Rapid catalyst identification for the synthesis of the pyrimidinone core of HIV integrase inhibitors

Article abstract of DOI:10.1002/anie.201201720

A microscale chemistry improvement engine: A pre-dosed microscale high-throughput experimentation additives platform enables rapid, serendipitous reaction improvement. This platform allowed one chemist to set up 475 experiments and analyze the results using MISER chromatography in a single day, thus resulting in two high-quality catalytic systems for the construction of the title compound 1. Support for a single-electron transfer mechanism was obtained. Copyright

Full text of DOI:10.1002/anie.201201720

Angewandte
Chemie
DOI: 10.1002/anie.201201720
High-Throughput Screening
Rapid Catalyst Identification for the Synthesis of the Pyrimidinone
Core of HIV Integrase Inhibitors**
Ana Bellomo, Nihan Celebi-Olcum, Xiaodong Bu, Nelo Rivera, Rebecca T. Ruck,
Christopher J. Welch, Kendall N. Houk,* and Spencer D. Dreher*
Microscale multiparallel chemistry platform technologies
enable modern synthetic chemists to rapidly apply known
potential solutions for a given synthetic transformation to
new, high-complexity synthetic problems. This paradigm is
typically most successful for solving problems where a rich
history of relevant literature precedents and a thorough
mechanistic understanding of the target reaction already
exist. For example, a Suzuki reaction platform employing an
array of known effective phosphine ligands, reaction solvents,
and inorganic bases can be screened to uncover optimal
substrate-specific reaction conditions with a high probability
of success. However, many synthetic problems arise for which
rationally designed platforms are not available, and cannot be
designed owing to a lack of precedent or understanding.
Nevertheless, microscale high-throughput experimentation
(HTE) tools can be used to maximize the opportunity to
serendipitously improve reaction performance, and several
recent, high-profile reports have highlighted some initial
progress in this area.[1–8]
Herein, we describe a broad-based, microscale additive-
screening platform that was designed to minimize complexity
and cost/time bottlenecks in the discovery of new ways to
improve the performance for a wide variety of reactions. By
using this approach, a single chemist was able to set up and
analyze 475 different reaction conditions in a single day to
identify new catalysts for the preparation of high-value
pyrimidinone heterocycles, which are at the core of HIV
Integrase inhibitors.[9,10] Follow-up investigations of identified
conditions combined with quantum mechanical calculations
led to the proposal of a single-electron transfer (SET)
activation mechanism for this catalysis, thus illustrating how
the pursuit of serendipitous solutions can sometimes lead to
improved understanding.
We envisioned creating a 96-well plate additives platform
containing 95 different, highly practical pre-dosed com-
pounds which might activate a reaction through a variety of
different mechanistic pathways; there was one empty well for
a comparative control reaction. The additives that were
chosen are displayed in Figure S1 of the Supporting Informa-
tion. Ideally, this reaction improvement engine would be
inexpensive and simple to use, while providing results that
would be a robust predictor of scale-up performance. To
minimize reagent cost, we utilized very small 250 mL HPLC-
vial inserts (microvials)[11] as reaction vials, thus permitting
reaction screening at approximately 20 mL of solvent (ca.
1 mg substrate per reaction). This is about 4–5 times smaller
than the 1 mL vials that are the current HTE standard in
industry and academic laboratories, and in our opinion
represents the present lower limit that still enables robust
chemistry development. The platform setup simply requires
weighing out the substrates, reagents, and solvent into a single
vial, then distributing this mixture to the additives plate using
a pipettor, a process that only takes a few minutes to
complete. Finally, after reaction incubation, we envisioned
using MISER chromatography[12,13] for sample analysis, as this
method enables an entire 96-well plate to be evaluated in
about 1 hour, and produces a convenient graphical output that
facilitates hit identification.
