Synthesis of ABT-378, an HIV protease inhibitor candidate: Avoiding the use of carbodiimides in a difficult peptide coupling

Article abstract of DOI:10.1021/op980214p

An alternative to carbodiimide-mediated peptide coupling protocols has been developed for a carboxylic acid prone to decomposition by polymerization. This method, involving the in situ generation of an acyl imidazolide, has been applied to the preparation of a lead clinical HIV protease inhibitor candidate, ABT-378. The nature of the polymerization and optimization of the new reaction conditions are presented.

Full text of DOI:10.1021/op980214p

Organic Process Research & Development 1999, 3, 145−148
Synthesis of ABT-378, an HIV Protease Inhibitor Candidate: Avoiding the Use
of Carbodiimides in a Difficult Peptide Coupling
Eric J. Stoner,* Peter J. Stengel, and Arthur J. Cooper
Process Research and DeVelopment, D54P, Abbott Laboratories, North Chicago, Illinois 60064
Abstract:
covery synthesis to a scale suitable for the preparation of
multikilogram quantities of drug. One transformation in the
synthesis which presented considerable complication is the
peptide coupling shown below, in which amine 37 and
L-valine-derived acid 48 are coupled, affording ABT-378 (2).
Being the final step in the synthetic sequence, this chemistry
needed to be high yielding and efficient and to produce 2 in
high purity with minimal racemization.
An alternative to carbodiimide-mediated peptide coupling
protocols has been developed for a carboxylic acid prone to
decomposition by polymerization. This method, involving the
in situ generation of an acyl imidazolide, has been applied to
the preparation of a lead clinical HIV protease inhibitor
candidate, ABT-378. The nature of the polymerization and
optimization of the new reaction conditions are presented.
The approval of the first HIV protease inhibitors in early
1996 provided the world with powerful new weapons in the
fight against HIV, the virus responsible for AIDS.1 HIV
protease is an enzyme critical to the life cycle of the virus,
and its inhibition disrupts viral replication, resulting in the
formation of immature, noninfectious viral particles.2 While
the protease inhibitors, combined together and in “drug
cocktails” with various reverse transcriptase inhibitors, can
be extremely potent in reducing blood levels of HIV, the
clinical benefit observed eventually degrades due to the
development of drug resistance, arising from predictable
mutations in the virus.3 As a result, research continues toward
the development of new and more powerful protease inhibi-
tors.4 Among this first generation of effective HIV protease
inhibitors is Abbott Laboratories’ Ritonavir (1)5 (Norvir).
Abbott’s next generation candidate, ABT-378 (2), is showing
considerable promise as well.6
The original discovery synthesis utilized 1-ethyl-3-(3′-
dimethylaminopropyl)carbodiimide hydrochloride (EDC)
activation of 4 to prepare the N-hydroxybenzotriazole
(HOBT) ester. While the HOBT ester of 4 was an efficient
coupling partner, leading to clean formation of 2, these two
reagents posed numerous problems for larger scale work.
The most prominent among these deterrent factors were the
toxicity, high cost, and limited availability of EDC.9 There-
fore, we set out to determine a more suitable method to
perform this coupling. Our goal was to develop a process as
high yielding and efficient as the EDC/HOBT process with
none of the hazards and excessive costs associated with those
reagents.
(5) Also referred to as ABT-538, see: Kempf, D. J.; Marsh, K. C.; Denissen,
J. F.; McDonald, E.; Vasavanonda, S.; Flentge, C. A.; Green, B. E.; Fino,
L.; Park, C. H.; Kong, X.-P.; Wideburg, N. E.; Saldivar, A.; Ruiz, L.; Kati,
W. M.; Sham, H. L.; Robins, T.; Stewart, K. D.; Hsu, A.; Plattner, J. J.;
Leonard, J. M.; Norbeck, D. W. Proc. Natl. Acad. Sci. U.S.A. 1995, 92
(7), 2484.
(6) Sham, H. L.; Norbeck, D. W.; Chen, X.; Betebenner, D. A.; Kempf, D. J.;
Herrin, T. R.; Kumar, G. N.; Condon, S. L.; Cooper, A. J.; Dickman, D.
A.; Hannick, S. M.; Kolaczkowski, L.; Oliver, P. A.; Plata, D. J.; Stengel,
P. J.; Stoner, E. J.; Tien, J.-H. J.; Liu, J.-H.; Patel, K. M. U.S. Patent Appl.
95-57226; Chem. Abstr. 1997, 127, 122001.
