: (a) 3,5-dimethylpyrazole,
KOH, 18-crown-6, acetonitrile, room temperature, 1 day, 41–70%;
or MW (80 °C, 15 min for 13; 120 °C, 30 min
for 14), 32–47%; or for 17: 3,5-dimethylpyrazole,
Cs2CO3, dioxane, MW 150 °C, 3 h, 64%; (b)
LiCl, DMF, MW 160 °C, 30 min (3 h for 21), 24–97%;
(c) LiOH–THF, rt, overnight, 87%; (d) NH3 (for 23) or MeNH2 (for 24) or Me2NH (for 25), MeOH, rt, overnight, 99–100%; (e)
Burgess reagent, THF, rt, overnight, 93%.
Synthesis
of Analogues of 2 from 2-Chloropyrimidine
Derivatives
Reagent and conditions: (a) 3,5-dimethylpyrazole,
KOH, 18-crown-6, acetonitrile, room temperature, 1 day, 41–70%;
or MW (80 °C, 15 min for 13; 120 °C, 30 min
for 14), 32–47%; or for 17: 3,5-dimethylpyrazole,
Cs2CO3, dioxane, MW 150 °C, 3 h, 64%; (b)
LiCl, DMF, MW 160 °C, 30 min (3 h for 21), 24–97%;
(c) LiOH–THF, rt, overnight, 87%; (d) NH3 (for 23) or MeNH2 (for 24) or Me2NH (for 25), MeOH, rt, overnight, 99–100%; (e)
Burgess reagent, THF, rt, overnight, 93%.The methyl ester analogue 17 provided another attractive
starting point for rapid hit expansion as it could be readily derivatized.
Hydrolysis provided the free acid 18, while displacement
with simple amines provided various amides 23–25. The primary amide was dehydrated to give the nitrile 26. Note that these simple chemical transformations allowed
us to generate various analogues with different physicochemical properties,
in particular, hydrogen bond donor/acceptor count, which allowed a
good assessment of SAR.The pyrazole moiety was modified by
replacing one of the methyl
groups with a hydroxyl group. This was performed by using aminoguanidine
as the precursor and reacting it with a β-keto ester[33] such as ethyl acetoacetate to give a pyrazolone
derivative 28 (Scheme ). The similar synthetic methodology gave the 4-methylpyrazolone
derivative 29. The β-keto ester precursor 27 was synthesized using a TiCl4-mediated condensation
reaction. The pyrazolone derivatives then underwent the ring formation
reaction to create the desired pyrimidine core (30 and 31).
Scheme 4
Synthesis of 3 and Analogues from Pyrazolone
Derivatives
Reagent and conditions: (a) TiCl4, triethylamine (TEA), DCM, 0 °C to rt, 4 h, 63%; (b)
ethyl acetoacetate (for 28) or 27 (for 29), NaOAc, H2O, overnight (or 7 days for 29), 52–64%; (c) ethyl acetoacetate, NaOEt, EtOH, reflux,
overnight, 28–38%.
Synthesis of 3 and Analogues from Pyrazolone
Derivatives
Reagent and conditions: (a) TiCl4, triethylamine (TEA), DCM, 0 °C to rt, 4 h, 63%; (b)
ethyl acetoacetate (for 28) or 27 (for 29), NaOAc, H2O, overnight (or 7 days for 29), 52–64%; (c) ethyl acetoacetate, NaOEt, EtOH, reflux,
overnight, 28–38%.The pyrazole moiety
was replaced entirely with cyclic saturated
examples (Scheme ).
Three cyclic saturated structures were used as examples: pyrrolidine,
2-methylpyrrolidine, and morpholine. To achieve this, unsubstituted
1H-pyrazole-1-carboximidamide was used as a guanylating
agent, attaching the guanyl group to the nitrogen atom of the cyclic
saturated examples. These guanyl derivatives then underwent the usual
ring formation reaction to give the analogues 35–37.
Scheme 5
Synthesis Analogues of 2 by Replacing
the Pyrazole with
Saturated Ring System
Reagent and conditions: (a) various
amines, TEA, MeCN, reflux, overnight, 74–85%; (b) ethyl acetoacetate,
NaOEt, EtOH, reflux, overnight, 52–72%.
Synthesis Analogues of 2 by Replacing
the Pyrazole with
Saturated Ring System
Reagent and conditions: (a) various
amines, TEA, MeCN, reflux, overnight, 74–85%; (b) ethyl acetoacetate,
NaOEt, EtOH, reflux, overnight, 52–72%.This ring formation strategy was useful in exploring the effect
of substitution on the 5-position of the pyrimidine core. The starting
compound 3,5-dimethyl-1H-pyrazole-1-carboximidamide
can be reacted with either β-keto ester 27 or acrylates
to give analogues 38, 39, and 42 (Scheme ). As with
in Scheme , basic
functional group transformation allowed the generation of analogues
with diverse physicochemical properties. It was interesting to note
that the ethyl ester 39 underwent transesterification
to give 40 upon treatment of methanolic ammonia; thus,
a different route was required to obtain the amide 41. Direct synthesis of 42 gave a side reaction, which
was documented in the literature,[34,35] thus an alternative
route was envisaged in which the amide 41 was dehydrated
to give 42.
Scheme 6
Exploration of 5-Substituted Pyrimidine
Analogues
Reagent and conditions: (a) 27 (for 38) or appropriate acrylates (39 and 42), NaOH, EtOH, 100 °C, overnight, 19–50%;
(b) NH3, MeOH, reflux, overnight, 102%; (c) NH4OH, 75 °C, overnight, 86%; (d) Burgess reagent, THF, rt, overnight,
20%.
Exploration of 5-Substituted Pyrimidine
Analogues
Reagent and conditions: (a) 27 (for 38) or appropriate acrylates (39 and 42), NaOH, EtOH, 100 °C, overnight, 19–50%;
(b) NH3, MeOH, reflux, overnight, 102%; (c) NH4OH, 75 °C, overnight, 86%; (d) Burgess reagent, THF, rt, overnight,
20%.Attempts were also made to change the
pyrimidine core for other
heterocycles. At first, pyridine was used with 2,6-dichloropyridine
derivatives as the starting point. The synthetic scheme (Scheme ) is similar to Scheme , but the hydroxylation
reaction required a copper-catalyzed reaction, modified from Yang
et al.[36] and Wang et al.[37] There were limited number of pyridine examples, reasons
of which will be discussed in the later sections.
Scheme 7
Pyridine Analogues
of 2
Reagent and conditions: (a) 3,5-dimethylpyrazole,
KOH, 18-crown-6, MeCN, MW 80 °C, 1.25–2.25 h, 50–55%;
(b) tetrabutylammonium hydroxide, CuI, 8-hydroxyquinoline 1-oxide,
H2O, MW 180 °C, 3 h, 58–87%.
Pyridine Analogues
of 2
Reagent and conditions: (a) 3,5-dimethylpyrazole,
KOH, 18-crown-6, MeCN, MW 80 °C, 1.25–2.25 h, 50–55%;
(b) tetrabutylammonium hydroxide, CuI, 8-hydroxyquinoline 1-oxide,
H2O, MW 180 °C, 3 h, 58–87%.All in all, these synthetic methodologies provided convenient
methods
to generate a diversity of analogues with varying pharmacophoric and
physicochemical properties (examples shown in Figure S4). Furthermore, these synthetic methods can be applied
in parallel synthesis, thus allowing rapid generation of analogues
for SAR exploration.
