An integrated chemical biology approach reveals the mechanism of action of HIV replication inhibitors

Article abstract of DOI:10.1016/j.bmc.2017.03.061

Continuous flow (microfluidic) chemistry was employed to prepare a small focused library of dihydropyrimidinone (DHPM) derivatives. Compounds in this class have been reported to exhibit activity against the human immunodeficiency virus (HIV), but their molecular target had not been identified. We tested the initial set of DHPMs in phenotypic assays providing a hit (1i) that inhibited the replication of the human immunodeficiency virus HIV in cells. Flow chemistry-driven optimization of 1i led to the identification of HIV replication inhibitors such as 1l with cellular potency comparable with the clinical drug nevirapine (NVP). Mechanism of action (MOA) studies using cellular and biochemical assays coupled with 3D fingerprinting and in silico modeling demonstrated that these drug-like probe compounds exert their effects by inhibiting the viral reverse transcriptase polymerase (RT). This led to the design and synthesis of the novel DHPM 1at that inhibits the replication of drug resistant strains of HIV. Our work demonstrates that combining flow chemistry-driven analogue refinement with phenotypic assays, in silico modeling and MOA studies is a highly effective strategy for hit-to-lead optimization applicable to the discovery of future therapeutic agents.

Full text of DOI:10.1016/j.bmc.2017.03.061

Accepted Manuscript
An Integrated Chemical Biology Approach Reveals the Mechanism of Action
of HIV Replication Inhibitors
Nicholas Pagano, Peter Teriete, Margrith Mattman, Li Yang, Beth A. Snyder,
Zhaohui Cai, Marintha L. Heil, Nicholas D.P. Cosford
PII:
S0968-0896(17)30657-0
DOI:
http://dx.doi.org/10.1016/j.bmc.2017.03.061
BMC 13659
Reference:
To appear in:
Bioorganic & Medicinal Chemistry
Received Date:
Revised Date:
Accepted Date:
23 December 2016
25 March 2017
29 March 2017
Please cite this article as: Pagano, N., Teriete, P., Mattman, M., Yang, L., Snyder, B.A., Cai, Z., Heil, M.L., Cosford,
N.D.P., An Integrated Chemical Biology Approach Reveals the Mechanism of Action of HIV Replication Inhibitors,
Bioorganic & Medicinal Chemistry (2017), doi: http://dx.doi.org/10.1016/j.bmc.2017.03.061
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

An Integrated Chemical Biology Approach Reveals the Mechanism
of Action of HIV Replication Inhibitors
a,†
Nicholas Pagano , Peter Teriete , Margrith Mattman , Li Yang , Beth A. Snyder , Zhaohui
a, †
a
a
b
b
b
Cai , Marintha L. Heil , and Nicholas D. P. Cosford
a,*
a
Cancer Metabolism & Signaling Networks Program and Conrad Prebys Center for Chemical
Genomics, Sanford Burnham Prebys Medical Discovery Institute, 10901 N. Torrey Pines Rd., La
Jolla, CA 92037.
b
Southern Research Institute, Drug Development Division, 431 Aviation Way, Frederick, MD
1701.
2
Corresponding Author: *To whom correspondence should be addressed. Tel: 858-646-3100.
†
Fax: 858-795-5225. e-mail: ncosford@sbpdiscovery.org. These authors contributed equally.

ABSTRACT
Continuous flow (microfluidic) chemistry was employed to prepare a small focused library of
dihydropyrimidinone (DHPM) derivatives. Compounds in this class have been reported to
exhibit activity against the human immunodeficiency virus (HIV), but their molecular target had
not been identified. We tested the initial set of DHPMs in phenotypic assays providing a hit (1i)
that inhibited the replication of the human immunodeficiency virus HIV in cells. Flow
chemistry-driven optimization of 1i led to the identification of HIV replication inhibitors such as
1
l with cellular potency comparable with the clinical drug nevirapine (NVP). Mechanism of
action (MOA) studies using cellular and biochemical assays coupled with 3D fingerprinting and
in silico modeling demonstrated that these drug-like probe compounds exert their effects by
inhibiting the viral reverse transcriptase polymerase (RT). This led to the design and synthesis of
the novel DHPM 1at that inhibits the replication of drug resistant strains of HIV. Our work
demonstrates that combining flow chemistry-driven analogue refinement with phenotypic assays,
in silico modeling and MOA studies is a highly effective strategy for hit-to-lead optimization
applicable to the discovery of future therapeutic agents.
Keywords: flow chemistry, microreactors, multistep synthesis, HIV, NNRTI, resistant virus
activity, dihydropyrimidinone

1
. Introduction
In broad terms the process of drug discovery can be separated into two different but well-defined
approaches. The target-based approach consists of identifying a suitable protein target and
screening chemical libraries of small molecules against this target to identify hits that may be
optimized into drugs. The phenotypic drug discovery approach involves screening in cells or
organisms where the target is unknown and the mechanism of action (MOA) of active
compounds is deduced later through a process of deconvolution. More recently the use of
targeted compound libraries based on so-called “privileged” scaffolds that are known to have
biological activity against one target class have been used to screen against other targets to
2
identify hits. All of these approaches are either target-based or phenotype-based and originate
-9
in searches for small molecule therapeutic agents for a specific disease, as discussed in detail by
6
Schenone and colleagues. We have refined a different approach. Our compound-centric
methodology that starts with highly promising scaffolds, identifies a putative therapeutic area
where these scaffolds may prove beneficial, and subsequently identies a molecular target for the
compound action. We have exemplified this novel strategy to generate new classes of bio-active
compounds based on our ability to rapidly generate diverse dihydropyrimidinone (DHPM)
1
0
derivatives.
DHPMs are well known as privileged scaffolds that exhibit favorable therapeutic and
1
1,12
pharmacological properties across a range of bioactivities in approved drugs.
A
comprehensive analysis of published bioactivities of DHPMs with some similarity to our
1
3
derivatives led us to investigate activity against human immunodeficiency virus (HIV),
a
therapeutic target that is amenable to phenotypic screening. HIV is the causative agent of
acquired immunodeficiency syndrome (AIDS), a disease for which there is no cure. Currently,
more than 1.1 million people in the USA are living with HIV infection. The prevalence of

