Combinatorial diversification of indinavir: In vivo mixture dosing of an HIV protease inhibitor library

Article abstract of DOI:10.1016/S0960-894X(00)00276-6

An efficient combination solution-phase/solid-phase route enabling the diversification of the P₁', P₂', and P₃ subsites of indinavir has been established. The synthetic sequence can facilitate the rapid generation of HIV protease inhibitors possessing more favorable pharmacokinetic properties as well as enhanced potencies. Multiple compound dosing in vivo may also accelerate the identification of potential drug candidates. (C) 2000 Elsevier Science Ltd. All rights reserved.

Full text of DOI:10.1016/S0960-894X(00)00276-6

Bioorganic & Medicinal Chemistry Letters 10 (2000) 1527±1530
Combinatorial Diversi®cation of Indinavir: In Vivo Mixture
Dosing of an HIV Protease Inhibitor Library
Thomas A. Rano,a,* Yuan Cheng,a Tracy T. Huening,a Fengqi Zhang,a
William A. Schleif,b Lori Gabryelski,b David B. Olsen,b Lawrence C. Kuo,b
Jiunn H. Lin,c Xin Xu,c Timothy V. Olah,c Debra A. McLoughlin,c
Richard King,c Kevin T. Chapmana and James R. Tataa
aDepartment of Medicinal Chemistry, Merck Research Laboratories, Rahway, NJ 07065, USA
bDepartment of Antiviral Research, Merck Research Laboratories, West Point, PA 19486, USA
cDepartment of Drug Metabolism, Merck Research Laboratories, West Point, PA 19486, USA
Received 14 March 2000; accepted 25 April 2000
0
0
AbstractÐAn ecient combination solution-phase/solid-phase route enabling the diversi®cation of the P1 , P2 , and P3 subsites of
indinavir has been established. The synthetic sequence can facilitate the rapid generation of HIV protease inhibitors possessing
more favorable pharmacokinetic properties as well as enhanced potencies. Multiple compound dosing in vivo may also accelerate
the identi®cation of potential drug candidates. # 2000 Elsevier Science Ltd. All rights reserved.
Introduction
Chemistry
The human immunode®ciency virus type 1 (HIV-1) is
the etiologic agent of acquired immunode®ciency syn-
drome (AIDS). The discovery of clinically e€ective HIV
protease inhibitors1 in the recent past has signi®cantly
improved the lifestyle of many individuals a‚icted with
the virus. Never-the-less, current protease inhibitors su€er
to some extent from issues not limited to ®rst-pass meta-
bolism, toxicities and food restrictions which often times
contribute to patient non-compliance. More recently,
the emergence of multi-drug resistant viral variants has
been con®rmed,2 further compromising the e€ectiveness
of current PI therapy. Indinavir, a potent and speci®c
orally bioavailable HIV protease inhibitor, is metabolized
by P450 isoforms in the CYP3A subfamily.3 In order to
further address the pharmacokinetic properties as well as
in vivo potencies of indinavir, we have initiated an inves-
tigation directed toward the examination of its metaboli-
cally labile sites via generation of an `indinavir-based'
combinatorial library.4 Depicted in Figure 1 are the major
sites of metabolism of indinavir.3 We postulated that
diversi®cation of these sites would lead to a second
generation HIV protease inhibitor possessing improved
pharmacokinetic and potency pro®les.
Our synthetic endeavor began with lactone 1 (Scheme 1)
as previously prepared by Dorsey et al.5 Hydrolysis of 1
employing aqueous LiOH in DME followed by removal
of the water and subsequent exposure to allyl bromide
provided hydroxyethylene isostere 2 (containing the `X'
dimension of the library). Tethering the secondary alcohol
of 2 to the resin was determined to provide the greatest
synthetic ¯exibility. Accordingly, esteri®cation of Rapp
TentaGel S CO2H resin6 employing 2 as the nucleophile
provided the polymer bound orthogonally protected
hydroxyethylene isostere 3. Noteworthy is the fact that
the orthogonal nature of the protecting groups permits
0
either the P2 or the P3 subsite to be spatially addressed.
*Corresponding author. Tel.: +1-732-594-5697. fax: +1-732-594-9464;
e-mail: thomas_rano@merck.com
Figure 1. Denotes metabolism by CYP3A4.
0960-894X/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.
PII: S0960-894X(00)00276-6