Biological Evaluation of 2 and
Its Analogues
The analogues synthesized (excluding intermediates 27–29 and 32–34) were then tested using the 10-point dose–response
whole
cell assay against S. aureus. Here,
we noted the challenges in evaluating biological activity for some
of these analogues. First, due to the low potency nature of small
fragment-like polar molecules, high concentrations were used during
the whole cell screening process. The dose–response assay was
performed with the modified top concentration of 1.8 mM, diluted threefold
across 10 concentration points, to better visualize the curve. As
we have a maximum in-house dimethyl sulfoxide (DMSO) concentration
limit of 1% in whole cell screening assays, the stock concentration
for such compounds was set to 200 mM. Unfortunately, analogues with
substituents on the 5-position of the pyrimidines (for example 16, 20, 39, 40, and 42) were found to be insoluble in DMSO at 200 mM.The
resynthesized 2 retained its inhibitory activity against S. aureus (Table ). Some analogues have shown greater potency of almost
a log fold than the original hit compound, in particular the benzyl
ether analogues. Removal of the substituent groups on the pyrimidine
core of the original hit 2 reduces the potency. The SAR
on the pyrazole core is limited, with no other substituents having
improved activity from the original 3,5-dimethyl substituents. The
replacement of the pyrazole core with the saturated ring system removed
any compound activity. The focused set of analogues that has been
made cover a range of positions in the molecule and are sufficient
to validate the hit and indicate potential ways to optimize the molecule.
The overall SAR can be summarized in Figure .
Table 1
Activity of 2 and Its
Analogues against S. aureus
compound
mean pEC50 (±SD)a
compound
mean pEC50 (±SD)
compound
mean pEC50 (±SD)
2
3.9 (±0.05)
15
NA
31
NA
3
3.8 (±0.07)
16
NSc
35
NA
4
3.2 (±0.05)
17
NA
36
NA
5
4.8 (±0.16)
18
NA
37
NA
6
4.7 (±0.01)
19
3.2 (±0.20)
38
2.8 (±0.03)
7
3.0 (±0.23)
20
NS
39
NS
8
3.6 (±0.70)
21
NS
40
NS
9
4.6 (±0.07)
22
NS
41
NA
10
4.4 (±0.04)
23
NS
42
NS
11
NAb
24
NA
43
NA
12
NA
25
NA
44
NA
13
3.0 (±0.17)
26
NA
45
NA
14
NA
30
NA
46
NA
Expressed
as the mean and its respective
standard deviation from three replicates.
NA = no activity at 1.8 mM.
NS = not soluble in DMSO stock solution
at 200 mM, hence not tested for compound activity.
Figure 4
Summary of the SAR of 2.
Summary of the SAR of 2.Expressed
as the mean and its respective
standard deviation from three replicates.NA = no activity at 1.8 mM.NS = not soluble in DMSO stock solution
at 200 mM, hence not tested for compound activity.As well as the potency of the compounds
against S. aureus, various metrics
were calculated. Ligand
efficiency index (LEI)[38] is normally calculated
for binding to a protein. LEI, defined as pEC50 divided
by heavy atom count (HAC), is the simplification of LE in which the
multiplier constant 1.37 was not applied in the equation. The value
of LEI describes the extent of the improvement of the potency when
more heavy atoms are added to the active compound. Originally used
in the target-based fragment drug discovery, it can also be applied
to whole cell screens where it measures how well each atom interacts
with the molecular target and determines if substituents contribute
significantly to the binding. Lipophilic ligand efficiency (LLE)[38] indicates whether the interactions are driven
by lipophilicity (Table ).
Table 2
Ligand Efficiency Analysis of Active
Compounds
compound
MW
HACa
clog P
mean pEC50
LEIb
LLEc
1d
249
17
3.0
3.6
0.21
0.6
2
204
15
1.0
3.9
0.26
2.9
3
223
15
1.6
3.8
0.24
2.2
4
243
15
2.1
3.2
0.21
1.1
5
324
24
3.1
4.8
0.20
1.7
6
324
24
3.1
4.7
0.20
1.6
7
324
24
3.4
3.0
0.13
–0.4
8
312
23
3.2
3.6
0.16
0.4
9
294
22
2.8
4.6
0.21
1.8
10
225
15
1.2
4.4
0.30
3.2
13
202
15
1.5
3.0
0.20
1.5
19
190
14
0.6
3.2
0.24
2.6
38
218
16
0.5
2.8
0.18
2.3
Heavy atom
count.
Ligand efficiency
index, defined
as pEC50/HAC.
Lipophilic ligand efficiency, defined
as pEC50 – clog P.
Compound 1 was repurchased
and retested for activity.
Heavy atom
count.Ligand efficiency
index, defined
as pEC50/HAC.Lipophilic ligand efficiency, defined
as pEC50 – clog P.Compound 1 was repurchased
and retested for activity.Although compound 5 was the most potent compound,
much of the binding compared to compound 2 was probably
driven by lipophilicity of the benzyl group, as shown by the drop
in LLE. Compound 10, on the other hand, had improved
LEI and LLE compared to compound 2 and is the most efficient
binder. It must be noted that compound 10 contains an
α-chloropyrimidine, which is potentially chemically reactive.
In addition, compound 10 is less lipophilic than compound 5.
Metabolomics Study of 2 and
Its Analogues and Other
Relevant Compounds
The original hit 2 and the
two selected compounds, 5 (highest potency) and 10 (highest LEI), were then subjected to metabolomics to elucidate
any potential mode of action. Additionally, the other hit compound
from the whole cell screening assay, 1, was included
in this metabolomics study. Two approved antibiotics, ciprofloxacin
and trimethoprim, were also included in this study for comparison.The metabolomics study was based on the kill-kinetic studies similar
to those used by Vincent et al.,[17] in which
the organism was exposed to an antibacterial agent and the effect
monitored over time. The nature of the kill kinetics can be affected
by the size of the starting inoculum and the dose of the compound
administered at the beginning of the study.S.
aureus RN6390, which has been
studied for its virulence and pathogenicity, was used, and it showed
similar whole cell response and sensitivity to the RN4220 strain (examples
in Table S1). Despite the subtle differences
between the two strains RN6390 and RN4220, both were classified as
methicillin sensitive and the lineage of these strains was traced
back to NCTC8325 (also known as RN1) strain. This gave a hint of the
whole cell similarity between the two strains and the capability of
these compounds to target metabolic pathways. which are common and
fundamental to these strains. The strain RN6390 has also been described
to be virulent in various infection models[39−41] and capable
of producing α-toxins. On the other hand, the strain RN4220,
although is useful in genetic manipulation studies, may contain mutations,
which may affect virulence factors and strain fitness,[42] and hence the change to the RN6390 strain for
the metabolomics study.The design of the metabolomics study
was based on that of Vincent
et al.[17] The minimal inhibitory concentration
(MIC) of each compound in this metabolomics study was determined by
the microdilution method (Table ).
Table 3
Comparison of MIC Values among Various
Compounds Showing Inhibitory Activity against S. aureus RN6390
compound
MIC (μg/mL)a
1
64
2
32
5
4
10
16
ciprofloxacin
0.25
trimethoprim
2
Determined through the microdilution
method in Luria–Bertani (LB) media.