HIV/AIDS is on the rise with about 50,000 new infections each year in the USA alone. The cost
of these new cases is estimated to be around $36.4 billion, comprised of both direct medical
1
4
costs ($6.7 billion) and loss of productivity ($29.7 billion). A comprehensive analysis of the
contemporary cost of HIV healthcare by Gebo and colleagues concluded: “HIV healthcare in the
United States continues to be expensive, with the majority of expenditures attributable to
medications. With improved HIV survival, costs may increase and should be monitored in the
1
5
16
future.” Despite the fact that there are over 25 FDA approved drugs for the treatment of HIV,
the well-documented ability of the virus to acquire resistance against established methods of
treatment requires the development of novel drug classes. There is therefore an ongoing
significant need to investigate probe molecules that could ultimately lead to drugs with the
1
7, 18
capacity to overcome such resistance.
Herein we describe how screening a focused library of small molecule compounds in a
phenotypic assay coupled with flow chemistry-driven hit-to-lead optimization, comprehensive
database and computational modeling led to the prediction and subsequent identification of the
MOA for a novel series of HIV replication inhibitors. Our efficient approach to target
identification and validation may provide the basis for an expanded drug discovery program.
Figure 1. 5-(Thiazol-2-yl)-3,4-dihydropyrimidin-2(1H)-one derivatives.

2
. Results and Discussion
We recently reported a highly efficient continuous flow method for the synthesis of substituted
1
0
dihydropyrimidinone (DHPM) derivatives (1). Our methodology provides access to a scaffold
with favorable drug-like characteristics and allows the generation of highly varied analogues.
The compounds prepared in this way are structurally related to thiazole derivatives such as
compound 2 that were initially disclosed in a patent application and subsequently reported in the
1
scientific literature (Figure 1). We were intrigued by the physical similarity between our
3
DHPM derivatives and the related structures 2, which were described to possess inhibitory
13
activity toward HIV replication in cells, although no MOA has been reported. We therefore
designed a biological testing funnel to narrow down the target of the DHPMs (Figure 2). We
initiated target identification studies by comparing antiviral activity in two commonly used
phenotypic screens. The MAGI-CCR5 antiviral assay is a single round antiviral assay employing
the MAGI-CCR5 reporter cell line that produces low levels of virus. In this assay, the virus
produced in an infected cell does not infect neighboring cells. MAGI-CCR5 cells can therefore
be used to assess the MOA of compounds that are active early in the replication cycle, such as
entry or viral reverse transcriptase polymerase (RT) inhibitors. We utilized peripheral blood
1
9
mononuclear cells (PBMC) in a multi-round assay as the comparator assay. In contrast to a
single-round assay, in the PBMC assay the neighboring cells are infected by the progeny virions
produced from the initial infection. Using control compounds as a reference, a difference of
greater than two-fold activity was required in order to convince us that the activity was localized
to early steps in the replication cycle. The result of this dual assay approach allowed for the
confirmation of antiviral activity while already narrowing the MOA. We tested an initial set of
analogues in cellular antiviral assays measuring HIV replication alongside cytotoxicity prior to

advancing to more in-depth and costly secondary studies to identify the MOA. The data
generated from the first set of compounds prompted us to investigate the SAR in more detail
utilizing the already established phenotypic assay and led to a new series of potent HIV
replication inhibitors with drug-like properties. The most promising compound was analyzed
relative to known HIV inhibitors and a hypothesis was generated that the DHPM target was the
HIV RT. Molecular modeling studies along with target specific biological assays confirmed this
hypothesis. Our approach to the discovery, rapid SAR and characterization of a series of new
HIV replication inhibitors is described in detail below.
Fig 2. Testing funnel summarizing the
biological assays employed in our study.
2
.1. Library construction
The multistep microfluidic synthesis of DHPM analogues 1is shown in Scheme 1 . At
1
0
the outset of our experiments, we elected to first vary the ꢀ-bromoketones 4 to investigate the
1
2
effects of changing the thiazole R substituent while keeping the DHPM R substituent constant
2
(
i.e. R = 3-hydroxyphenyl). Thus, 0.75 M solutions of thioamide 3 and ꢀ-bromoketones 4 in
DMF were pumped (32.5 µL/min) into a 250 µL reactor heated to 150 °C for 3.75 minutes. The
stream exiting the first microreactor containing each newly constructed ketothiazole intermediate
7
was introduced to a separate stream (32.5 µL/min) of 3-hydroxybenzaldehyde 5a and urea (6a)
(
0.90 M, DMF). The combined flow of reactants (97.5 µL/min) was pumped into a 1000 µL

reactor heated to 200 °C to generate new DHPM derivatives (1a-1k). Overall, the continuous
two-chip microfluidic sequence required less than one hour for completion from start to finish
(
injection, reaction, and 1250 µL collection). The yields for the three-step (thiazole
formation/deprotection/Biginelli reaction), two-chip sequence were high (39%-48%) (Table 1)
and provided sufficient quantities of compounds 1a-1k for characterization and evaluation in
cellular antiviral assays (Table 1). Of the eleven entries, only the extremely electron poor ꢀ-
bromoketone 4e (4-trifluoromethylphenyl) failed to furnish product using this methodology and
thus 1e was prepared using a one-pot batch procedure using conditions optimized using our flow
method (Scheme 4). It is notable that the reaction conditions developed for the flow method
translated very successfully into batch mode.