We were interested in applying this platform to the
synthesis of 2-substituted-5-hydroxy-6-oxo-1,6-dihydropyri-
midine-4-carboxylic acid (pyrimidinone) derivatives of type
2 (Scheme 1), an important class of compounds with well-
documented biological activity.[9,10,14–16] Currently, the most
effective approach for the synthesis of pyrimidinones employs
a two-step process which involves a Michael addition of an N-
hydroxy amidine to an acetylynic diester with a subsequent
[*] Dr. A. Bellomo
Department of Chemistry, University of Pennsylvania
Philadelphia, PA 19104-6323 (USA)
Dr. N. Celebi-Olcum, Prof. Dr. K. N. Houk
Department of Chemistry and Biochemistry, University of California
Los Angeles, CA 90095 (USA)
E-mail: houk@chem.ucla.edu
Dr. X. Bu, Dr. N. Rivera, Dr. R. T. Ruck, Dr. C. J. Welch,
Dr. S. D. Dreher
Department of Process Chemistry, Merck Research Laboratories
P.O. Box 2000, Rahway, NJ 07065 (USA)
E-mail: spencer_dreher@merck.com
[**] We thank the following individuals from the Merck Research
Laboratories: T. J. Novak for his help in obtaining HRMS data and
Rosemary Marques for assistance with IR. A.B. thanks the NSF
GOALI program for financial support.
Scheme 1. Catalyst screening for the synthesis of pyrimidinone 2a
from Michael adduct 1a using the additive platform. The pyrimidone
core is highlighted in red. A MISER chromatogram for the 96 reactions
run in 1,4-dioxane reveals which additives resulted in significant
product formation (highlighted in green).
Supporting information for this article (general synthetic proce-
dures, 1H and 13C NMR and HPLC data for compounds listed in all
tables, and MISER data) is available on the WWW under http://dx.
doi.org/10.1002/anie.201201720.
Angew. Chem. Int. Ed. 2012, 51, 1 – 5
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
1
These are not the final page numbers!

.
Angewandte
Communications
Table 1: Best leads obtained using the additive plate in 1,4-dioxane.[a]
thermal dissociation/recombination event to afford the key
pyrimidinone scaffold, typically in 15–50% overall yield.[15–20]
This thermal rearrangement requires elevated temperatures
(1408C) to convert both E and Z isomers into product, thus
leading to significant decomposition and low to moderate
yields. To date, there has been no report of using catalysis to
improve the yields of these reactions, or to decrease the
reaction temperatures.
To use the additives-screening platform for this project we
first established, through temperature screening, that the
highest temperature at which no pyrimidinone product 2a
was formed thermally from 1a in a variety of solvents was
608C. The additive screening was carried out by dosing the
substrate 1a (2.5 mmol) and internal standard (0.25 mmol
biphenyl) in 20 mL of five different solvents (o-xylene, 1,4-
dioxane, methanol, NMP and DMF) to five of the 250 mL
additives plates containing 20–50% of the pre-dosed addi-
tives. Each reaction was then heated for 4 hours at 608C.
Upon completion, the reactions were diluted with MeOH and
the five plates were subjected to HPLC MISER analysis (see
the Supporting Information).[12,13] Thus, with this platform
and workflow in hand, 475 different reactions were evaluated
with just 370 mg of material in a single day.
Well
Additive
2 a
AY [%]
B08
B09
B10
C05
C06
C07
C08
C09
C10
C11
D11
H07
H09
H11
FeCl2
FeCl3
[Fe(acac)2]
[NiCl2(PPh3)2]
CuCl
CuCl2
CuBr
CuI
32
35
51
12
48
20
44
29
36
74
34
2
[Cu(acac)2]
dibromo-(1,10-phenanthroline)/CuII
Cu powder <75 micron
dppf
XPhos
8
19
phenanthroline
[a] Wells B03, G08 and G11 indicated mass hits by MISER analysis, but
did not contain the desired product by follow-up HPLC analysis.
acac=acetylacetonate, dppf=1,1’-bis(diphenylphosphanyl)ferrocene,
XPhos=2-dicyclohexylphosphino-2’,4’,6’-triisopropylbiphenyl.
Table 2: Reaction optimization for substrate 1a using the most active
additive on 2.5 mmol scale.