Determined through the microdilution
method in Luria–Bertani (LB) media.Using 4 times the MIC as the starting dose, the bacteria
were inoculated
and incubated in fresh LB media for 2 h prior to the administration
of the compound, thus coinciding with the exponential growth phase
where bacterial metabolic activity is most active. The metabolome
was sampled at intervals of 90 min to account for S.
aureus doubling time. At later time points, the bacteria
are likely to be undergoing gross changes in metabolites due to cell
death. The earlier time points are more likely to be representative
of the specific mechanism of action.The metabolome extraction
protocol was adapted from that described
by Vincent et al.,[17] by including bead
beating alongside chemical extraction (with 1:3:1 chloroform–methanol–water
as the solvent) to obtain the internal metabolome.[43] Glass beads were used as previously described[43−46] for the extraction of metabolites from S. aureus.Metabolome samples were analyzed by Glasgow Polyomics, University
of Glasgow. The levels of metabolites and the identification of metabolites
were determined based on the methods by Vincent et al.[17] The comparison of metabolite levels was initially
performed using principal component analysis (PCA). A PCA plot is
useful to reduce multidimensional data such as the data gathered in
this study, as each data point consists of the various intensities
of metabolites detected in each sample. By reducing multidimensional
data to two- or three-dimensional data, a PCA plot presents a clear
visual representation to identify any clustering of data, which gives
an indication of similarity with one another. Figure shows the effect of various compounds exhibiting
antibacterial activity on the bacterial metabolome, visualized using
PCA plots.
Figure 5
PCA plot of drug treatments on S. aureus RN6390. Raw intensities were glog transformed and the plot was generated
using the MetaboAnalyst software.[47] Time
points are combined for clarity and one outlier from the treatment
group 5 was removed as it had much lower intensities
across the entire spectrum.
PCA plot of drug treatments on S. aureus RN6390. Raw intensities were glog transformed and the plot was generated
using the MetaboAnalyst software.[47] Time
points are combined for clarity and one outlier from the treatment
group 5 was removed as it had much lower intensities
across the entire spectrum.Ciprofloxacin has previously been shown to cause drastic changes
to the nucleotide pool,[46] and we also noted
similar changes, in particular, in cytidine and guanine pool (Table S2). Trimethoprim inhibits dihydrofolate
reductase (DHFR), but the folate pool was of insufficient abundance
to be identified in the liquid chromatography–mass spectrometry
platform we used. A significant increase in uridine is likely related
to an accumulation of deoxyuridine monophosphate as its methylation
to deoxythymidine monophosphate (not identified) is prevented as DHFR
is inhibited depriving the cells of the key methyl group donor used
in DNA synthesis.The observation that 2 and its
two analogues have
distinct clusters from the rest suggests that these compounds have
the same/similar mode of action. However, although changes of metabolites
were observed upon treatment with 2 and its two analogues,
no metabolic pathways stood out as specific targets for these compounds
based on clear changes in abundance across a single enzyme. This could
indicate that 2 and its analogues act against a target,
which has pleiotropic effects, against several targets, or the compound
could act against targets with metabolites poorly represented in the
database.As ciprofloxacin and trimethoprim target nucleic acid
metabolism
(DNA and folate biosynthesis, respectively), it was interesting to
observe that this general similarity in their mode of action translated
to similar clustering in the PCA plot in Figure . Based on these clustering patterns, it
can be suggested that the modes of action of 2 and its
analogues are distinct compared to ciprofloxacin and trimethoprim.
In addition, the clustering also suggests that compound 1 may have a distinct mode of action from the rest of the compounds
tested. Nevertheless, the results from these early metabolomics studies
must be supported by other orthogonal biological studies to generate
or further strengthen a hypothesis on the mode of action.These
early results presented the challenge of determining the
right moment in an early antibacterial drug discovery program for
these compounds to be subjected to metabolomics study. As these compounds
(excluding marketed antibiotics) are small polar molecules with low
potency, there is always the possibility of their exerting activity
at multiple targets across bacterial metabolism. In this case, during
hit expansion to improve the potency of the analogues, the mode of
action can shift. As seen in Figure , the more potent compound 5 seems to
have a different metabolic fingerprint from its original hit 2. This may indicate that metabolomics can be useful to monitor
whether SAR exploration may lead to a switch in the mechanism of action.
Conclusions
We have introduced the notion of small polar
molecules as an attractive
starting point in antibacterial drug discovery. This requires a well-constructed
library to provide new chemical scaffolds with good physicochemical
properties and novelty to allow efficient chemical expansion and optimization.The whole cell screening process was designed to be amenable to
high-throughput conditions, with straightforward set-up and easy-to-measure
end points. From our in-house whole cell screening assay using our
small polar library, we have identified two promising hit compounds, 1 and 2. The hit expansion of 2 has
been illustrated in detail in this article, with diverse chemistry
to probe the effect of various substitutions on the pyrazoloylpyrimidine
core. This diverse chemistry can be amenable to parallel combinatorial
chemistry to allow quick synthesis of analogues as well. Some initial
SAR can be observed from these analogues synthesized, with compounds
such as 5 and 10 gaining potency from the
original hit. Although the original low-molecular-weight hit had relatively
weak activity (but good ligand efficiency), it was possible to optimize
it to a compound with much greater potency, not much less potent than
trimethoprim, a clinically used antibiotic. As such, we believe that
this series has the potential for further rounds of chemical optimization
to improve potency and physicochemical properties, using ligand efficiency
metrics as indicators. As physicochemical properties of these compounds
are important in ensuring not only optimal entry into the bacteria
but also pharmacokinetic properties, we would suggest using ligand
efficiency metrics such as LEI and LLE to guide the direction of the
chemical optimization.Early metabolomics study indicated that 2 and its
analogues, alongside the other whole cell hit 1, have
distinct mode of actions compared to other antibiotics tested (ciprofloxacin
and trimethoprim), even though the mode of actions for 2 and its analogues (alongside 1) cannot be precisely
determined. It is also interesting to note the shift in the mode of
action as the compound is optimized. This is perhaps not surprising.
A weak, although ligand efficient, low-molecular-weight compound may
inhibit several targets weakly. However, during compound optimization,
the profile of which molecular targets is inhibited and by how much,
may change.Determining the precise mode of action of a molecular
target can
be very challenging. A future study to derive the metabolomic fingerprint
of key antibiotics will be important as a reference to compare to
new compounds would be very valuable. Orthogonal mode of action studies
such as chemical proteomics or resistance generation followed by whole
genome sequencing can also be proposed to support any hypothesis from
this untargeted metabolomics approach.Despite the limitations
discussed, we are optimistic that whole
cell screening coupled with metabolomic can be an attractive route
to finding new starting points with novel modes of action in early
antibacterial drug discovery. The work of Pogliano has shown that
there is a distinct phenotype associated with compounds, which target
particular pathways within bacteria.[48] There
is a scope for extending this current work, by seeing if there is
a correlation between the metabolome fingerprint and the morphological
effect on the bacteria. This whole cell-based platform highlights
the capability for finding new hits with cellular activity and potential
new mode of action. The mode of action can then be validated at the
early stages of hit discovery, which then increases the odds of success
for the drug development process.
Experimental Section
Whole
Cell Screening
The whole cell screening was performed
using the in-house small polar library. The composition of the library
and some illustrated examples of synthesis of such molecules were
described by Ray et al.[22] Compounds from
the library were stamped into each well from column 1 to column 22
of 384-well plates with a Echo 550 Liquid Handler (Labcyte Inc.). S. aureus RN4220 (Division of Molecular Microbiology,
University of Dundee) was used for whole cell screening. All cultures,
which require shaking, were shaken at 200 rpm. An overnight culture
was prepared by culturing a colony in 3 mL (for S.
aureus RN4220) of fresh LB media (Central Technical
Services, University of Dundee) at 37 °C for 12–15 h.