Scheme 1. Microfluidic setup for the synthesis of DHPM analogues 1.
To evaluate the activity of this set of DHPMs, we assessed antiviral activity in the single
round MAGI-CCR5 reporter cells and the PBMC assay. The structures of DHPMs 1a-1k and
their antiviral activity against HIV-1Ba-L replication in both MAGI-CCR5 reporter cells and also
in PBMCs are presented in Table 1. Compounds that dose-dependently reduced MAGI-CCR5 ꢁ-
galactoside activity with a therapeutic index (TI) ꢂ 10 (i.e. IC50 ꢃ 10 ꢄM and TC50 ꢂ 100 ꢄM)
were tested along with inactive and sub-optimally active compounds in the PBMC assay to
confirm activity and the trend observed in the MAGI-CCR5 assay. For these experiments, the
non-nucleoside RT inhibitor (NNRTI) nevirapine (NVP) was used as a reference standard. We
observed that the difference in compound activity in the assays was between 2-4 fold for most

compounds in the series suggesting that the MOA of the DHPM series involved viral pathways
and genes associated with virus entry or reverse transcription.
We were pleased to observe that one of our initial compounds, 1i (4-cyanophenyl),
proved to be a sub-micromolar inhibitor with an IC50 value of 0.44 ± 0.03 ꢄM in the MAGI-
CCR5 assay. In fact, when compared to the other compounds in this initial library, 1i proved to
be far superior with one of the lowest IC90 values (2.08 ± 0.32 ꢄM) as well. In addition, and most
encouraging, 1i maintained high levels of cell viability while exhibiting a satisfactory TI of 231.3
±
15.9. This activity of 1i was confirmed in PBMCs demonstrating specific viral reduction and
limited cell toxicity after an extended exposure in culture. Halogen substitution (entries 6-8) also
yielded fairly potent compounds (1f-1h); however, this was accompanied by increased levels of
cellular toxicity. Interestingly, DHPMs bearing hydrogen-bonding functional groups such as
methoxy and hydroxy (1b and 1c) were poor inhibitors with IC50 values of 66.4 ± 17.7 ꢄM and
6
8.3 ± 3.9 ꢄM, respectively in MAGI-CCR5 and 29.6 ꢄM and 39.9 ± 6.9 ꢄM, respectively in
PBMCs. These results are especially notable because the 4-cyanophenyl moiety present in 1i is
significantly different, both sterically and electronically, from the cyclohexyl ring present at the
corresponding position of compound 2 (Figure 1). Thus, we had discovered a previously
unidentified structural motif that imparts potent HIV inhibitory activity.

Table 1. DHPMs 1a-k and Antiviral Activity Against HIV-1 MAGI-CCR5 Cells or PBMCs.
b
MAGI-CCR5
b
1
R
Yield
a
PBMC
Entry Product
(
%)
IC (µM)
IC (µM) TC (µM)
TI
IC (µM)
IC (µM)
TC (µM)
TI
5
0
90
50
50
90
50
c
4
6
3
5
1
2
3
4
5
1a
1b
1c
1d
1e
79 ± 7.4
66 ± 17.7
68 ± 3.9
31 ± 6.9
5.2 ± 1.6
> 100
> 100
1.3
37 ± 8.0
81 ± 8.7
76 ± 13.1
2.0
c
c
4
4
e
30
e
78
e
91
e
> 100
> 100
1.5
1.3
2.0
7.7
3.0
1.4
4.1
9.7
98 ± 0.8
80 ± 2.4
23 ± 0.5
89 ± 0.6
61 ± 1.7
40 ± 4.2
40 ± 6.9
85 ± 2.0
55 ± 7.4
65 ± 2.9
57 ± 1.7
c
44
16 ± 4.0 48 ± 11.6
H C
3
d
61
5.9 ± 1.3
8.0 ± 4.4
18 ± 4.9
18 ± 6.7
c
6
7
8
9
1f
1g
1h
39
48
3.8 ± 0.6
1.9 ± 0.3
2.3 ± 2.9
50 ± 5.4
11 ± 2.8
15 ± 2.9
60 ± 1.4
55 ± 2.2
60 ± 0.8
16
29
26
58 ± 6.8
65 ± 2.3
56 ± 2.5
7.3
50
16
1.3 ± 0.2 4.1 ± 0.3
3.6 ± 11.6 12 ± 4.4
c
43
46
1i
1j
0.44 ± 0.03 2.1 ± 0.3
98 ± 2.1
55 ± 0.6
> 100
10
223 0.19 ± 0.05 0.66 ± 0.07
84 ± 8.2
47 ± 6.0
NT
442
3.6
NT
c
1
0
1
2
43
45
--
43 ± 1.1
49 ± 2.8
84 ± 0.4
> 100
1.3
2.0
13 ± 1.5
NT
29 ± 1.8
NT
c
1
1
1k
NVP
--
0.04 ± 0.00 0.03 ± 0.00
250 0.06 ± 0.01 0.18 ± 0.03
10
167
IC50 – compound concentration that reduces viral replication by 50%. IC90 – compound concentration that reduces viral replication by 90%. TC50
compound concentration that reduces cell viability by 50%. Therapeutic Index (TI) = TC50/IC50. NT = not tested. NVP = nevirapine. Isolated yields
–
a
b
based on 1250 µL collection volumes and following silica gel chromatography. An average of at least three independent experiments. Flow synthesis
c
1
0, 20 d
e
Prepared using one-pot batch procedure, see Scheme 2 for details. n = 1.
previously reported.

1
Having discovered an R substituent (4-cyanophenyl) that imparts submicromolar potency, as in
2
i, we next focused our efforts on probing the R substituent on the DHPM ring. We therefore
1
adopted the same flow-based synthetic protocol above and shown in Scheme 1 but now with
variation in the aryl aldehyde component 5. In addition to exploring different oxo-substituted
benzaldehyde precursors, we also wished to investigate whether para-cyano substitution was
indeed optimal as well as the effect that N-methylation would have on the potency of our
compounds by using methylurea (6b). Thus, 0.45 M solutions of thioamide 3 and ꢀ-
bromoketones 4i (4-cyanophenyl) or 4l (3-cyanophenyl) in DMF were pumped (32.5 µL/min)
into a 250 µL reactor heated to 150 °C for 3.75 minutes. Then, each newly constructed
ketothiazole intermediate 7 was introduced to a separate stream (32.5 µL/min) of aldehydes 5
3
and either urea (6a, R = H) or methylurea (6b, R = Me) (0.54 M, DMF). The combined flow
3
(
(
97.5 µL/min) was pumped into a 1000 µL reactor heated to 200 °C. Unlike our previous setup
Scheme 1), the concentration of the reaction was lowered due to variable precipitation issues at
0
4
.75 M. The isolated yields for the three-step, two-chip sequence were somewhat modest (22%-
5%) (Table 2), the poorest results observed when using resonance-donating benzaldehydes
(
entries 1, 2, 4, 6, and 9). This decrease in multistep efficiency is most likely a result of less
favorable imine formation in the Biginelli reaction with the electron-rich aldehydes.