Figure S1 in the Supporting Information shows the
MISER chromatograms for the 96 reactions run in 1,4-
dioxane. The reactions giving product are easily identified
by the presence of a semiquantitative peak at the relevant
mass. As expected, the control well (A01), which contains no
additive, showed no product formation. Rewardingly, product
2a was formed in significant quantities with a number of
practical catalysts (Cu, Fe, Ni) and also several ligands. The
assay yields (AY) of the successful leads were then deter-
mined by conventional quantitative HPLC analysis, with the
best leads tabulated in Table 1.
Entry
Catalyst, loading
T [8C]
Solvent
AY [%] 2
1
2
3, 50 mol%
3, 5 mol%
60
60
1,4-dioxane
1,4-dioxane
74
77
The best catalyst lead, dibromo(1,10-phenanthroline)/CuII
(3; 50 mol%), from the initial screening in 1,4-dioxane
afforded 74% AY of the desired pyrimidinone 2a (under
noncatalyzed conditions at 1408C, 50% AY was obtained for
product 2a). We decided to further optimize this lead using
multiparallel screening in 250 mL vials (Table 2). First, we
found that the loading of 3 could be reduced from 50 mol% to
5 mol% in 1,4-dioxane without impacting the reaction yield
(entries 1 and 2). Subsequently, we found that the dichloro
analogue 4 performed identically to 3 (entry 3), which had
limited commercial availability. We next evaluated several
environmentally friendly solvents as alternatives to 1,4-
dioxane (entries 4–8), and found that 2-methyltetrahydro-
furan (2-MeTHF) and cyclopentylmethyl ether (CPME) were
effective green replacements, with 2-MeTHF affording 2a in
84% AY (entry 8). We next evaluated the effects of temper-
ature and loading with catalyst 4 on the reaction performance
(entries 9–13). Reactions at 608C, 408C, and room temper-
ature, at 5 mol% and 1 mol% catalyst loadings showed that
the best reaction conditions in terms of yield, loading, and
time were using 5 mol% of catalyst for 4 hours at 408C, thus
giving 90% HPLC AY (entry 10). Remarkably, overnight
reactions at room temperature (entry 12) afforded complete
conversion, and hence we had identified a practical catalyst
that was able to reduce the reaction temperature by 1208C.
3
4
5
6
7
8
9
10
11
12
13
4, 5 mol%
4, 5 mol%
4, 5 mol%
4, 5 mol%
4, 5 mol%
4, 5 mol%
4, 1 mol%
4, 5 mol%
4, 1 mol%
4, 5 mol%
4, 1 mol%
60
60
60
60
60
60
60
40
40
RT
RT
1,4-dioxane
CPME
THF
75
79
56
40
42
84
74
90
78
84
70
MeOH
NMP[a]
2-MeTHF
2-MeTHF
2-MeTHF
2-MeTHF
2-MeTHF
2-MeTHF
[a] NMP=N-methylpyrrolidone.
The substrate scope for pyrimidinone formation was next
explored with a series of electronically diverse and pyridine-
containing Michael reaction adducts (1b–g), each prepared
by the method described by Culbertson (Table 3).[20] The
optimized procedure using catalyst 4 produced compounds 2b
and 2c in excellent yields (entries 1 and 2). This catalyst also
promoted the reaction with the remaining substrates 1d–g at
608C, however the yields were not as high (entries 3–8).
To improve the yields of 2d–g, we next evaluated solvent,
temperature, and catalyst loading effects on these substrates
2
www.angewandte.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2012, 51, 1 – 5
These are not the final page numbers!

Angewandte
Chemie
Table 3: Best reaction conditions for the synthesis of compounds 2b–g
using catalyst 4 on 2.5 mmol scale.
Scheme 2. High-yielding reaction conditions for the synthesis of com-
pound 2 f using catalyst 5.
Entry
Substrate
Catalyst loading
T [8C]
Solvent
2
AY [%]
Table 4: Reaction conditions for substrates 1a–g using the two catalysts
discovered by comparing yields obtained in microscale screening and
scale-up results.