The culture was dispensed into white-walled, clear flat-bottom 384-well
plates (Greiner Bio-One) by a Multidrop Combi Reagent Dispenser (Thermo
Scientific). Plates were read using PHERAstar FS (BMG LABTECH). 50
μL of the culture (prepared by diluting an overnight culture
of S. aureus RN4220 with fresh LB media
at 1:400 ratio) was added to columns of each well, depending on the
type of study: single-point study (columns 1–22 of the plate)
or (columns 1–10 and 13–22 of the plate). The positive
control (gentamicin 0.5 mg/mL) was filled into column 23 (for single-point
study) or columns 11 and 23 (for dose–response study), while
the negative control (no drug) was filled in the remaining columns.
The plates were sealed with air-permeable seals (AeraSeal, Excel Scientific,
Inc.) and incubated still at 37 °C for 6 h. The plate was then
put into a plate reader to measure OD600nm. All calculations
were performed on ActivityBase XE (IDBS) using the raw data obtained
from the plate reader. The raw data correspond to the OD600nm measured on each well. PA positive control is defined as a completely
active compound with 100% inhibitory effect against bacteria. The
negative control is defined as a completely inactive compound with
0% inhibitory effect against bacteria. Hit analysis was performed
and visualized using Vortex (Dotmatics Limited). Graphic visualization
and statistical analysis were performed and visualized using Vortex
and GraphPad Prism 6 (GraphPad Software Inc.).
Chemistry
General Procedures
Chemicals and solvents were purchased
from Aldrich Chemical Co., Acros, TCI, and Fluorochem. Analytical
thin-layer chromatography (TLC) was performed on precoated TLC plates
(layer 0.20 mm silica gel 60 with fluorescent indicator UV 254, from
Merck). Developed plates were air-dried and analyzed under a UV lamp
(UV 254/365 nm) or sprayed with a potassium permanganate solution
to visualize oxidizable substances. Flash column chromatography was
performed using prepacked silica gel cartridges (Teledyne ISCO) using
CombiFlash Rf200i (Teledyne ISCO). Microwave-assisted chemistry was
conducted using a Biotage initiator microwave synthesizer. 1H and 13C NMR spectra were recorded on a Bruker ARX-500
spectrometer (500, 470, and 125 MHz for 1H, 19F, and 13C NMR, respectively). Spectra were acquired using
DMSO-d6, D2O, and CDCl3 as solvents. Chemical shifts (δ) are reported in parts
per million relative to the residual solvent peak as the internal
reference, and coupling constants (J) are reported
in hertz (Hz). Coupling constants are assumed to be H–H coupling
constants unless stated otherwise. The spin multiplicity is reported
as s = singlet, s(b) = broad singlet, d = doublet, t = triplet, q
= quartet, m = multiplet. LC–MS analyses were performed with
either an Agilent HPLC 1100 series connected to a Bruker Daltonics
MicrOTOF or an Agilent Technologies 1200 series HPLC connected to
an Agilent Technologies 6130 quadrupole spectrometer, where both instruments
were connected to an Agilent diode array detector. LC–MS chromatographic
separations were conducted with a Waters X bridge C18 column, 50 mm
× 2.1 mm, 3.5 μm particle size; mobile phase, H2O/MeCN + 0.1% HCOOH, or H2O/MeCN + 0.1% NH3; linear gradient from 80:20 to 5:95 over 3.5 min and then held for
1.5 min; flow rate of 0.5 mL/min. High-resolution electrospray measurements
were performed on a Bruker Daltonics MicrOTOF mass spectrometer. All
assay compounds had a measured purity of ≥95% (by TLC and UV)
as determined using both LC–MS and NMR spectroscopy. Melting
points were determined on the Griffin melting point apparatus and
were uncorrected.
General Procedure A for the Coupling of Chloropyrimidines
and
3,5-Dimethylpyrazoles
Method 1 (Compounds 3, 4, and 17)
To a solution of appropriate chloropyrimidine (1 equiv) in dioxane (0.2
M) was added 3,5-dimethylpyrazole (1.05 equiv) and was subjected to
microwave irradiation at 150 °C for 3 h. The solvent was removed,
and the crude residue was purified using flash column chromatography
(DCM/heptane or EtOAc/heptane) to give the intended product.
Method
2 (Compounds 12–16)
To a
solution of appropriate chloropyrimidine (1 equiv) in acetonitrile
(0.33 M) were added 3,5-dimethylpyrazole (1 equiv), KOH (1 equiv),
and 18-crown-6 (0.1 equiv). The reaction mixture was left to stir
at rt for 1 day or was subjected to microwave irradiation at various
temperatures. The reaction mixture was quenched with 2 M HCl, and
the aqueous layer was extracted with ethyl acetate. The organic layer
was then separated and concentrated in vacuo to give the residue,
which was then purified using flash column chromatography (0–25%
EtOAc in heptane) to give the intended product.
To a solution of 4,6-dichloro-2-(3,5-dimethyl-1H-pyrazol-1-yl)pyrimidine (3) (179 mg, 0.80
mmol) in THF (4 mL) was added 1 M NaOH (4 mL) and then subjected to
microwave irradiation at 150 °C for 30 min. Excess base was neutralized
with 2 M HCl, and the reaction mixture was extracted with ethyl acetate.
The organic phase was separated and concentrated in vacuo. The residue
was then purified using flash column chromatography (0–3% MeOH
in DCM) to give 3 as a white solid (148 mg, 90%): mp
126–126 °C (lit. mp 135–137 °C[49]); 1H NMR (CDCl3, 500 MHz)
δ: 10.40 (s(b), 1H, OH), 6.08 (s, 1H, Ar H), 6.05 (s, 1H, Ar
H), 2.69 (s, 3H, CH3), 2.30 (s, 3H, CH3), 2.27
(s, 3H, CH3); 13C NMR (CDCl3, 125
MHz) δ: 165.1 (C), 161.9 (C), 151.9 (C), 147.5 (C), 143.6 (C),
111.5 (CH), 108.9 (CH), 24.0 (CH3), 15.1 (CH3), 13.6 (CH3); MS (ESI) m/z 205 [M + H]+; HRMS (ESI) m/z: calcd for C10H13N4O+ [M + H]+, 205.1084; found, 205.1091.
Alternative
Procedure Using TFA–DCM
To 2-(3,5-dimethyl-1H-pyrazol-1-yl)-4-((4-methoxybenzyl)oxy)-6-methylpyrimidine
(5) (49 mg, 0.15 mmol) was added excess volume of trifluoroacetic
acid (200 equiv) and an equivalent volume of DCM. The reaction mixture
was left to stir for 1 h at rt. The reaction mixture was evaporated
in vacuo, and the residue was purified by reverse-phase UV-HPLC to
give 2 as a white solid (22 mg, 72%).
To a solution of 4,6-dichloro-2-(3,5-dimethyl-1H-pyrazol-1-yl)pyrimidine (4) (100 mg, 0.41
mmol) in THF (2.05 mL) were added 1 M NaOH (0.82 mL) and additional
water (1.23 mL) and then subjected to microwave irradiation at 150
°C for 30 min. Excess base was neutralized with 2 M HCl, and
the reaction mixture was extracted with ethyl acetate. The organic
phase was separated and concentrated in vacuo. The residue was then
purified using flash column chromatography (DCM) to give 10 as a white solid (63 mg, 68%): mp 132–134 °C; 1H NMR (CDCl3, 500 MHz) δ: 10.50 (s(b), 1H, OH),
6.29 (s, 1H, Ar H), 6.04 (s, 1H, Ar H), 2.69 (s, 3H, CH3), 2.28 (s, 3H, CH3); 13C NMR (CDCl3, 125 MHz) δ: 160.4 (C), 158.8 (C), 153.2 (C), 147.4 (C), 144.3
(C), 112.3 (CH), 109.3 (CH), 15.0 (CH3), 13.6 (CH3); MS (ESI) m/z 225 [M + H]+ (35Cl); 227 [M + H]+ (37Cl); HRMS (ESI) m/z: calcd for
C9H10N4ClO+ [M + H]+, 225.0538; found, 225.0497.