1528
T. A. Rano et al. / Bioorg. Med. Chem. Lett. 10 (2000) 1527±1530
Examination of historical data suggested that the P3 sub-
stituent plays an important role in maintaining activity in
the HIV viral spread assay. Therefore, we initially chose to
spatially address P3.
Table 1. In vitro enzyme inhibition and potency of pools/compounds
Pool/compound
Enzyme
inhibition
IC50, nMa
Inhibition of
viral spread
(IIIb) CIC95, nMa
Z1
Z2
Z3
2.5
4.3
21.8
0.59b
1.1
<160
1250
10,000
50.5c
200
Toward that end, deprotection of the allyl ester employ-
ing Pd(PPh3)4 and 1,3-dimethylbarbituric acid provided
the carboxylic acid.7 Amide formation using standard
carbodiimide coupling incorporated the requisite amino
indanol moiety.8 Completion of the resin bound indinavir
molecule involved acid catalyzed removal of the Boc
group followed by reductive amination. Finally, release
of the protease inhibitor from the resin was achieved by
mild base catalyzed transesteri®cation,6 providing indina-
vir in 63% yield and 87% purity as determined by
HPLC.9
Indinavir (X1Y1Z1)
X1Y1Z2
X1Y1Z3
1.6
400
aExcept where noted, values are for n=1.
bAverage for n=178.
cAverage for n=106.
would be used to probe modest amino indanol mod-
i®cations.10 Figure 2 depicts the generic structure of the
library and the X, Y, and Z subunits employed.
With a viable solid-phase synthesis of indinavir now
secured, we turned our attention toward construction of a
0
small library to demonstrate proof of concept. The P2 and
Preparation of the library was as follows: the 5 resin
bound hydroxyethylene isostere fragments were archived
and mixed. The allyl group was removed on a bulk two-
thirds of the resin and this material further split into 4
separate pools in order to attach the Y subunits. After
doing so, the 4 pools were archived, mixed and the Boc
group was removed. This resin was then split into 2 sepa-
rate pools followed by reductive amination employing the
carboxaldehyde Z subunits. Finally, the protease inhibi-
tors were released from the resin by gentle warming in
10% TEA/MeOH.
P3 ligands could easily be diversi®ed by simply varying
the amines and aldehydes employed in those respective
0
several steps to the lithium enolate progenitor of lactone 1.
By varying the alkylating agent, diversi®cation of P1
positions. However, diversifying P1 required retreating
0
could be achieved. The requisite hydroxyethylene isosteres
were then attached to the solid phase to begin construc-
tion of the library. The selection of these X fragments
0
was directed toward either ®lling the hydrophobic S1
pocket or blocking metabolism. With the di€erentiation
0
selection of the `Y' (P2 ) and `Z' (P3) dimensions. Based on
of P1 now complete, e€orts were directed toward the
0
A separate investigation into the sulfonylation of the
piperazine was also ongoing. We demonstrated that spa-
tially addressing the Y dimension was also possible by ®rst
removing the Boc group of 3 with TFA followed by sulfo-
nylation11 of the resultant secondary amine (Scheme 2).
historical data, the Z dimension was selected to control
the potency of the target protease inhibitors (see Table 1
for the IC50/CIC95 values of single pure compounds
X1Y1Z1, X1Y1Z2, and X1Y1Z3) whilst the Y dimension
Scheme 1. Solid-phase synthesis of indinavir.

