Dilevalol via resin-mediated epimerization: A case study. Reaction mechanism to reactor design to a viable process

Article abstract of DOI:10.1021/op990185y

A case study on a novel commercial process for the preparation of dilevalol, 5-[1-hydroxy-2-[(1-methyl-3-phenylpropyl)-amino]-ethyl]salicylamide, (R,R)-1, from crude (S,R)-1, is described. This covers all of the work from the initial laboratory observation of an acid-induced racemization of the benzylic carbinol (epimerization) of pure (R,R)-1, to the identification and optimization of an acidic resin that facilitated the epimerization of the core-process-generated crude (S,R)-1 on a solid support, and the subsequent reactor modification culminating in a cost-efficient, industrial-scale, semi-continuous batch process for the preparation of dilevalol.

Full text of DOI:10.1021/op990185y

Organic Process Research & Development 2000, 4, 107−121
Dilevalol via Resin-Mediated Epimerization: A Case Study. Reaction
Mechanism to Reactor Design to a Viable Process
J. Cerami, D. Gala,* D. Hou, S. Kalliney, J. L. Mas, P. Nyce, L. Peer, and G. Wu
Schering-Plough Research Institute, 1011 Morris AVenue, Union, New Jersey 07083, U.S.A.
Scheme 1
Abstract:
A case study on a novel commercial process for the preparation
of dilevalol, 5-[1-hydroxy-2-[(1-methyl-3-phenylpropyl)-amino]-
ethyl]salicylamide, (R,R)-1, from crude (S,R)-1, is described.
This covers all of the work from the initial laboratory observa-
tion of an acid-induced racemization of the benzylic carbinol
(epimerization) of pure (R,R)-1, to the identification and
optimization of an acidic resin that facilitated the epimerization
of the core-process-generated crude (S,R)-1 on a solid support,
and the subsequent reactor modification culminating in a cost-
efficient, industrial-scale, semi-continuous batch process for the
preparation of dilevalol.
Dilevalol (R,R)-1‚HCl, the eutomer of labetalol [racemic-
1] with â1-receptor blockade and significantly reduced
R-receptor blockade activity, is a potent bronchodilator.1,2
Work towards a nonproprietary, commercial synthesis is
depicted in Scheme 1. This involved the reduction of ketone
2 to the alcohol 3a with NaBH4 or to 3b via Zn(BH4)2. In
either case the R,R-isomer was crystallized via the use of
dibenzoyl-L-tartaric acid (L-DBTA). To enhance the utility
of the chiral amine moiety after isolation of 1, the conversion
of the S,R-isomer in the process mother liquors to the desired
R,R-isomer was of interest. This paper describes an efficient
process for such a conversion. This emerged from the initial
incidental laboratory observation which, after rigorous
laboratory optimization, followed by engineering equipment
design was converted into an industrially feasible process.
Asymmetric Induction. Hartley3 had shown that reduc-
tions of aminoketone 2 (or its N-methyl derivative) with
NaBH4 proceed with little, if any (0-8% de), 1,4-asymmetric
induction from the existing chiral center incipient in the
R-amine moiety. However, he found for the N-benzyl
derivative of 2 that de’s as high as 52% could be obtained
using NaBH4 and with a bulky reducing reagent such as
L-Selectride the de increased to 64%.2 To account for these
results, Hartley proposed that the reduction of N-benzyl
aminoketone-2 proceeded through a nonchelated transition
state and the observed selectivity “is mainly a consequence
of steric approach control to a molecule in which rotation is
restricted by the N-benzyl group.”
Scheme 2
suitable reducing reagent/complex could be found that would
allow for high de for the reduction of aminoketone 2, since
we wished to avoid the added number of steps, and more
importantly, the additional costs required for introducing and
removing the N-benzyl group. Although Hartley reported
only a modest de (8%) for the reduction of 2 using NaBH4,
it was thought that either the use of a more bulky reducing
reagent or the addition of a reagent that would allow for
formation of a transient metal complex might be able to
mimic the effect of the N-benzyl group. For the latter case,
the newly formed N-chiral center can adopt either the R- or
S-configuration, as illustrated by the two Newman projections
depicted in Scheme 2. However, the R-isomer is thermody-
namically more favored since the large M group is situated
between the small H and medium methyl group. This, in
turn, should direct the hydride attack from the backside,
leading to the R,R-isomer.
Encouraged by these results for the reduction of N-benzyl
aminoketone-2, a program was undertaken to determine if a
(1) Walker, D. In Dilevalol Hydrochloride. Development of a Commercial
process. Process Chemistry in the Pharmaceutial Industry; Gadamasetti,
K. G.. Ed.; Mercell Decker, NY, 1999; p 125.
(2) Gold, E. J.; Chang, W.; Cohen, M.; Baum, T.; Ehrreich, S.; Johnson, G.;
Prioli, N.; Sybertz, E. J. J. Med. Chem. 1982, 25, 1363.
(3) Hartley, D. Chem. Ind. 1981, 551.
A number of nonchiral reducing reagents were screened
and these results are listed in Table 1. In all cases the de’s
were determined by GC.4 The bulky reducing reagents did
not show much improvement, as the best result was with
10.1021/op990185y CCC: $19.00 © 2000 American Chemical Society and The Royal Society of Chemistry
Published on Web 03/02/2000
Vol. 4, No. 2, 2000 / Organic Process Research & Development
•
107

Table 1. Asymmetric reduction via 1,4-induction, non-chiral reducing reagentsa,b
number
reagent
solvent
temperature
RR:SR
1
2
3
4
5
6
7
8
9
NaBH4‚Ni(OAc)2
MeOH
0 °C f rt
0 °C f rt
0 °C f rt
-15 °C f rt
rt
51:49
51:49
54:46
52:48
55:45
54:46
50:50
55:45
53:47
58:42
40:60
52:48
54:46
53:47
59:41
63:37
61:39
53:47
60:40
54:46
50:50
58:42
60:40
56:44
53:47
54:46
55:45
55:45
42:58
29:71
46:54
46:54
52:48
51:49
50:50
52:48
51:49
52:48
51:49
48:52
45:55
45:55
NaBH4‚Cu(OAc)2
MeOH
MeOH
EtOH/H2O
DME3
THF
NaBH4‚CuCl2
NaBH4‚CeCl3
NaBH4‚CoCl2
NaBH4‚Al2O3
rt
NaBH4‚AlEt2Cl
THF
-78 °C f rt
0 °C f rt
0 °C f rt
rt
-78 °C f rt
0 °C f rt
rt
NaBH4‚SnCl2‚2H2O
NaBH4‚ZnCl2
MeOH
MeOH
DME
THF
MeOH
Et2O
THF
THF
THF
DME
CH3CN
THF
THF
THF
THF
THF
THF
i-PrOH
CH2Cl2
THF
CH2Cl2
CH2Cl2
CH2Cl2
toluene
Et2O
CH2Cl2
CH2Cl2
CH2Cl2
THF
THF
CH2Cl2
Et2O
THF
CH2Cl2
CH2Cl2
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
31
32
33
34
35
36
37
38
39
40
41
NaBH4‚Zn(OPiv)2
NaBH4/Ti(O-i-Pr)4
NaBH4/Ti(O-i-Pr)4
NaCNBH3‚ZnCl2
BaBH4‚Cu(OAc)2‚H2O
Zn(BH4)2
Zn(BH4)2
rt
0 °C
-10 °C
Zn(BH4)2
-10 °C
Zn(BH4)2‚1.5DMF
Zn(BH4)2‚1.5DMF
K-Selectride
KS-Selectride
L-Selectride
LS-Selectride
LiEt3BH
Al(O-i-Pr) 3
rt
-10 °C
-40 °C f rt
0 °C
-78 °C f rt
-78 °C f rt
-78 °C f rt
reflux
-78 °C f rt
0 °C
(i-Pr)2NH‚BH3
(i-Pr)2NH‚BH3
(i-Pr)2NH‚BH3/Mg(OCOCF3)2
(i-Pr)2NH‚BH3/Mg(OCOCF3)2
(i-Pr)2NH‚BH3/Mg(OCOCF3)2
(i-Pr)2NH‚BH3/Mg(OCOCF3)2
(i-Pr)2NH‚BH3/Mg(OCOCF3)2
(i-Pr)2NH‚BH3/Li(OCOCF3)2
(i-Pr)2NH‚BH3/CF3CO2H
(i-Pr)2NH‚BH3/MgCl 2
(i-Pr)2NH‚BH3/Cu(OAc)2‚H2O
(i-Pr)2NH‚BH3/CoCl2‚6H2O
(i-Pr)2NEt‚BH3/Mg(OCOCF3)2
t-BuNH2‚BH3/BF3‚Et2O
t-BuNH2‚BH3/Mg(OCOCF3)2
t-BuNH2‚BH3/Mg(OCOCF3)2
t-BuNH2‚BH3/Ti(O-i-Pr)4
0 °C
-40 °C
-78 °C
-78 °C
0 °C
-78 °C f rt
-78 °C f rt
-78 °C f rt
-78 °C f rt
-78 °C
-78 °C
-78 °C f rt
-78 °C f rt
-78 °C f rt
-78 °C f rt
a In general, the reductions were carried out by dissolving 2 in the specified solvent and temperatures and adding 1-3 molar equiv of the reducing reagent,
depending on whether the free base or the HCl salt of 2 was used. The reactions samples were quenched with MeOH and the diastereomeric ratio was determined by
capillary GC.4 b Zn(BH4)2 was prepared in situ by mixing NaBH4 and ZnCl2 in 1:0.5 molar ratios in THF and the final concentration of Zn(BH4)2 was ∼0.7 M. c DME
) 1,2-dimethoxyethane.
LS-Selectride, in which case the de increased to 20%.
Similarly, Valery-Meerwein-Pondorf conditions or Li and
Ba borohydrides also did not show any significant improve-
ment (6-8% de). In general, for the sodium borohydride‚
metal complexes the reduction did not occur with any
significant enhancement, as typically the de’s ranged from
0 to 16%. However, for the Ti(O-i-Pr)4 complex a reversal
in selectivity was observed (entry 11), as it showed a
preference for the S,R-diastereomer using THF as the solvent.
Noteworthy is the fact that the selectivity shifts back to the
normal slight preference for the R,R-diastereomer when
MeOH was used as a solvent (entry 12). The best results
were obtained using Zn(BH4)2, as de’s as high as 26% were
obtained for the R,R-diastereomer (entry 16).
Turning away from the metal borohydrides, a number of
amine boranes or their complexes were also investigated.
Interestingly, for (i-Pr)2NH‚BH3 itself, the typical results were
obtained, with a slight preference for the R,R-diastereomer.
Although addition of additives such as CoCl2 and Cu(OAc)2
did not have any appreciable effect, the addition of Mg-
(OCOCF3)2 did have a pronounced effect on the reduction,
as de’s as high as 42% were obtained, but for the S,R-
diastereomer (Entry 30). That the effect was due to the
specific combination of the (i-Pr)2NH‚BH3/Mg(OCOCF3)2
(4) GC: (a) Fused silica capillary column (4 m × 0.25 mm) coated with SP-
2250 with film thickness of 0.2 µm, FID detector. Injector and detector
ports at 250 °C, column oven at 240 °C. Helium as a carrier gas with 20-
30 cm/sec/linear velocity. Inject 1 µL. Run time of 10 min. Sample prepared
by derivatization with methylboronic acid (GLC grade) in pyridine and
allowed to stand at room temperature for 20 min; or (b) in an alternative
GC method the column used was 10% OV-17 on 100/120 mesh Gas Chrom
Q packed in a 6 ft × 1/4 in glass column, or equivalent. The carrier gas
was helium at a flow rate of 30 mL/min. Both injector and detector
temperature were about 340 °C. The oven temperature was about 300 °C.
Detection was with flame ionization. Derivatization of the amino alcohols
were made with 1-butaneboronic acid in dry pyridine (over 4 Å molecular
sieves) before injection on the column.
108
•
Vol. 4, No. 2, 2000 / Organic Process Research & Development