General Procedure B for
the Formation of Benzyl Ether Analogues
(Compounds 5–9)
To a solution
of 4-chloro-2-(3,5-dimethyl-1H-pyrazol-1-yl)-6-methylpyrimidine
(3) (1 equiv) in acetonitrile (0.3 M) were added the
appropriate benzyl alcohol (1.2 equiv) and Cs2CO3 (1.2 equiv). The reaction mixture was then refluxed overnight. The
reaction mixture was then cooled to rt and diluted with ethyl acetate.
The organic phase was then washed with brine, dried, and concentrated
in vacuo. The crude product was then purified using flash column chromatography
(0–50% ethyl acetate in heptane) to give the desired product.
To a solution of 2-(3,5-dimethyl-1H-pyrazol-1-yl)-6-methylpyrimidin-4-ol (2)
(51 mg, 0.25 mmol) and K2CO3 (69 mg, 0.50 mmol)
in DMF (2.77 mL) was added methyl iodide (62 μL, 1.00 mmol).
The reaction was then refluxed overnight. Water was added into the
reaction mixture, and the reaction mixture was washed with ethyl acetate.
The organic phase was then separated, dried, and concentrated in vacuo
to give a residue, which was purified using flash column chromatography
(10–25% ethyl acetate in heptane) to give 11 as
a off-brown solid (41 mg, 75%): 1H NMR (CDCl3, 500 MHz) δ: 6.24 (s, 1H, Ar H), 5.94 (s, 1H, Ar H), 3.34
(s, 3H, NCH3), 2.28 (s, 3H, CH3), 2.22 (s, 3H,
CH3), 2.20 (s, 3H, CH3); 13C NMR
(CDCl3, 125 MHz) δ: 163.1 (C), 162.1 (C), 151.5 (C),
148.3 (C), 142.4 (C), 111.4 (CH), 108.1 (CH), 32.2 (CH3), 23.4 (CH3), 13.6 (CH3), 11.8 (CH3); MS (ESI) m/z 219 [M + H]+; HRMS (ESI) m/z: calcd
for C11H15N4O+ [M + H]+, 219.1240; found, 219.1243.
To a solution of methyl 2-(3,5-dimethyl-1H-pyrazol-1-yl)-6-methylpyrimidine-4-carboxylate (17) (231 mg, 0.94 mmol) in 1:1 THF/H2O (14 mL)
was added lithium hydroxide (79 mg, 1.88 mmol), and the reaction mixture
was allowed to stir overnight at rt. The reaction mixture was concentrated
in vacuo, and the residue was purified using mass-directed HPLC to
give 18 as a white solid (190 mg, 87%): 1H
NMR (CDCl3, 500 MHz) δ: 7.82 (s, 1H, Ar H), 7.07
(s(b), 1H, OH), 6.10 (s, 1H, Ar H), 2.74 (s, 3H, CH3),
2.70 (s, 3H, CH3), 2.35 (s, 3H, CH3); 13C NMR (CDCl3, 125 MHz) δ: 172.3 (C), 164.3 (C),
156.5 (C), 155.5 (C), 152.0 (C), 143.7 (C), 116.2 (CH), 111.0 (CH),
24.6 (CH3), 15.6 (CH3), 13.6 (CH3); MS (ESI) m/z 233 [M + H]+; HRMS (ESI) m/z: calcd
for C11H14N4O2+ [M + H]+, 233.1033; found, 233.1056.
General
Procedure C for the O-Demethylation of Methoxypyrimidines
Using LiCl (Compounds 19–22)
To a solution of the suitable methoxypyrimidines (1 equiv) in DMF
(0.2 M) was added LiCl (3 equiv unless otherwise stated). The reaction
mixture was subjected to microwave irradiation at 160 °C for
30 min (unless otherwise stated). The reaction was then acidified
with dilute HCl. The aqueous layer was then extracted with DCM. The
organic phase was then collected and concentrated in vacuo to give
the residue, which was purified using UV-directed HPLC to yield the
appropriate product.
General procedure C, from 5-chloro-2-(3,5-dimethyl-1H-pyrazol-1-yl)-4-methoxypyrimidine (16) (33
mg, 0.14 mmol) to give 22 as white solid (38 mg, 86%):
mp 151–152 °C; 1H NMR (CDCl3, 500
MHz) δ: 10.76 (s(b), 1H, OH), 8.00 (s, 1H, Ar H), 6.08 (s, 1H,
Ar H), 2.66 (s, 3H, CH3), 2.27 (s, 3H, CH3); 13C NMR (CDCl3, 125 MHz) δ: 157.3 (C), 152.6
(CH), 151.0 (C), 147.0 (C), 143.8 (C), 119.3 (C), 112.0 (CH), 14.9
(CH3), 13.6 (CH3); MS (ESI) m/z 225 [M + H]+ (35Cl); 227
[M + H]+ (37Cl); HRMS (ESI) m/z: calcd for C9H10N4OCl+ [M + H]+, 225.0538; found, 225.0536.
General Procedure D for the Formation of Substituted Amides
(Compounds 23–25)
Methyl
2-(3,5-dimethyl-1H-pyrazol-1-yl)-6-methylpyrimidine-4-carboxylate
(17) (1 equiv) was dissolved in suitable amines (5 equiv)
in methanol, and the reaction mixture was left to stir overnight at
rt. The reaction mixture was concentrated in vacuo to give the intended
product.
To a solution of 2-(3,5-dimethyl-1H-pyrazol-1-yl)-6-methylpyrimidine-4-carboxamide (23) (50 mg, 0.22 mmol) in THF (1 mL) was added Burgess reagent
(258 mg, 1.08 mmol) in two portions, and the reaction mixture was
allowed to stir overnight at rt. Water was added to the reaction,
and the reaction mixture was washed with DCM. The organic phase was
separated, dried, and concentrated in vacuo to give a residue, which
was purified using flash column chromatography (0–50% ethyl
acetate in heptane) to give 26 as a white solid (43 mg,
93%): 1H NMR (CDCl3, 500 MHz) δ: 7.34
(s, 1H, Ar H), 6.10 (s, 1H, Ar H), 2.72 (s, 3H, CH3), 2.69
(s, 3H, CH3), 2.37 (s, 3H, CH3); 13C NMR (CDCl3, 125 MHz) δ: 172.3 (C), 157.5 (C),
152.8 (C), 143.9 (C), 141.6 (C), 120.2 (CH), 115.3 (C), 111.5 (CH),
24.7 (CH3), 15.6 (CH3), 14.0 (CH3); MS (ESI) m/z 214 [M + H]+; HRMS (ESI) m/z: calcd
for C11H12N5+ [M + H]+, 214.1087; found, 214.1101.
Ethyl 2-Methyl-3-oxopropanoate
(27)
To
a stirred solution of ethyl propionate (2.30 mL, 2.04 g, 20 mmol),
methyl formate (3.70 mL, 3.60 g, 60 mmol) in DCM (20 mL) at 0 °C
was added 1.0 M TiCl4 solution in DCM dropwise (40 mL).