1
2
3
4
5
6
7
8
1b
1c
1d
1e
1e
1 f
5 mol%
5 mol%
5 mol%
5 mol%
5 mol%
5 mol%
5 mol%
50 mol%
60
60
60
60
40
60
60
60
2-MeTHF
2-MeTHF
2-MeTHF
2-MeTHF
DCE
2-MeTHF
2-MeTHF
2-MeTHF
94
88
51
16
71
17
51
75
Conditions A[a]
Conditions B[b]
Microvials Scale-up
Microvials
Scale-up
AY [%]
AY [%]
AY [%]
AY [%]
2a
2b
2c
2d
2e
2 f
2g
90
94
88
51
71
17
75
94 (85)
92 (83)
83 (80)
55 (50)
71 (73)
25
quant.
quant.
94
49
41
90 (88)
99 (94)
74 (74)
34 (30)
44 (40)
quant. (91)[c]
43
1g
1g
88
46
and found that the yield of compound 2e could be improved
to 71% by using 5 mol% of 4 in 1,2-dichloroethane (DCE) at
408C (entry 5). Similarly, the yield of compound 2g was
improved to 75% when we increased the loading of catalyst 4
back to 50 mol% (entry 8). However, we were unsuccessful in
improving the yields for compounds 2d and 2 f in this way.[21]
Also present in the initial screen of additives, were several
ligands (dppf, XPhos, and phenanthroline), which gave some
conversion of 1a into 2a. Follow-up experiments revealed
that these ligands gave product slowly and did not turnover as
catalysts. We hoped, however that we might be able to
identify a ligand that could act as an effective catalyst, so we
employed a pre-dosed microscale ligand-screening approach
(again in 250 mL vials) with which we could rapidly evaluate
112 sterically and electronically diverse, mono- and bidentate
achiral phosphine ligands. The phosphine screen was con-
ducted for 4 hours at 608C in total reaction volume of 20 mL
of 1,4-dioxane containing 2.5 mmol of 1a and 5 mol% of
ligand per reaction. In this way a single ligand, 1,1’-(ferroce-
nediyl)phenylphosphine (5) was identified to have significant
catalytic activity. This ligand gave excellent conversion and
78% HPLC AY of 2a at a 5 mol% catalyst loading (see the
Supporting Information).
We then applied this ligand to the synthesis of compounds
2a–g while also evaluating seven solvents, and found that each
substrate had a different solvent that was optimal (see the
Supporting Information). Most importantly, we were able to
identify a new catalyst that afforded 2 f in a very high, 88%
AY (Scheme 2).
With identification of two different, novel catalytic
systems, using 4 and 5, for the synthesis of pyrimidinones
we were interested in determining how these small-scale
screening results would translate to a 0.4 mmol (160 ꢀ) scale.
Indeed, the small-scale assay yields proved to be highly
predictive of scale-up performance, as demonstrated in
Table 4.
77 (75)[c]
[a] Conditions A: 4 (5 mol%), 2-MeTHF, 608C; 2a: 24 h, 408C; 2b: 48 h;
2c: 24 h; 2d: 24 h; 2e: 24 h, DCE, 408C; 2 f: 48 h (56% conv.); 2g: 24 h,
50 mol% additive. [b] Conditions B: 5 (5 mol%), 608C; 2a: 2-MeTHF,
24 h; 2b: DME, 24 h; 2c: NMP, 48 h (74% conv.); 2d: NMP, 48 h (55%
conv.); 2e: NMP, 24 h (95% conv.); 2 f: NMP, 6 h; 2g: NMP, 48 h (87%
conv.). Numbers within parentheses correspond to yields of isolated
products. Control reactions with no catalyst were run at 608C and 408C
and no product formation was observed. [c] 2 f and 2g could not be
purified in good yield by silica gel chromatography because streaking of
the products on the column and therefore the yields of isolated products
shown here correspond to their characterization as benzoates 6 f and 6g
(see the Supporting Information). Compound 2g was always obtained
with variable amounts of impurities.
establish a mechanism for the reaction. It was previously
proposed and supported through density functional calcula-
tions, that the thermal synthesis of the pyrimidinones involves
À
a direct N O cleavage to form a polar radical pair (PRP) with
a substantial preference for recombination.[22] With our new
catalysts, we surmised that the Cu and Fe may be working to
reduce the energy of the transition state through a SET
reduction mechanism. Quantum mechanical calculations[23] at
the B3LYP/6-31G(d) level showed that indeed the dissocia-
tion and recombination of the radical anion should be much
more facile than that of the neutral species (Figure 1).