(7) Prepared from diamino alcohol intermediate used in Ritonavir synthesis,
see: (a) Stuk, T. L.; Haight, A. R.; Scarpetti, D.; Allen, M. S.; Menzia, J.
A.; Robbins, T. A.; Parekh, S. I.; Langridge, D. C.; Tien, J.-H. J.; Pariza,
R. J.; Kerdesky, F. A. J. J. Org. Chem. 1994, 59, 4040. (b) Haight, A. R.;
Stuk, T. L.; Menzia, J. A.; Robbins, T. A. Tetrahedron Lett. 1997, 38 (24),
4191 and references therein.
(8) Prepared from phenoxycarbonyl-L-valine and chloropropylamine hydro-
chloride, see: Stengel, P. J.; Cooper, A. J.; Oliver, P. A.; Patel, K. M.,
manuscript in preparation.
(9) For an excellent review of this reagent, see: Encyclopedia of Reagents for
Organic Synthesis; Paquette, L. A., Ed.; Wiley: New York, 1995; Vol. 4,
p 2430.
The development of ABT-378 (2) has been justifiably
rapid, and, as might be expected, numerous difficulties were
encountered in making the fast-paced transition from dis-
(1) Deeks, S. G.; Smith, M.; Holodniy, M.; Kahn, J. O. J. Am. Med. Assoc.
1997, 277 (2), 145.
(2) Moyle, G.; Gazzard, B. Drugs 1996, 51 (5), 701.
(3) (a) Molla, A.; Korneyeva, M.; Gao, Q.; Vasavanonda, S.; Schipper, P. J.;
Mo, H.-M.; Markowitz, M.; Chernyavskiy, T.; Niu, P.; Lyonns, N.; Hsu,
A.; Granneman, G. R.; Ho, D. D.; Boucher, C. A. B.; Leonard, J. M.;
Norbeck, D. W.; Kempf, D. J. Nat. Med. (N. Y.) 1996, 2 (7), 760. (b) Schmit,
J.-C.; Ruiz, L.; Clotet, B.; Raventos, A.; Tor, J.; Leonard, J.; Desmyter, J.;
De Clercq, E.; Vandamme, A.-M. AIDS 1996, 10, 995.
(4) Sham, H. L.; Chen, X. Exp. Opin. InVest. Drugs 1996, 5 (8), 977.
10.1021/op980214p CCC: $18.00 © 1999 American Chemical Society and Royal Society of Chemistry
Published on Web 02/26/1999
Vol. 3, No. 2, 1999 / Organic Process Research & Development
•
145

Our initial efforts involved the screening of common
peptide coupling methods. Of those examined, couplings with
propyl phosphonic anhydride10 in the presence of N-meth-
ylmorpholine proved the most viable. Unfortunately, this
methodology suffered by comparison to the original process
in two respects. First, propyl phosphonic anhydride/N-
methylmorpholine failed to give full conversion of amine 3
to 2, and second, these reagents appeared to cause some
degradation of 3, which made the isolation and purification
of 2 more difficult.
The extent of conversion of amine 3 to product 2 was a
critical factor in the development of this step of the synthesis.
We found that levels of unreacted amine 3 above a few
percent made the final purification of 2 by crystallization
exceedingly difficult, and substantial losses in yield were
observed in efforts to achieve the purities needed for clinical
use. Therefore, a very high conversion of amine 3 to inhibitor
candidate 2 was considered an essential feature of any
workable coupling protocol.
Our subsequent efforts were hampered by our inability
to prepare active esters of 4 without using carbodiimides with
their attendant waste and toxicity. Treatment of 4 with oxalyl
chloride, acyl chlorides, and numerous chloroformates,
followed by reaction with N-hydroxysuccinimide, N-hy-
droxybenzotriazole, and the like, failed to yield the desired
activated esters in any appreciable quantity. In these experi-
ments, no unreacted acid remained, having been converted
into an intractable semisolid mass. It became clear that the
activation of 4 achieved with oxalyl chloride and other
reagents generated an intermediate species so unstable that
it degraded nearly instantaneously (even at low temperatures).
This made the subsequent discovery that acyl chloride 5
could be prepared with SOCl2 or POCl3 and then isolated as
a stable crystalline solid all the more surprising.11 It was this
discovery that led us to evaluate the direct coupling of acyl
chloride 5 with amine 3 as the most desirable approach. The
fact that acyl chloride 5 could be isolated provided evidence
that it was not responsible for the degradation observed
earlier.
would never achieve our desired goal of complete conversion
with this base. We were resistant to improving the conversion
ratio by further increasing the amount of acyl chloride 5 used,
due to the fact that acid 4 represented a significant contributor
to the overall cost of the synthesis. Although we had
examined various methods for the removal of unreacted 3
with aqueous washes, these proved unreliable and problem-
atic.