Table 2. Asymmetric reduction via 1,4-induction, chiral reducing reagentsa,b
number
reagent
solvent
temperature
rt
rt
rt
rt
rt
rt
rt
rt
rt
rt
rt
rt
rt
rt
rt
rt
rt
rt
rt
rt
rt
rt
RR:SR
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
26
27
28
29
30
31
32
NaBH4‚(L)-penicillamine
NaBH4‚(D)-penicillamine
NaBH4‚(D)-mandelic acid
NaBH4‚(D)-leucine
THF
THF
THF
THF
THF
THF
THF
THF
THF
THF
THF
THF
THF
THF
THF
THF
THF
THF
THF
THF
THF
THF
THF
THF
THF
Et2O
Et2O
Et2O
MeOH
THF
69:31
38:62
55:45
60:40
72:28
41:59
62:38
51:49
49:51
59:41
55:45
60:40
51:49
56:44
56:44
59:41
55:45
37:73
70:30
63:37
40:60
57:43
54:46
15:85
81:19
20:80
66:34
48:52
52:48
55:45
NaBH4‚(D)-tert-leucine
NaBH4‚(L)-isoleucine
NaBH4‚(D)-phenylalanine
NaBH4‚(L)-proline
NaBH4‚(D)-proline
NaBH4‚(D)-tryptophan
NaBH4‚(D)-tyrosine
NaBH4‚(L)-threonine
NaBH4‚(L)-cysteine
NaBH4‚(L)-cystine
NaBH4‚(L)-lactic acid
NaBH4‚(L)-aspartic acid
NaBH4‚(L)-tartaric acid
NaBH4‚(L)-valine
NaBH4‚(D)-valine
NaBH4‚(L)-valine‚ Mg(OCOCF3)2
NaBH4‚(D)-valine‚ Mg(OCOCF3)2
NaBH4‚(D)-valinol
Diisopinocampheylborane
LiAlH4‚(R)-BINOL‚ 2 eq EtOH
LiAlH4‚(S)-BINOL‚ 2 eq EtOH
LiAlH4‚Chirald
-78 °C f rt
-78 °C
-78 °C
-78 °C
-78 °C f rt
0 °C f rt
0 °C f rt
-78 °C f rt
LiAlH4‚(S)-2-anilinomethyl)pyrrolidine
LiAlH4‚(l)-menthol
t-BuNH2‚BH3/Mg-(+)-tartrate
L-selectride/Mg-(+)-tartrate
Hydrogenation Conditions:c
33
34
35
36
37
(R)-(S)-BPPFOH‚[Rh(COD)Cl]2, 1 eq TEA
(S,S)-BPPM‚[Rh(COD)Cl]2, 1 eq TEA
(-)-DIOP‚[Rh(COD)Cl]2
EtOH
EtOH
EtOH
EtOH
EtOH
rt
rt
rt
rt
rt
73:27
69:31
60:40
60:40
61:39
(S)-BINAP‚[Rh(COD)Cl]2
(R)-(+)-PROPHOS‚[Rh(COD)Cl]2
a In general, the reductions were carried out by dissolving 2 in the specified solvent and temperatures and adding 1-3 molar equiv of the reducing reagent,
depending on whether the free base or the HCl salt of 2 was used. b The reactions samples were quenched with MeOH, and the diastereomeric ratio was determined
by capillary GC.4 c All hydrogenations were carried out at 50 psi H2 with substrate:catalyst ratio of 50:1. Hydrogenation times were 1-6 days.
complex was clearly demonstrated by the fact that Li-
(OCOCF3)2, MgCl2, CF3CO2H or the substitution of (i-
Pr)2NEt‚BH3 for (i-Pr)2NH‚BH3, under identical reaction
conditions, gave the normal selectivity (0-4% de) for the
R, R-diastereomer. A similar, but smaller selectivity for the
S,R-diastereomer was also observed for the t-BuNH2‚BH3
complexes with Mg(OCOCF3)2 and Ti(O-i-Pr)4.
A possible explanation for the unexpected reversal of
selectivity for these cases, as well as the previously men-
tioned example with NaBH4/Ti(O-i-Pr)4 in THF, may be that
the reduction occurs from the same face (“front-side”) attack
as the metal in the N-R-configuration five-membered transi-
tion state depicted in Scheme 2 (assuming the chelated
transition state). Use of more polar solvents, such as MeOH
or higher temperatures, which presumably breaks up the
complex, would then account for the selectivity favoring the
R,R-diastereomer. In the cases with NaBH4, Zn(BH4)2 and
the bulky Selectride reducing reagents, the reduction may
be proceeding via a nonchelated transition state as proposed
by Hartley.3 The possibility remains that competitive reduc-
tion mechanisms at different rates could be occurring in all
cases, leading to the above results.
Since the nonchiral reducing reagents did not give
appreciable selectivity for the desired R,R-diastereomer, a
number of chiral reducing reagents were also investigated
and these results are listed in Table 2. Itoh et al.5 had reported
that complexes formed from amino acids such as (D)-proline
and NaBH4 were effective for asymmetric reduction of amino
ketones. A number of these types of complexes were
examined (entries 1-21) and the best results were obtained
using either the NaBH4‚(D)-valine or the NaBH4‚(D)-tert-
leucine complexes in which case de’s of 40-44% were
obtained. Interestingly, the addition of Mg(OCOCF3)2 had
a similar effect as when added to the achiral reducing
reagents. As before, the selectivity was reversed (Compare
entries 19 and 21).
Several other chiral reducing reagents, such as Brown’s6
chiral isopinocampheylchloroborane reagent, as well as the
chiral LiAlH4 complexes formed from BINOL,7 Chirald,8
(5) Umino, N.; Iwakuma, T.; Itoh, N. Chem. Pharm. Bull. 1979, 27, 1479.
(6) Chandrasekharan, J.; Ramachandran, P. V.; Brown, H. C. J. Org. Chem.
1985, 50, 5446.
(7) BINOL is 1,1′-bi-2-naphthol. See Noyori, R.; Tomino, I.; Tanimoto, Y.;
Nishizawa, M. J. Am. Chem. Soc. 1984, 106, 6709.
Vol. 4, No. 2, 2000 / Organic Process Research & Development
•
109