While copper has a well-documented legacy as a SET
catalyst,[24] we were interested in determining whether the
ligand 5 also proceeded through SET activation or whether it
proceeded through a new mechanism. Experimentally, we
determined that ferrocene itself is not a catalyst in this
reaction, and also that other ferrocenyl ligands (dppf, dippf,
dtbpf) gave product but were significantly less reactive than 5
(see the Supporting Information). The calculated oxidation
potential of the catalyst 5 (E0calc Fc(PPh)+/Fc(PPh) = À0.62 V
in acetonitrile vs. Fc+/Fc) suggests that 5 may be a better
reducing agent than ferrocene itself, thus supporting its role as
a SET catalyst. We intend to study the specific catalytic
With these successful synthetic results identified through
serendipity, we were interested in determining if we could
Angew. Chem. Int. Ed. 2012, 51, 1 – 5
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
3
These are not the final page numbers!

.
Angewandte
Communications
[7] J. Kobayashi, Y. Mori, K. Okamoto, R. Akiyama, M. Ueno, T.
Kitamori, S. Kobayashi, Science 2004, 304, 1305.
[8] O. Trapp, S. K. Weber, S. Bauch, W. Hofstadt, Angew. Chem.
2007, 119, 7447; Angew. Chem. Int. Ed. 2007, 46, 7307.
[9] G. R. Humphrey, P. J. Pye, Y.-L. Zhong, R. Angelaud, D. B.
Askin, P. E. Maligres, D. E. Mancheno, R. A. Miller, R. A.
Reamer, S. A. Weissman, Org. Process Res. Dev. 2011, 15, 73.
[10] V. Summa, A. Petrocchi, F. Bonelli, B. Crescenzi, M. Donghi, M.
Ferrara, F. Fiore, C. Gardelli, O. Gonzalez Paz, D. J. Hazuda
et al., J. Med. Chem. 2008, 51, 5843.
[11] Previous experiments run in our lab using 250 mL vials for
a variety of academic and drug development projects showed
excellent scalability. S. Goble, S. Berritt, S. D. Dreher, unpub-
lished results.
[12] C. J. Welch, X. Gong, W. Schaefer, E. C. Pratt, T. Brkovic, Z.
Prizada, J. F. Cuff, B. Kosjek, Tetrahedron: Asymmetry 2010, 21,
1674.
[13] High-throughput Analysis in the Pharmaceutical Industry (Ed.:
P. G. Wang), CRC Press, 2009.
Figure 1. Energy profile for SET catalysis mechanism. B3LYP/6-31G(d)
free energies calibrated with G3(MP2)B3 (DG, kcalmolÀ1) for the Z
and E isomers (values for E isomers in parenthesis) in the gas phase.
[14] D. E. OꢂBrien, L. T. Weinstock, R. H. Springer, C. C. Cheng, J.
Heterocycl. Chem. 1967, 4, 49.
[15] V. Summa, A. Petrocchi, V. G. Matassa, M. Taliani, R. Laufer,
R. D. Francessco, S. Altmura, P. Pace, J. Med. Chem. 2004, 47,
5336.
[16] E. Wagner, L. Becan, E. Nowakowska, Bioorg. Med. Chem. Lett.
2004, 12, 265.
[17] S. D. Dreher, N. Ikemoto, V. Gresham, J. Liu, P. G. Dormer, J.
Balsells, D. Mathre, T. J. Novak, J. D. I. Armstrong, Tetrahedron
Lett. 2004, 45, 6023.
activity in more detail and to pursue other potential
applications.