We screened a large array of amine and inorganic bases
in this reaction, including pyridine, N,N-dimethylamino-
pyridine, Na2CO3, NaHCO3, N-methylmorpholine, and tri-
ethylamine, but failed to find a suitable set of reaction
conditions. Reaction of 5 with imidazole produced the
expected acyl imidazolide, which we found to be superior
to anything we had tried. We were also pleased to find that
the addition of 5 to amine 3 and imidazole produced even
higher conversions and cleaner reactions. Subsequently, a
“first generation” optimized procedure was developed which
involved the addition of a DMF solution of 1.1-1.3 equiv
of acyl chloride 5 to a mixture of 1.0 equiv of amine 3 and
1.3-1.5 equiv of imidazole in EtOAc, producing 2 in very
high purities, with conversions of greater than 95%. This
process proved useful enough to prepare multikilogram
quantities of 2.
While pleased with our working process, we remained
committed to further improvement, especially in light of a
second pivotal discovery. We had initially assumed that a
slight excess of 5 was needed to cover losses due to
hydrolysis which had occurred during processing. The
byproduct of this hydrolysis, acid 4, was probably being
removed during the aqueous workup. During the quench of
these coupling reactions, a fine solid appeared at the interface
of the aqueous and organic layers. In larger scale reactions,
this solid remained in suspension and made further process-
ing (layer separations, etc.) extremely difficult. Filtration to
remove this solid was possible but problematic owing to the
extremely small particle size, which slowed processing and
hampered throughput.12
To our surprise, isolation and characterization of this
material revealed it to be a polymer comprised of acid 4
monomer units. Based on our analytical data, we believe 6
best represents the structure of this polymer, although we
cannot completely eliminate structure 7 or structures consist-
ing of both types of polymer linkages. The “head-to-tail”
orientation of polymer 6 is supported principally by the IR
spectrum, which lacks the presence of ester carbonyl
stretches, instead showing a single large carbonyl absorption
at 1678 cm-1, and the 13C NMR spectrum, which shows two
major carbonyl peaks at 173.1 and 157.6 ppm in CDCl3.13
As expected, the range of polymer molecular weights
varies with experimental parameters. Polymers isolated from
a number of laboratory and early pilot plant runs have
molecular weight distributions varying from around 800 to
over 4000 amu. The most common molecular weights
correspond to 14-16 monomer units. Mass spectral data and
With a convenient synthesis of 5 in hand, many activated
esters of 4 were now accessible to us; however, these
compounds remained less desirable than any direct coupling
protocol. Our initial attempts to couple acyl chloride 5 with
amine 3 in the presence of pyridine in a variety of solvents
were promising. The reactions were clean but failed to
produce complete reaction of 3 to 2, with conversions on
the order of 15-60% being common. While the conversion
ratio (and therefore the yield) could be improved by using
larger excesses of pyridine, it was quickly apparent that we
(10) For an excellent review of this reagent, see: Encyclopedia of Reagents for
Organic Synthesis; Paquette, L. A., Ed.; Wiley: New York, 1995; Vol. 6,
p 4336.
(12) Crude reaction mixtures containing large amounts of this polymer have
the consistency of latex paint.
(13) We have assigned these two peaks as the amide carbonyl and the urea
carbonyl, respectively.
(11) Xi, N.; Ciufolini, M. A. Tetrahedron Lett. 1995, 36 (17), 6595.
146
•
Vol. 3, No. 2, 1999 / Organic Process Research & Development

elemental analysis further suggest that only acid 4 units are
incorporated in the polymer, as shown. Hydrolysis of these
polymers with strong base regenerates acid 4.
available alkylated and nitrated imidazole derivatives, finding
them to be inferior to imidazole with respect to this reaction.
Conclusion
Presently we have achieved a coupling protocol which is
suitable for bulk drug production. Our non-carbodiimide-
mediated coupling of polymerization-prone carboxylic acid
4 and amine 3 leading to ABT-378 is efficient and
inexpensive and relies on readily available raw materials.
We have avoided a number of safety and environmental
issues by the substitution of thionyl chloride and imidazole
for EDC and HOBT. Furthermore, while we have a viable,
reproducible, large scale process, work continues to further
refine the chemistry. This process has been run on a
multikilogram scale numerous times, and a version of this
protocol involving a different amine partner has been run
successfully on over a 200-kg scale.