After 30 min, trimethylamine (6.97 mL, 5.06 g, 50 mmol) was added
dropwise. After 1 h, the reaction mixture was warmed to rt over the
period of 3 h. The reaction was then quenched with water and extracted
with DCM. The organic phase was then separated, dried, and concentrated
in vacuo to give a residue, which was purified using flash column
chromatography (DCM) to give 27 as oil with fruity odor
(1.63 g, 63%). NMR spectra indicated a mixture of the aldehyde and
enol tautomers with the ratio of approximately 1.5:1. 1H NMR (CDCl3, 500 MHz) δ: Aldehyde tautomer: 9.81
(d, J = 1.5 Hz, 1H, aldehyde H), 4.27 (q, J = 7.2 Hz, 2H, CHCH3), 3.41 (qd, J1 =
7.2 Hz, J2 = 1.5 Hz, 1H, CH), 1.38 (d, J = 7.2 Hz, 3H, CH3), 1.33 (t, J = 7.2 Hz, 3H, CH2CH). Enol tautomer: 11.3 (d, J = 12.5 Hz, 1H,
OH), 7.02 (dq, J1 = 12.5 Hz, J2 = 1.2 Hz, 1H, CH), 4.27 (q, J = 7.2
Hz, 2H, CHCH3), 1.70 (d, J = 1.2 Hz, 3H, CH3), 1.35
(t, J = 7.1 Hz, 3H, CH2CH). 13C NMR (CDCl3, 125 MHz) δ: Aldehyde tautomer: 197.4 (CH), 169.7 [C, estimated
from heteronuclear multiple bond correlation (HMBC)], 61.6 (CH2), 52.7 (CH), 14.2 (CH3), 10.3 (CH3).
Enol tautomer: 172.4 (C, estimated from HMBC), 160.1 (CH), 100.2 (C),
60.4 (CH2), 14.2 (CH3), 12.4 (CH3). NMR spectra in concordance with those reported by El-Mansy et
al.[50]
General Procedure E for
the Formation of Pyrazolone Derivatives
(Compounds 28 and 29)
To a solution
of aminoguanidine hemisulfate (1 equiv) in water (2.5 M) were added
NaOAc (1 equiv) and the appropriate β-keto ester or β-aldehydoester
(1 equiv). The reaction mixture was left to stir at rt. The precipitate
was then filtered and washed with minimum volume of cold water to
give the intended product.
General procedure E, using ethyl 2-methyl-3-oxopropanoate
(27) (381 mg, 2.93 mmol) and stirred for 7 days to give 29 as as off-white solid (214 mg, 52%): 1H NMR
(DMSO-d6, 500 MHz) δ: 10.5 (s(b),
1H, OH), 7.83 (s(b), 3H, NH2 and NH), 7.28 (s, 1H, Ar H),
1.64 (s, 3H, CH3); 13C NMR (DMSO-d6, 125 MHz) δ: 165.8 (C), 154.0 (C), 148.5 (CH),
88.8 (C), 8.0 (CH3); HRMS (ESI) m/z: calcd for C5H9N4O+ [M + H]+, 141.0771; found, 141.0775.
General
Procedure F for the Formation of Carboximidamides (Compounds 32–34)
To a solution of appropriate
amine (1.01 equiv) in acetonitrile (2.3 M) were added triethylamine
(1.01 equiv) and 1H-pyrazole-1-carboximidamide (1
equiv). The reaction mixture was refluxed overnight. The reaction
mixture was then cooled down to rt, and the precipitate was filtered
from the reaction mixture to give the product as a hydrochloride salt.
Pyrrolidine-1-carboximidamide Hydrochloride (32)
General procedure F, using pyrrolidine (132 μL,
113 mg, 1.58 mmol) to give 32 as a light yellow solid
(177 mg, 75%): 1H NMR (DMSO-d6, 500 MHz) δ: 7.35 (s(b), 4× NH), 3.34–3.31 (m,
4H, 2× CH2), 1.92–1.89 (m, 4H, 2× CH2); 13C NMR (DMSO-d6, 125 MHz) δ: 155.1 (C), 47.4 (2× CH2), 25.2
(2× CH2); HRMS (ESI) m/z: calcd for C5H12N3+ [M
+ H]+, 114.1026; found, 114.1042.
General procedure F, using
2-methylpyrrolidine (530 μL,
442 mg, 5.19 mmol) to give 33 as a light yellow solid
(717 mg, 85%): 1H NMR (DMSO-d6, 500 MHz) δ: 7.35 (s(b), 4× NH), 4.09–4.03 (m,
1H, CH), 3.47–3.43 (m, 1H, CH from CH2), 3.28–3.23
(m, 1H, CH from CH2), 2.06–1.88 (m, 3H, CH2 and CH from CH2), 1.69–1.62 (m, 1H, CH from CH2), 1.10 (d, J = 6.3 Hz, 3H, CH3); 13C NMR (DMSO-d6, 125 MHz)
δ: 154.6 (C), 53.9 (CH), 47.3 (CH2), 32.5 (CH2), 22.8 (CH2), 19.2 (CH3); HRMS (ESI) m/z: calcd for C6H14N3+ [M + H]+, 128.1182; found, 128.1193.
Morpholine-4-carboximidamide Hydrochloride (34)
General procedure F, using morpholine (451 μL, 456 mg, 5.23
mmol) to give 34 as a light yellow solid (633 mg, 74%): 1H NMR (DMSO-d6, 500 MHz) δ:
7.66 (s(b), 4× NH), 3.65–3.63 (m, 4H, 2× CH2), 3.44–3.42 (m, 4H, 2× CH2); 13C NMR (DMSO-d6, 125 MHz) δ: 157.1
(C), 65.7 (2× CH2), 45.6 (2× CH2);
HRMS (ESI) m/z: calcd for C5H12N3O+ [M + H]+, 130.0975; found, 130.0975.
General Procedure G for
the Ring Formation of Substituted Pyrimidines
from Carboximidamides (Compounds 30, 31, 33, 36, and 37)
To a solution of the suitable carboximidamides
(1 equiv) in ethanol (0.6 M) were added NaOEt (1 equiv) and ethyl
acetoacetate (1 equiv). The reaction mixture was refluxed overnight.
Work-up and purification described individually in each example gave
the intended product.
General procedure G, from 5-hydroxy-4-methyl-1H-pyrazole-1-carboximidamide (29) (98 mg, 0.70
mmol). The reaction mixture was concentrated in vacuo, and the residue
was purified using mass-directed HPLC to give 31 as an
off-yellow solid (41 mg, 28%): 1H NMR (CDCl3, 500 MHz) δ: 9.19 (s(b), 2H, 2× OH), 7.32 (s, 1H, Ar
H), 5.98 (s, 1H, Ar H), 2.21 (s, 3H, CH3), 1.85 (s, 3H,
CH3); 13C NMR (CDCl3, 125 MHz) δ:
162.4 (C), 161.3 (C), 148.8 (C), 144.8 (C), 132.5 (CH), 109.0 (CH),
98.9 (C), 23.3 (CH3), 6.5 (CH3); MS (ESI) m/z 207 [M + H]+; HRMS (ESI) m/z: calcd for C9H11N4O2+ [M + H]+, 207.0877;
found, 207.0888.
6-Methyl-2-(pyrrolidin-1-yl)pyrimidin-4-ol
(35)
General procedure G, from pyrrolidine-1-carboximidamide
hydrochloride
(32) (173 mg, 1.16 mmol). Water was added to the reaction,
and the mixture was washed with DCM. The organic phase was then separated,
dried, and concentrated in vacuo to give a residue, which was purified
using flash column chromatography (0–5% MeOH/NH3 in DCM) to give 35 as an off-white solid (133 mg, 64%): 1H NMR (CDCl3, 500 MHz) δ: 11.2 (s(b), 1H,
OH), 5.61 (s, 1H, Ar H), 3.60–3.58 (m, 4H, 2× CH2), 2.18 (s, 3H, CH3), 2.03–2.00 (m, 4H, 2×
CH2); 13C NMR (CDCl3, 125 MHz) δ:
167.8 (C), 165.8 (C), 152.3 (C), 99.5 (CH), 46.8 (2× CH2), 25.4 (2× CH2), 24.5 (CH3); MS (ESI) m/z 180 [M + H]+; HRMS (ESI) m/z: calcd for C9H13N3O+ [M + H]+, 180.1131; found,
180.1136.