(l)-menthol9 and (S)-2-(anilinomethyl)pyrrolidine10 were tried
(entries 23-32). Although both the Chirald and (l)-menthol
complexes showed selectivity for the undesired S,R-diaste-
reomer, the remaining reagents showed selectivity for the
desired R,R-diastereomer, with the (S)-BINOL‚LiAlH4 com-
plex affording the best result yet observed for the reduction
of amino ketone 2, with a de of 62%.
Scheme 3
Hiyashi et al.11 reported that catalytic hydrogenation of a
secondary amino ketone (similar to 2) proceeded in excellent
yield with 95% ee using the (R)-(S)-BPPFOH‚Rh chiral
complex. This ligand, as well as a number of other chiral
hydrogenation catalysts were also tried, including (S,S)-
BPPM,10 (-)-DIOP,11 (S)-BINAP,12 and (R)-(+)-PRO-
PHOS,13 and these results are also depicted in Table 2 (entries
33-37). The best results were obtained using the BPPF-
OH ligand, in which case a 46% de was observed.
were anticipated to introduce economic burden and create
complications17 in the isolation of acceptable quality 1,
especially when tight specifications for the final product were
already set.
Although the selectivity for reduction of aminoketone 2
was as high as 63:37 with the achiral reducing reagents and
as high as 81:19 with the chiral reducing reagents, the fact
remained that recrystallization of 1 with L-DBTA in aqueous
butanol or ethanol was necessary to improve the enantiomeric
and chemical purities leading to acceptable quality 1.
However, under the best crystallization conditions the
recovery of the desired product was partial, and it left 4 (with
∼1:3 mixture of R,R- to S,R-isomer) in the mother liquor.
To make this process more economical by utilizing the chiral
amine moiety more efficiently, a route for obtaining ad-
ditional 1 from the mother liquors was desirable. For this
purpose three options were considered: (A) A preferential
conversion of the S,R-isomer to the R,R-isomer; (B) the
oxidation of the benzylic alcohols of 4 to the preexisting
penultimate intermediate 2 which could then be reduced; or
(C) a nonselective, complete racemization of the benzylic
carbinol, that is, epimerization, of 4 (1:3 mixture of RR:SR
isomers) to 3a (1:1 mixture of RR:SR). Although less
efficient than the first alternative, the last two allow for a
convenient recycling via the existing process.
For option A the previously known ephedrine to ψ-ephe-
drine conversion18 was appealing since if it could selectively
invert one of the two isomers (ideally inversion of the S
center of the S,R-isomer), it could lead to the most desirable
process. As a control, this inversion was applied to (R,R)-1.
At the end of the reaction sequence (Scheme 3, shown for
ephedrine), instead of recovering either unchanged (R,R)-1
(a desirable outcome) or the inverted product (S,R)-1 (an
undesirable outcome), compound 3a (1:1 mixture of RR:SR)
was obtained! Thus, this approach, where the epimerization
of the benzylic alcohol had taken place, was undesirable for
a one-step conversion of 4 to 1.
For option B, in control experiments, the oxidation of 1,
a â-amino alcohol, to the corresponding aminoketone 2 was
investigated. This oxidation proved difficult. Under a variety
of oxidation conditions19-32 significant discoloration along
(17) Protection of the phenolic OH, and the amine nitrogen could have overcome
some of these problems and the ones perceived for the oxidations. However,
from a commercial point of view this need for extra protection/deprotection
was unattractive.
(18) Welsh, L. H. J. Am. Chem. Soc. 1949, 71, 3500.
An additional approach involving a Mitsunobu type
inversion16 which could invert both alcohols of 4 to a mixture
of 3:1::RR:SR isomers, was ruled out for the following
reasons. The cost of Mitsunobu reagents and the removal of
the byproducts resulting from the use of Mitsunobu reagents
as well as the one resulting from the participation of the
amine or the phenolic group during the inversion reaction
(19) (a) Mancuso, A. J.; Huang, S.-L.; Swern, D. J. Org. Chem. 1978, 43, 2480.
(b) Barton, H. D. R.; Garner, B. J.; Wightman, R. H. J. Chem. Soc. 1964,
1855.
(20) (a) Firouzabadi, H.; Ghadedri, E. Tetrahedron Lett. 1978, 839. (b) Garigipati,
R. S.; Freyer, A. J.; Whittle, R. R.; Weinreb, S. M. J. Am. Chem. Soc.
1984, 106, 7861.
(21) (a) Frechet, J. M. J.; Darling, P.; Farrall, M. J. J. Org. Chem. 1981, 46,
1728. (b) Frechet, J. M. J.; Warnock, J.; Farrall, M. J. J. Org. Chem. 1978,
43, 2618. (c) McKillop, A.; Young, D. W. Synthesis 1979, 401.
(22) Cainelli, G.; Cardillo, G.; Orena, M.; Sandri, S.J. Am. Chem. Soc. 1976,
98, 6737.
(8) Chirald is (2S,3R)-(+)-4-(dimethylamino)-1,2-diphenyl-3-methyl-2-butanol.
See Yamaguchi, S.; Mosher, H. S. J. Org. Chem. 1973, 38, 1870.
(9) Andisano, R.; Angeloni, A. S.; Marzocchi, S. Tetrahedron 1973, 29, 913.
(10) Mukaiyama, T.; Asami, M.; Hanna, J.-i.; Kobayashi, S. Chem. Lett. 1977,
783.
(11) (R)-(S)-BPPF-OH is (R)-R-[(S)-2-bis(diphenylphosphine)ferrocenyl]ethanol.
See Hayashi, T.; Katsumura, A.; Kinishi, M.; Kumada, M. Tetrahedron
Lett. 1979, 20, 425.
(12) (S,S)-BPPM is (2S,4S)-(-)-1-tert-butoxycarbonyl-4-(diphenylphosphino)-
2-(diphenylphosphinomethyl)pyrrolidine. See Achiwa, K.; Kagure, T.;
Ojima, I. Tetrahedron Lett. 1977, 18, 4431.
(13) (-)-DIOP is (2R,3R)-(-)-2,3-O-isopropylidene-2,3-dihydroxy-1,4-bis(di-
phenylphosphino)butane. See Kagan, H. B.; Dang, T.-P. J. Am. Chem.. Soc.
1972, 94, 6429.
(23) (a) Highet, R. J.; Wildman, W. C. J. Am. Chem. Soc. 1955, 77, 4399. (b)
Minor, J. J.; Vanderwerf, C. A. J. Org. Chem. 1952, 17, 1425.
(24) (a) Cacchi, S.; La Torre, F.; Misiti, D. Synthesis 1979, 356. (b) Gelbard,
G.; Brunelet, T.; Jouittbau, C. Tetrahedron Lett. 1980, 21, 4653. (c)
Santainello, E.; Ferraboschi, P. Synth. Commun. 1980, 10, 75.
(25) (a) Brown, H. C.; Garg, C. P. J. Am. Chem. Soc. 1961, 83, 2952. (b) Rao,
Y. S.; Filler, R. J. Org. Chem. 1974, 39, 3304.
(26) (a) Djerassi, C. Org. Reactions. 1951, 6, 207. (b) Easthan, J. F.; Teranishi,
R. Organic Syntheses; Wiley & Sons: New York, 1963; Collect. Vol. IV,
p 192.
(27) McKillop, A.; Ford, M. E. Synth. Commun. 1972, 2, 307.
(28) Cella, J. A.; McGrath, J. P. Tetrahedron Lett. 1975, 4115.
(29) Menger, F. M.; Lee, C. J. Org. Chem. 1979, 44, 3446.
(30) (a) Singh, R. P.; Subbarao, H. N.; Dev, S. Tetrahedron 1979, 35, 1789. (b)
Santaniello, E.; Ponti, F.; Manzocchi, A. Synthesis 1978, 534.
(31) (a) Bowden, K.; Heilbron, I. M.; Jones, E. R. H.; Weedon, B. C. L. J. Chem.
Soc. 1946, 39. (b) Bowers, A. Halsall, T. G., Jones, E. R. H.; Lemin, J. A.
J. Chem. Soc. 1953, 2548.
(14) (S)-BINAP is (S)-(-)-2, 2′-bis(diphenylphosphino)-1, 1′-binaphthyl. See
Takaya, H.; Noyori, R. et al. J. Org. Chem. 1986, 51, 629.
(15) (R)-PROPHOS is (2R)-(+)-bis(1,2-diphenylphosphino)propane. See Fryzuk,
M. D.; Bosnich, B. J. Am. Chem. Soc. 1978, 100, 5491.
(16) Mitsunobu, O. Synthesis 1981, 1.
(32) Parikh, J. R.; Doering W. J. Am. Chem. Soc. 1967, 89, 5505.
110
•
Vol. 4, No. 2, 2000 / Organic Process Research & Development