The formation of this polymer was, in our minds, another
indication of the instability of certain activated derivatives
of 4. While we have not completely determined the cause
of this polymerization, we were fortunate in that the cure
was quickly discovered. When 3 equiv or more of imidazole
was used, the polymerization was completely prevented. The
added benefits of this simple and inexpensive change are
numerous and include easier processing, nearly quantitative
conversion of 3 to 2 (typically greater than 99%), cleaner
crude reaction mixtures, and higher yields. While the
reactions become slower as more imidazole is used, it is an
acceptable cost relative to the benefits achieved.
Experimental Section
All reactions were performed under a positive pressure
of nitrogen. Solvent concentration was accomplished with a
Bu¨chi rotary evaporator at ca. 15 mmHg pressure using a
vacuum pump (not water aspiration).14 Commercial grade
anhydrous solvents and reagents were used without further
purification. HPLC purities were determined by peak area
percent.
Discussion and Final Optimization
The polymerization of acyl chloride 5 is intriguing. At
lower imidazole concentrations (<1 equiv), polymerization
is the main reaction pathway of 5, while when a sufficient
excess of imidazole is used, it is a virtually nonexistant
process. Our analysis of this phenonmenon is based on the
dual nature of imidazole as base and nucleophile. We know
from experience that the coupling of acyl chlorides with
amines is a fast process, producing HCl at a correspondingly
fast rate. If this acid is not removed quickly enough, it is
our belief that HCl-activated 5 becomes prone to polymer-
ization and a competition is set up between coupling, removal
of acid by imidazole, and polymerization. As expected,
couplings with less than 2 equiv of imidazole are extremely
rapid.
When, however, more than 3 equiv of imidazole is
employed in a coupling reaction, there is sufficient base
present to compensate for any initial coupling and HCl
generation which take place. Additionally, imidazole now
becomes a competitive nucleophile, allowing for the forma-
tion of the acyl imidazolide derivative of 4, which couples
with 3 relatively slowly and is quite stable. We have observed
an experimental “gray area” which occurs when imidazole
is present at roughly 1.7-3.0 equiv. Experimentally, this
region produces results intermediate to the two extremes,
with yields varying with other experimental parameters such
as temperature and agitation.
Preparation of Acyl Chloride 5. Acid 4 (25.8 g, 0.129
mol, 1.1 equiv) and 590 mL of THF were charged to a 1-L,
three-necked, round-bottomed flask and cooled to 0 °C. Neat
thionyl chloride (17.4 g, 10.7 mL, 1.25 equiv) was added
dropwise over 15 min via a pressure-equalizing addition
funnel. When the addition was complete, the reaction mixture
was warmed to room temperature and stirred for an additional
1 h. HPLC analysis of samples prepared by quenching into
MeOH showed no unreacted 4 at this time. The reaction
mixture was then evaporated to dryness in vacuo and 150
mL of heptane added to facilitate the removal of unreacted
thionyl chloride azeotropically. The heptane slurry was
evaporated to dryness in vacuo, affording acyl chloride 5 as
a hygroscopic off-white solid (HPLC analysis, >98% pure).
The crude 5 was immediately slurried in 180 mL of
anhydrous DMF and held for use in the next step.15
Acyl Chloride 5. 1H NMR (300 MHz, CD3CN): δ 12.5
(v br s, 1H), 4.61 (d, J ) 3 Hz, 1H), 3.61 (app t, 4H), 2.44
(14) Water aspiration can lead to partial hydrolysis of acyl chloride 5 during
reduction in volume.
(15) Acyl chloride 5 slowly degrades in DMF solution at room temperature
and should be used within 2-3 h for best results; solid 5 is stable under
nitrogen indefinitely.
We subsequently screened a large number of other bases,
including 1,2,4-triazole, thiazole, and various commercially
Vol. 3, No. 2, 1999 / Organic Process Research & Development
•
147

(d sept, J ) 3, 6.5 Hz, 1H), 2.16 (m, 2H), 1.16 (d, J ) 6.5
Hz, 3H), 1.01 (d, J ) 6.5 Hz, 3H). 13C NMR (75 MHz, CD3-
CN): δ 165.2, 156.0, 65.5, 41.9, 39.0, 29.0, 19.0, 16.7, 16.4.
MS (DCI, NH3) 215 (M + H)+.
colored solid, 71.5 g (97%, HPLC purity >98%, although
DMF and ethyl acetate are present).