T. A. Rano et al. / Bioorg. Med. Chem. Lett. 10 (2000) 1527±1530
1529
nitrogen which increases aqueous solubility. The latter
is critical in cell based assay potency allowing inhibitors to
cross the cell membrane.5 Although large and highly lipo-
philic ligands are well tolerated at P3, they display lower
potencies for the reasons stated above, as exempli®ed by
the Z2 pool. Replacement of the `arylmethyl' substituents
with a sulfonyl moiety resulted in a pool of moderate
potency in the binding assay, however, the potency in the
cell based spread assay was severely compromised. This
is almost certainly due to the dramatic decrease in the
pKa of the N4 piperazine nitrogen resulting in inhibitors
of lower aqueous solubility, ultimately restricting the
potential of these inhibitors to penetrate the cell mem-
brane.5 Notably, the intrinsic potencies as well as the
inhibition of the spread of viral infection achieved by
each pool exactly conforms to our predicted results, as
shown in Table 1. Prioritization of the pools for in vivo
analysis would thus be centered on the viral spread
assay since that assay is clearly more predictive of the in
vivo results.
Figure 2. Library and subunits. Y1 R=H; Y2 R=Cl; Y3 R=Br; Y4
R=F.
Pharmacokinetics
The next challenge we faced was the multiple component
in vivo dosing of our library (0.5 mpk/compound; 20
compounds per dog [n=2], 0.05 M citric acid solution).
We postulated that individual compounds possessing
optimal pharmacokinetic properties would be readily
discernible within the mixture.15 This strategy would
accelerate the identi®cation of those compounds using
only 2 dogs rather than the 40 dogs required to dose 20
individual compounds. However, if accurate pharmaco-
kinetic data were to be interpreted from the results, it
was imperative that the components of each pool be as
close to equimolar as possible. Clearly, it would be
unreasonable to establish whether the lack of PK data
for one or more compound in the mixture was due to
either poor bioavailability or absence of that component
in the mixture. In order to determine the composition of
pool Z1, all 20 single pure protease inhibitors were indi-
vidually prepared via the route described in Scheme 1.9
From that array of compounds, an equimolar 20 com-
ponent pool was reconstituted and delivered for LC/
MS/MS analysis along with the 20 individual com-
pounds. By comparison, it was concluded that the Z1
pool was indeed an equal mixture of 20 compounds
suitable for in vivo dosing. Analysis of the anticipated
data required a thorough understanding of the P450
substrate-structure±activity relationship. Incorporation
of the P3 ligand of indinavir (which is responsible for
P450 inhibition) as the Z ligand renders all of the com-
ponents in the Z1 pool as P450 inhibitors/substrates,
therefore averaging the 3A4 inhibitory contribution of
any one compound.16 Chart 1 depicts the results of the 20
compound combinatorial mixture (Z1 pool) dosed in
dogs. Close examination of Chart 1 indicates that a sig-
ni®cant amount of SAR can be gleaned from the result-
ing data. It is clearly evident that the set of compounds
Scheme 2. Sulfonylation of N4 piperazine.
Removal of the allyl group of 7, amide formation and
cleavage cleanly provided the `tosylated' protease inhibitor
(not shown). This strategy was utilized for processing the
remaining one-third of the resin. Removal of the Boc
group was followed by exposing the resin to 10 equiv of
MsCl in 1:1 pyridine:DMF (2Â1 h). The resin was split
into 4 pools, the allyl group was removed and the Y sub-
units coupled. Cleavage from the resin provided the mix-
ture of mesylated protease inhibitors now spatially
addressed in the Y dimension.12 In order to comply with
our original design of spatially addressing the Z dimension,
equimolar portions of each of these mixtures were
combined and grouped with the previously prepared
wells for submission to our HIV viral spread assay.13
Results
The biological activities of the 3 pools are displayed in
Table 1. The compounds were tested for their ability to
prevent cleavage of a substrate by the protease enzyme
(IC50) and to inhibit the spread of viral infection in
MT4 human T-lymphoid cells infected with the IIIb
isolate (CIC95).14 The results in Table 1 indicate that
mixtures had no deleterious e€ects on our assays. As
anticipated, the pool containing the Z1 ligand (present
in indinavir) displayed the highest anity for the HIV
protease enzyme in the binding assay and elicited the best
potency in the viral spread assay. This is largely due to
the 3-pyridylmethyl group providing both lipophilicity
for binding to the protease as well as a weakly basic
0
X2Y1-4Z1 (isobutyl in P1 ) achieved the highest Cmax levels
and AUC's in both dogs, followed by the set of com-
0
pounds X1Y1 4Z1 (benzyl in P1 ). However, only modest
di€erences in the t1=2's were realized. Furthermore, the