Scheme 4
conditions as determined by entry 2. The other decomposition
products were subsequently identified as shown in entries
3,4 below. Realizing that the instability of L-DBTA to
aqueous mineral acids would not allow for a practical kinetic
resolution or epimerization, removal of L-DBTA was deemed
necessary for the conversion of 4 to 1. Thus, after the removal
of L-DBTA as a Na salt, crude 4 in water/alcohol was
subjected to acidic epimerization (Table 4, entries 3,4). In
both cases up to 20% of n-butyl ethers 13a formed which
cocrystallized with 1‚L-DBTA at the end. A confirmation
(MS, NMR) of these structures was made via the synthesis
of these compounds in alcohol as a solvent as shown below
in Scheme 6.36 The use of milder acids, such as pTSA or
CF3COOH in place of H2SO4 also led to the formation of
these ethers. The use of EtOH in place of n-butanol resulted
in the ethyl ethers 13b. Thus, the removal of alcohols and
L-DBTA from the mother liquors, that is, formation of free
base 4, became necessary for the epimerization to generate
good quality product.
With free base 4 as the substrate, essentially all of the
successful conditions listed in Table 3 became a viable
option. Although HCl led to a faster epimerization under
the conditions evaluated, the use of HCl was less desirable
than H2SO4 as it would restrict the type of equipment that
could be used in the plant, and furthermore, the recovery of
the product was somewhat low as it generated 12 (HCl-
induced amide hydrolysis is well-documented in the litera-
ture). On the basis of the purity profile of the epimerized
mixture, the rate, and the cost of reagents, it was determined
that sulfuric acid was the acid of preference at this stage,
and the subsequent optimization work was done with it. In
experiments 22-24 (Table 3) the addition of 1% glacial
acetic acid helped solubilize the substrate without affecting
the rate or the recovery, whereas the addition of 1 mol equiv
of 2.5 N HCl improved the rate of epimerization while
typically lowering the recovery of 3 by ∼3-5%. Further
optimization work revealed that with aqueous H2SO4 the
reaction concentration of 5-25% (some suspension) had no
significant effect on the rate of epimerization of pure 1 or
on the recovery of 3a. Naturally, higher reaction concentra-
tions meant more throughput and less neutralization at the
end of the reaction, and hence it was preferred. These
conditions were evaluated for the epimerization of free base
4 (Table 4, entries 5-7). Since crude, free base 4 was more
with a mixture of products resulted. Although the aminoke-
tone was present in most of these reactions, its isolation
without chromatographic purification was difficult.17 The
inability of this process to produce the desired product which
could be readily isolated forced a reinvestigation of the
epimerization observed above.
As for option C, observations were made earlier in the
process33 that the formation of hydrochloride salt of the
(R,R)-1 under nonambient conditions with aqeous HCl had
led to a lowering of de’s by a partial epimerization of the
alcohol center. This fact, coupled with the epimerization
observed under option A, indicated that it was probably the
aqueous acidic conditions (a step common to both proce-
dures) that caused the epimerization of the alcohol center
through the intermediacy of the quinonemethide species10
and its tautomer 9, (Scheme 4) with or without the involve-
ment of the aziridine species 11.34
This prompted the development of option C, a complete
epimerization of the amino alcohol 4 in the mother liquor to
3a. To develop the above observation into a process, the
efforts were divided in two parts. These were (i) the
identification of ideal parameters for epimerization of 1, and
(ii) the application of these parameters to the crude mixture
of the isomers in the core process mother liquor (which
contained excess L-DBTA and butanol) for the conversion
of 4 to 1 via 3a.
Epimerization Parameters. To establish the parameters
for acid-induced epimerization, initial small-scale laboratory
control experiments were conducted on pure 1, and the results
are summarized in Table 3.
It was clear from the above experiments that aqueous
conditions, lower pH, and moderately high temperature were
necessary for an efficient epimerization. This is in agreement
with the proposed quinonemethide-based mechanism.
Application of Epimerization for the Preparation of 1
from 4. Although a preferential conversion of the S,R-isomer
to R,R-isomer was not possible, theoretically a kinetic
resolution of 4 to 1 as depicted in Scheme 5 was possible.
Since a crystallization of 1 from 3a with L-DBTA can be
accomplished in aqueous butanol (or ethanol), additional
L-DBTA was added to the mother liquor containing 4 prior
to acid-catalyzed epimerization (entries 1,2 of Table 4). In
these cases only a trace of solid was visible at the end of
the reaction. HPLC35 and GC4 analysis of the reaction
mixtures showed epimerization along with a large number
of new products. Some of these new compounds were the
result of decomposition of L-DBTA under the experimental
(34) It is possible that this process may be favored by protonation of the amine
under the acidic conditions which could enhance the proton transfer to the
benzylic hydroxy group, accelerating the quinonemethide formation.
(35) HPLC: (a) µ-Bondapak C-18 or equivalent, (30 cm × 4 mm), Methanol:
0.07 M KH2PO4 (55:45) adjusted to pH 7.5 with 1 N aqueous NaOH, flow
rate 1 mL/min, 254 nm detection. Internal standard: o-phenylphenol; or (b)
Waters phenyl column 25 cm × 4 mm; 40% acetonitrile/water containing
0.01% sodium decanesulfonic acid; flow rate 1 mL/min; 254 nm.
(36) HPLC: Under the system described in ref 35 above, the butyl ethers
(resolved) had an average RRt of 3.95, whereas the ethyl ethers (a broad
hump) had a RRt of 1.8 (compared to 1 with Rr of 1.0). GC: Under the
system described in ref 4a, the butyl ethers had a RRt of 1.28, whereas the
ethyl ethers had RRt of 0.84 (compared to 1 with Rr of 1.0).
(33) Personal communication with Dr. Esther Babad. The author (D.G.) greatly
appreciates many stimulating discussions with Dr. Babad on this
project.
Vol. 4, No. 2, 2000 / Organic Process Research & Development
•
111

Table 3. Epimerization of pure 1
no.
substrate
conditionsa
RR:SR ratiob
comments
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
1‚HCl
1‚HCl
1‚HCl
1‚HCl
1‚HCl
1‚HCl
1‚HCl
1‚HCl
1‚HCl
1‚HCl
1‚HCl
1‚HCl
1‚HCl
1‚HCl
1‚HCl
1‚HCl
1‚HCl
1
1
1
1
1
1
1
1
1
EtOH/H2O, reflux, 3 days
i-PrOH/H2O, reflux, 3 days
n-BuOH, reflux, 3 h
cat. pTSA, EtOAc, reflux, 3 h
cat. pTSA, iPrOH, reflux, 3 h
gl. CH3COOH, 85°, 8 h
50% aq. CH3COOH, 90°, 2 days
50% aq. CH3COOH, 110°, 50 psi, 21 h
50% aq. CH3COOH, 130°, 65 psi, 15 h
CF3COOH, 55 °C, 8 h
<5% epimerization
<5% epimerization
<5% epimerization
<5% epimerization
<5% epimerization
<5% epimerization
1:1
6:4
1:1
53:47
1:1
1:1
1:1
55:45
1:1
1:1
1:1
1:1
53:47
1:1
1:1
1:1
9:1
95:5
56:44
60:40
73:27
1:1
dec
trc. dec
80% yield
50% aq. CF3COOH, 90 °C, 1 day
50% aq. HCOOH, reflux, 32 h
90% HCOOH, reflux, 18 h
40% yield
87% yield
89% yield
89% yield
82% yieldc
78% yieldc
83% yield
98% yield
95% yield
92% yield
50% aq. CF3COOH, 85°, 8 h
50% (20N) aq. H2SO4, 45°, 3 h
25% (10N) aq. H2SO4, 50°, 3 h
10 N(83%) aq. HCl, 50°, 3 h
10 N (83% aq.) HCl, 60°, 5 h
6 N (50% aq.) HCl, 60°, 5 h
20 N (50%) aq. H2SO4, 40°, 8 h
20 N (50%) aq. H2SO4, 50°, 6 h
10 N (25%) aq. H2SO4, 55°, 3 h
5 N (12%) aq. H2SO4, 55°, 2 h
2.5 N (6%) aq. H2SO4, 50°, 6 h
40% aq. HNO3, 40°, 3 h
concentrated HNO3, rt, 1 h
50% aq. H3PO4, 40°, 26 h
10 N (25%) aq. H2SO4, 65°, 2.5 h
10 N (25%) aq. H2SO4, 70°, 2 h
10 N (25%) aq. H2SO4, 100°, 1 h
58% yield
67% yield
92% yield
88% yield
75% yield
44% yield
1
1
1
1
1:1
1:1
a Screening experiments were done at 2-5% reaction concentration under the conditions described above. Scale up of the H2SO4 epimerization were done at up to
25% reaction concentration which is described in the text. Where applicable, the RR:SR ratios were determined with GC,4 and quantitations performed with HPLC.35
b Ideally a complete epimerization within a period of 4-5 h was desirable. Thus, recovery and yield determination was conducted only for selected experiments.
c 5-10% of compound 12 formed in these reactions.
Scheme 5
process labor intensive as well as volumetrically inefficient
overall. Hence, a more practical alternative was sought.
It appeared that, if the substrate, a secondary base, could
be retained on a sulfonic acid resin, it would allow for a
thorough washing of the resin with n-butanol followed by
water for the removal of L-DBTA and n-BuOH, respectively
(both of which could be recovered and recycled as detailed
later). Additionally 4 could be epimerized on the resin by
heating with steam/hot water (to attain 50-60 °C). A pH
adjustment of the resin with a minimum amount of base
could elute the epimerized product from the resin which
could then be regenerated, effectively serving as a catalyst.
Thus, the use of resin could lead to an elegant, environmen-
tally friendly, economical epimerization process. To evaluate
this possibility the following control experiments were
conducted. When an aqueous solution of 1 was treated with
Amberlite IR-120 resin, none of the substrate remained in
the aqueous layer as judged by HPLC. Naturally, the same
was true for the aqeous n-BuOH containing substrate.
Encouraged by these results we designed a systematic
program to convert this observation into a viable process.
Salient features of this work are described below.
soluble than pure 1, no additional reagents were needed for
its solubilization.
The outcome for the acid-induced epimerization of 4
paralleled the results obtained when pure 1 was used as a
substrate for epimerization. After epimerization and workup,
1‚L-DBTA was isolated, leaving second generation 4 in the
mother liquors. This second generation 4, after removal of
L-DBTA and n-butanol was subjected to a second epimer-
ization with > 80% overall recovery of 3a for the second
cycle. This established the much desired recyclability of the
process.
In principle, on the basis of the quality of isolated 1 and
the mass balance during epimerization, the above process
appeared attractive. However, the need for the removal of
L-DBTA and n-butanol from the mother liquors, the neu-
tralization of H2SO4 after epimerization followed by extrac-
tion of 3a, and finally isolation of the desired product by a
partial crystallization of 1 as a L-DBTA salt made this
112
•
Vol. 4, No. 2, 2000 / Organic Process Research & Development