The structures of DHPMs 1l-1w and their antiviral activity against HIV-1Ba-L in MAGI-CCR5
reporter cells and PBMCs are presented in Table 2. Overall, most benzaldehyde components
2
leading to para-oxygen substituted R moieties yielded compounds with improved levels of
potency with IC50 values < 0.15 µM (1l, 1m, 1p and 1q). We were pleased to observe that only
one of these compounds (1m) decreased cell viability to any significant extent as measured by
the MAGI-CCR5 TC50 value. 4-Cyanophenyl substitution proved superior to 3-cyanophenyl
substitution, which led to decreased potency and increased toxicity levels (1t and 1u), while
3
methyl substitution at R (6b) also provided much poorer inhibitory activity in cells (1v and 1w).
2
Lastly, these data helped us to develop two hypotheses. The first was that relatively bulky R
substituents were well tolerated. Both 1p (phenoxy) and 1q (benzyloxy) proved to be among our
best compounds up to this point. The second was that, at a minimum, a hydrogen-bond acceptor
is important for attaining increased levels of potency. The molecular modeling studies described
below support both of these ideas.

Table 2. DHPMs 1l-1w and Antiviral Activity Against HIV-1 MAGI-CCR5 Cells or PBMCs.
b
b
Yield
(
2
R
1
R
3
R
MAGI-CCR5
IC (µM) TC (µM)
PBMC
IC (µM) IC (µM) TC (µM)
Entry Product
a
%)
IC (µM)
TI
TI
5
0
90
50
50
90
50
1
2
1l
22
4-CNPh
4-CNPh
H
H
0.04 ± 0.00 0.52 ± 0.15
> 10.00
250 0.04 ± 0.00 0.12 ± 0.02 > 100.00
2500
1m
27
0.08 ± 0.01 0.37 ± 0.07 7.0 ± 1.3
88
0.12 ± 0.02 0.30 ± 0.03 76.6 ± 23.4
638
3
4
1n
1o
35
22
4-CNPh
4-CNPh
H
H
0.38 ± 0.04
1.8 ± 0.4
> 100.00
263 0.56 ± 0.16 1.6 ± 0.5
> 100.00
179
667
0.38 ± 0.06 37.8 ± 31.1 > 100.00
263 0.15 ± 0.01 0.42 ± 0.06 > 100.00
5
6
7
1p
1q
1r
33
30
40
4-CNPh
4-CNPh
4-CNPh
H
H
H
0.12 ± 0.02 0.62 ± 0.04
0.13 ± 0.01 0.75 ± 0.01
> 10.00
> 10.00
> 100.00
83
77
63
0.53 ± 0.08 1.8 ± 0.4
> 100.00
189
476
12
0.21 ± 0.10 0.60 ± 0.30 > 100.00
8.6 ± 3.3 21.0 ± 6.8 > 100.00
1.6 ± 0.3
7.6 ± 0.9
8
9
1s
1t
45
27
4-CNPh
3-CNPh
H
H
2.4 ± 0.6
4.4 ± 0.6
69.8 ± 30.2 > 100.00
42
16
41.2 ± 21.9 > 100.00
> 100.00
2.4
27
40.9 ± 1.5 69.6 ± 4.8
38.1 ± 2.7 51.8 ± 5.9
3.3 ± 1.2 10.2 ± 1.7 87.5 ± 12.5
10
1u
1v
26
35
3-CNPh
3-CNPh
H
8.7 ± 1.0
55.4 ± 2.4
2.4 ± 0.9
6.0
1.8
42
5.0 ± 1.4 15.1 ± 4.1 74.0 ± 16.2
14.5 ± 1.5 59.6 ± 4.6 64.2 ± 10.3
15
11
Me
89.2 ± 0.5
13.8 ± 6.2
> 100.00
> 100.00
4.4
1
2
3
1w
28
--
4-CNPh
--
Me
--
3.2 ± 0.1
8.5 ± 0.1
> 100.00
31
1
NVP
--
0.04 ± 0.00 0.42 ± 0.03 10.0 ± 0.0
250 0.06 ± 0.01 0.19 ± 0.03 10.0 ± 0.0
167
IC50 – compound concentration that reduces viral replication by 50%. IC90 – compound concentration that reduces viral replication by 90%. TC50 – compound
concentration that reduces cell viability by 50%. TI = TC50/IC50. NT = not tested. NVP = nevirapine. Isolated yields based on 1250 µL collection volumes and
a
b
following silica gel chromatography. An average of at least three independent experiments.

2
We next investigated the incorporation of heteroatoms into the aryl substituent R by selecting
the appropriate aldehyde building blocks 5. While we were encouraged by both the potency and
toxicity profiles of our best compounds up to this point, the majority of them remained relatively
hydrophobic and we therefore aimed to improve the drug-like properties of the scaffold.
1
For the microfluidic synthesis of this set of analogues ꢀ-bromoketone 4i (R = 4-cyanophenyl)
was used in combination with various heteroaryl aldehydes 5. While the overall yield for this
continuous sequence was high (34%-46%) (Table 3), only pyridine substitution proved viable
under these conditions. Even at decreased reaction concentrations (0.45 M), extended
heterocyclic substrates failed to yield products (1aa-1ad) often precipitating within the second
microreactor chip. Thus, similar to 1e, these four compounds were prepared using a one-pot,
batch-mode procedure (Scheme 2). Overall, these unfunctionalized heterocycles possessed
relatively good potency, with N-methylindole 1ad reaching the nanomolar range with an IC50
value of 0.74 ꢄM ± 0.17. Furthermore, only one entry displayed any level of cellular toxicity
(
1ac).