General procedure G, from 2-methylpyrrolidine-1-carboximidamide
hydrochloride (33) (195 mg, 1.19 mmol). Water was added
to the reaction, and the mixture was washed with DCM. The organic
phase was then separated, dried, and concentrated in vacuo to give
a residue, which was purified using flash column chromatography (0–5%
MeOH/NH3 in DCM) to give 36 as a light brown
solid (167 mg, 72%): 1H NMR (CDCl3, 500 MHz)
δ: 11.0 (s(b), 1H, OH), 5.60 (s, 1H, Ar H), 4.40–4.34
(m, 1H, CH), 3.72–3.67 (m, 1H, CH from CH2), 3.50–3.45
(m, 1H, CH from CH2), 2.17 (s, 3H, CH3), 2.14–1.97
(m, 2H, 2× CH from CH2), 1.75–1.70 (m, 1H,
CH from CH2), 1.24 (d, J = 6.3 Hz, 3H,
CH3); 13C NMR (CDCl3, 125 MHz) δ:
167.8 (C), 165.9 (C), 151.9 (C), 99.3 (CH), 53.6 (CH), 46.8 (CH2), 32.6 (CH2), 24.6 (CH3), 23.2 (CH2), 19.4 (CH3); MS (ESI) m/z 194 [M + H]+; HRMS (ESI) m/z: calcd for C10H15N3O+ [M + H]+, 194.1288; found, 194.1309.
6-Methyl-2-morpholinopyrimidin-4-ol (37)
General
procedure G, from morpholine-4-carboximidamide hydrochloride
(34) (97 mg, 0.59 mmol). Water was added to the reaction,
and the mixture was washed with DCM. The organic phase was then separated,
dried, and concentrated in vacuo to give a residue, which was purified
using flash column chromatography (0–5% MeOH/NH3 in DCM) to give 37 as an off-white solid (59 mg, 52%): 1H NMR (CDCl3, 500 MHz) δ: 12.0 (s(b), 1H,
OH), 5.66 (s, 1H, Ar H), 3.81–3.79 (m, 4H, 2× CH2), 3.76–3.74 (m, 4H, 2× CH2), 2.18 (s, 3H,
CH3); 13C NMR (CDCl3, 125 MHz) δ:
167.7 (C), 166.2 (C), 153.8 (C), 100.8 (CH), 53.6 (CH), 66.5 (2×
CH2), 44.8 (2× CH2), 24.4 (CH3); MS (ESI) m/z 196 [M + H]+; HRMS (ESI) m/z: calcd
for C9H14N3O2+ [M + H]+, 196.1081; found, 196.1094.
General
Procedure H for the Formation of 5-Substituted Pyrimidine
Analogues (Compounds 38, 39, and 42)
To a solution of 3,5-dimethyl-1H-pyrazole-1-carboximidamide
nitrate (1 equiv) in ethanol (0.65 M) were added the appropriate 1,3-dicarbonyls
or acrylates (1.1 equiv) and NaOH (1.1 equiv). The reaction mixture
was stirred overnight at 100 °C. Work-up and purification described
individually in each example gave the intended product.
General procedure H, using ethyl 2-methyl-3-oxopropanoate
(27) (260 mg, 2.00 mmol). The reaction mixture was then
cooled down to rt and concentrated in vacuo to give a residue, which
was then purified with mass-directed HPLC to give 38 as
a white solid (77 mg, 19%): 1H NMR (CDCl3, 500
MHz) δ: 10.6 (s(b), 1H, OH), 7.70 (s, 1H, Ar H), 6.04 (s, 1H,
Ar H), 2.66 (s, 3H, CH3), 2.26 (s, 3H, CH3),
2.09 (s, 3H, CH3); 13C NMR (CDCl3, 125 MHz) δ: 162.0 (C), 151.6 (C), 150.5 (CH), 146.8 (C),
143.3 (C), 121.0 (C), 111.2 (CH), 14.9 (CH3), 13.6 (CH3), 12.8 (CH3); MS (ESI) m/z 216 [M + H]+; HRMS (ESI) m/z: calcd for C10H13N4O+ [M + H]+, 205.1084; found, 205.1111.
General procedure H, using ethyl 2-cyano-3-ethoxyacrylate
(433 mg, 2.56 mmol). The reaction mixture was then cooled down to
rt, and the precipitate was filtered and washed with a minimal volume
of cold ethanol. The precipitate was then redissolved in methanol
and the solution was passed through an ion exchange column to remove
any side products and give 42 as an off-white solid (142
mg, 28%): 1H NMR (CDCl3, 500 MHz) δ: 10.9
(s(b), 1H, OH), 8.31 (s, 1H, Ar H), 6.15 (s, 1H, Ar H), 2.68 (s, 3H,
CH3), 2.30 (s, 3H, CH3); 13C NMR
(CDCl3, 125 MHz) δ: 162.1 (CH), 157.5 (C), 154.7
(C), 150.5 (C), 145.0 (C), 113.6 (C), 113.4 (CH), 15.2 (CH3), 13.7 (CH3); MS (ESI) m/z 216 [M + H]+; HRMS (ESI) m/z: calcd for C10H10N5O+ [M + H]+, 216.0880; found, 216.0900.
Alternative
Procedure Using Burgess Reagent
To a solution
of 2-(3,5-dimethyl-1H-pyrazol-1-yl)-4-hydroxypyrimidine-5-carboxamide
(41) (112 mg, 0.48 mmol) in THF (1 mL) was added Burgess
reagent (286 mg, 1.20 mmol). The reaction mixture was stirred overnight
at rt. The reaction mixture was concentrated in vacuo, and the residue
was purified using mass-directed HPLC to give 42 as a
white solid (21 mg, 20%).
Ethyl 2-(3,5-dimethyl-1H-pyrazol-1-yl)-4-hydroxypyrimidine-5-carboxylate (39) (183 mg, 0.70 mmol) was suspended in NH4OH (3 mL). The
suspension was then stirred at 75 °C overnight, of which the
suspension became clear. The reaction mixture was then cooled to rt
and was concentrated in vacuo to give 41 as a white solid
(140 mg, 86%): 1H NMR (MeOD, 500 MHz) δ: 8.77 (s,
1H, Ar H), 6.04 (s, 1H, Ar H), 2.60 (s, 3H, CH3), 2.24
(s, 3H, CH3); 13C NMR (D2O, 125 MHz)
δ: 174.4 (C), 170.7 (C), 159.8 (CH), 159.5 (C), 151.5 (C), 144.2
(C), 111.9 (C), 110.2 (CH), 14.7 (CH3), 13.5 (CH3); MS (ESI) m/z 234 [M + H]+; HRMS (ESI) m/z: calcd
for C10H12N5O2+ [M + H]+, 234.0986; found, 234.0996.
General
Procedure I for the Coupling of Chloropyridines and
3,5-Dimethylpyrazoles (Compounds 43 and 44)
To a solution of appropriate chloropyridine (1 equiv)
in acetonitrile (0.6 M) were added 3,5-dimethylpyrazole (1 equiv),
KOH (1 equiv), and 18-crown-6 (0.1 equiv). The reaction mixture was
subjected to microwave irradiation at 80 °C. The reaction mixture
was quenched with 2 M HCl, and the aqueous layer was extracted with
ethyl acetate. The organic layer was then separated and concentrated
in vacuo to give the residue, which was then purified using flash
column chromatography (0–10% EtOAc in heptane) to give the
intended product.