The solid was dissolved in 350 mL of ethyl acetate at 70
°C, and 350 mL of heptane was slowly added.19 After being
reheated to 70 °C, the solution was slowly cooled to 10 °C.
The product was collected by filtration; the wet product cake
was washed with 100 mL of 0 °C 1:1 (v/v) heptane/ethyl
acetate. The solid was dried in vacuo at 60 °C for 24 h,
affording 63.65 g (86.2%) of product as a colorless solid. A
second crop of off-white crystals was isolated from the
mother liquors, 4.30 g (5.8%), for a total of 67.95 g (92%
yield). Both crops of product 2 assayed as >99% pure by
HPLC.
Characterization Data for ABT-378, 2. Mp (EtOAc):
124-127 °C (uncorrected).20 IR (KBr): 3413, 3335, 3289,
3060, 2966, 1671, 1650, 1624, 1545, 1520, 1453, 1189, 701
cm-1. 1H NMR (300 MHz, CDCl3): δ 7.30-7.13 (m, 10H),
7.02-6.92 (m, 3H), 6.86 (v br s, 1H), 5.68 (br s, 1H), 4.25
(m, 1H), 4.19 (app d, J ) 10 Hz, 2H), 4.19 (m, 2H), 3.78
(m, app d sept, 1H), 3.12 (m, 1H), 3.06 (m, 2H), 2.97 (d, J
) 7.6 Hz, 2H), 2.88 (m, 1H), 2.81 (app ABX dd, J ) 14,
5.2 Hz, 1H), 2.68 (app ABX, dd, J ) 14, 9.5 Hz, 1H), 2.23
(m, 1H), 2.18 (s, 6H), 1.83 (s, 1H), 1.74 (m, 2H), 1.53 (m,
1H), 1.28 (m, 2H), 0.83 (app t, J ) 7 Hz, 6H). 13C NMR
(75 MHz, CDCl3): δ 170.7, 168.8, 156.5, 154.2, 138.1,
138.0, 130.3, 129,3, 129.2, 129.0, 128.4, 128.2, 126.3, 126.0,
124.6, 70.2, 69.7, 63.1, 54.4, 48.7, 41.8, 41.1, 40.8, 40.0,
38.2, 25.4, 21.7, 19.6, 18.7, 16.1. MS (ESI): 629 (M + H)+,
651 (M + Na)+. Anal. Calcd for C37H48N4O5: C, 70.66; H,
7.69; N 8.91. Found: C, 70.26; H, 7.73; N, 8.79.
Coupling Procedure: Synthesis of ABT-378, 2. A dry
2-L, three-necked, jacketed, round-bottomed flask equipped
with a mechanical stirrer, nitrogen inlet adapter, and pressure-
equalizing addition funnel was charged with amine 3 (52.4
g, 0.117 mol, 1.0 equiv), imidazole (26.3 g, 3.3 equiv), and
590 mL of ethyl acetate, and the contents were cooled to 0
°C. To this reaction mixture was slowly added the DMF
slurry of acyl chloride 5 at such a rate as to maintain the
internal temperature below 5 °C (about 30 min).16 The
reaction mixture was stirred for an additional 30 min at 0
°C and then warmed to room temperature overnight. HPLC
analysis revealed complete consumption of starting material
(amine 3) after 12 h.17
The reaction mixture was quenched by the addition of
0.2 N aqueous HCl (1 L) and 100 mL of ethyl acetate and
allowed to stir for 30 min. The layers were separated,18 and
the organic layer was washed once with 5% (w/w) aqueous
NaHCO3 (500 mL) and twice with 500 mL of distilled water.
The organic layer was dried over MgSO4, filtered, and
evaporated to dryness in vacuo, affording crude 2 as a cream-
(16) Even with very efficient cooling, this rate of addition is recommended to
avoid polymer formation and to ensure complete conversion of amine 3 to
product.
(17) Typical reaction times vary from 12 to 24 h.
(18) If polymer is formed, it will appear as a finely divided solid in suspension.
It can be removed at this stage by filtration through diatomaceous earth.
(19) The recovery of product is greatly dependent upon the water content of
the recrystallization solution. A water content of this solution above 0.2%
will result in a significant loss of product.
(20) Product crystallized in this manner contains approximately 2% residual
ethyl acetate which cannot be removed by further drying. Removal of this
residual solvent requires additional processing.
Received for review October 15, 1998.
OP980214P
148
•
Vol. 3, No. 2, 1999 / Organic Process Research & Development