Table 3. DHPMs 1x-1ad and Antiviral Activity Against HIV-1 MAGI-CCR5 Cells or PBMCs.
c
c
Yield
%)
2
R
MAGI-CCR5
PBMC
EntryProduct
(
IC (µM)
IC (µM) TC (µM)
TI
IC (µM)
IC (µM)
TC (µM)
TI
5
0
90
50
50
90
50
a
1
2
3
4
1x
1y
34
39
46
2.0 ± 0.5
2.9 ± 0.1
1.4 ± 0.4
> 100.00 > 100.00
> 100.00 > 100.00
> 100.00 > 100.00
50
1.3 ± 0.5
1.5 ± 0.4
4.4 ± 2.1
> 100.00
> 100.00
77
a
a
35
71
26
14.3 ± 5.5
67
125
24
1z
0.80 ± 0.20 34.6 ± 32.7 > 100.00
b
1aa
51
3.8 ± 1.0 98.3 ± 1.7 > 100.00
1.5 ± 0.3 12.7 ± 3.6 > 100.00
2.9 ± 0.9
1.3 ± 0.2
10.5 ± 4.0 68.3 ± 22.1
b
5
6
1ab
1ac
56
67
24
5.5 ± 0.8
> 100.00
77
20
b
38
2.1 ± 0.4 10.4 ± 1.4 50.5 ± 7.3
3.5 ± 1.8
1.1 ± 0.3
10.8 ± 4.8 69.8 ± 16.5
b
7
8
1ad
20
0.7 ± 0.1
3.7 ± 0.7 > 100.00
143
3.2 ± 0.5
> 100.00
91
NVP
--
--
0.04 ± 0.00 0.42 ± 0.03 10.0 ± 0.0 250
0.06 ± 0.01 0.19 ± 0.03 10.0 ± 0.0
167
IC90 – compound concentration that reduces viral replication by 90%. TC50 – compound concentration that reduces cell viability by 50%. TI = TC50/IC50. NVP
a
nevirapine. Isolated yields based on 1250 µL collection volumes and following silica gel chromatography. Prepared using one-pot procedure, see Scheme
b
=
2
c
for details. An average of at least three independent experiments. IC50 – compound concentration that reduces viral replication by 50%.

Although the latest round of cellular data (Table 3) suggested that indole derivatives (1ac and
ad) showed promising HIV inhibitory activity we elected to further investigate additional SAR
1
around the pyridine series. Fortunately, a considerable number of substituted pyridyl aldehydes
are commercially available allowing us to access these structures rapidly. Thus we constructed
DHPMs 1ae-1as all from ꢀ-bromoketone 4i (R1 = 4-cyanophenyl) and the results along with
their antiviral activity against HIV-1 MAGI-CCR5 cells are presented in Table 4. The overall
yield for this three-step, two-chip microfluidic sequence was high (42%-64%) with only six
examples (entries 6, 10-13, 15) requiring a one-pot batch preparation due to precipitation inside
the microchip. Initially, perhaps most impressive is that all 15 pyridyl entries do not display any
measureable levels of cellular toxicity. The combination of the nitrogen heteroatom with either
para alkoxy- (1an ) or aryloxy- (1as) yielded several sub-micromolar inhibitors.

Table 4. DHPMs 1ae-1at and Antiviral Activity Against HIV-1 MAGI-CCR5 Cells or PBMCs.
MAGI-CCR5
c
c
PBMC
Yield
(
2
R
Entry Product
a
%)
IC (µM)
IC (µM)
TC (µM)
TI
IC (µM)
IC (µM)
TC (µM)
TI
5
0
90
50
50
90
50
a
1
1ae
45
1.1 ± 0.1
0.7 ± 0.1
5.4 ± 0.7
> 100.00
> 100
91
0.59 ± 0.13
2.0 ± 0.3
> 100
170
a
2
3
4
1af
1ag
1ah
64
49
59
> 100
> 100
> 100
143
227
1.6
0.53 ± 0.17
0.26 ± 0.04
NT
1.3 ± 0.6
0.83 ± 0.03
NT
> 100
> 100
NT
189
385
a
0.44 ± 0.07 2.6 ± 0.4
63.4 ± 20.0 > 100.00
a
a
5
1ai
47
0.48± 0.08
2.4 ± 0.6
> 100
208
0.38 ± 0.06
0.94 ± 0.03
> 100
263
b
a
6
7
1aj
54
38
2.2 ± 0.5 85.5 ± 14.5
> 100
> 100
46
31
1.5 ± 0.7
4.4 ± 1.4
8.8 ± 6.2
> 100
> 100
67
23
1ak
3.2 ± 1.0
> 100.00
44.5 ± 28.2
a
8
9
1al
42
47
0.86 ± 0.02 5.8 ± 0.4
> 100
> 100
116
286
1.9 ± 0.7
5.7 ± 1.5
99.1 ± 0.9
> 100
52
a
1am
0.35 ± 0.02 1.4 ± 0.3
0.10 ± 0.02 0.58 ± 0.13
0.21 ± 0.09
0.57 ± 0.19
476
833
b
b
10
1an
1ao
48
48
> 10.0
> 100
100
9.6
0.12 ± 0.01
NT
0.43 ± 0.09
NT
> 100
NT
11
10.4 ± 3.0
> 100.00
b
b
a
12
13
14
1ap
1aq
1ar
57
37
54
3.8 ± 0.8 66.0 ± 20.4
> 100
> 100
> 100
26
63
91
2.6 ± 0.6
1.6 ± 0.7
1.2 ± 0.8
7.3 ± 1.2
5.9 ± 1.5
4.6 ± 1.9
> 100
> 100
> 100
39
63
83
1.6 ± 0.3
1.1 ± 0.3
14.4 ± 3.0
7.3 ± 0.8

b
1
5
6
1as
33
0.64 ± 0.04 2.8 ± 0.1
0.39 ± 0.05 1.9 ± 0.3
> 100
> 100
156
1.2 ± 0.3
3.4 ± 0.8
67.5 ± 17.8
56
b
d
d
d
1
1at
--
30
256
250
0.65
2.17
> 100
154
167
NVP
--
--
0.04 ± 0.00 0.42 ± 0.03 10.0 ± 0.0
0.06 ± 0.01
0.19 ± 0.03
10 ± 0.0
IC50 – compound concentration that reduces viral replication by 50%. IC90 – compound concentration that reduces viral replication by 90%. TC50 – compound
concentration that reduces cell viability by 50%. TI = TC50/IC50. NT = not tested. NVP = nevirapine. Isolated yields based on 1250 µL collection volumes and
a
b
c
d
following silica gel chromatography. Prepared using one-pot batch procedure, see Scheme 2 for details. An average of at least three independent experiments. n
1.
=