General procedure I, from 2,6-dichloropyridine
(444 mg, 3.0 mmol), and subjected to microwave irradiation for 2.25
h. Chromatographic purification gave compound 44 as a
white solid (345 mg, 55%): mp 68–69 °C; 1H
NMR (CDCl3, 500 MHz) δ: 7.83 (d, J = 8.1 Hz, 1H, Ar H), 7.73 (dd, J1 =
7.9 Hz, J2 = 7.9 Hz, 1H, Ar H), 7.16 (d, J = 7.7 Hz, 1H, Ar H), 6.01 (s, 1H, Ar H), 2.69 (s, 3H,
CH3), 2.31 (s, 3H, CH3); 13C NMR
(CDCl3, 125 MHz) δ: 153.1 (C), 150.5 (C), 148.7 (C),
142.2 (C), 140.5 (CH), 120.3 (CH), 113.2 (CH), 109.7 (CH), 14.8 (CH3), 13.6 (CH3); MS (ESI) m/z 208 [M + H]+; HRMS (ESI) m/z: calcd for C10H11N3Cl+ [M + H]+, 208.0636; found, 208.0642.
General Procedure J for the Copper-Catalyzed Hydroxylation of
Aryl Halides (Compounds 45 and 46)
To a solution of the suitable aryl halide (1.0 equiv) in water (0.06
M) were added 55% tetrabutylammonium hydroxide (4.0 equiv), CuI (0.1
equiv), and 8-hydroxyquinoline 1-oxide (0.4 equiv). The reaction mixture
was then subjected to microwave irradiation at 180 °C for 3 h.
The reaction mixture was washed with ethyl acetate, and the organic
phase was separated, dried, and concentrated in vacuo to give a residue,
which was purified using flash column chromatography (0–25%
EtOAc in heptane) to give the desired product.
Compounds tested were obtained
from commercial
suppliers (marketed antibiotics) or in-house (hit compounds and analogues). S. aureus RN6390 was from in-house laboratory collection.
LB were prepared in-house. All procedures were performed under aseptic
conditions.
Determination of MIC via Microdilution Method
A series
of dilution was prepared from a stock solution of compounds (dissolved
in water/DMSO for known antibacterial drugs, DMSO for other synthetic
in-house compounds) by diluting the stock solution with LB media.
From each dilution, 100 μL was added into the wells of a 96-well
plate. At the same time, the inoculum was prepared by making a 0.5
McFarland standard culture by diluting an overnight culture of S. aureus RN6390 with the appropriate media. The
standard culture was then further diluted 100-fold with the same media.
This diluted culture (100 μL) was added into the wells of a
96-well plate, bringing the final volume to 200 μL. The plate
was covered with the lid provided and was incubated at 35–37
°C in air for 18–20 h. The MIC of the compound (averaged
from two independent replicates) was defined as the lowest concentration
of the compound, at which there was no visible growth.
Metabolomics
Study of S. aureus RN6390 upon Exposure
to Various Compounds with Antibacterial Activity
The metabolite
extraction was adapted from Vincent et al.[17] Briefly, a 0.5 mL overnight culture of S. aureus RN6390 was inoculated into 49.5 mL of LB
and the culture was incubated with shaking at 200 rpm at 37 °C.
At 2 h, antibacterial compounds were added at 4× MIC. Samples
were taken at 0, 1.5, and 3 h after the addition in 10 mL aliquots.
Samples were cooled to 4 °C in a dry-ice-ethanol bath to quench
metabolism before they were temporarily stored in ice. Cells (10 mL)
were pelleted at 3894 relative centrifugal force (RCF), washed in
10 mL of cold 0.85% saline, and then resuspended in 1 mL of 0.85%
saline. The OD600nm of the cell suspension was taken and
adjusted to 1. A 1 mL aliquot of cells was pelleted and resuspended
in 200 μL of chloroform–methanol–water mixture
(1:3:1, by volume) in a microcentrifuge tube. Acid-washed glass beads
(200 μL) with the particle size ≤106 μm (Sigma-Aldrich)
were added to the suspension, and the tube was shaken at 2000 rpm
at 4 °C for 45 min (Eppendorf ThermoMixer C). The tube was subjected
to a final centrifugation at 15871 RCF of which the supernatant was
then taken and kept at −80 °C prior to data acquisition.
Nine independent samples were generated for the untreated control,
and three independent samples were generated for each treatment group.
The data acquisition was similar to Vincent et al.[17] A 10 μL aliquot of each sample was run in a randomized
order on a ZIC-pHILIC (polymeric hydrophilic interaction chromatography)
column (Merck SeQuant) coupled to an Orbitrap mass spectrometer (Thermo
Scientific) according to previously published methods.[51] Data analysis was performed using the MzMatch[52] and IDEOM[53] software
packages for untargeted, initial analyses. PiMP[54] was used for further pathway analysis, generating log2 fold changes and peak verification. The PCA plot was generated
using the MetaboAnalyst software.[47] Metabolomics
data have been deposited to the EMBL-EBI MetaboLights database (DOI:
10.1093/nar/gks1004. PubMed PMID: 23109552) with the identifier MTBLS788.
Authors: Ramanan Laxminarayan; Adriano Duse; Chand Wattal; Anita K M Zaidi; Heiman F L Wertheim; Nithima Sumpradit; Erika Vlieghe; Gabriel Levy Hara; Ian M Gould; Herman Goossens; Christina Greko; Anthony D So; Maryam Bigdeli; Göran Tomson; Will Woodhouse; Eva Ombaka; Arturo Quizhpe Peralta; Farah Naz Qamar; Fatima Mir; Sam Kariuki; Zulfiqar A Bhutta; Anthony Coates; Richard Bergstrom; Gerard D Wright; Eric D Brown; Otto Cars Journal: Lancet Infect Dis Date: 2013-11-17 Impact factor: 25.071
Authors: Dan Chen; James C Errey; Laura H Heitman; Fiona H Marshall; Adriaan P Ijzerman; Gregg Siegal Journal: ACS Chem Biol Date: 2012-10-12 Impact factor: 5.100
Authors: Richard A Scheltema; Andris Jankevics; Ritsert C Jansen; Morris A Swertz; Rainer Breitling Journal: Anal Chem Date: 2011-03-14 Impact factor: 6.986
Authors: Melanie Gertz; Frank Fischer; Giang Thi Tuyet Nguyen; Mahadevan Lakshminarasimhan; Mike Schutkowski; Michael Weyand; Clemens Steegborn Journal: Proc Natl Acad Sci U S A Date: 2013-07-09 Impact factor: 11.205
Authors: Peter C Ray; Michael Kiczun; Margaret Huggett; Andrew Lim; Federica Prati; Ian H Gilbert; Paul G Wyatt Journal: Drug Discov Today Date: 2016-10-26 Impact factor: 7.851
Authors: Yoann Gloaguen; Fraser Morton; Rónán Daly; Ross Gurden; Simon Rogers; Joe Wandy; David Wilson; Michael Barrett; Karl Burgess Journal: Bioinformatics Date: 2017-12-15 Impact factor: 6.937
Figure 1
Figure 2
Figure 3
Scheme 1
Scheme 2
Scheme 3
Scheme 4
Scheme 5
Scheme 6
Scheme 7
Figure 4
Figure 5