Table 4. Epimerizationa of the material 4 from core process
mother liquors: (SR:RR::∼75:25)
Table 5. Loading capacities of various sulfonic acid resins
saturation
H+
capacity
no.
substrate
conditions
RR:SR comments
amino
capacity g/L of
eq/L of wet resin
% cross alcohol
1
4‚L-DBTA,
n-BuOH
4‚L-DBTA
aq. TFA, 90°, 8 h
aq. TFA, 70°, 6 h
1:1
1:1
dec
dec
resin typea
IR 120
IR 118
IR 200
XF8 43280
XUR 0525-L87-109-1 4.5
Ionac 241-T
XFS 43279
XFS
linkage loaded g/Lb wet resin
(g)
2
3
4 aq. n-BuOHb 10 N (25% aq.) H2SO4, 1:2d
40°, 2 h
8.0
8.0
8.0
6.0
100
100
114
165
162
206
220
170
1.9
1.3
1.7
2.1
1.1
1.1
1.1
0.5
624
427
558
690
361
361
361
165
4
5
6
7
4 aq. n-BuOH 10 N (25% aq.) H2SO4, 1:1d 63% yield
40°, 4 h
free base 4c
free base 4
free base 4
10 N (25% aq.) H2SO4, 1:1
55°, 2 h
5 N (12% aq.) H2SO4,
55°, 6 h
10 N (87% aq.) HCl,
55°, 3 h
87% yield
4.0
3.5
2
1:1
83% yield
1:1e 70% yield
a Resins were stirred with excess of amino alcohol 4 (FW 328.4) for ∼18 h
at room temperature. b The concentration of amino alcohol was determined by
HPLC.35b
a Epimerizations conditions were similar to the ones described under Table
3. b Aqueous n-BuOH solution were prepared as follows: The mother liquors
from core process (n-BuOH saturated with water) were treated with sat. aq.
Na2CO3 to pH 8-9. The resultant layers were separated, the aq. layer was
extracted with EtOAc and combined with the organic layer. The combined
organic layer was concentrated to remove EtOAc and the remaining n-BuOH
solution of 3a was used for epimerization. c Free base 4 was prepared as follows:
The n-BuOH mother liquors were concentrated to a semisolid under vacuum at
35 °C. To this was added EtOAc followed by sat. aq. Na2CO3 such that the pH
of the aq. layer remained at 8-9. The EtOAc layer was separated and dried
under vacuum to a light tan solid which was the used for epimerization. d n-
Butyl ethers 13a formed. e Amide hydrolysis to the carboxylic acid 12 accounted
for the mass balance.
perature, the rate of addition of mother liquors to the resin,
the rate of agitation of the resin, can also influence the mass
transport and hence the degree as well as the rates of
adsorption and epimerization. All except the rates of agitation
(kept at minimum to allow for a column process mode, and
to minimize physical degradation of the resin) were studied.
The Resin Loading Capacity As a Function Of Cross
Linkage and H+ Capacity. The resin loading capacity was
determined by gently stirring the commercially available wet
resins with a large excess of typical core process mother
liquor at room temperature for an overnight period. The total
amino alcohol in solution before and after resin treatment
was determined with HPLC. The results of this study are
shown in Table 5.
Scheme 6
The data indicated that the extent of the amino alcohol
loading on the resin was primarily influenced by the degree
of cross-linkage, whereas the H+ capacity of the resin played
a minor role. The resins with 6-8% cross linkage resulted
in inefficient loading. On the other hand XFS appeared
attractive from the loading perspective, but with all of its
acidic sites occupied it would not allow for epimerization
without exogenous acid. Additionally, mass balance deter-
mination in this case indicated adsorption of n-BuOH on the
resin as well, which could be detrimental as it could form
the butyl ethers. Thus, XFS 43279, Ionac 241-T, and the
XUR resins were chosen for further optimization. After some
additional work on these resins (similar to the one described
below for XFS 43279) and commercial considerations, XFS
43279 was chosen as the resin of choice. The rest of this
manuscript details the use of this resin for the development
of the epimerization process.
Rate of Amino Alcohol Loading on the Resin. Experi-
ments were undertaken to determine the optimum loading
parameters for Dowex XFS 43279(H+). Here the resin was
gently agitated with a large amount of mother liquor
containing excess amino alcohol 4 at 23°, 30°, 50°, 60°, 70°,
80°, and 90 °C in a batch mode. This study indicated that
the loading was temperature-dependent with optimum load-
ing attained at 40 °C. Further increase in loading temperature
either decreased the loading, or at 70 °C and above led to
ethers formation. The relevant results from this study are
summarized in Table 6 and additional information is provided
Resin Studies. To minimize the large number of experi-
ments needed to determine the optimum resin, as well as
the parameters that could be used with the selected resin(s)
for the desired process, some practical assumptions were
made up front. It was decided that although the core process
was still being optimized, the typical mother liquors to be
used were essentially of similar composition (within (8%
variation in the numbers described here, i.e., consisting of
85% n-butanol/water with the R,R-amino alcohol, the S,R-
amino alcohol, and the L-DBTA components in approxi-
mately 12, 38, and 55 g/L amounts, respectively) to the one
that would be eventually used. Only the commercially (bulk)
available37 sulfonic acid resins would be screened for this
study. For these resins, the following interdependent proper-
ties, available from the manufacturer of the resins, were
considered likely to influence the epimerization process. In
general, the resin capacity, degree of cross linkage, particle
size, the size of the cavities (resulting from the cross-linkage
of the polymer in the resin bed; in wet resins the hydrated
sulfonate groups swell, whereas the presence of alcohols
depletes the water in the cavities, resulting in shrinkage of
the resin) all influence the mass transport as well as
adsorption of amino alcohol 4 on the resin. Furthermore, the
parameters controlled in the laboratory, for example, tem-
(37) The IR resins were obtained from Rohm and Haas Co, Philadelphia, Pa.
The rest of the resins were obtained from the Dow Chemical Co, Larkin
Laboratory, Midland, Michigan.
Vol. 4, No. 2, 2000 / Organic Process Research & Development
•
113