As mentioned previously, a “one-pot” batch-mode procedure was necessary for the
production of certain target DHPMs 1. Even at reduced concentrations (0.45 M), several reaction
sequences suffered from irregular precipitation inside the second microchip. Thus, we developed
an in-flask batch procedure to access these compounds and this one-pot process is shown in
Scheme 2. Unlike our automated flow procedure, the batch process does not facilitate rapid high-
throughput synthesis. However, the batch method does allow us to conduct transformations using
heterogeneous conditions at much higher concentrations to target specific analogues as needed.
In summary, utilizing alternative methodologies allowed us to quickly access most substitution
patterns of interest.
Scheme 2. “One-pot” batch synthesis of DHPMs 1.
At this stage we had synthesized more than forty DHPMs and tested them extensively in
cellular assays measuring the inhibition of HIV replication and found several analogues with
submicromolar potency. We next tested some of the most promising compounds to assess their
drug-like properties in vitro using absorption, distribution, metabolism, and excretion (ADME)
assays. The results of these studies are shown in Table 5. Permeability was assessed via a parallel
artificial membrane permeability assay that estimates blood brain barrier penetration (BBB-
PAMPA). The results suggested that the DHPMs are likely to exhibit poor CNS permeability in

2
1
vivo. The data for plasma and microsomal stability were very encouraging, with all compounds
exhibiting good to excellent parameters. Taken together, the in vitro ADME data suggest that
compounds in this series are likely to have drug-like properties suitable for in vivo proof-of-
concept studies.

Table 5. In vitro ADME data for selected DHPMs.
2
PAMPAa
Plasmba
Stability (%)
Microsomal
Stability (%)
Entry
Cpd
R
b
(LogPe)
1
1i
-7.84 ± 0.22
88.46 ± 10.73
100.00 ± 8.50
95.97 ± 1.36
89.42 ± 2.93
52.35 ± 0.71
100.00 ± 4.60
63.15 ± 4.43
75.55 ± 7.86
2
3
4
1l
1m
1y
-7.82 ± 0.14
-7.11 ± 0.06
-7.50 ± 0.10
5
1ad
-7.05 ± 0.12
95.63 ± 4.19
62.87 ± 1.40
6
7
8
9
1ai
1ae
1an
1am
-7.40 ± 0.03
-6.24 ± 0.07
-7.25 ± 0.05
-7.82 ± 0.05
91.65 ± 2.56
100.00 ± 6.37
100.00 ± 18.70
98.29 ± 8.08
63.40 ± 0.73
49.38 ± 1.00
69.42 ± 1.94
67.61 ± 1.35
a
Permeability is monitored by measuring the amount of compound that can diffuse through a polar brain
lipid membrane to predict BBB permeability. Percent remaining after incubation for 60 min. at 37 °C.
b

2
.2. Mechanism of Action, Biological Relevance and the In Silico-Guided Design of a New
Inhibitor (1at).
Given the potency, ability to rapidly produce derivatives and identification of unique
structural motifs, we concluded that molecular modeling and secondary biological studies to
identify the direct protein target of the DHPMs were warranted. Comparison of our compounds
with the NNRTIs rilpivirine and etravirine, using in silico modeling, suggested some apparent
similarities in the structures and possible overlap of pharmacophoric features (Figure 3; Table 6).
We therefore sought experimental confirmation by first probing activity against viruses with
mutations in the NNRTI binding pocket followed by assessment of DHPMs binding to the viral
RT.
Compound Rilpivirine
Etravirine
0.776
(
R)-1i
S)-1i
R)-1an
S)-1an
0.807
0.805
0.820
0.813
(
0.796
(
0.780
(
0.798
Table 6. Molecular steric alignment similarity
between the stereoisomers of two DHPM
analogues and the NNRTIs rilpivirine and
1
etravirine in their active conformation
Figure 3. Small molecule steric alignment based on the
active conformation of rilpivirine (green) and
compound 1i (orange) (Discovery Studio 3.0; Accelrys,
San Diego, CA).
(Discovery Studio 3.0; Accelrys, San Diego,
CA).
The advent of highly active anti-retroviral therapy (HAART) for the chronic suppression of
HIV replication in AIDS patients has dramatically increased the mean survival time and
improved the quality of life for individuals infected with HIV. Prophylactic administration of

anti-retroviral therapy has been shown to reduce the risk of acquiring an infection in healthy
2
2-29
individuals.
Furthermore, administration of the anti-retroviral drug nevirapine to pregnant
women intrapartum and neonatally has been shown to successfully prevent transmission to
3
0
infants. In an effort to eliminate new infections in children, this treatment was expanded and it
is now recommended that HIV infected pregnant women are provided lifelong anti-retroviral
therapy as well as postpartum administration of a single dose of nevirapine or AZT from birth
3
1-34
through age 4-6 weeks for the child.
In 2012, the prophylactic administration of anti-
retroviral therapy was approved for the prevention of infection of healthy individuals at high risk
3
5
of contracting the disease. Now more than ever, millions of infected and uninfected people are
relying on lifelong anti-retroviral agents until a sterilizing cure can be identified.
The HIV RT is an RNA-dependent DNA polymerase that transcribes the viral RNA into
double-stranded proviral DNA copy. The RT has three functions: (1) transcription of the viral
RNA into DNA; (2) cleavage of the viral RNA by Ribonuclease H; and (3) transcription of the
second DNA strand. The RT is a heterodimer composed of two subunits p66 and p51. The 440
N-terminal amino acids of the p66 subunit functions as the viral polymerase and the C-terminal
1
20 amino acids function as the ribonuclease. The p51 subunit contains the same 440 N-terminal
amino acids found in the p66 subunit. NNRTIs are allosteric inhibitors of RT polymerization.
Structural studies of RT in complex with NNRTIs reveal that inhibitors bind into a pocket
primarily composed of amino acids residues Lys101, Lys103, Tyr181, Tyr188, Phe227, Trp229,
1
, 36-40
and Tyr318.
NNRTI binding is non-catalytic and non-competitive; therefore, the
combination of the small molecule surface contact and flexibility influence the potency of the
41-44
inhibitors.