Herein, we demonstrated how a pre-dosed, microscale
additives platform can enable a single chemist to evaluate
475 discrete reactions in a single day, thus providing a con-
venient and powerful “improvement engine” for organic
synthesis. Employing this platform and subsequent microscale
optimization techniques rapidly led to the identification of
robust substrate-specific procedures for a range of biologi-
cally important pyrimidinones which are at the core of HIV
Integrase inhibitors. We envision that similar, generally useful
platform tools will soon become more widely available, thus
dramatically impacting chemistry development and enabling
increased access to chemical diversity and lower-cost syn-
thesis. Most importantly, we believe that such platforms will
lead to the discovery of new and potentially useful chemical
reactivity and reaction mechanisms.
[18] V. A. Boyd, J. Mason, P. Hanumesh, J. Price, C. J. Russell, T. R.
Webb, J. Comb. Chem. 2009, 11, 1100.
[19] Y.-L. Zhong, H. Zhou, D. R. Gauthier, J. Askin, Tetrahedron
Lett. 2006, 47, 1315.
[20] T. P. Culbertson, J. Heterocycl. Chem. 1979, 16, 1423.
[21] In the case of compound 2d the moderate HPLC AY obtained
was due in part to the formation of an imidazole by-product
which was tentatively identified by LC-MS. Generation of this
type of by-product was also described in Ref. [9] of this paper.
This by-product was also observed for compound 2e. To attempt
to find high-yielding reaction conditions for the synthesis of
pyrimidinones 2d and 2e we went back to our original 96-
additive plate and we evaluated the corresponding Z/E adducts
using 2-MeTHF as the solvent. No more active additives other
than 4 were found for both substrates.
Received: March 2, 2012
Revised: April 16, 2012
Published online: && &&, &&&&
Keywords: heterocycles · high-throughput screening ·
.
homogeneous catalysis · P ligands · synthetic methods
[1] I. W. Davies, C. J. Welch, Science 2009, 325, 701.
[2] S. D. Dreher, P. G. Dormer, D. L. Sandrock, G. A. Molander, J.
Am. Chem. Soc. 2008, 130, 9257.
[22] P. J. Pye, Y.-L. Zhong, G. O. Jones, R. A. Reamer, K. N. Houk, D.
Askin, Angew. Chem. 2008, 120, 4202; Angew. Chem. Int. Ed.
2008, 47, 4134.
[23] Gaussian 09 (Revision A.02), M. J. Frisch, et al.; see the
Supporting Information.
[24] A. E. King, L. M. Huffman, A. Casitas, M. Costas, X. Ribas, S. S.
Stahl, J. Am. Chem. Soc. 2010, 132, 12068, and references
therein.
[3] G. A. Molander, S. L. J. Trice, S. D. Dreher, J. Am. Chem. Soc.
2010, 132, 17701.
[4] D. W. Robbins, J. F. Hartwig, Science 2011, 333, 1423.
[5] A. McNally, C. K. Prier, D. W. C. MacMillan, Science 2011, 334,
1114.
[6] M. T. Reetz, K. M. Kꢁhling, A. Deege, H. Hinrichs, D. Belder,
Angew. Chem. 2000, 112, 4049; Angew. Chem. Int. Ed. 2000, 39,
3891.
4
www.angewandte.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2012, 51, 1 – 5
These are not the final page numbers!

Angewandte
Chemie
Communications
High-Throughput Screening
A. Bellomo, N. Celebi-Olcum, X. Bu,
N. Rivera, R. T. Ruck, C. J. Welch,
K. N. Houk,*
S. D. Dreher*
&&&& — &&&&
Rapid Catalyst Identification for the
Synthesis of the Pyrimidinone Core of HIV
Integrase Inhibitors
A microscale chemistry improvement
engine: A pre-dosed microscale high-
throughput experimentation additives
platform enables rapid, serendipitous
reaction improvement. This platform
allowed one chemist to set up 475
experiments and analyze the results using
MISER chromatography in a single day,
thus resulting in two high-quality catalytic
systems for the construction of the title
compound 1. Support for a single-elec-
tron transfer mechanism was obtained.
Angew. Chem. Int. Ed. 2012, 51, 1 – 5
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
5
These are not the final page numbers!