Table 6. Temperature effect on amino alcohol loading (g/L)
at 90 °C for 2 h resulted in lower (84%) recovery of 3
compared to 95% recovery of 3 after 1 h at 90 °C. Thus,
with the resin loaded at ∼50% of the capacity, a temperature
of 85-90 °C for a period of about 1 h was considered
optimum for epimerization.
for XFS 43279a
time (h)
23 °C
30 °C
40 °C
50 °C
60 °C
0
0
105
111
127
124
133
140
144
148
154
160
163
160
0
139
148
149
153
160
162
166
167
174
178
177
179
0
150
155
162
170
171
173
180
179
181
182
180
181
0
167
170
171
172
170
171
170
168
167
164
160
160
0
158
154
154
151
151
149
150
151
150
150
149
151
1.0
1.25
1.5
2
2.5
3
3.5
4.0
5
Resin Loading vs Epimerization. During the above resin
loading recovery study it became clear that the resin not only
allowed for the removal of n-BuOH as well as the expensive
L-DBTA, by selectively absorbing the amino alcohols, but
it also allowed for the removal of the remainder of the
impurities formed during the preceding portion of synthesis
of 3a. In reality it was also observed that XFS 43279 allowed
for the above even when it was loaded near its saturation
point with respect to its H+ capacity. Thus, in the next set
of experiments, the entire process of loading, epimerization,
and recovery was evaluated on the above resin using under-
loaded to over-loaded resin. As shown in Table 9, over-
loaded resin led to a very slow epimerization upon heating
of the resin. Prolonged heating of 90% loaded resin did not
drive the epimerization to completion. It was obvious at this
stage that the addition of (sulfuric) acid could enhance the
rate of epimerization. The commercial feasibility of such a
process was also seriously evaluated, and the results of this
work are discussed below.
These experiments also demonstrated that the remainder
of the mass balance under prolonged or over-heated condi-
tions in the resin-induced epimerization of the amino alcohol
was accounted for by the formation of 12, a compound not
seen during aqueous H2SO4 epimerization. Compound 12 is
not readily absorbed on the resin and does not crystallize as
a L-DBTA salt under the conditions in which 1‚L-DBTA is
crystallized. Thus, although its formation during this process
lowers the yield, it did not interfere with isolation of
acceptable quality 1.
High-Loading Exogenous Acid Epimerization. It ap-
peared that 90% loading offered an opportunity to increase
the throughput only if epimerization could be completed.
With 90% of the acidic sites occupied with the amino
alcohol, the remaining 10% acidic sites afforded limited
opportunity for epimerization. The addition of a limited
amount of an exogenous acid such as H2SO4 (or less
preferably, HCl), could lead to complete epimerization
through synergism. In control experiments, introduction of
either HCl or H2SO4 resulted in complete epimerization.
For the use of H2SO4 three variables were examined:
temperature, acid concentration, and reaction time. The
results of this study are summarized in Table 10. Note that
for these experiments the plots of the log of enantiomeric
excess vs time were linear. Such plots led to the determi-
nation of epimerization half-life which is listed in the table.
The linearity of reaction rates to acid strength further supports
a simple acid-catalyzed quinonemethide mechanism for the
epimerization reaction.
6
7
8
a The results were obtained using the analytical procedures described under
Table 5.
in Figure 1. From these experiments, an optimum loading
period of 3.5 h at a temperature of 40 °C was selected for
∼50% resin saturation (which would allow for epimerization
without additional acid).
Recovery Of Amino Alcohol Loaded on the Resin.
Although the resin loading was determined by measuring
the decrease in the amino alcohol content of the mother liquor
after resin treatment, it was also important to know the
quality and the ease of recovery of the amino alcohol from
the resin. A systematic study was undertaken to optimize
the recovery of the absorbed compound from the resin. The
results of this work are summarized in Table 7.
The total recovery as a function of NaOH molarity and
bed-volumes is shown in Figure 2 below. Note that the first
bed-volume of 3 M NaOH recovered slightly more amino
alcohol than 2 M NaOH. Between 2 and 3 M NaOH, (both
of which led to excellent recovery of the amino alcohol), 2
M NaOH was desirable as it meant less shock to the resin,
and less acid use for the subsequent pH adjustment. This
eventually resulted in a better crystallization of 1‚L-DBTA
(discussed later) compared to 3 M NaOH.
The Rates Of Epimerization. Having identified the
optimized loading and recovery parameters, the rate of
epimerization on the loaded column was next investigated.
Here two possibilities were explored. Epimerization without
additional acid was the first choice as it could be better suited
for resin stability, and also minimize unit operations. On the
other hand, epimerization with additional acid would improve
the throughput.
Epimerization Without Additional Acid. On the basis
of the observations made with sulfuric acid, it was clear that
temperature would significantly influence the rate of epimer-
ization of the amino alcohol loaded on the Dowex XFS
43279 resin. Hence the rate of epimerization as a function
of temperature was studied. The results are summarized in
Table 8 and depicted in Figure 3.
From these data, epimerization conditions of 1.2 N
H2SO4/80 °C/2.5 h and 0.6 N H2SO4/90 °C/1.5 h were
considered to be reasonable conditions for large-scale work
and were chosen for yield determination. At 90% capacity
loading, epimerization at 90 °C with 0.6 N H2SO4 gave 82%
This study indicated that at 60 °C the epimerization was
incomplete even after prolonged heating (Figure 3), whereas
at 70, 80, and 90 °C, a complete epimerization was
accomplished in 9, 4, and 1 h, respectively. In separate side-
by-side comparisons it was determined that extended heating
114
•
Vol. 4, No. 2, 2000 / Organic Process Research & Development

Figure 1. Temperature effect on loading.
Table 7. Per cent recovery of free base vs molarity of sodium hydroxide in n-butanola
cycle
layer
1 M
2 M
3 M
4 M
5 M
1
2
3
4
5
n-butanol
aqueous
n-butanol
aqueous
n-butanol
aqueous
n-butanol
aqueous
n-butanol
aqueous
51% (1 b.vb)
0.8% (1 b.v)
38%(1 b.v)
0.5%(1 b.v)
3.7%(1 b.v)
0.2%(1 b.v)
0%(0.5 b.v)
0%(0.5 b.v)
0%(0.5 b.v)
0%(0.5 b.v)
86%(1 b.v)
1.1%(1 b.v)
7.4%(1 b.v)
1.0%(1 b.v)
4.9%(1 b.v)
0%(1 b.v)
0%(0.5 b.v)
0%(0.5 b.v)
0%(0.5 b.v)
0%(0.5 b.v)
87.6%(1 b.v)
0.4%(1 b.v)
4.8%(0.5 b.v)
0%(1 b.v)
1.6%(0.5 b.v)
0%(0.5 b.v)
0.9%(0.5 b.v)
0%(0.5 b.v)
0.3%(0.5 b.v)
0%(0.5 b.v)
86.3% (1 b.v)
0.2%(1 b.v)
4.9%(0.5b.v)
0%(0.5 b.v)
1.3%(0.5 b.v)
0%(0.5 b.v)
0.7%(0.5 b.v)
0%(0.5 b.v)
0.2%(0.5 b.v)
0%(0.5 b.v)
81.6%(1 b.v)
0%(1 b.v)
4.3%(0.5 b.v)
0%(0.5 b.v)
1.0%(0.5 b.v)
0%(0.5 b.v)
0.6%(0.5 b.v)
0%(0.5 b.v)
0%(0.5 b.v)
0%(0.5 b.v)
total recovered
93.8%
100%
95.3%
93.6%
87.5%
a For these experiments, saturated n-butanol solutions mixed with aqueous sodium hydroxide solutions of known strength were used to elute the resins in batch
process mode with minimum agitation under nitrogen atmosphere at room temperature for 45 min. The resin was filtered. The pH of the filtrate was adjusted to 8.3
with 6 N H2SO4. The layers were split and analyzed separately by the HPLC procedure listed under Table 5. This cycle was repeated as necessary. b b.v. represents
bed volume.
recovery yield, compared with an 86% yield at 80 °C with
1.2 N acid. While these yields are comparable to that
obtained with 60% loading and no added acid epimerization,
the throughput was increased by approximately by 1.5 times.
In either case, 1‚L-DBTA could be crystallized38 from 3a
obtained at any one of the above resin loadings.
Transferring the Process to the Plant. The experimental
data was used as a foundation to conduct a series of pilot
plant trials to further define the manufacturing process for
the recovery of dilevalol‚L-DBTA. The primary objectives
of the trials were to demonstrate, under plant conditions, the
loading, epimerization, and elution characteristics of the
process; to determine the lifetime of the ion-exchange resin;
and to study the crystallization and filtration characteristics
of the dilevalol‚L-DBTA isolated via the recovery process.
The development effort was primarily focused on imple-
menting the process in a multi-purpose bulk pharmaceutical
chemical facility, utilizing as much existing process equip-
ment as possible to minimize the capital investment. Initially
the pilot plant-scale epimerizations of mother liquors were
conducted using a 12 in diameter × 10 ft ion-exchange
column, with the intention of scaling up to an existing 36 in
diameter column, both of which were available at the time
(Figure 4).
Mother liquors from the core process containing 25 (
5% of the R,R-isomer were used for this work. The mother
liquors were loaded on the resin at ambient temperature (15-
25 °C)39 at a flow rate of about 2 L/min. The resin was loaded
to approximately 160 g/L during the pilot trials (less than
50% loading), with leakage during loading minimized (0-
3%) by reducing the flow rate to the column when the pH
(38) In a typical crystallization, 1.1 mol of L-DBTA and the amino alcohol in
15-20 volumes of n-BuOH saturated with water was heated at 50 °C (pH
of solution was 3.4-3.6). The mixture was cooled to 25 °C with agitation
and filtered. Typically the crystallized product (33-37% yield) contained
>97: 3 ratio of the isomers as determined by GC.
(39) Laboratory optimum of 40 °C would necessitate capital investment, which
was ruled out from commercial point of view. The necessary loading was
achieved by controlling the rate of loading.
Vol. 4, No. 2, 2000 / Organic Process Research & Development
•
115