Several DHPMs with potent activity in cells and low toxicity were selected for more
extensive biological and in silico modeling studies. For the biological studies, we tested 1l, 1m,
1
ad, 1ae, 1am, 1an and 1ar against A17, a virus that has amino acid variations at positions 103
and 181 (K103N and Y181C) in the NNRTI binding pocket relative to wild type. The asparagine
variation at residue 103 and the cysteine variation at residue 181 are known to alter the non-
nucleoside inhibitor binding-pocket sufficiently to disrupt the binding of NNRTIs such as
nevirapine. Therefore, loss of antiviral activity against the virus would suggest the DHPM was
binding the RT. As shown in Table 7, the K103N and Y181C mutations in the A17 virus greatly
affect if not abolish the antiviral activity of the DHPMs tested in cells.
Table 7. DHPMs and inhibitory activity against HIV wildtype (NL4-3) and A17 (K103N,
Y181C) virus in MAGI-CCR5 cells.
Compound
Virus
NL4-3
A17
IC (µM)
IC (µM)
TC (µM)
TI
765
1.6
5
0
90
50
0.06 ± 0.01
28 ± 1.4
0.05 ± 0.01
3.4 ± 0.1
0.52 ± 0.15
> 100
0.62 ± 0.09
95 ± 2.6
0.62 ± 0.04
24 ± 8.0
5.0 ± 2.3
> 100
1
l
46 ± 6.1
5.5 ± 1.0
> 100
NL4-3
A17
110
1.6
1
m
NL4-3
A17
192
1
1
ad
ae
NL4-3
A17
1.1 ± 0.2
> 100
7.6 ± 0.2
> 100
> 91
NA
1
> 100
NL4-3
A17
0.50 ± 0.14
> 100
3.2 ± 1.0
> 100
> 200
1
1
am
an
ar
NVP
> 100
NL4-3
A17
< 0.32 ± 0.0
> 100
1.9 ± 0.9
> 100
> 313
1
1
> 100
NL4-3
A17
0.86 ± 0.24
> 100
7.7 ± 1.1
> 100
> 116
NA
1
> 100
NL4-3
A17
0.04 ± 0.0
> 10
0.54 ± 0.01
> 10
> 250
1
> 10
IC50 - compound concentration that reduces viral replication by 50%. IC90 - compound concentration that reduces
viral replication by 90%. TC50 - compound concentration that reduces cell viability by 50%. TI = TC50/IC50. NVP =
nevirapine.

For in silico modeling studies, compounds 1i, 1l, 1m, 1y, 1ad, 1ae, 1ai, and 1an were
docked into the NNRTI binding pocket of wild-type (PDB code 3MEC, 3MEE, 2ZD1) as well as
Y181C, K103N (PDB code 4I2Q and 4I2P) mutant forms of the RT to identify other potential
interactions between RT and the DHPMs as well as gain insight into the possible orientation of
4
5
the scaffold (data not shown). The results of these studies indicated that the docked poses of the
DHPMs aligned well with the orientation of
etravirine and rilpivirine in the NNRTI binding
pocket, matching the steric alignment performed
previously (Figure 3; Table 7).
A single chiral center in the DHPM structure
gives rise to two enantiomers in the racemate. In
Figure 4. Small molecule alignment focusing on
order to predict differences in activity of each
pharmacophore features between rilpivirine (green)
and the two stereo-isomers of compound 1i (R =
orange and S = purple; Discovery Studio 3.0;
Accelrys, San Diego, CA). The VDW surface is
shown in transparent gray. Yellow-green sphere:
hydrophobic centroids, including aromatic rings.
Blue spheres: hydrogen-bond donors. Purple
enantiomer we performed another alignment in
silico,
this
time
taking
additional
spheres: hydrogen-bond acceptors. Blue-Purple pharmacophoric features into account. The
spheres: hydrogen-bond acceptors and donors.
differences between (R)-1i and (S)-1i when
aligned with rilpivirine are shown in Figure 4. It
is noteworthy here that the steric as well as
pharmacophoric alignments for the (R)-1i result
Figure 5. Molecular interaction between ripilvirine
or (R)-1i and the NNRTI-site. Green shading
Compound
R)-1i
S)-1i
R)-1an
S)-1an
ReRank Score
-132.1
represents hydrophobic region. Blue shading
represents hydrogen bond acceptor. Cyan arrows
represent hydrogen bonds (donor to acceptor). Grey
parabolas represents accessible surface for large
areas. Broken thick line around ligand shape
indicates accessible surface. Size of residue ellipse
represents the strength of the contact (Table ST3).
(
(
-94.5
(
-136.4
(
-83.1
Table 8. Docking results from the stereoisomers of
compounds 1i and 1an in the NNRTI binding site of
HIV-RT from the rilpivirine structure (PDB code
1
2D distance between residue label and ligand
represents proximity (ICM, Molsoft, San Diego)
3MEE (Molegro MVD, Aarhus, Denmark).

in a near identical conformation. Subsequently, both enantiomers were docked into the NNRTI
binding-site of wild type (WT) reverse transcriptase taken from the rilpivirine PDB submission
(
code 3MEE) (Molegro MVD, Aarhus, Denmark). The resulting poses were ranked based on the
docking score and conformational similarity to the alignment in Figure 4 (Table 8). The results
suggest a clear separation of activity between the two enantiomers, with (R)-1i being strongly
favored. In support of these data the relevant amino acid interactions of ripilvirine compared
with (R)-1i within the non-nucleoside binding site of the HIV RT are shown in Figure 5.
In order to support our mechanistic hypothesis, the activities of selected DHPM derivatives
4
6
were confirmed in a biochemical RT assay using a method described previously (Table 9).
Using a chiral column, the enantiomers of compound 1i were separated and individually tested in
the MAGI-CCR5 cellular assay. Each enantiomer, (R)-1i and (S)-1i, was also evaluated in the
PBMC cellular assay and an HIV RT biochemical assay. Compound (R)-1i was found to inhibit
virus replication and polymerization of the HIV RT in a dose-dependent manner whereas (S)-1i
had no activity (Table 9). These results further support our hypothesis that the MOA of the
DHPMs is through inhibition of the viral RT at the NNRTI site. The data also support the
proposed active conformations for the DHPMs determined by in silico modeling. All together,
these studies demonstrate that, like nevirapine, the MOA of the DHPMs is through inhibition of
the HIV RT and strongly suggest that the NNRTI binding pocket is the site of compound
binding.