1530
T. A. Rano et al. / Bioorg. Med. Chem. Lett. 10 (2000) 1527±1530
J. H. Biochem. Pharmacol. 1997, 53, 1187. (c) Lin, J. H.; Chiba,
C.; Balani, S. K.; Chen, I.-W.; Kwei, G. Y.-S.; Vastag, K. J.;
Nishime, J. A. Drug Metab. Dispos. 1996, 24, 1111.
4. The solid-phase synthesis of HIV protease inhibitors has
been previously reported: (a) Baker, C. T.; Salituro, F. G.;
Court, J. J.; Deininger, D. D.; Kim, E. E.; Li, B.; Novak, P.
M.; Rao, B. G.; Pazhanisamy, S.; Schairer, W. C.; Tung, R. D.
Bioorg. Med. Chem. Lett. 1998, 8, 3631. (b) Wang, G. T.; Li,
S.; Wideburg, N.; Kra€t, G. A.; Kempf, J. D. J. Med. Chem
1995, 38, 2995. (c) Kick, E. D.; Ellman, J. A. J. Med. Chem.
1995, 38, 1427 and references sited therein.
5. Dorsey, B. D.; Levin, R. B.; McDaniel, S. L.; Vacca, J. P.;
Guare, J. P.; Darke, P. L.; Zugay, J. A.; Emini, E. A.; Schleif,
W. A.; Quintero, J. C.; Lin, J. H.; Chen, I.-W.; Holloway, M.
K.; Fitzgerald, P. M. D.; Axel, M. G.; Ostovic, D.; Anderson,
P. S.; Hu€, J. R. J. Med. Chem. 1994, 37, 3443.
6. Cheng, Y.; Lu, Z.; Chapman, K. T.; Tata, J. R. J. Comb.
Chem. Submitted for publication.
7. Washing the resin with glacial acetic acid eciently
removed palladium based residues. Standard washing proto-
cols throughout the synthesis typically involved various com-
binations of DMF, DCM, THF, and IPA. Final resin washing
included either DCM or DMF, depending on the next reaction
sequence.
8. The initial amide coupling conditions employed were amine
(10 equiv), EDC (15 equiv), and HOBT (15 equiv) for 24 h.
The conditions were later improved by employing HBTU (2.5
equiv), HOBT (3.75 equiv), DIEA (5 equiv), and amine (2.5
equiv) for 1 h.
9. Analysis by HPLC (4 Â 250 mm LiChrospher 100 RP-18 5
mm particle size. Gradient elution 25/75 to 90/10 MeCN/H2O
containing 0.1% TFA, 1.0 mL/min for 15 min) by area inte-
gration at 210 nm and 254 nm. The structure assigned to each
new compound is in accord with its 400 MHz NMR spectrum
as well as appropriate ion identi®cation by mass spectrometry.
10. The synthesis of these amino indanol analogues will be
reported elsewhere.
Chart 1. Pharmacokinetic pro®le of Z1 pool. Pool was delivered orally
as a solution in 0.05 M citric acid.
more highly lipophilic compounds (sets X3Y1 4Z1 and
X4Y1 4Z1) elicited the lowest AUC's. This is most likely
due to the e€ects of decreased water solubility. The fact
that compound X1Y1Z1 (indinavir) displayed the ®fth
highest Cmax and AUC con®rms our hypothesis that
pharmacokinetically attractive protease inhibitors would
not escape detection in a mixture of 20 compounds.
Overall, the results in Chart 1 do provide valuable infor-
mation on the in vivo dosing of multiple component
mixtures.
Conclusion
11. The resin was neutralized by washing with 30% TEA/
THF, then THF or DCM prior to initiating the sulfonylation
reaction.
12. Spatially addressing the Y dimension is the subject of a
subsequent library and will be reported in due course.
13. Overall, this strategy a€orded 3 distinct pools diversi®ed
at X, Y, and spatially addressed at Z.
14. Vacca, J. P.; Dorsey, B. D.; Schleif, W. A.; Levin, R. B.;
McDaniel, S. L.; Darke, P. L.; Zugay, J.; Quintero, J. C.;
Blahy, O. M.; Roth, E.; Sardana, V. V.; Schlabach, A. J.;
Graham, P. I.; Condra, J. H.; Gotlib, L.; Holloway, M. K.;
Lin, J.; Chen, I.-W.; Vastag, K.; Ostovic, D.; Anderson, P. S.;
Emini, E. A.; Hu€, J. R. Proc. Nat. Acad. Sci. USA 1994, 91,
4096.
15. The use of quantitative LC/MS/MS to detect minute
quantities of closely related mixtures of drug substances in the
extracts of biological ¯uids has been reported: Olah, T. V.;
McLoughlin, D. A.; Gilbert, J. D. Rapid Commun. Mass
Spectrom. 1997, 11, 17 and references sited therein. For other
publications involving mixture dosing, see: Allen, M. C.; Shah,
T. S.; Day, W. W. Pharm. Res. 1998, 15, 93.
In summary, we have established a ¯exible0 solid-phase
0
synthesis enabling the diversi®cation of the P1 , P2 , and P3
subsites of indinavir. The synthetic route can be utilized in
the generation of an `indinavir-based' library or libraries
directed toward identi®cation of a second generation
HIV protease inhibitor possessing improved metabolic
and potency pro®les. We have also demonstrated that in
vitro potency can be measured accurately on mixtures
obtained from libraries. Furthermore, reliable phar-
macokinetic data can be acquired via oral dosing of
combinatorial mixtures in dogs. The design, preparation
and analysis of much larger libraries of HIV protease
inhibitors is currently underway and will be reported in
the near future.
References and Notes
1. Vacca, J. P.; Condra, J. H. Drug Discov. Today 1997, 2, 261.
2. Condra, J. H. Drug Resis. Updates 1998, 1, 1.
3. The hepatic/intestinal metabolism and pharmacokinetics of
indinavir has been extensively investigated: (a) Lin, J. H. Adv.
Drug Del. Rev. 1999, 39, 211. (b) Chiba, M.; Hensleigh, M.; Lin,
16. Future libraries would incorporate a well absorbed CYP3A4
substrate in each pool as a standard to monitor P450 activity. An
increase in the Cmax of this substrate would indicate the presence
of a CYP3A4 inhibitor in that particular pool.