Figure 2. Recovery of amino alcohol vs NaOH.
Table 8. Rates of epimerization of amino alcohol loaded on
Dowex XFS-43279 (%RR)a,b
considered to be complete. After allowing the column to cool
to ∼45 °C, elution was conducted by pumping 2 N sodium
hydroxide saturated with 12% v/v n-butanol through the
column at a flow rate of about 2 L/min. The first one or two
bed-volumes of the column effluent contained no product.
When the pH increased to above 8 the eluate was collected
for product recovery. The next two bed-volumes of the eluate
contained about 60% of the epimerized amino alcohols.
However, an additional four or five bed-volumes were
necessary to increase the recovery to above 80%. Typical
elution profiles from the pilot trials are shown in Figure 6
for epimerizations at both 80 °C and 90 °C.
One of the difficulties in the column elution was the
precipitation and oiling of the product in the column effluent,
presumably due to high localized concentrations of amino
alcohol in the resin bed. Additional n-butanol during elution
did not significantly increase product solubility and recovery
from the column process. A further complication in the
column scale-up effort was the variability of the process yield
with respect to the epimerization temperature. In the column
operation, the yield for epimerization at 80 °C was 88-
100%, whereas at 90 °C it was 73-92%. Some of this was
due to the formation of 12, as seen in the laboratory batches,
where as the rest appeared to be resulting from inefficiencies
in eliminating oxygen and n-butanol from the system, causing
degradation of the product on the column.
time (h)
60 °C
70 °C
80 °C
90 °C
0
1
2
3
4
5
6
7
8
33.0
35.0
36.0
36.7
38.9
40.0
40.9
41.5
42.0
42.6
42.9
43.3
43.6
33.0
35.5
38.9
41.9
44.4
46.9
48.0
48.9
49.0
49.3
49.5
49.7
49.7
33.0
46.0
47.9
49.0
48.9
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
33.0
49.6
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
9
10
11
12
a In a batch operation mode, the resin loaded with amino alcohol was filtered.
(The filtrates were saved for the recovery of L-DBTA and n-butanol, which is
described later in this manuscript). The resin was washed with n-BuOH and
water to remove the residual amounts of L-DBTA and n-butanol, respectively.
Additional water was added, and the resin was heated under nitrogen at the
indicated temperatures in the graphs. At the time points, resins were removed
and treated as described under the recovery section to elute the amino alcohol.
GC procedure4b was used to determine RR:SR ratios. b The need of nitrogen
atmosphere for the epimerization process was known from the following
experiments which had been conducted prior to the above study. A resin
sample loaded with amino alcohol was split into two equal portions. One-half
was epimerized under air and the other under nitrogen at 80 °C for 5 h. The
yield of amino alcohol recovered from the resin under nitrogen was 98%
compared to 78% under air. The presence of air results in the amino alcohol
degradation.
After adding n-butanol to the combined eluants to ensure
the proper amino alcohol concentration for crystallization
of the dilevalol‚L-DBTA, the pH of the eluate was adjusted
from a pH g 12 to pH 8.5 ( 0.1 using sulfuric acid, and the
product was extracted into the n-butanol layer for crystal-
lization. After the addition of L-DBTA, the dilevalol‚L-DBTA
that crystallized from the epimerization process was more
difficult to filter than that from the core process. Investigation
of this difficulty revealed that since large volumes of eluates
were generated from the column, they required correspond-
of the spent mother liquors began to rise, indicative of
increasing concentration of the amino alcohol in solution.
The typical leakage of amino alcohol during mother liquor
loading is shown in Figure 5.
After the resin was washed with n-BuOH and water (to
remove L-DBTA and n-BuOH, respectively), epimerization
was achieved by passing heated deionized water through the
resin column at temperatures between 80 °C to 90 °C. When
the isomer ratio of the product reached 50 ( 3% RR, the
heating was stopped as at this stage epimerization was
116
•
Vol. 4, No. 2, 2000 / Organic Process Research & Development

Figure 3. Temperature effect on rates of Epimerization.
Table 9. Resin loading vs epimerizationa
loading
under-loading normal loading over-loading
% of H+ capacity
elution yield
SR:RR
30%
70%
54:46
60%
81%
51:49
90%
92%
38:62
a Epimerization was conducted at 80-90 °C for 2.5 h to allow for maximum
epimerization at higher loading. Mass balance indicated formation of 12 in 19,
9, and 2% amounts for 30, 60, and 90% loading, respectively. Thus, prolonged
heating past epimerization under the “normal”, i.e., 50% loading, is undesirable.
Table 10. Overloaded resin exogenous acid-catalyzed
epimerization rate studya
H2SO4 concd
temperature
time, ratio
4.0 N
0.3 N
0.6 N
1.2 N
0.6 N
60 °C
80 °C
80 °C
80 °C
90 °C
RR:SR RR:SR RR:SR RR:SR RR:SR
0
23:77
24:76
-
-
-
29:71
31:69
-
36:64
37:63
27:73
30:70
31:69
34:66
37:63
39:61
-
44:56
45:55
46:54
28:72
-
30:70
38:61
43:57
-
48:52
52:48
50:50
-
24:76
37:63
42:58
-
46:54
49:51
-
-
-
-
25:75
38:62
46:54
50:50
-
-
-
-
-
-
0.5
1.0
1.5
2.0
2.5
3.0
4.0
5.0
6.0
Figure 4. Resin column.
The resin was initially regenerated to its hydrogen form
using 20% w/w sulfuric acid. The concentration of the
sulfuric acid used for regeneration was later decreased to
7% w/w to minimize precipitation of calcium sulfate in the
resin bed; the calcium being present as a result of the resin
deionizing the city water content of the core process mother
liquors during the loading process. The resin was successfully
regenerated to 96-100% of its original capacity for seven
consecutive trials. However, after the seventh column trial
the adsorbed amino alcohols could not be epimerized without
first reactivating the resin by either an acidic wash or
conducting the epimerization in 5% w/w sulfuric acid. In
this case an additional loss of product (14%) was noticed
during epimerization in the presence of sulfuric acid at 80
°C.
t1/2
322
136
74
36
24
a Resin aliquots were removed from the epimerization slurries with time as
described above. These samples were assayed by GC for RR:SR ratio as described
earlier.
ingly larger amounts of sulfuric acid to neutralize the sodium
hydroxide present. As a result, the dilevalol‚L-DBTA crystal-
lization slurries filtered poorly due to the presence and
coprecipitation of sodium sulfate, occasionally resulting in
a two-phase crystallization and filtration in the plant.
Regardless of these difficulties, these slurries were success-
fully filtered using either a rotary pressure filter or centrifuge.
In addition to some difficulties described above, several
additional factors contributed to the decision to proceed with
the batch operation recovery process for full-scale production.
Whereas the batch process cycle times were primarily mass-
Vol. 4, No. 2, 2000 / Organic Process Research & Development
•
117

Figure 5. Leakage during loading.
Figure 6. Elution recovery.
transfer dependent, the processing times associated with the
column operation were determined by the fluid mechanics
of the system. The column cycle time was significantly
extended since only very low flow rates could be achieved
even with a large pressure drop across the column. This
physical limitation was exacerbated by significant particle
attrition occurring within the resin bed over the course of
the column trials. It also became difficult to ensure complete
epimerization after several runs: this was the result of an
“overloading” condition observed at the top of the column
as well as the affinity for the resin to deionize the 15% city
water present in the mother liquors. Both factors resulted in
fewer acid sites being available on the resin to achieve
complete epimerization. Finally, the largest contributing
factor was the elution profile of the product from the column.
Between six and seven bed-volumes were required for
complete elution of the product from the column.
in the eluate volume not only improved the reproducibility
of the product crystallization, but also increased the through-
put of the process in the existing plant equipment. The main
challenge was isolating the solutions from the ion-exchange
resin. Initial attempts involved filtering the resin between
the loading, epimerization, and elution cycles; however, in
full-scale production this method would increase the potential
for personnel exposure and require additional labor and
equipment for the process. To better conduct pilot plant trials
of the batch process, 316 L stainless steel filter cylinders
were purchased and installed on a dip pipe in the resin reactor
to filter the process solutions from the resin. Since the reactor
had a large diameter with a relatively small overall height
of resin, the pressure drop when filtering through the resin
was much less severe than that in the 10 ft column, affording
much quicker removal of the process solutions for further
processing and causing a smaller amount of particle attrition.
As reported previously, the mother liquor loading opera-
tion for the batch process was temperature-dependent, so that
this was conducted between 40 and 60 °C to maintain leakage
at e4%. In this temperature range, a desired resin loading
Scale-up of the batch process was more straightforward
compared to utilizing the existing plant columns. For the
batch process only four bed-volumes of the eluants were
necessary for product recovery. This significant reduction
118
•
Vol. 4, No. 2, 2000 / Organic Process Research & Development