Table 9. DHPM enantiomers (R)-1i and (S)-1i and their respective inhibitory activity against
HIV-1 reverse transcriptase (RT).
RT
Biochemical
PBMC Antiviral Assay
Product
IC (µM)
IC (µM)
TC (µM)
TI
IC (µM) ± SD
5
0
90
50
50
Nevirapine
R)-1i
S)-1i
0.04 ± 0.003
0.11 ± 0.05
22 ± 1.5
0.17 ± 0.12
0.42 ± 0.24
77 ± 3.4
> 10
> 100
> 100
> 250
> 909
> 4.5
0.18 ± 0.01
0.60 ± 0.09
Not active
(
(
Further investigation of DHPM poses docked on the crystal structure of WT RT bound to
1 47
etravirine and rilpivirine respectively (PDB code 3MEC, 3MEE; and PDB code 2ZD1 )
highlighted several key interactions observed in almost all calculated poses for 1i, 1l, 1m, 1y,
1
ad, 1ae, 1ai, 1am, and 1an. First, the carbonyl group of the pyrimidinone ring interacts with the
backbone amide of K103 or N103, respectively. Second, the nitrogen in the 5 position of the
pyrimidinone forms an interaction with the backbone carbonyl of residue 103, while the nitrogen
in the 1-position forms an interaction with the side chain carbonyl of E138 from the p51 subunit.
Third, the thiazole ring shows a consistent interaction with the β-methylenes of Y181 and the
non-polar side chain of V179. Fourth, the cyanophenyl ring interacts with the NNRTI binding
pocket at amino acids Y181, Y188, F227, W229 and V106 (Figure S1; see Supplementary Data).
This part of the NNRTI binding pocket can accommodate reasonably bulky groups, such as the
cyclohexyl group in compound 2. However, the planar shape of the phenyl ring in our series of
DHPMs adds an additional ꢅ-stacking interaction with the aromatic sidechain of Y188, while the
polar property of the cyano group in the 4-position can contribute through its interaction with the
solvent space in the DNA binding groove. Comparing this to etravirine and rilpivirine bound to
the same structures it is evident that the bulk of the DHPM binding is driven by ligand to
backbone interactions. Upon closer investigation, we concluded that the NNRTI binding pocket

could be further filled by introduction of a methyl group on the 4–cyanophenyl moiety of the
DHPM. We postulated that this methyl group would interact with the β-methylenes of tyrosine
1
83 and tryptophan 229 enabling the compound to bind to the RT with some possible variations
at the residues in these positions. Based on the in silico modeling studies 3-methyl-4-(2-(6-
methyl-4-(1-methyl-1H-indol-2-yl)-2-oxo-1,2,3,4-tetrahydropyrimidin-5-yl)thiazol-4-
yl)benzonitrile (1at) was synthesized. Thus, as shown in Scheme 3, the preparation of 4-acetyl-3-
methylphenyl trifluoromethanesulfonate from 1-(4-hydroxy-2-methylphenyl)ethan-1-one
proceeded under standard conditions. Introduction of the nitrile was achieved through palladium
catalyzed coupling of zinc cyanide with the triflate to afford 4-acetyl-3-methylbenzonitrile.
Bromination of this material then provided 4-(2-bromoacetyl)-3-methylbenzonitrile which was
employed in the batch synthesis of 1at as shown in Scheme 4. The data for the initial biological
evaluation of 1at are included in Table 4, indicating that the compound exhibits sub-micromolar
potency in both of the primary cellular assays.
Scheme 3. Batch synthesis of 4-(2-bromoacetyl)-3-methylbenzonitrile.
Scheme 4. Batch synthesis of 3-methyl-4-(2-(6-methyl-4-(1-methyl-1H-indol-2-yl)-2-oxo-
1
,2,3,4-tetrahydropyrimidin-5-yl)thiazol-4-yl)benzonitrile (1at).

Activity against NNRTI resistant virus is an important feature of a compound with promise
for further development. Mutations at the two amino acid positions, Y181C and K103N, have
48
been associated with resistance to most NNRTIs. The docking of the DHPM 1at containing the
4
-cyanophenyl moiety into the crystal structures of Y181C, K103N RT bound to rilpivirine and
4
two analogues (Figure S2; see Supplementary Data) (PDB code 3BGR and PDB codes 4I2Q,
7
4
9
I2P ) show that favorable interactions between Y181C, K103N and the DHPM would still be
4
feasible. Consistent with our hypothesis, we found that addition of a methyl group on the
cyanophenyl ring (as in 1at) resulted in activity against NNRTI resistant virus (A17) while
maintaining potency against wild type virus (see Table 10). Importantly, compound 1at exhibits
micromolar activity against the A17 strain whereas nevirapine is essentially inactive. Based on
these results we expanded testing to include cell lines with another, highly clinically relevant,
mutant form of RT (E138A) and were able to detect activity against this form of RT as well
(
Table ST1; see Supplementary Data). It is noteworthy that our approach allowed the rapid
design of analogues with an activity profile matching that of nevirapine, a standard of care drug
5
0
still commonly used in triple combination therapy. . Taken together, the results of these
experiments suggest that the DHPMs are small molecule HIV replication inhibitors with the
potential to treat resistant strains of virus.