Table 11. Representative batch process plant results
resin
1000 L
dilevalol M/L
loaded on resin
spent
loss to n-butanol wash
loss to water wash
total leakage
176.8 kg
169.0 kg
6.3 kg (3.6%)
1.2 kg
0.27 kg
7.9 kg (4.4%)
eluate 1
eluate 2
n-butanol wash
total eluate
loss to water washes
acid from hydrolysis
total mass balance
124.6 kg
23.5 kg
6.1 kg
154.2 kg
2.0 kg
12.5 kg
175 kg (99%)
Figure 7. Resin reactor.
of between 175 g/L and 200 g/L for the batch process was
realized. After filtering the spent mother liquor, the resin
was washed with n-butanol and deionized water prior to
epimerization. Difficulties similar to the column process in
achieving complete epimerization were experienced during
the batch trials; therefore, a 5% w/w sulfuric acid wash was
added to ensure a reproducible process. Epimerization of the
amino alcohol was accomplished by agitating the resin in
deionized water for 2 to 3 h at 90 °C. An inert atmosphere
was maintained in the reactor during epimerization by
sparging nitrogen through the filter cylinders. Recovery of
the epimerized product was accomplished in two two-bed-
volume elutions with 2 N sodium hydroxide saturated with
12% v/v n-butanol at temperatures below 50 °C. Tempera-
tures above 50 °C caused degradation of the product in the
eluates. One hour for each elution was necessary to maximize
recovery of the epimerized product from the resin. The
eluates were transferred to other vessels for pH adjustment
and phase separation, while the resin was regenerated with
dilute sulfuric acid for its next cycle.
Over the course of the first 50 batches, a slow particle
attrition of the resin was observed. Initially, the majority of
the resin was in the 150-350 µm range, but by the 50th
batch almost half was smaller than 150 µm. Titration of the
regenerated resin to determine its activity was based on its
volume, and on this basis no loss in resin activity was
observed. The particle attrition resulted in a reduction in the
overall volume of resin in the reactor, due to a reduction in
the void volume per liter of resin and due to some physical
losses of resin as well. To make up for these losses,
approximately 15% additional resin was necessary to bring
the volume of resin in the reactor back to the initial vol-
ume.
Resin Reactor Design For a Semicontinuous Com-
mercial Batch Process. Having proven the cost-effectiveness
of the reproducible resin process in generating good quality
1, we focused our efforts on scaling up this process on
production scale. Vendors with expertise in fabricating the
filtration assemblies for large water deionizing systems were
contacted to assist in the design of the resin reactor. Use of
filter cylinders installed in the bottom of the reactor was
necessary to remove the liquid streams from the resin as
shown in Figure 7. The full-scale manufacturing process
required the purchase and installation of a 2000-gallon glass-
lined reactor to accommodate the resin. Rather than the
standard curved-blade impeller, a glass-lined pitched-blade
turbine agitator was installed in the reactor to improve the
mixing of the resin. A hub assembly with two radially
oriented tiers of four 1 in × 12 in filter cylinders was
fabricated for the production reactor. The filtration assembly
was supported from the bottom outlet of the reactor. A
Teflon40 annular sleeve and Teflon protection rings for the
ends of the filter cylinders were necessary to protect the glass
lining against scratches. The filtration assembly and media
were constructed in 316 L stainless steel for the start-up of
the manufacturing process.
The Manufacturing process. The plant design for the
final process includes the recovery of L-DBTA and n-butanol
as well, as shown in Figure 7. Flow rates and processing
times were balanced to maximize the throughput of both the
dilevalol and L-DBTA recovery processes. In the final
process, water was added to the spent mother liquors and
their pH adjusted to between 7 and 8 to partition the L-DBTA
into the aqueous layer for its recovery. Residual n-butanol
was removed from this aqueous stream by vacuum distilla-
tion, and the L-DBTA was recovered as a solid upon the
addition of dilute sulfuric acid. The crystallization of the
recovered L-DBTA was almost instantaneous, affording the
opportunity to operate the L-DBTA crystallizer in a semi-
continuous mode while collecting the product on a centrifuge.
The amino alcohols in n-BuOH were loaded on the column
for epimerization. The remaining n-butanol streams from the
The recovered dilevalol‚L-DBTA was crystallized inde-
pendently in a batch operation, with the product collected
and washed on a centrifuge. Due to the smaller eluate
volumes, the sodium sulfate burden was significantly reduced
in the dilevalol‚L-DBTA crystallizations from the batch
epimerization process, significantly improving the filterability
of the slurries. The mother liquors from the recovered
dilevalol‚L-DBTA crystallization were combined with those
from the core process for further epimerization. Typical plant
results are shown in Table 11.
(40) Teflon is a registered trademark of E. I. du Pont de Nemours and
Company.
Vol. 4, No. 2, 2000 / Organic Process Research & Development
•
119

Figure 8. Plant design for dilevalol recovery processes.
spent mother liquors, as well as various n-butanol washes
from the process, were collected for solvent recovery using
a wiped-film evaporator.
where higher (90%) resin-loaded amino alcohol was epimer-
ized with the help of additional acid (sulfuric acid in that
case) was successfully scaled up in the pilot plant on a 100
kg scale. Thus, with proper equipment resistant to acid
corrosion, this process could lead to enhanced throughput,
meeting the increased demands. For further improvement in
the commercial process, specifications for an alternate
filtration assembly constructed in Hastelloy C-276 were
generated. This was done to allow for implementing the
proposed process, which involved the potential use of dilute
sulfuric acid with or without the help of a controlled amount
of hydrochloric acid in the process (to take advantage of the
better solubility of sodium chloride versus sodium sulfate,
and faster epimerization). These changes were not put into
practice as a commercial decision was made not to continue
with this product.
The batch process was successfully demonstrated on a
manufacturing scale approximately 70 times, recovering
several tonnes of dilevalol‚L-DBTA which was converted
to good quality (meeting or exceeding analytical specifica-
tions set for the material obtained from the core process)
dilevalol hydrochloride salt.41 The average recovery of the
epimerized amino alcohols was 83% for the loading, epimer-
ization, elution, and pH adjustment steps. The recovered
dilevalol‚L-DBTA crystallization step yield was ∼33%,
resulting in an overall yield of ∼25%. Additionally, 83%
of the L-DBTA (expensive reagent at this stage of this
product life) present in the original mother liquors was
recovered as part of the process and recycled into the
crystallization of dilevalol‚L-DBTA. The overall plant de-
sign for the dilevalol recovery process is shown in Fig-
ure 8.
(41) Dilevalol hydrochloride was prepared from the L-DBTA salt as follows:
To a slurry of dilevalol dibenzoyl-L-tartrate in methyl-isobutyl ketone
(MIBK) and water was added aqueous NaOH to a pH of 8.4-9.8. The
MIBK layer containing the free base was combined with fresh water and
then treated with concentrated HCl to pH e 1, maintaining a temperature
of 50-60 °C. The resultant slurry was cooled to about 20 °C, filtered,
washed with MIBK and water, and dried at 75 °C. The yield was 85-95%.
Typical purity of this material was >97%, producing quality material that
passed a battery of tests.
Improved Throughput Commercial Process Plans.
Anticipating increased demand for the product and encour-
aged by the stability of the resin to dilute acid-catalyzed
epimerization, we sought to increase the throughput of this
process. The experiment described earlier in this manuscript,
120
•
Vol. 4, No. 2, 2000 / Organic Process Research & Development

In summary, through a combination of understanding the
reaction mechanism, an optimization of reagent, (a sulfonic
acid resin in this case), and appropriate modifications to the
reactor, a novel, fully reproducible, commercial epimerization
process for the preparation of dilevalol, from what would
otherwise have been an expensive waste, was developed. This
process was scaled up to prepare several tonnes of dilev-
alol.
Acknowledgment
We thank Dr. Derek Walker for his continuous encour-
agement and for many useful discussions on this project as
well as on this manuscript.
Received for review February 8, 1999.
OP990185Y
Vol. 4, No. 2, 2000 / Organic Process Research & Development
•
121